net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 143
 144 static inline int rtm_msgindex(int msgtype)
 145 {
 146         int msgindex = msgtype - RTM_BASE;
 147
 148         /*
 149          * msgindex < 0 implies someone tried to register a netlink
 150          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 151          * the message type has not been added to linux/rtnetlink.h
 152          */
 153         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 154
 155         return msgindex;
 156 }
 157
 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 159 {
 160         struct rtnl_link **tab;
 161
 162         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 163                 protocol = PF_UNSPEC;
 164
 165         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 166         if (!tab)
 167                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 168
 169         return tab[msgtype];
 170 }
 171
 172 static int rtnl_register_internal(struct module *owner,
 173                                   int protocol, int msgtype,
 174                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 175                                   unsigned int flags)
 176 {
 177         struct rtnl_link *link, *old;
 178         struct rtnl_link __rcu **tab;
 179         int msgindex;
 180         int ret = -ENOBUFS;
 181
 182         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 183         msgindex = rtm_msgindex(msgtype);
 184
 185         rtnl_lock();
 186         tab = rtnl_msg_handlers[protocol];
 187         if (tab == NULL) {
 188                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 189                 if (!tab)
 190                         goto unlock;
 191
 192                 /* ensures we see the 0 stores */
 193                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 194         }
 195
 196         old = rtnl_dereference(tab[msgindex]);
 197         if (old) {
 198                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         } else {
 202                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 203                 if (!link)
 204                         goto unlock;
 205         }
 206
 207         WARN_ON(link->owner && link->owner != owner);
 208         link->owner = owner;
 209
 210         WARN_ON(doit && link->doit && link->doit != doit);
 211         if (doit)
 212                 link->doit = doit;
 213         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 214         if (dumpit)
 215                 link->dumpit = dumpit;
 216
 217         link->flags |= flags;
 218
 219         /* publish protocol:msgtype */
 220         rcu_assign_pointer(tab[msgindex], link);
 221         ret = 0;
 222         if (old)
 223                 kfree_rcu(old, rcu);
 224 unlock:
 225         rtnl_unlock();
 226         return ret;
 227 }
 228
 229 /**
 230  * rtnl_register_module - Register a rtnetlink message type
 231  *
 232  * @owner: module registering the hook (THIS_MODULE)
 233  * @protocol: Protocol family or PF_UNSPEC
 234  * @msgtype: rtnetlink message type
 235  * @doit: Function pointer called for each request message
 236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link **tab, *link;
 290         int msgindex;
 291
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293         msgindex = rtm_msgindex(msgtype);
 294
 295         rtnl_lock();
 296         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 297         if (!tab) {
 298                 rtnl_unlock();
 299                 return -ENOENT;
 300         }
 301
 302         link = tab[msgindex];
 303         rcu_assign_pointer(tab[msgindex], NULL);
 304         rtnl_unlock();
 305
 306         kfree_rcu(link, rcu);
 307
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_unregister);
 311
 312 /**
 313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 314  * @protocol : Protocol family or PF_UNSPEC
 315  *
 316  * Identical to calling rtnl_unregster() for all registered message types
 317  * of a certain protocol family.
 318  */
 319 void rtnl_unregister_all(int protocol)
 320 {
 321         struct rtnl_link **tab, *link;
 322         int msgindex;
 323
 324         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 325
 326         rtnl_lock();
 327         tab = rtnl_msg_handlers[protocol];
 328         if (!tab) {
 329                 rtnl_unlock();
 330                 return;
 331         }
 332         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 333         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 334                 link = tab[msgindex];
 335                 if (!link)
 336                         continue;
 337
 338                 rcu_assign_pointer(tab[msgindex], NULL);
 339                 kfree_rcu(link, rcu);
 340         }
 341         rtnl_unlock();
 342
 343         synchronize_net();
 344
 345         kfree(tab);
 346 }
 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 348
 349 static LIST_HEAD(link_ops);
 350
 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 352 {
 353         const struct rtnl_link_ops *ops;
 354
 355         list_for_each_entry(ops, &link_ops, list) {
 356                 if (!strcmp(ops->kind, kind))
 357                         return ops;
 358         }
 359         return NULL;
 360 }
 361
 362 /**
 363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 364  * @ops: struct rtnl_link_ops * to register
 365  *
 366  * The caller must hold the rtnl_mutex. This function should be used
 367  * by drivers that create devices during module initialization. It
 368  * must be called before registering the devices.
 369  *
 370  * Returns 0 on success or a negative error code.
 371  */
 372 int __rtnl_link_register(struct rtnl_link_ops *ops)
 373 {
 374         if (rtnl_link_ops_get(ops->kind))
 375                 return -EEXIST;
 376
 377         /* The check for setup is here because if ops
 378          * does not have that filled up, it is not possible
 379          * to use the ops for creating device. So do not
 380          * fill up dellink as well. That disables rtnl_dellink.
 381          */
 382         if (ops->setup && !ops->dellink)
 383                 ops->dellink = unregister_netdevice_queue;
 384
 385         list_add_tail(&ops->list, &link_ops);
 386         return 0;
 387 }
 388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 389
 390 /**
 391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 392  * @ops: struct rtnl_link_ops * to register
 393  *
 394  * Returns 0 on success or a negative error code.
 395  */
 396 int rtnl_link_register(struct rtnl_link_ops *ops)
 397 {
 398         int err;
 399
 400         /* Sanity-check max sizes to avoid stack buffer overflow. */
 401         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 402                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 403                 return -EINVAL;
 404
 405         rtnl_lock();
 406         err = __rtnl_link_register(ops);
 407         rtnl_unlock();
 408         return err;
 409 }
 410 EXPORT_SYMBOL_GPL(rtnl_link_register);
 411
 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 413 {
 414         struct net_device *dev;
 415         LIST_HEAD(list_kill);
 416
 417         for_each_netdev(net, dev) {
 418                 if (dev->rtnl_link_ops == ops)
 419                         ops->dellink(dev, &list_kill);
 420         }
 421         unregister_netdevice_many(&list_kill);
 422 }
 423
 424 /**
 425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 426  * @ops: struct rtnl_link_ops * to unregister
 427  *
 428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 429  * integrity (hold pernet_ops_rwsem for writing to close the race
 430  * with setup_net() and cleanup_net()).
 431  */
 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 433 {
 434         struct net *net;
 435
 436         for_each_net(net) {
 437                 __rtnl_kill_links(net, ops);
 438         }
 439         list_del(&ops->list);
 440 }
 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 442
 443 /* Return with the rtnl_lock held when there are no network
 444  * devices unregistering in any network namespace.
 445  */
 446 static void rtnl_lock_unregistering_all(void)
 447 {
 448         struct net *net;
 449         bool unregistering;
 450         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 451
 452         add_wait_queue(&netdev_unregistering_wq, &wait);
 453         for (;;) {
 454                 unregistering = false;
 455                 rtnl_lock();
 456                 /* We held write locked pernet_ops_rwsem, and parallel
 457                  * setup_net() and cleanup_net() are not possible.
 458                  */
 459                 for_each_net(net) {
 460                         if (net->dev_unreg_count > 0) {
 461                                 unregistering = true;
 462                                 break;
 463                         }
 464                 }
 465                 if (!unregistering)
 466                         break;
 467                 __rtnl_unlock();
 468
 469                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 470         }
 471         remove_wait_queue(&netdev_unregistering_wq, &wait);
 472 }
 473
 474 /**
 475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 476  * @ops: struct rtnl_link_ops * to unregister
 477  */
 478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 479 {
 480         /* Close the race with setup_net() and cleanup_net() */
 481         down_write(&pernet_ops_rwsem);
 482         rtnl_lock_unregistering_all();
 483         __rtnl_link_unregister(ops);
 484         rtnl_unlock();
 485         up_write(&pernet_ops_rwsem);
 486 }
 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 488
 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 490 {
 491         struct net_device *master_dev;
 492         const struct rtnl_link_ops *ops;
 493         size_t size = 0;
 494
 495         rcu_read_lock();
 496
 497         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 498         if (!master_dev)
 499                 goto out;
 500
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops || !ops->get_slave_size)
 503                 goto out;
 504         /* IFLA_INFO_SLAVE_DATA + nested data */
 505         size = nla_total_size(sizeof(struct nlattr)) +
 506                ops->get_slave_size(master_dev, dev);
 507
 508 out:
 509         rcu_read_unlock();
 510         return size;
 511 }
 512
 513 static size_t rtnl_link_get_size(const struct net_device *dev)
 514 {
 515         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 516         size_t size;
 517
 518         if (!ops)
 519                 return 0;
 520
 521         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 522                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 523
 524         if (ops->get_size)
 525                 /* IFLA_INFO_DATA + nested data */
 526                 size += nla_total_size(sizeof(struct nlattr)) +
 527                         ops->get_size(dev);
 528
 529         if (ops->get_xstats_size)
 530                 /* IFLA_INFO_XSTATS */
 531                 size += nla_total_size(ops->get_xstats_size(dev));
 532
 533         size += rtnl_link_get_slave_info_data_size(dev);
 534
 535         return size;
 536 }
 537
 538 static LIST_HEAD(rtnl_af_ops);
 539
 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 541 {
 542         const struct rtnl_af_ops *ops;
 543
 544         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 545                 if (ops->family == family)
 546                         return ops;
 547         }
 548
 549         return NULL;
 550 }
 551
 552 /**
 553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 554  * @ops: struct rtnl_af_ops * to register
 555  *
 556  * Returns 0 on success or a negative error code.
 557  */
 558 void rtnl_af_register(struct rtnl_af_ops *ops)
 559 {
 560         rtnl_lock();
 561         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 562         rtnl_unlock();
 563 }
 564 EXPORT_SYMBOL_GPL(rtnl_af_register);
 565
 566 /**
 567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 568  * @ops: struct rtnl_af_ops * to unregister
 569  */
 570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 571 {
 572         rtnl_lock();
 573         list_del_rcu(&ops->list);
 574         rtnl_unlock();
 575
 576         synchronize_rcu();
 577 }
 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 579
 580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 581                                     u32 ext_filter_mask)
 582 {
 583         struct rtnl_af_ops *af_ops;
 584         size_t size;
 585
 586         /* IFLA_AF_SPEC */
 587         size = nla_total_size(sizeof(struct nlattr));
 588
 589         rcu_read_lock();
 590         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 591                 if (af_ops->get_link_af_size) {
 592                         /* AF_* + nested data */
 593                         size += nla_total_size(sizeof(struct nlattr)) +
 594                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 595                 }
 596         }
 597         rcu_read_unlock();
 598
 599         return size;
 600 }
 601
 602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 603 {
 604         struct net_device *master_dev;
 605         bool ret = false;
 606
 607         rcu_read_lock();
 608
 609         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 610         if (master_dev && master_dev->rtnl_link_ops)
 611                 ret = true;
 612         rcu_read_unlock();
 613         return ret;
 614 }
 615
 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 617                                      const struct net_device *dev)
 618 {
 619         struct net_device *master_dev;
 620         const struct rtnl_link_ops *ops;
 621         struct nlattr *slave_data;
 622         int err;
 623
 624         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 625         if (!master_dev)
 626                 return 0;
 627         ops = master_dev->rtnl_link_ops;
 628         if (!ops)
 629                 return 0;
 630         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 631                 return -EMSGSIZE;
 632         if (ops->fill_slave_info) {
 633                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 634                 if (!slave_data)
 635                         return -EMSGSIZE;
 636                 err = ops->fill_slave_info(skb, master_dev, dev);
 637                 if (err < 0)
 638                         goto err_cancel_slave_data;
 639                 nla_nest_end(skb, slave_data);
 640         }
 641         return 0;
 642
 643 err_cancel_slave_data:
 644         nla_nest_cancel(skb, slave_data);
 645         return err;
 646 }
 647
 648 static int rtnl_link_info_fill(struct sk_buff *skb,
 649                                const struct net_device *dev)
 650 {
 651         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 652         struct nlattr *data;
 653         int err;
 654
 655         if (!ops)
 656                 return 0;
 657         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 658                 return -EMSGSIZE;
 659         if (ops->fill_xstats) {
 660                 err = ops->fill_xstats(skb, dev);
 661                 if (err < 0)
 662                         return err;
 663         }
 664         if (ops->fill_info) {
 665                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 666                 if (data == NULL)
 667                         return -EMSGSIZE;
 668                 err = ops->fill_info(skb, dev);
 669                 if (err < 0)
 670                         goto err_cancel_data;
 671                 nla_nest_end(skb, data);
 672         }
 673         return 0;
 674
 675 err_cancel_data:
 676         nla_nest_cancel(skb, data);
 677         return err;
 678 }
 679
 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 681 {
 682         struct nlattr *linkinfo;
 683         int err = -EMSGSIZE;
 684
 685         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 686         if (linkinfo == NULL)
 687                 goto out;
 688
 689         err = rtnl_link_info_fill(skb, dev);
 690         if (err < 0)
 691                 goto err_cancel_link;
 692
 693         err = rtnl_link_slave_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         nla_nest_end(skb, linkinfo);
 698         return 0;
 699
 700 err_cancel_link:
 701         nla_nest_cancel(skb, linkinfo);
 702 out:
 703         return err;
 704 }
 705
 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 707 {
 708         struct sock *rtnl = net->rtnl;
 709         int err = 0;
 710
 711         NETLINK_CB(skb).dst_group = group;
 712         if (echo)
 713                 refcount_inc(&skb->users);
 714         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 715         if (echo)
 716                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 717         return err;
 718 }
 719
 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 721 {
 722         struct sock *rtnl = net->rtnl;
 723
 724         return nlmsg_unicast(rtnl, skb, pid);
 725 }
 726 EXPORT_SYMBOL(rtnl_unicast);
 727
 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 729                  struct nlmsghdr *nlh, gfp_t flags)
 730 {
 731         struct sock *rtnl = net->rtnl;
 732         int report = 0;
 733
 734         if (nlh)
 735                 report = nlmsg_report(nlh);
 736
 737         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 738 }
 739 EXPORT_SYMBOL(rtnl_notify);
 740
 741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 742 {
 743         struct sock *rtnl = net->rtnl;
 744
 745         netlink_set_err(rtnl, 0, group, error);
 746 }
 747 EXPORT_SYMBOL(rtnl_set_sk_err);
 748
 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 750 {
 751         struct nlattr *mx;
 752         int i, valid = 0;
 753
 754         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 755         if (metrics == dst_default_metrics.metrics)
 756                 return 0;
 757
 758         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 759         if (mx == NULL)
 760                 return -ENOBUFS;
 761
 762         for (i = 0; i < RTAX_MAX; i++) {
 763                 if (metrics[i]) {
 764                         if (i == RTAX_CC_ALGO - 1) {
 765                                 char tmp[TCP_CA_NAME_MAX], *name;
 766
 767                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 768                                 if (!name)
 769                                         continue;
 770                                 if (nla_put_string(skb, i + 1, name))
 771                                         goto nla_put_failure;
 772                         } else if (i == RTAX_FEATURES - 1) {
 773                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 774
 775                                 if (!user_features)
 776                                         continue;
 777                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 778                                 if (nla_put_u32(skb, i + 1, user_features))
 779                                         goto nla_put_failure;
 780                         } else {
 781                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 782                                         goto nla_put_failure;
 783                         }
 784                         valid++;
 785                 }
 786         }
 787
 788         if (!valid) {
 789                 nla_nest_cancel(skb, mx);
 790                 return 0;
 791         }
 792
 793         return nla_nest_end(skb, mx);
 794
 795 nla_put_failure:
 796         nla_nest_cancel(skb, mx);
 797         return -EMSGSIZE;
 798 }
 799 EXPORT_SYMBOL(rtnetlink_put_metrics);
 800
 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 802                        long expires, u32 error)
 803 {
 804         struct rta_cacheinfo ci = {
 805                 .rta_error = error,
 806                 .rta_id =  id,
 807         };
 808
 809         if (dst) {
 810                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 811                 ci.rta_used = dst->__use;
 812                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 813         }
 814         if (expires) {
 815                 unsigned long clock;
 816
 817                 clock = jiffies_to_clock_t(abs(expires));
 818                 clock = min_t(unsigned long, clock, INT_MAX);
 819                 ci.rta_expires = (expires > 0) ? clock : -clock;
 820         }
 821         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 822 }
 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 824
 825 static void set_operstate(struct net_device *dev, unsigned char transition)
 826 {
 827         unsigned char operstate = dev->operstate;
 828
 829         switch (transition) {
 830         case IF_OPER_UP:
 831                 if ((operstate == IF_OPER_DORMANT ||
 832                      operstate == IF_OPER_UNKNOWN) &&
 833                     !netif_dormant(dev))
 834                         operstate = IF_OPER_UP;
 835                 break;
 836
 837         case IF_OPER_DORMANT:
 838                 if (operstate == IF_OPER_UP ||
 839                     operstate == IF_OPER_UNKNOWN)
 840                         operstate = IF_OPER_DORMANT;
 841                 break;
 842         }
 843
 844         if (dev->operstate != operstate) {
 845                 write_lock_bh(&dev_base_lock);
 846                 dev->operstate = operstate;
 847                 write_unlock_bh(&dev_base_lock);
 848                 netdev_state_change(dev);
 849         }
 850 }
 851
 852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 853 {
 854         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 855                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 856 }
 857
 858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 859                                            const struct ifinfomsg *ifm)
 860 {
 861         unsigned int flags = ifm->ifi_flags;
 862
 863         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 864         if (ifm->ifi_change)
 865                 flags = (flags & ifm->ifi_change) |
 866                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 867
 868         return flags;
 869 }
 870
 871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 872                                  const struct rtnl_link_stats64 *b)
 873 {
 874         a->rx_packets = b->rx_packets;
 875         a->tx_packets = b->tx_packets;
 876         a->rx_bytes = b->rx_bytes;
 877         a->tx_bytes = b->tx_bytes;
 878         a->rx_errors = b->rx_errors;
 879         a->tx_errors = b->tx_errors;
 880         a->rx_dropped = b->rx_dropped;
 881         a->tx_dropped = b->tx_dropped;
 882
 883         a->multicast = b->multicast;
 884         a->collisions = b->collisions;
 885
 886         a->rx_length_errors = b->rx_length_errors;
 887         a->rx_over_errors = b->rx_over_errors;
 888         a->rx_crc_errors = b->rx_crc_errors;
 889         a->rx_frame_errors = b->rx_frame_errors;
 890         a->rx_fifo_errors = b->rx_fifo_errors;
 891         a->rx_missed_errors = b->rx_missed_errors;
 892
 893         a->tx_aborted_errors = b->tx_aborted_errors;
 894         a->tx_carrier_errors = b->tx_carrier_errors;
 895         a->tx_fifo_errors = b->tx_fifo_errors;
 896         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 897         a->tx_window_errors = b->tx_window_errors;
 898
 899         a->rx_compressed = b->rx_compressed;
 900         a->tx_compressed = b->tx_compressed;
 901
 902         a->rx_nohandler = b->rx_nohandler;
 903 }
 904
 905 /* All VF info */
 906 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 907                                    u32 ext_filter_mask)
 908 {
 909         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 910                 int num_vfs = dev_num_vf(dev->dev.parent);
 911                 size_t size = nla_total_size(0);
 912                 size += num_vfs *
 913                         (nla_total_size(0) +
 914                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 915                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 916                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 917                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 918                          nla_total_size(MAX_VLAN_LIST_LEN *
 919                                         sizeof(struct ifla_vf_vlan_info)) +
 920                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 921                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 922                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 923                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 924                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 925                          nla_total_size(sizeof(struct ifla_vf_trust)));
 926                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 927                         size += num_vfs *
 928                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 929                                  /* IFLA_VF_STATS_RX_PACKETS */
 930                                  nla_total_size_64bit(sizeof(__u64)) +
 931                                  /* IFLA_VF_STATS_TX_PACKETS */
 932                                  nla_total_size_64bit(sizeof(__u64)) +
 933                                  /* IFLA_VF_STATS_RX_BYTES */
 934                                  nla_total_size_64bit(sizeof(__u64)) +
 935                                  /* IFLA_VF_STATS_TX_BYTES */
 936                                  nla_total_size_64bit(sizeof(__u64)) +
 937                                  /* IFLA_VF_STATS_BROADCAST */
 938                                  nla_total_size_64bit(sizeof(__u64)) +
 939                                  /* IFLA_VF_STATS_MULTICAST */
 940                                  nla_total_size_64bit(sizeof(__u64)) +
 941                                  /* IFLA_VF_STATS_RX_DROPPED */
 942                                  nla_total_size_64bit(sizeof(__u64)) +
 943                                  /* IFLA_VF_STATS_TX_DROPPED */
 944                                  nla_total_size_64bit(sizeof(__u64)));
 945                 }
 946                 return size;
 947         } else
 948                 return 0;
 949 }
 950
 951 static size_t rtnl_port_size(const struct net_device *dev,
 952                              u32 ext_filter_mask)
 953 {
 954         size_t port_size = nla_total_size(4)            /* PORT_VF */
 955                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 956                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 957                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 958                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 959                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 960         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 961         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 962                 + port_size;
 963         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 964                 + port_size;
 965
 966         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 967             !(ext_filter_mask & RTEXT_FILTER_VF))
 968                 return 0;
 969         if (dev_num_vf(dev->dev.parent))
 970                 return port_self_size + vf_ports_size +
 971                         vf_port_size * dev_num_vf(dev->dev.parent);
 972         else
 973                 return port_self_size;
 974 }
 975
 976 static size_t rtnl_xdp_size(void)
 977 {
 978         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 979                           nla_total_size(1) +   /* XDP_ATTACHED */
 980                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 981                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 982
 983         return xdp_size;
 984 }
 985
 986 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 987                                      u32 ext_filter_mask)
 988 {
 989         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 990                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 991                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 992                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 993                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 994                + nla_total_size(sizeof(struct rtnl_link_stats))
 995                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 996                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 997                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 998                + nla_total_size(4) /* IFLA_TXQLEN */
 999                + nla_total_size(4) /* IFLA_WEIGHT */
1000                + nla_total_size(4) /* IFLA_MTU */
1001                + nla_total_size(4) /* IFLA_LINK */
1002                + nla_total_size(4) /* IFLA_MASTER */
1003                + nla_total_size(1) /* IFLA_CARRIER */
1004                + nla_total_size(4) /* IFLA_PROMISCUITY */
1005                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1006                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1007                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1008                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1009                + nla_total_size(1) /* IFLA_OPERSTATE */
1010                + nla_total_size(1) /* IFLA_LINKMODE */
1011                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1012                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1013                + nla_total_size(4) /* IFLA_GROUP */
1014                + nla_total_size(ext_filter_mask
1015                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1016                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1017                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1018                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1019                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1020                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1021                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1022                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1023                + rtnl_xdp_size() /* IFLA_XDP */
1024                + nla_total_size(4)  /* IFLA_EVENT */
1025                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1026                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1027                + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1028                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1029                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1030                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1031                + nla_total_size(4)  /* IFLA_MIN_MTU */
1032                + nla_total_size(4)  /* IFLA_MAX_MTU */
1033                + 0;
1034 }
1035
1036 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1037 {
1038         struct nlattr *vf_ports;
1039         struct nlattr *vf_port;
1040         int vf;
1041         int err;
1042
1043         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1044         if (!vf_ports)
1045                 return -EMSGSIZE;
1046
1047         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1048                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1049                 if (!vf_port)
1050                         goto nla_put_failure;
1051                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1052                         goto nla_put_failure;
1053                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1054                 if (err == -EMSGSIZE)
1055                         goto nla_put_failure;
1056                 if (err) {
1057                         nla_nest_cancel(skb, vf_port);
1058                         continue;
1059                 }
1060                 nla_nest_end(skb, vf_port);
1061         }
1062
1063         nla_nest_end(skb, vf_ports);
1064
1065         return 0;
1066
1067 nla_put_failure:
1068         nla_nest_cancel(skb, vf_ports);
1069         return -EMSGSIZE;
1070 }
1071
1072 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1073 {
1074         struct nlattr *port_self;
1075         int err;
1076
1077         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1078         if (!port_self)
1079                 return -EMSGSIZE;
1080
1081         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1082         if (err) {
1083                 nla_nest_cancel(skb, port_self);
1084                 return (err == -EMSGSIZE) ? err : 0;
1085         }
1086
1087         nla_nest_end(skb, port_self);
1088
1089         return 0;
1090 }
1091
1092 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1093                           u32 ext_filter_mask)
1094 {
1095         int err;
1096
1097         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1098             !(ext_filter_mask & RTEXT_FILTER_VF))
1099                 return 0;
1100
1101         err = rtnl_port_self_fill(skb, dev);
1102         if (err)
1103                 return err;
1104
1105         if (dev_num_vf(dev->dev.parent)) {
1106                 err = rtnl_vf_ports_fill(skb, dev);
1107                 if (err)
1108                         return err;
1109         }
1110
1111         return 0;
1112 }
1113
1114 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1115 {
1116         int err;
1117         struct netdev_phys_item_id ppid;
1118
1119         err = dev_get_phys_port_id(dev, &ppid);
1120         if (err) {
1121                 if (err == -EOPNOTSUPP)
1122                         return 0;
1123                 return err;
1124         }
1125
1126         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1127                 return -EMSGSIZE;
1128
1129         return 0;
1130 }
1131
1132 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1133 {
1134         char name[IFNAMSIZ];
1135         int err;
1136
1137         err = dev_get_phys_port_name(dev, name, sizeof(name));
1138         if (err) {
1139                 if (err == -EOPNOTSUPP)
1140                         return 0;
1141                 return err;
1142         }
1143
1144         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1145                 return -EMSGSIZE;
1146
1147         return 0;
1148 }
1149
1150 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1151 {
1152         struct netdev_phys_item_id ppid = { };
1153         int err;
1154
1155         err = dev_get_port_parent_id(dev, &ppid, false);
1156         if (err) {
1157                 if (err == -EOPNOTSUPP)
1158                         return 0;
1159                 return err;
1160         }
1161
1162         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1163                 return -EMSGSIZE;
1164
1165         return 0;
1166 }
1167
1168 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1169                                               struct net_device *dev)
1170 {
1171         struct rtnl_link_stats64 *sp;
1172         struct nlattr *attr;
1173
1174         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1175                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1176         if (!attr)
1177                 return -EMSGSIZE;
1178
1179         sp = nla_data(attr);
1180         dev_get_stats(dev, sp);
1181
1182         attr = nla_reserve(skb, IFLA_STATS,
1183                            sizeof(struct rtnl_link_stats));
1184         if (!attr)
1185                 return -EMSGSIZE;
1186
1187         copy_rtnl_link_stats(nla_data(attr), sp);
1188
1189         return 0;
1190 }
1191
1192 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1193                                                struct net_device *dev,
1194                                                int vfs_num,
1195                                                struct nlattr *vfinfo,
1196                                                u32 ext_filter_mask)
1197 {
1198         struct ifla_vf_rss_query_en vf_rss_query_en;
1199         struct nlattr *vf, *vfstats, *vfvlanlist;
1200         struct ifla_vf_link_state vf_linkstate;
1201         struct ifla_vf_vlan_info vf_vlan_info;
1202         struct ifla_vf_spoofchk vf_spoofchk;
1203         struct ifla_vf_tx_rate vf_tx_rate;
1204         struct ifla_vf_stats vf_stats;
1205         struct ifla_vf_trust vf_trust;
1206         struct ifla_vf_vlan vf_vlan;
1207         struct ifla_vf_rate vf_rate;
1208         struct ifla_vf_mac vf_mac;
1209         struct ifla_vf_broadcast vf_broadcast;
1210         struct ifla_vf_info ivi;
1211
1212         memset(&ivi, 0, sizeof(ivi));
1213
1214         /* Not all SR-IOV capable drivers support the
1215          * spoofcheck and "RSS query enable" query.  Preset to
1216          * -1 so the user space tool can detect that the driver
1217          * didn't report anything.
1218          */
1219         ivi.spoofchk = -1;
1220         ivi.rss_query_en = -1;
1221         ivi.trusted = -1;
1222         /* The default value for VF link state is "auto"
1223          * IFLA_VF_LINK_STATE_AUTO which equals zero
1224          */
1225         ivi.linkstate = 0;
1226         /* VLAN Protocol by default is 802.1Q */
1227         ivi.vlan_proto = htons(ETH_P_8021Q);
1228         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1229                 return 0;
1230
1231         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1232
1233         vf_mac.vf =
1234                 vf_vlan.vf =
1235                 vf_vlan_info.vf =
1236                 vf_rate.vf =
1237                 vf_tx_rate.vf =
1238                 vf_spoofchk.vf =
1239                 vf_linkstate.vf =
1240                 vf_rss_query_en.vf =
1241                 vf_trust.vf = ivi.vf;
1242
1243         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1244         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1245         vf_vlan.vlan = ivi.vlan;
1246         vf_vlan.qos = ivi.qos;
1247         vf_vlan_info.vlan = ivi.vlan;
1248         vf_vlan_info.qos = ivi.qos;
1249         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1250         vf_tx_rate.rate = ivi.max_tx_rate;
1251         vf_rate.min_tx_rate = ivi.min_tx_rate;
1252         vf_rate.max_tx_rate = ivi.max_tx_rate;
1253         vf_spoofchk.setting = ivi.spoofchk;
1254         vf_linkstate.link_state = ivi.linkstate;
1255         vf_rss_query_en.setting = ivi.rss_query_en;
1256         vf_trust.setting = ivi.trusted;
1257         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1258         if (!vf)
1259                 goto nla_put_vfinfo_failure;
1260         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1261             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1262             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1263             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1264                     &vf_rate) ||
1265             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1266                     &vf_tx_rate) ||
1267             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1268                     &vf_spoofchk) ||
1269             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1270                     &vf_linkstate) ||
1271             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1272                     sizeof(vf_rss_query_en),
1273                     &vf_rss_query_en) ||
1274             nla_put(skb, IFLA_VF_TRUST,
1275                     sizeof(vf_trust), &vf_trust))
1276                 goto nla_put_vf_failure;
1277         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1278         if (!vfvlanlist)
1279                 goto nla_put_vf_failure;
1280         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1281                     &vf_vlan_info)) {
1282                 nla_nest_cancel(skb, vfvlanlist);
1283                 goto nla_put_vf_failure;
1284         }
1285         nla_nest_end(skb, vfvlanlist);
1286         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1287                 memset(&vf_stats, 0, sizeof(vf_stats));
1288                 if (dev->netdev_ops->ndo_get_vf_stats)
1289                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1290                                                           &vf_stats);
1291                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1292                 if (!vfstats)
1293                         goto nla_put_vf_failure;
1294                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1295                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1296                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1297                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1298                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1299                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1300                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1301                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1302                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1303                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1304                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1305                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1306                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1307                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1308                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1309                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1310                         nla_nest_cancel(skb, vfstats);
1311                         goto nla_put_vf_failure;
1312                 }
1313                 nla_nest_end(skb, vfstats);
1314         }
1315         nla_nest_end(skb, vf);
1316         return 0;
1317
1318 nla_put_vf_failure:
1319         nla_nest_cancel(skb, vf);
1320 nla_put_vfinfo_failure:
1321         nla_nest_cancel(skb, vfinfo);
1322         return -EMSGSIZE;
1323 }
1324
1325 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1326                                            struct net_device *dev,
1327                                            u32 ext_filter_mask)
1328 {
1329         struct nlattr *vfinfo;
1330         int i, num_vfs;
1331
1332         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1333                 return 0;
1334
1335         num_vfs = dev_num_vf(dev->dev.parent);
1336         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1337                 return -EMSGSIZE;
1338
1339         if (!dev->netdev_ops->ndo_get_vf_config)
1340                 return 0;
1341
1342         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1343         if (!vfinfo)
1344                 return -EMSGSIZE;
1345
1346         for (i = 0; i < num_vfs; i++) {
1347                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo, ext_filter_mask))
1348                         return -EMSGSIZE;
1349         }
1350
1351         nla_nest_end(skb, vfinfo);
1352         return 0;
1353 }
1354
1355 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1356 {
1357         struct rtnl_link_ifmap map;
1358
1359         memset(&map, 0, sizeof(map));
1360         map.mem_start   = dev->mem_start;
1361         map.mem_end     = dev->mem_end;
1362         map.base_addr   = dev->base_addr;
1363         map.irq         = dev->irq;
1364         map.dma         = dev->dma;
1365         map.port        = dev->if_port;
1366
1367         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1368                 return -EMSGSIZE;
1369
1370         return 0;
1371 }
1372
1373 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1374 {
1375         const struct bpf_prog *generic_xdp_prog;
1376
1377         ASSERT_RTNL();
1378
1379         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1380         if (!generic_xdp_prog)
1381                 return 0;
1382         return generic_xdp_prog->aux->id;
1383 }
1384
1385 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1386 {
1387         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1388 }
1389
1390 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1391 {
1392         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1393                                XDP_QUERY_PROG_HW);
1394 }
1395
1396 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1397                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1398                                u32 (*get_prog_id)(struct net_device *dev))
1399 {
1400         u32 curr_id;
1401         int err;
1402
1403         curr_id = get_prog_id(dev);
1404         if (!curr_id)
1405                 return 0;
1406
1407         *prog_id = curr_id;
1408         err = nla_put_u32(skb, attr, curr_id);
1409         if (err)
1410                 return err;
1411
1412         if (*mode != XDP_ATTACHED_NONE)
1413                 *mode = XDP_ATTACHED_MULTI;
1414         else
1415                 *mode = tgt_mode;
1416
1417         return 0;
1418 }
1419
1420 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1421 {
1422         struct nlattr *xdp;
1423         u32 prog_id;
1424         int err;
1425         u8 mode;
1426
1427         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1428         if (!xdp)
1429                 return -EMSGSIZE;
1430
1431         prog_id = 0;
1432         mode = XDP_ATTACHED_NONE;
1433         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1434                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1435         if (err)
1436                 goto err_cancel;
1437         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1438                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1439         if (err)
1440                 goto err_cancel;
1441         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1442                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1443         if (err)
1444                 goto err_cancel;
1445
1446         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1447         if (err)
1448                 goto err_cancel;
1449
1450         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1451                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1452                 if (err)
1453                         goto err_cancel;
1454         }
1455
1456         nla_nest_end(skb, xdp);
1457         return 0;
1458
1459 err_cancel:
1460         nla_nest_cancel(skb, xdp);
1461         return err;
1462 }
1463
1464 static u32 rtnl_get_event(unsigned long event)
1465 {
1466         u32 rtnl_event_type = IFLA_EVENT_NONE;
1467
1468         switch (event) {
1469         case NETDEV_REBOOT:
1470                 rtnl_event_type = IFLA_EVENT_REBOOT;
1471                 break;
1472         case NETDEV_FEAT_CHANGE:
1473                 rtnl_event_type = IFLA_EVENT_FEATURES;
1474                 break;
1475         case NETDEV_BONDING_FAILOVER:
1476                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1477                 break;
1478         case NETDEV_NOTIFY_PEERS:
1479                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1480                 break;
1481         case NETDEV_RESEND_IGMP:
1482                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1483                 break;
1484         case NETDEV_CHANGEINFODATA:
1485                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1486                 break;
1487         default:
1488                 break;
1489         }
1490
1491         return rtnl_event_type;
1492 }
1493
1494 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1495 {
1496         const struct net_device *upper_dev;
1497         int ret = 0;
1498
1499         rcu_read_lock();
1500
1501         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1502         if (upper_dev)
1503                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1504
1505         rcu_read_unlock();
1506         return ret;
1507 }
1508
1509 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1510                           bool force)
1511 {
1512         int ifindex = dev_get_iflink(dev);
1513
1514         if (force || dev->ifindex != ifindex)
1515                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1516
1517         return 0;
1518 }
1519
1520 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1521                                               struct net_device *dev)
1522 {
1523         char buf[IFALIASZ];
1524         int ret;
1525
1526         ret = dev_get_alias(dev, buf, sizeof(buf));
1527         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1528 }
1529
1530 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1531                                   const struct net_device *dev,
1532                                   struct net *src_net, gfp_t gfp)
1533 {
1534         bool put_iflink = false;
1535
1536         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1537                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1538
1539                 if (!net_eq(dev_net(dev), link_net)) {
1540                         int id = peernet2id_alloc(src_net, link_net, gfp);
1541
1542                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1543                                 return -EMSGSIZE;
1544
1545                         put_iflink = true;
1546                 }
1547         }
1548
1549         return nla_put_iflink(skb, dev, put_iflink);
1550 }
1551
1552 static int rtnl_fill_link_af(struct sk_buff *skb,
1553                              const struct net_device *dev,
1554                              u32 ext_filter_mask)
1555 {
1556         const struct rtnl_af_ops *af_ops;
1557         struct nlattr *af_spec;
1558
1559         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1560         if (!af_spec)
1561                 return -EMSGSIZE;
1562
1563         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1564                 struct nlattr *af;
1565                 int err;
1566
1567                 if (!af_ops->fill_link_af)
1568                         continue;
1569
1570                 af = nla_nest_start_noflag(skb, af_ops->family);
1571                 if (!af)
1572                         return -EMSGSIZE;
1573
1574                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1575                 /*
1576                  * Caller may return ENODATA to indicate that there
1577                  * was no data to be dumped. This is not an error, it
1578                  * means we should trim the attribute header and
1579                  * continue.
1580                  */
1581                 if (err == -ENODATA)
1582                         nla_nest_cancel(skb, af);
1583                 else if (err < 0)
1584                         return -EMSGSIZE;
1585
1586                 nla_nest_end(skb, af);
1587         }
1588
1589         nla_nest_end(skb, af_spec);
1590         return 0;
1591 }
1592
1593 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1594                             struct net_device *dev, struct net *src_net,
1595                             int type, u32 pid, u32 seq, u32 change,
1596                             unsigned int flags, u32 ext_filter_mask,
1597                             u32 event, int *new_nsid, int new_ifindex,
1598                             int tgt_netnsid, gfp_t gfp)
1599 {
1600         struct ifinfomsg *ifm;
1601         struct nlmsghdr *nlh;
1602         struct Qdisc *qdisc;
1603
1604         ASSERT_RTNL();
1605         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1606         if (nlh == NULL)
1607                 return -EMSGSIZE;
1608
1609         ifm = nlmsg_data(nlh);
1610         ifm->ifi_family = AF_UNSPEC;
1611         ifm->__ifi_pad = 0;
1612         ifm->ifi_type = dev->type;
1613         ifm->ifi_index = dev->ifindex;
1614         ifm->ifi_flags = dev_get_flags(dev);
1615         ifm->ifi_change = change;
1616
1617         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1618                 goto nla_put_failure;
1619
1620         qdisc = rtnl_dereference(dev->qdisc);
1621         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1622             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1623             nla_put_u8(skb, IFLA_OPERSTATE,
1624                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1625             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1626             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1627             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1628             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1629             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1630             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1631             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1632             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1633             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1634 #ifdef CONFIG_RPS
1635             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1636 #endif
1637             put_master_ifindex(skb, dev) ||
1638             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1639             (qdisc &&
1640              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1641             nla_put_ifalias(skb, dev) ||
1642             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1643                         atomic_read(&dev->carrier_up_count) +
1644                         atomic_read(&dev->carrier_down_count)) ||
1645             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1646             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1647                         atomic_read(&dev->carrier_up_count)) ||
1648             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1649                         atomic_read(&dev->carrier_down_count)))
1650                 goto nla_put_failure;
1651
1652         if (event != IFLA_EVENT_NONE) {
1653                 if (nla_put_u32(skb, IFLA_EVENT, event))
1654                         goto nla_put_failure;
1655         }
1656
1657         if (rtnl_fill_link_ifmap(skb, dev))
1658                 goto nla_put_failure;
1659
1660         if (dev->addr_len) {
1661                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1662                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1663                         goto nla_put_failure;
1664         }
1665
1666         if (rtnl_phys_port_id_fill(skb, dev))
1667                 goto nla_put_failure;
1668
1669         if (rtnl_phys_port_name_fill(skb, dev))
1670                 goto nla_put_failure;
1671
1672         if (rtnl_phys_switch_id_fill(skb, dev))
1673                 goto nla_put_failure;
1674
1675         if (rtnl_fill_stats(skb, dev))
1676                 goto nla_put_failure;
1677
1678         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1679                 goto nla_put_failure;
1680
1681         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1682                 goto nla_put_failure;
1683
1684         if (rtnl_xdp_fill(skb, dev))
1685                 goto nla_put_failure;
1686
1687         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1688                 if (rtnl_link_fill(skb, dev) < 0)
1689                         goto nla_put_failure;
1690         }
1691
1692         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1693                 goto nla_put_failure;
1694
1695         if (new_nsid &&
1696             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1697                 goto nla_put_failure;
1698         if (new_ifindex &&
1699             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1700                 goto nla_put_failure;
1701
1702
1703         rcu_read_lock();
1704         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1705                 goto nla_put_failure_rcu;
1706         rcu_read_unlock();
1707
1708         nlmsg_end(skb, nlh);
1709         return 0;
1710
1711 nla_put_failure_rcu:
1712         rcu_read_unlock();
1713 nla_put_failure:
1714         nlmsg_cancel(skb, nlh);
1715         return -EMSGSIZE;
1716 }
1717
1718 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1719         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1720         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1721         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1722         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1723         [IFLA_MTU]              = { .type = NLA_U32 },
1724         [IFLA_LINK]             = { .type = NLA_U32 },
1725         [IFLA_MASTER]           = { .type = NLA_U32 },
1726         [IFLA_CARRIER]          = { .type = NLA_U8 },
1727         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1728         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1729         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1730         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1731         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1732         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1733         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1734         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1735          * allow 0-length string (needed to remove an alias).
1736          */
1737         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1738         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1739         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1740         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1741         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1742         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1743         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1744         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1745         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1746         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1747         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1748         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1749         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1750         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1751         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1752         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1753         [IFLA_XDP]              = { .type = NLA_NESTED },
1754         [IFLA_EVENT]            = { .type = NLA_U32 },
1755         [IFLA_GROUP]            = { .type = NLA_U32 },
1756         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1757         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1758         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1759         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1760         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1761 };
1762
1763 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1764         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1765         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1766         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1767         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1768 };
1769
1770 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1771         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1772         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1773         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1774         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1775         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1776         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1777         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1778         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1779         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1780         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1781         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1782         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1783         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1784 };
1785
1786 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1787         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1788         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1789                                     .len = PORT_PROFILE_MAX },
1790         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1791                                       .len = PORT_UUID_MAX },
1792         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1793                                     .len = PORT_UUID_MAX },
1794         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1795         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1796
1797         /* Unused, but we need to keep it here since user space could
1798          * fill it. It's also broken with regard to NLA_BINARY use in
1799          * combination with structs.
1800          */
1801         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1802                                     .len = sizeof(struct ifla_port_vsi) },
1803 };
1804
1805 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1806         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1807         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1808         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1809         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1810 };
1811
1812 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1813 {
1814         const struct rtnl_link_ops *ops = NULL;
1815         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1816
1817         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1818                 return NULL;
1819
1820         if (linfo[IFLA_INFO_KIND]) {
1821                 char kind[MODULE_NAME_LEN];
1822
1823                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1824                 ops = rtnl_link_ops_get(kind);
1825         }
1826
1827         return ops;
1828 }
1829
1830 static bool link_master_filtered(struct net_device *dev, int master_idx)
1831 {
1832         struct net_device *master;
1833
1834         if (!master_idx)
1835                 return false;
1836
1837         master = netdev_master_upper_dev_get(dev);
1838         if (!master || master->ifindex != master_idx)
1839                 return true;
1840
1841         return false;
1842 }
1843
1844 static bool link_kind_filtered(const struct net_device *dev,
1845                                const struct rtnl_link_ops *kind_ops)
1846 {
1847         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1848                 return true;
1849
1850         return false;
1851 }
1852
1853 static bool link_dump_filtered(struct net_device *dev,
1854                                int master_idx,
1855                                const struct rtnl_link_ops *kind_ops)
1856 {
1857         if (link_master_filtered(dev, master_idx) ||
1858             link_kind_filtered(dev, kind_ops))
1859                 return true;
1860
1861         return false;
1862 }
1863
1864 /**
1865  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1866  * @sk: netlink socket
1867  * @netnsid: network namespace identifier
1868  *
1869  * Returns the network namespace identified by netnsid on success or an error
1870  * pointer on failure.
1871  */
1872 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1873 {
1874         struct net *net;
1875
1876         net = get_net_ns_by_id(sock_net(sk), netnsid);
1877         if (!net)
1878                 return ERR_PTR(-EINVAL);
1879
1880         /* For now, the caller is required to have CAP_NET_ADMIN in
1881          * the user namespace owning the target net ns.
1882          */
1883         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1884                 put_net(net);
1885                 return ERR_PTR(-EACCES);
1886         }
1887         return net;
1888 }
1889 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1890
1891 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1892                                       bool strict_check, struct nlattr **tb,
1893                                       struct netlink_ext_ack *extack)
1894 {
1895         int hdrlen;
1896
1897         if (strict_check) {
1898                 struct ifinfomsg *ifm;
1899
1900                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1901                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1902                         return -EINVAL;
1903                 }
1904
1905                 ifm = nlmsg_data(nlh);
1906                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1907                     ifm->ifi_change) {
1908                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1909                         return -EINVAL;
1910                 }
1911                 if (ifm->ifi_index) {
1912                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1913                         return -EINVAL;
1914                 }
1915
1916                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1917                                                      IFLA_MAX, ifla_policy,
1918                                                      extack);
1919         }
1920
1921         /* A hack to preserve kernel<->userspace interface.
1922          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1923          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1924          * what iproute2 < v3.9.0 used.
1925          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1926          * attribute, its netlink message is shorter than struct ifinfomsg.
1927          */
1928         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1929                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1930
1931         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
1932                                       extack);
1933 }
1934
1935 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1936 {
1937         struct netlink_ext_ack *extack = cb->extack;
1938         const struct nlmsghdr *nlh = cb->nlh;
1939         struct net *net = sock_net(skb->sk);
1940         struct net *tgt_net = net;
1941         int h, s_h;
1942         int idx = 0, s_idx;
1943         struct net_device *dev;
1944         struct hlist_head *head;
1945         struct nlattr *tb[IFLA_MAX+1];
1946         u32 ext_filter_mask = 0;
1947         const struct rtnl_link_ops *kind_ops = NULL;
1948         unsigned int flags = NLM_F_MULTI;
1949         int master_idx = 0;
1950         int netnsid = -1;
1951         int err, i;
1952
1953         s_h = cb->args[0];
1954         s_idx = cb->args[1];
1955
1956         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
1957         if (err < 0) {
1958                 if (cb->strict_check)
1959                         return err;
1960
1961                 goto walk_entries;
1962         }
1963
1964         for (i = 0; i <= IFLA_MAX; ++i) {
1965                 if (!tb[i])
1966                         continue;
1967
1968                 /* new attributes should only be added with strict checking */
1969                 switch (i) {
1970                 case IFLA_TARGET_NETNSID:
1971                         netnsid = nla_get_s32(tb[i]);
1972                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
1973                         if (IS_ERR(tgt_net)) {
1974                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
1975                                 return PTR_ERR(tgt_net);
1976                         }
1977                         break;
1978                 case IFLA_EXT_MASK:
1979                         ext_filter_mask = nla_get_u32(tb[i]);
1980                         break;
1981                 case IFLA_MASTER:
1982                         master_idx = nla_get_u32(tb[i]);
1983                         break;
1984                 case IFLA_LINKINFO:
1985                         kind_ops = linkinfo_to_kind_ops(tb[i]);
1986                         break;
1987                 default:
1988                         if (cb->strict_check) {
1989                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
1990                                 return -EINVAL;
1991                         }
1992                 }
1993         }
1994
1995         if (master_idx || kind_ops)
1996                 flags |= NLM_F_DUMP_FILTERED;
1997
1998 walk_entries:
1999         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2000                 idx = 0;
2001                 head = &tgt_net->dev_index_head[h];
2002                 hlist_for_each_entry(dev, head, index_hlist) {
2003                         if (link_dump_filtered(dev, master_idx, kind_ops))
2004                                 goto cont;
2005                         if (idx < s_idx)
2006                                 goto cont;
2007                         err = rtnl_fill_ifinfo(skb, dev, net,
2008                                                RTM_NEWLINK,
2009                                                NETLINK_CB(cb->skb).portid,
2010                                                nlh->nlmsg_seq, 0, flags,
2011                                                ext_filter_mask, 0, NULL, 0,
2012                                                netnsid, GFP_KERNEL);
2013
2014                         if (err < 0) {
2015                                 if (likely(skb->len))
2016                                         goto out;
2017
2018                                 goto out_err;
2019                         }
2020 cont:
2021                         idx++;
2022                 }
2023         }
2024 out:
2025         err = skb->len;
2026 out_err:
2027         cb->args[1] = idx;
2028         cb->args[0] = h;
2029         cb->seq = net->dev_base_seq;
2030         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2031         if (netnsid >= 0)
2032                 put_net(tgt_net);
2033
2034         return err;
2035 }
2036
2037 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2038                              struct netlink_ext_ack *exterr)
2039 {
2040         const struct ifinfomsg *ifmp;
2041         const struct nlattr *attrs;
2042         size_t len;
2043
2044         ifmp = nla_data(nla_peer);
2045         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2046         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2047
2048         if (ifmp->ifi_index < 0) {
2049                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2050                                     "ifindex can't be negative");
2051                 return -EINVAL;
2052         }
2053
2054         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2055                                     exterr);
2056 }
2057 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2058
2059 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2060 {
2061         struct net *net;
2062         /* Examine the link attributes and figure out which
2063          * network namespace we are talking about.
2064          */
2065         if (tb[IFLA_NET_NS_PID])
2066                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2067         else if (tb[IFLA_NET_NS_FD])
2068                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2069         else
2070                 net = get_net(src_net);
2071         return net;
2072 }
2073 EXPORT_SYMBOL(rtnl_link_get_net);
2074
2075 /* Figure out which network namespace we are talking about by
2076  * examining the link attributes in the following order:
2077  *
2078  * 1. IFLA_NET_NS_PID
2079  * 2. IFLA_NET_NS_FD
2080  * 3. IFLA_TARGET_NETNSID
2081  */
2082 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2083                                                struct nlattr *tb[])
2084 {
2085         struct net *net;
2086
2087         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2088                 return rtnl_link_get_net(src_net, tb);
2089
2090         if (!tb[IFLA_TARGET_NETNSID])
2091                 return get_net(src_net);
2092
2093         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2094         if (!net)
2095                 return ERR_PTR(-EINVAL);
2096
2097         return net;
2098 }
2099
2100 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2101                                              struct net *src_net,
2102                                              struct nlattr *tb[], int cap)
2103 {
2104         struct net *net;
2105
2106         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2107         if (IS_ERR(net))
2108                 return net;
2109
2110         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2111                 put_net(net);
2112                 return ERR_PTR(-EPERM);
2113         }
2114
2115         return net;
2116 }
2117
2118 /* Verify that rtnetlink requests do not pass additional properties
2119  * potentially referring to different network namespaces.
2120  */
2121 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2122                                     struct netlink_ext_ack *extack,
2123                                     bool netns_id_only)
2124 {
2125
2126         if (netns_id_only) {
2127                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2128                         return 0;
2129
2130                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2131                 return -EOPNOTSUPP;
2132         }
2133
2134         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2135                 goto invalid_attr;
2136
2137         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2138                 goto invalid_attr;
2139
2140         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2141                 goto invalid_attr;
2142
2143         return 0;
2144
2145 invalid_attr:
2146         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2147         return -EINVAL;
2148 }
2149
2150 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2151 {
2152         if (dev) {
2153                 if (tb[IFLA_ADDRESS] &&
2154                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2155                         return -EINVAL;
2156
2157                 if (tb[IFLA_BROADCAST] &&
2158                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2159                         return -EINVAL;
2160         }
2161
2162         if (tb[IFLA_AF_SPEC]) {
2163                 struct nlattr *af;
2164                 int rem, err;
2165
2166                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2167                         const struct rtnl_af_ops *af_ops;
2168
2169                         rcu_read_lock();
2170                         af_ops = rtnl_af_lookup(nla_type(af));
2171                         if (!af_ops) {
2172                                 rcu_read_unlock();
2173                                 return -EAFNOSUPPORT;
2174                         }
2175
2176                         if (!af_ops->set_link_af) {
2177                                 rcu_read_unlock();
2178                                 return -EOPNOTSUPP;
2179                         }
2180
2181                         if (af_ops->validate_link_af) {
2182                                 err = af_ops->validate_link_af(dev, af);
2183                                 if (err < 0) {
2184                                         rcu_read_unlock();
2185                                         return err;
2186                                 }
2187                         }
2188
2189                         rcu_read_unlock();
2190                 }
2191         }
2192
2193         return 0;
2194 }
2195
2196 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2197                                   int guid_type)
2198 {
2199         const struct net_device_ops *ops = dev->netdev_ops;
2200
2201         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2202 }
2203
2204 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2205 {
2206         if (dev->type != ARPHRD_INFINIBAND)
2207                 return -EOPNOTSUPP;
2208
2209         return handle_infiniband_guid(dev, ivt, guid_type);
2210 }
2211
2212 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2213 {
2214         const struct net_device_ops *ops = dev->netdev_ops;
2215         int err = -EINVAL;
2216
2217         if (tb[IFLA_VF_MAC]) {
2218                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2219
2220                 if (ivm->vf >= INT_MAX)
2221                         return -EINVAL;
2222                 err = -EOPNOTSUPP;
2223                 if (ops->ndo_set_vf_mac)
2224                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2225                                                   ivm->mac);
2226                 if (err < 0)
2227                         return err;
2228         }
2229
2230         if (tb[IFLA_VF_VLAN]) {
2231                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2232
2233                 if (ivv->vf >= INT_MAX)
2234                         return -EINVAL;
2235                 err = -EOPNOTSUPP;
2236                 if (ops->ndo_set_vf_vlan)
2237                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2238                                                    ivv->qos,
2239                                                    htons(ETH_P_8021Q));
2240                 if (err < 0)
2241                         return err;
2242         }
2243
2244         if (tb[IFLA_VF_VLAN_LIST]) {
2245                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2246                 struct nlattr *attr;
2247                 int rem, len = 0;
2248
2249                 err = -EOPNOTSUPP;
2250                 if (!ops->ndo_set_vf_vlan)
2251                         return err;
2252
2253                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2254                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2255                             nla_len(attr) < NLA_HDRLEN) {
2256                                 return -EINVAL;
2257                         }
2258                         if (len >= MAX_VLAN_LIST_LEN)
2259                                 return -EOPNOTSUPP;
2260                         ivvl[len] = nla_data(attr);
2261
2262                         len++;
2263                 }
2264                 if (len == 0)
2265                         return -EINVAL;
2266
2267                 if (ivvl[0]->vf >= INT_MAX)
2268                         return -EINVAL;
2269                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2270                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2271                 if (err < 0)
2272                         return err;
2273         }
2274
2275         if (tb[IFLA_VF_TX_RATE]) {
2276                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2277                 struct ifla_vf_info ivf;
2278
2279                 if (ivt->vf >= INT_MAX)
2280                         return -EINVAL;
2281                 err = -EOPNOTSUPP;
2282                 if (ops->ndo_get_vf_config)
2283                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2284                 if (err < 0)
2285                         return err;
2286
2287                 err = -EOPNOTSUPP;
2288                 if (ops->ndo_set_vf_rate)
2289                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2290                                                    ivf.min_tx_rate,
2291                                                    ivt->rate);
2292                 if (err < 0)
2293                         return err;
2294         }
2295
2296         if (tb[IFLA_VF_RATE]) {
2297                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2298
2299                 if (ivt->vf >= INT_MAX)
2300                         return -EINVAL;
2301                 err = -EOPNOTSUPP;
2302                 if (ops->ndo_set_vf_rate)
2303                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2304                                                    ivt->min_tx_rate,
2305                                                    ivt->max_tx_rate);
2306                 if (err < 0)
2307                         return err;
2308         }
2309
2310         if (tb[IFLA_VF_SPOOFCHK]) {
2311                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2312
2313                 if (ivs->vf >= INT_MAX)
2314                         return -EINVAL;
2315                 err = -EOPNOTSUPP;
2316                 if (ops->ndo_set_vf_spoofchk)
2317                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2318                                                        ivs->setting);
2319                 if (err < 0)
2320                         return err;
2321         }
2322
2323         if (tb[IFLA_VF_LINK_STATE]) {
2324                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2325
2326                 if (ivl->vf >= INT_MAX)
2327                         return -EINVAL;
2328                 err = -EOPNOTSUPP;
2329                 if (ops->ndo_set_vf_link_state)
2330                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2331                                                          ivl->link_state);
2332                 if (err < 0)
2333                         return err;
2334         }
2335
2336         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2337                 struct ifla_vf_rss_query_en *ivrssq_en;
2338
2339                 err = -EOPNOTSUPP;
2340                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2341                 if (ivrssq_en->vf >= INT_MAX)
2342                         return -EINVAL;
2343                 if (ops->ndo_set_vf_rss_query_en)
2344                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2345                                                            ivrssq_en->setting);
2346                 if (err < 0)
2347                         return err;
2348         }
2349
2350         if (tb[IFLA_VF_TRUST]) {
2351                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2352
2353                 if (ivt->vf >= INT_MAX)
2354                         return -EINVAL;
2355                 err = -EOPNOTSUPP;
2356                 if (ops->ndo_set_vf_trust)
2357                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2358                 if (err < 0)
2359                         return err;
2360         }
2361
2362         if (tb[IFLA_VF_IB_NODE_GUID]) {
2363                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2364
2365                 if (ivt->vf >= INT_MAX)
2366                         return -EINVAL;
2367                 if (!ops->ndo_set_vf_guid)
2368                         return -EOPNOTSUPP;
2369                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2370         }
2371
2372         if (tb[IFLA_VF_IB_PORT_GUID]) {
2373                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2374
2375                 if (ivt->vf >= INT_MAX)
2376                         return -EINVAL;
2377                 if (!ops->ndo_set_vf_guid)
2378                         return -EOPNOTSUPP;
2379
2380                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2381         }
2382
2383         return err;
2384 }
2385
2386 static int do_set_master(struct net_device *dev, int ifindex,
2387                          struct netlink_ext_ack *extack)
2388 {
2389         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2390         const struct net_device_ops *ops;
2391         int err;
2392
2393         if (upper_dev) {
2394                 if (upper_dev->ifindex == ifindex)
2395                         return 0;
2396                 ops = upper_dev->netdev_ops;
2397                 if (ops->ndo_del_slave) {
2398                         err = ops->ndo_del_slave(upper_dev, dev);
2399                         if (err)
2400                                 return err;
2401                         netdev_update_lockdep_key(dev);
2402                 } else {
2403                         return -EOPNOTSUPP;
2404                 }
2405         }
2406
2407         if (ifindex) {
2408                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2409                 if (!upper_dev)
2410                         return -EINVAL;
2411                 ops = upper_dev->netdev_ops;
2412                 if (ops->ndo_add_slave) {
2413                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2414                         if (err)
2415                                 return err;
2416                 } else {
2417                         return -EOPNOTSUPP;
2418                 }
2419         }
2420         return 0;
2421 }
2422
2423 #define DO_SETLINK_MODIFIED     0x01
2424 /* notify flag means notify + modified. */
2425 #define DO_SETLINK_NOTIFY       0x03
2426 static int do_setlink(const struct sk_buff *skb,
2427                       struct net_device *dev, struct ifinfomsg *ifm,
2428                       struct netlink_ext_ack *extack,
2429                       struct nlattr **tb, char *ifname, int status)
2430 {
2431         const struct net_device_ops *ops = dev->netdev_ops;
2432         int err;
2433
2434         err = validate_linkmsg(dev, tb);
2435         if (err < 0)
2436                 return err;
2437
2438         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2439                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2440                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2441                                                             tb, CAP_NET_ADMIN);
2442                 if (IS_ERR(net)) {
2443                         err = PTR_ERR(net);
2444                         goto errout;
2445                 }
2446
2447                 err = dev_change_net_namespace(dev, net, pat);
2448                 put_net(net);
2449                 if (err)
2450                         goto errout;
2451                 status |= DO_SETLINK_MODIFIED;
2452         }
2453
2454         if (tb[IFLA_MAP]) {
2455                 struct rtnl_link_ifmap *u_map;
2456                 struct ifmap k_map;
2457
2458                 if (!ops->ndo_set_config) {
2459                         err = -EOPNOTSUPP;
2460                         goto errout;
2461                 }
2462
2463                 if (!netif_device_present(dev)) {
2464                         err = -ENODEV;
2465                         goto errout;
2466                 }
2467
2468                 u_map = nla_data(tb[IFLA_MAP]);
2469                 k_map.mem_start = (unsigned long) u_map->mem_start;
2470                 k_map.mem_end = (unsigned long) u_map->mem_end;
2471                 k_map.base_addr = (unsigned short) u_map->base_addr;
2472                 k_map.irq = (unsigned char) u_map->irq;
2473                 k_map.dma = (unsigned char) u_map->dma;
2474                 k_map.port = (unsigned char) u_map->port;
2475
2476                 err = ops->ndo_set_config(dev, &k_map);
2477                 if (err < 0)
2478                         goto errout;
2479
2480                 status |= DO_SETLINK_NOTIFY;
2481         }
2482
2483         if (tb[IFLA_ADDRESS]) {
2484                 struct sockaddr *sa;
2485                 int len;
2486
2487                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2488                                                   sizeof(*sa));
2489                 sa = kmalloc(len, GFP_KERNEL);
2490                 if (!sa) {
2491                         err = -ENOMEM;
2492                         goto errout;
2493                 }
2494                 sa->sa_family = dev->type;
2495                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2496                        dev->addr_len);
2497                 err = dev_set_mac_address_user(dev, sa, extack);
2498                 kfree(sa);
2499                 if (err)
2500                         goto errout;
2501                 status |= DO_SETLINK_MODIFIED;
2502         }
2503
2504         if (tb[IFLA_MTU]) {
2505                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2506                 if (err < 0)
2507                         goto errout;
2508                 status |= DO_SETLINK_MODIFIED;
2509         }
2510
2511         if (tb[IFLA_GROUP]) {
2512                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2513                 status |= DO_SETLINK_NOTIFY;
2514         }
2515
2516         /*
2517          * Interface selected by interface index but interface
2518          * name provided implies that a name change has been
2519          * requested.
2520          */
2521         if (ifm->ifi_index > 0 && ifname[0]) {
2522                 err = dev_change_name(dev, ifname);
2523                 if (err < 0)
2524                         goto errout;
2525                 status |= DO_SETLINK_MODIFIED;
2526         }
2527
2528         if (tb[IFLA_IFALIAS]) {
2529                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2530                                     nla_len(tb[IFLA_IFALIAS]));
2531                 if (err < 0)
2532                         goto errout;
2533                 status |= DO_SETLINK_NOTIFY;
2534         }
2535
2536         if (tb[IFLA_BROADCAST]) {
2537                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2538                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2539         }
2540
2541         if (ifm->ifi_flags || ifm->ifi_change) {
2542                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2543                                        extack);
2544                 if (err < 0)
2545                         goto errout;
2546         }
2547
2548         if (tb[IFLA_MASTER]) {
2549                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2550                 if (err)
2551                         goto errout;
2552                 status |= DO_SETLINK_MODIFIED;
2553         }
2554
2555         if (tb[IFLA_CARRIER]) {
2556                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2557                 if (err)
2558                         goto errout;
2559                 status |= DO_SETLINK_MODIFIED;
2560         }
2561
2562         if (tb[IFLA_TXQLEN]) {
2563                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2564
2565                 err = dev_change_tx_queue_len(dev, value);
2566                 if (err)
2567                         goto errout;
2568                 status |= DO_SETLINK_MODIFIED;
2569         }
2570
2571         if (tb[IFLA_GSO_MAX_SIZE]) {
2572                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2573
2574                 if (max_size > GSO_MAX_SIZE) {
2575                         err = -EINVAL;
2576                         goto errout;
2577                 }
2578
2579                 if (dev->gso_max_size ^ max_size) {
2580                         netif_set_gso_max_size(dev, max_size);
2581                         status |= DO_SETLINK_MODIFIED;
2582                 }
2583         }
2584
2585         if (tb[IFLA_GSO_MAX_SEGS]) {
2586                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2587
2588                 if (max_segs > GSO_MAX_SEGS) {
2589                         err = -EINVAL;
2590                         goto errout;
2591                 }
2592
2593                 if (dev->gso_max_segs ^ max_segs) {
2594                         dev->gso_max_segs = max_segs;
2595                         status |= DO_SETLINK_MODIFIED;
2596                 }
2597         }
2598
2599         if (tb[IFLA_OPERSTATE])
2600                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2601
2602         if (tb[IFLA_LINKMODE]) {
2603                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2604
2605                 write_lock_bh(&dev_base_lock);
2606                 if (dev->link_mode ^ value)
2607                         status |= DO_SETLINK_NOTIFY;
2608                 dev->link_mode = value;
2609                 write_unlock_bh(&dev_base_lock);
2610         }
2611
2612         if (tb[IFLA_VFINFO_LIST]) {
2613                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2614                 struct nlattr *attr;
2615                 int rem;
2616
2617                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2618                         if (nla_type(attr) != IFLA_VF_INFO ||
2619                             nla_len(attr) < NLA_HDRLEN) {
2620                                 err = -EINVAL;
2621                                 goto errout;
2622                         }
2623                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2624                                                           attr,
2625                                                           ifla_vf_policy,
2626                                                           NULL);
2627                         if (err < 0)
2628                                 goto errout;
2629                         err = do_setvfinfo(dev, vfinfo);
2630                         if (err < 0)
2631                                 goto errout;
2632                         status |= DO_SETLINK_NOTIFY;
2633                 }
2634         }
2635         err = 0;
2636
2637         if (tb[IFLA_VF_PORTS]) {
2638                 struct nlattr *port[IFLA_PORT_MAX+1];
2639                 struct nlattr *attr;
2640                 int vf;
2641                 int rem;
2642
2643                 err = -EOPNOTSUPP;
2644                 if (!ops->ndo_set_vf_port)
2645                         goto errout;
2646
2647                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2648                         if (nla_type(attr) != IFLA_VF_PORT ||
2649                             nla_len(attr) < NLA_HDRLEN) {
2650                                 err = -EINVAL;
2651                                 goto errout;
2652                         }
2653                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2654                                                           attr,
2655                                                           ifla_port_policy,
2656                                                           NULL);
2657                         if (err < 0)
2658                                 goto errout;
2659                         if (!port[IFLA_PORT_VF]) {
2660                                 err = -EOPNOTSUPP;
2661                                 goto errout;
2662                         }
2663                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2664                         err = ops->ndo_set_vf_port(dev, vf, port);
2665                         if (err < 0)
2666                                 goto errout;
2667                         status |= DO_SETLINK_NOTIFY;
2668                 }
2669         }
2670         err = 0;
2671
2672         if (tb[IFLA_PORT_SELF]) {
2673                 struct nlattr *port[IFLA_PORT_MAX+1];
2674
2675                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2676                                                   tb[IFLA_PORT_SELF],
2677                                                   ifla_port_policy, NULL);
2678                 if (err < 0)
2679                         goto errout;
2680
2681                 err = -EOPNOTSUPP;
2682                 if (ops->ndo_set_vf_port)
2683                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2684                 if (err < 0)
2685                         goto errout;
2686                 status |= DO_SETLINK_NOTIFY;
2687         }
2688
2689         if (tb[IFLA_AF_SPEC]) {
2690                 struct nlattr *af;
2691                 int rem;
2692
2693                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2694                         const struct rtnl_af_ops *af_ops;
2695
2696                         rcu_read_lock();
2697
2698                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2699
2700                         err = af_ops->set_link_af(dev, af);
2701                         if (err < 0) {
2702                                 rcu_read_unlock();
2703                                 goto errout;
2704                         }
2705
2706                         rcu_read_unlock();
2707                         status |= DO_SETLINK_NOTIFY;
2708                 }
2709         }
2710         err = 0;
2711
2712         if (tb[IFLA_PROTO_DOWN]) {
2713                 err = dev_change_proto_down(dev,
2714                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2715                 if (err)
2716                         goto errout;
2717                 status |= DO_SETLINK_NOTIFY;
2718         }
2719
2720         if (tb[IFLA_XDP]) {
2721                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2722                 u32 xdp_flags = 0;
2723
2724                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2725                                                   tb[IFLA_XDP],
2726                                                   ifla_xdp_policy, NULL);
2727                 if (err < 0)
2728                         goto errout;
2729
2730                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2731                         err = -EINVAL;
2732                         goto errout;
2733                 }
2734
2735                 if (xdp[IFLA_XDP_FLAGS]) {
2736                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2737                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2738                                 err = -EINVAL;
2739                                 goto errout;
2740                         }
2741                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2742                                 err = -EINVAL;
2743                                 goto errout;
2744                         }
2745                 }
2746
2747                 if (xdp[IFLA_XDP_FD]) {
2748                         err = dev_change_xdp_fd(dev, extack,
2749                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2750                                                 xdp_flags);
2751                         if (err)
2752                                 goto errout;
2753                         status |= DO_SETLINK_NOTIFY;
2754                 }
2755         }
2756
2757 errout:
2758         if (status & DO_SETLINK_MODIFIED) {
2759                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2760                         netdev_state_change(dev);
2761
2762                 if (err < 0)
2763                         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",
2764                                              dev->name);
2765         }
2766
2767         return err;
2768 }
2769
2770 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2771                         struct netlink_ext_ack *extack)
2772 {
2773         struct net *net = sock_net(skb->sk);
2774         struct ifinfomsg *ifm;
2775         struct net_device *dev;
2776         int err;
2777         struct nlattr *tb[IFLA_MAX+1];
2778         char ifname[IFNAMSIZ];
2779
2780         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2781                                      ifla_policy, extack);
2782         if (err < 0)
2783                 goto errout;
2784
2785         err = rtnl_ensure_unique_netns(tb, extack, false);
2786         if (err < 0)
2787                 goto errout;
2788
2789         if (tb[IFLA_IFNAME])
2790                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2791         else
2792                 ifname[0] = '\0';
2793
2794         err = -EINVAL;
2795         ifm = nlmsg_data(nlh);
2796         if (ifm->ifi_index > 0)
2797                 dev = __dev_get_by_index(net, ifm->ifi_index);
2798         else if (tb[IFLA_IFNAME])
2799                 dev = __dev_get_by_name(net, ifname);
2800         else
2801                 goto errout;
2802
2803         if (dev == NULL) {
2804                 err = -ENODEV;
2805                 goto errout;
2806         }
2807
2808         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2809 errout:
2810         return err;
2811 }
2812
2813 static int rtnl_group_dellink(const struct net *net, int group)
2814 {
2815         struct net_device *dev, *aux;
2816         LIST_HEAD(list_kill);
2817         bool found = false;
2818
2819         if (!group)
2820                 return -EPERM;
2821
2822         for_each_netdev(net, dev) {
2823                 if (dev->group == group) {
2824                         const struct rtnl_link_ops *ops;
2825
2826                         found = true;
2827                         ops = dev->rtnl_link_ops;
2828                         if (!ops || !ops->dellink)
2829                                 return -EOPNOTSUPP;
2830                 }
2831         }
2832
2833         if (!found)
2834                 return -ENODEV;
2835
2836         for_each_netdev_safe(net, dev, aux) {
2837                 if (dev->group == group) {
2838                         const struct rtnl_link_ops *ops;
2839
2840                         ops = dev->rtnl_link_ops;
2841                         ops->dellink(dev, &list_kill);
2842                 }
2843         }
2844         unregister_netdevice_many(&list_kill);
2845
2846         return 0;
2847 }
2848
2849 int rtnl_delete_link(struct net_device *dev)
2850 {
2851         const struct rtnl_link_ops *ops;
2852         LIST_HEAD(list_kill);
2853
2854         ops = dev->rtnl_link_ops;
2855         if (!ops || !ops->dellink)
2856                 return -EOPNOTSUPP;
2857
2858         ops->dellink(dev, &list_kill);
2859         unregister_netdevice_many(&list_kill);
2860
2861         return 0;
2862 }
2863 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2864
2865 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2866                         struct netlink_ext_ack *extack)
2867 {
2868         struct net *net = sock_net(skb->sk);
2869         struct net *tgt_net = net;
2870         struct net_device *dev = NULL;
2871         struct ifinfomsg *ifm;
2872         char ifname[IFNAMSIZ];
2873         struct nlattr *tb[IFLA_MAX+1];
2874         int err;
2875         int netnsid = -1;
2876
2877         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2878                                      ifla_policy, extack);
2879         if (err < 0)
2880                 return err;
2881
2882         err = rtnl_ensure_unique_netns(tb, extack, true);
2883         if (err < 0)
2884                 return err;
2885
2886         if (tb[IFLA_IFNAME])
2887                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2888
2889         if (tb[IFLA_TARGET_NETNSID]) {
2890                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2891                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2892                 if (IS_ERR(tgt_net))
2893                         return PTR_ERR(tgt_net);
2894         }
2895
2896         err = -EINVAL;
2897         ifm = nlmsg_data(nlh);
2898         if (ifm->ifi_index > 0)
2899                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2900         else if (tb[IFLA_IFNAME])
2901                 dev = __dev_get_by_name(tgt_net, ifname);
2902         else if (tb[IFLA_GROUP])
2903                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2904         else
2905                 goto out;
2906
2907         if (!dev) {
2908                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2909                         err = -ENODEV;
2910
2911                 goto out;
2912         }
2913
2914         err = rtnl_delete_link(dev);
2915
2916 out:
2917         if (netnsid >= 0)
2918                 put_net(tgt_net);
2919
2920         return err;
2921 }
2922
2923 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2924 {
2925         unsigned int old_flags;
2926         int err;
2927
2928         old_flags = dev->flags;
2929         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2930                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2931                                          NULL);
2932                 if (err < 0)
2933                         return err;
2934         }
2935
2936         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2937                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2938         } else {
2939                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2940                 __dev_notify_flags(dev, old_flags, ~0U);
2941         }
2942         return 0;
2943 }
2944 EXPORT_SYMBOL(rtnl_configure_link);
2945
2946 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
2947                                     unsigned char name_assign_type,
2948                                     const struct rtnl_link_ops *ops,
2949                                     struct nlattr *tb[],
2950                                     struct netlink_ext_ack *extack)
2951 {
2952         struct net_device *dev;
2953         unsigned int num_tx_queues = 1;
2954         unsigned int num_rx_queues = 1;
2955
2956         if (tb[IFLA_NUM_TX_QUEUES])
2957                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2958         else if (ops->get_num_tx_queues)
2959                 num_tx_queues = ops->get_num_tx_queues();
2960
2961         if (tb[IFLA_NUM_RX_QUEUES])
2962                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2963         else if (ops->get_num_rx_queues)
2964                 num_rx_queues = ops->get_num_rx_queues();
2965
2966         if (num_tx_queues < 1 || num_tx_queues > 4096) {
2967                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
2968                 return ERR_PTR(-EINVAL);
2969         }
2970
2971         if (num_rx_queues < 1 || num_rx_queues > 4096) {
2972                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
2973                 return ERR_PTR(-EINVAL);
2974         }
2975
2976         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2977                                ops->setup, num_tx_queues, num_rx_queues);
2978         if (!dev)
2979                 return ERR_PTR(-ENOMEM);
2980
2981         dev_net_set(dev, net);
2982         dev->rtnl_link_ops = ops;
2983         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2984
2985         if (tb[IFLA_MTU]) {
2986                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2987                 int err;
2988
2989                 err = dev_validate_mtu(dev, mtu, extack);
2990                 if (err) {
2991                         free_netdev(dev);
2992                         return ERR_PTR(err);
2993                 }
2994                 dev->mtu = mtu;
2995         }
2996         if (tb[IFLA_ADDRESS]) {
2997                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2998                                 nla_len(tb[IFLA_ADDRESS]));
2999                 dev->addr_assign_type = NET_ADDR_SET;
3000         }
3001         if (tb[IFLA_BROADCAST])
3002                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3003                                 nla_len(tb[IFLA_BROADCAST]));
3004         if (tb[IFLA_TXQLEN])
3005                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3006         if (tb[IFLA_OPERSTATE])
3007                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3008         if (tb[IFLA_LINKMODE])
3009                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3010         if (tb[IFLA_GROUP])
3011                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3012         if (tb[IFLA_GSO_MAX_SIZE])
3013                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3014         if (tb[IFLA_GSO_MAX_SEGS])
3015                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3016
3017         return dev;
3018 }
3019 EXPORT_SYMBOL(rtnl_create_link);
3020
3021 static int rtnl_group_changelink(const struct sk_buff *skb,
3022                 struct net *net, int group,
3023                 struct ifinfomsg *ifm,
3024                 struct netlink_ext_ack *extack,
3025                 struct nlattr **tb)
3026 {
3027         struct net_device *dev, *aux;
3028         int err;
3029
3030         for_each_netdev_safe(net, dev, aux) {
3031                 if (dev->group == group) {
3032                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3033                         if (err < 0)
3034                                 return err;
3035                 }
3036         }
3037
3038         return 0;
3039 }
3040
3041 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3042                           struct nlattr **attr, struct netlink_ext_ack *extack)
3043 {
3044         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3045         unsigned char name_assign_type = NET_NAME_USER;
3046         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3047         const struct rtnl_link_ops *m_ops;
3048         struct net_device *master_dev;
3049         struct net *net = sock_net(skb->sk);
3050         const struct rtnl_link_ops *ops;
3051         struct nlattr *tb[IFLA_MAX + 1];
3052         struct net *dest_net, *link_net;
3053         struct nlattr **slave_data;
3054         char kind[MODULE_NAME_LEN];
3055         struct net_device *dev;
3056         struct ifinfomsg *ifm;
3057         char ifname[IFNAMSIZ];
3058         struct nlattr **data;
3059         int err;
3060
3061 #ifdef CONFIG_MODULES
3062 replay:
3063 #endif
3064         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3065                                      ifla_policy, extack);
3066         if (err < 0)
3067                 return err;
3068
3069         err = rtnl_ensure_unique_netns(tb, extack, false);
3070         if (err < 0)
3071                 return err;
3072
3073         if (tb[IFLA_IFNAME])
3074                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3075         else
3076                 ifname[0] = '\0';
3077
3078         ifm = nlmsg_data(nlh);
3079         if (ifm->ifi_index > 0) {
3080                 dev = __dev_get_by_index(net, ifm->ifi_index);
3081         } else if (ifm->ifi_index < 0) {
3082                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3083                 return -EINVAL;
3084         } else {
3085                 if (ifname[0])
3086                         dev = __dev_get_by_name(net, ifname);
3087                 else
3088                         dev = NULL;
3089         }
3090
3091         master_dev = NULL;
3092         m_ops = NULL;
3093         if (dev) {
3094                 master_dev = netdev_master_upper_dev_get(dev);
3095                 if (master_dev)
3096                         m_ops = master_dev->rtnl_link_ops;
3097         }
3098
3099         err = validate_linkmsg(dev, tb);
3100         if (err < 0)
3101                 return err;
3102
3103         if (tb[IFLA_LINKINFO]) {
3104                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3105                                                   tb[IFLA_LINKINFO],
3106                                                   ifla_info_policy, NULL);
3107                 if (err < 0)
3108                         return err;
3109         } else
3110                 memset(linkinfo, 0, sizeof(linkinfo));
3111
3112         if (linkinfo[IFLA_INFO_KIND]) {
3113                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3114                 ops = rtnl_link_ops_get(kind);
3115         } else {
3116                 kind[0] = '\0';
3117                 ops = NULL;
3118         }
3119
3120         data = NULL;
3121         if (ops) {
3122                 if (ops->maxtype > RTNL_MAX_TYPE)
3123                         return -EINVAL;
3124
3125                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3126                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3127                                                           linkinfo[IFLA_INFO_DATA],
3128                                                           ops->policy, extack);
3129                         if (err < 0)
3130                                 return err;
3131                         data = attr;
3132                 }
3133                 if (ops->validate) {
3134                         err = ops->validate(tb, data, extack);
3135                         if (err < 0)
3136                                 return err;
3137                 }
3138         }
3139
3140         slave_data = NULL;
3141         if (m_ops) {
3142                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3143                         return -EINVAL;
3144
3145                 if (m_ops->slave_maxtype &&
3146                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3147                         err = nla_parse_nested_deprecated(slave_attr,
3148                                                           m_ops->slave_maxtype,
3149                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3150                                                           m_ops->slave_policy,
3151                                                           extack);
3152                         if (err < 0)
3153                                 return err;
3154                         slave_data = slave_attr;
3155                 }
3156         }
3157
3158         if (dev) {
3159                 int status = 0;
3160
3161                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3162                         return -EEXIST;
3163                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3164                         return -EOPNOTSUPP;
3165
3166                 if (linkinfo[IFLA_INFO_DATA]) {
3167                         if (!ops || ops != dev->rtnl_link_ops ||
3168                             !ops->changelink)
3169                                 return -EOPNOTSUPP;
3170
3171                         err = ops->changelink(dev, tb, data, extack);
3172                         if (err < 0)
3173                                 return err;
3174                         status |= DO_SETLINK_NOTIFY;
3175                 }
3176
3177                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3178                         if (!m_ops || !m_ops->slave_changelink)
3179                                 return -EOPNOTSUPP;
3180
3181                         err = m_ops->slave_changelink(master_dev, dev, tb,
3182                                                       slave_data, extack);
3183                         if (err < 0)
3184                                 return err;
3185                         status |= DO_SETLINK_NOTIFY;
3186                 }
3187
3188                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3189         }
3190
3191         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3192                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3193                         return rtnl_group_changelink(skb, net,
3194                                                 nla_get_u32(tb[IFLA_GROUP]),
3195                                                 ifm, extack, tb);
3196                 return -ENODEV;
3197         }
3198
3199         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3200                 return -EOPNOTSUPP;
3201
3202         if (!ops) {
3203 #ifdef CONFIG_MODULES
3204                 if (kind[0]) {
3205                         __rtnl_unlock();
3206                         request_module("rtnl-link-%s", kind);
3207                         rtnl_lock();
3208                         ops = rtnl_link_ops_get(kind);
3209                         if (ops)
3210                                 goto replay;
3211                 }
3212 #endif
3213                 NL_SET_ERR_MSG(extack, "Unknown device type");
3214                 return -EOPNOTSUPP;
3215         }
3216
3217         if (!ops->setup)
3218                 return -EOPNOTSUPP;
3219
3220         if (!ifname[0]) {
3221                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3222                 name_assign_type = NET_NAME_ENUM;
3223         }
3224
3225         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3226         if (IS_ERR(dest_net))
3227                 return PTR_ERR(dest_net);
3228
3229         if (tb[IFLA_LINK_NETNSID]) {
3230                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3231
3232                 link_net = get_net_ns_by_id(dest_net, id);
3233                 if (!link_net) {
3234                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3235                         err =  -EINVAL;
3236                         goto out;
3237                 }
3238                 err = -EPERM;
3239                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3240                         goto out;
3241         } else {
3242                 link_net = NULL;
3243         }
3244
3245         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3246                                name_assign_type, ops, tb, extack);
3247         if (IS_ERR(dev)) {
3248                 err = PTR_ERR(dev);
3249                 goto out;
3250         }
3251
3252         dev->ifindex = ifm->ifi_index;
3253
3254         if (ops->newlink) {
3255                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3256                 /* Drivers should call free_netdev() in ->destructor
3257                  * and unregister it on failure after registration
3258                  * so that device could be finally freed in rtnl_unlock.
3259                  */
3260                 if (err < 0) {
3261                         /* If device is not registered at all, free it now */
3262                         if (dev->reg_state == NETREG_UNINITIALIZED ||
3263                             dev->reg_state == NETREG_UNREGISTERED)
3264                                 free_netdev(dev);
3265                         goto out;
3266                 }
3267         } else {
3268                 err = register_netdevice(dev);
3269                 if (err < 0) {
3270                         free_netdev(dev);
3271                         goto out;
3272                 }
3273         }
3274         err = rtnl_configure_link(dev, ifm);
3275         if (err < 0)
3276                 goto out_unregister;
3277         if (link_net) {
3278                 err = dev_change_net_namespace(dev, dest_net, ifname);
3279                 if (err < 0)
3280                         goto out_unregister;
3281         }
3282         if (tb[IFLA_MASTER]) {
3283                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3284                 if (err)
3285                         goto out_unregister;
3286         }
3287 out:
3288         if (link_net)
3289                 put_net(link_net);
3290         put_net(dest_net);
3291         return err;
3292 out_unregister:
3293         if (ops->newlink) {
3294                 LIST_HEAD(list_kill);
3295
3296                 ops->dellink(dev, &list_kill);
3297                 unregister_netdevice_many(&list_kill);
3298         } else {
3299                 unregister_netdevice(dev);
3300         }
3301         goto out;
3302 }
3303
3304 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3305                         struct netlink_ext_ack *extack)
3306 {
3307         struct nlattr **attr;
3308         int ret;
3309
3310         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3311         if (!attr)
3312                 return -ENOMEM;
3313
3314         ret = __rtnl_newlink(skb, nlh, attr, extack);
3315         kfree(attr);
3316         return ret;
3317 }
3318
3319 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3320                                   const struct nlmsghdr *nlh,
3321                                   struct nlattr **tb,
3322                                   struct netlink_ext_ack *extack)
3323 {
3324         struct ifinfomsg *ifm;
3325         int i, err;
3326
3327         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3328                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3329                 return -EINVAL;
3330         }
3331
3332         if (!netlink_strict_get_check(skb))
3333                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3334                                               ifla_policy, extack);
3335
3336         ifm = nlmsg_data(nlh);
3337         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3338             ifm->ifi_change) {
3339                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3340                 return -EINVAL;
3341         }
3342
3343         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3344                                             ifla_policy, extack);
3345         if (err)
3346                 return err;
3347
3348         for (i = 0; i <= IFLA_MAX; i++) {
3349                 if (!tb[i])
3350                         continue;
3351
3352                 switch (i) {
3353                 case IFLA_IFNAME:
3354                 case IFLA_EXT_MASK:
3355                 case IFLA_TARGET_NETNSID:
3356                         break;
3357                 default:
3358                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3359                         return -EINVAL;
3360                 }
3361         }
3362
3363         return 0;
3364 }
3365
3366 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3367                         struct netlink_ext_ack *extack)
3368 {
3369         struct net *net = sock_net(skb->sk);
3370         struct net *tgt_net = net;
3371         struct ifinfomsg *ifm;
3372         char ifname[IFNAMSIZ];
3373         struct nlattr *tb[IFLA_MAX+1];
3374         struct net_device *dev = NULL;
3375         struct sk_buff *nskb;
3376         int netnsid = -1;
3377         int err;
3378         u32 ext_filter_mask = 0;
3379
3380         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3381         if (err < 0)
3382                 return err;
3383
3384         err = rtnl_ensure_unique_netns(tb, extack, true);
3385         if (err < 0)
3386                 return err;
3387
3388         if (tb[IFLA_TARGET_NETNSID]) {
3389                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3390                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3391                 if (IS_ERR(tgt_net))
3392                         return PTR_ERR(tgt_net);
3393         }
3394
3395         if (tb[IFLA_IFNAME])
3396                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3397
3398         if (tb[IFLA_EXT_MASK])
3399                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3400
3401         err = -EINVAL;
3402         ifm = nlmsg_data(nlh);
3403         if (ifm->ifi_index > 0)
3404                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3405         else if (tb[IFLA_IFNAME])
3406                 dev = __dev_get_by_name(tgt_net, ifname);
3407         else
3408                 goto out;
3409
3410         err = -ENODEV;
3411         if (dev == NULL)
3412                 goto out;
3413
3414         err = -ENOBUFS;
3415         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3416         if (nskb == NULL)
3417                 goto out;
3418
3419         err = rtnl_fill_ifinfo(nskb, dev, net,
3420                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3421                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3422                                0, NULL, 0, netnsid, GFP_KERNEL);
3423         if (err < 0) {
3424                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3425                 WARN_ON(err == -EMSGSIZE);
3426                 kfree_skb(nskb);
3427         } else
3428                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3429 out:
3430         if (netnsid >= 0)
3431                 put_net(tgt_net);
3432
3433         return err;
3434 }
3435
3436 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3437 {
3438         struct net *net = sock_net(skb->sk);
3439         struct net_device *dev;
3440         struct nlattr *tb[IFLA_MAX+1];
3441         u32 ext_filter_mask = 0;
3442         u16 min_ifinfo_dump_size = 0;
3443         int hdrlen;
3444
3445         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3446         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3447                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3448
3449         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3450                 if (tb[IFLA_EXT_MASK])
3451                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3452         }
3453
3454         if (!ext_filter_mask)
3455                 return NLMSG_GOODSIZE;
3456         /*
3457          * traverse the list of net devices and compute the minimum
3458          * buffer size based upon the filter mask.
3459          */
3460         rcu_read_lock();
3461         for_each_netdev_rcu(net, dev) {
3462                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3463                                              if_nlmsg_size(dev,
3464                                                            ext_filter_mask));
3465         }
3466         rcu_read_unlock();
3467
3468         return nlmsg_total_size(min_ifinfo_dump_size);
3469 }
3470
3471 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3472 {
3473         int idx;
3474         int s_idx = cb->family;
3475         int type = cb->nlh->nlmsg_type - RTM_BASE;
3476         int ret = 0;
3477
3478         if (s_idx == 0)
3479                 s_idx = 1;
3480
3481         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3482                 struct rtnl_link **tab;
3483                 struct rtnl_link *link;
3484                 rtnl_dumpit_func dumpit;
3485
3486                 if (idx < s_idx || idx == PF_PACKET)
3487                         continue;
3488
3489                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3490                         continue;
3491
3492                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3493                 if (!tab)
3494                         continue;
3495
3496                 link = tab[type];
3497                 if (!link)
3498                         continue;
3499
3500                 dumpit = link->dumpit;
3501                 if (!dumpit)
3502                         continue;
3503
3504                 if (idx > s_idx) {
3505                         memset(&cb->args[0], 0, sizeof(cb->args));
3506                         cb->prev_seq = 0;
3507                         cb->seq = 0;
3508                 }
3509                 ret = dumpit(skb, cb);
3510                 if (ret)
3511                         break;
3512         }
3513         cb->family = idx;
3514
3515         return skb->len ? : ret;
3516 }
3517
3518 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3519                                        unsigned int change,
3520                                        u32 event, gfp_t flags, int *new_nsid,
3521                                        int new_ifindex)
3522 {
3523         struct net *net = dev_net(dev);
3524         struct sk_buff *skb;
3525         int err = -ENOBUFS;
3526         size_t if_info_size;
3527
3528         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3529         if (skb == NULL)
3530                 goto errout;
3531
3532         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3533                                type, 0, 0, change, 0, 0, event,
3534                                new_nsid, new_ifindex, -1, flags);
3535         if (err < 0) {
3536                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3537                 WARN_ON(err == -EMSGSIZE);
3538                 kfree_skb(skb);
3539                 goto errout;
3540         }
3541         return skb;
3542 errout:
3543         if (err < 0)
3544                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3545         return NULL;
3546 }
3547
3548 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3549 {
3550         struct net *net = dev_net(dev);
3551
3552         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3553 }
3554
3555 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3556                                unsigned int change, u32 event,
3557                                gfp_t flags, int *new_nsid, int new_ifindex)
3558 {
3559         struct sk_buff *skb;
3560
3561         if (dev->reg_state != NETREG_REGISTERED)
3562                 return;
3563
3564         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3565                                      new_ifindex);
3566         if (skb)
3567                 rtmsg_ifinfo_send(skb, dev, flags);
3568 }
3569
3570 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3571                   gfp_t flags)
3572 {
3573         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3574                            NULL, 0);
3575 }
3576
3577 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3578                          gfp_t flags, int *new_nsid, int new_ifindex)
3579 {
3580         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3581                            new_nsid, new_ifindex);
3582 }
3583
3584 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3585                                    struct net_device *dev,
3586                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3587                                    int type, unsigned int flags,
3588                                    int nlflags, u16 ndm_state)
3589 {
3590         struct nlmsghdr *nlh;
3591         struct ndmsg *ndm;
3592
3593         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3594         if (!nlh)
3595                 return -EMSGSIZE;
3596
3597         ndm = nlmsg_data(nlh);
3598         ndm->ndm_family  = AF_BRIDGE;
3599         ndm->ndm_pad1    = 0;
3600         ndm->ndm_pad2    = 0;
3601         ndm->ndm_flags   = flags;
3602         ndm->ndm_type    = 0;
3603         ndm->ndm_ifindex = dev->ifindex;
3604         ndm->ndm_state   = ndm_state;
3605
3606         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
3607                 goto nla_put_failure;
3608         if (vid)
3609                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3610                         goto nla_put_failure;
3611
3612         nlmsg_end(skb, nlh);
3613         return 0;
3614
3615 nla_put_failure:
3616         nlmsg_cancel(skb, nlh);
3617         return -EMSGSIZE;
3618 }
3619
3620 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
3621 {
3622         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3623                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
3624                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3625                0;
3626 }
3627
3628 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3629                             u16 ndm_state)
3630 {
3631         struct net *net = dev_net(dev);
3632         struct sk_buff *skb;
3633         int err = -ENOBUFS;
3634
3635         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
3636         if (!skb)
3637                 goto errout;
3638
3639         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3640                                       0, 0, type, NTF_SELF, 0, ndm_state);
3641         if (err < 0) {
3642                 kfree_skb(skb);
3643                 goto errout;
3644         }
3645
3646         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3647         return;
3648 errout:
3649         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3650 }
3651
3652 /*
3653  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3654  */
3655 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3656                      struct nlattr *tb[],
3657                      struct net_device *dev,
3658                      const unsigned char *addr, u16 vid,
3659                      u16 flags)
3660 {
3661         int err = -EINVAL;
3662
3663         /* If aging addresses are supported device will need to
3664          * implement its own handler for this.
3665          */
3666         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3667                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3668                 return err;
3669         }
3670
3671         if (vid) {
3672                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3673                 return err;
3674         }
3675
3676         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3677                 err = dev_uc_add_excl(dev, addr);
3678         else if (is_multicast_ether_addr(addr))
3679                 err = dev_mc_add_excl(dev, addr);
3680
3681         /* Only return duplicate errors if NLM_F_EXCL is set */
3682         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3683                 err = 0;
3684
3685         return err;
3686 }
3687 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3688
3689 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3690                          struct netlink_ext_ack *extack)
3691 {
3692         u16 vid = 0;
3693
3694         if (vlan_attr) {
3695                 if (nla_len(vlan_attr) != sizeof(u16)) {
3696                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3697                         return -EINVAL;
3698                 }
3699
3700                 vid = nla_get_u16(vlan_attr);
3701
3702                 if (!vid || vid >= VLAN_VID_MASK) {
3703                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3704                         return -EINVAL;
3705                 }
3706         }
3707         *p_vid = vid;
3708         return 0;
3709 }
3710
3711 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3712                         struct netlink_ext_ack *extack)
3713 {
3714         struct net *net = sock_net(skb->sk);
3715         struct ndmsg *ndm;
3716         struct nlattr *tb[NDA_MAX+1];
3717         struct net_device *dev;
3718         u8 *addr;
3719         u16 vid;
3720         int err;
3721
3722         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3723                                      extack);
3724         if (err < 0)
3725                 return err;
3726
3727         ndm = nlmsg_data(nlh);
3728         if (ndm->ndm_ifindex == 0) {
3729                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3730                 return -EINVAL;
3731         }
3732
3733         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3734         if (dev == NULL) {
3735                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3736                 return -ENODEV;
3737         }
3738
3739         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3740                 NL_SET_ERR_MSG(extack, "invalid address");
3741                 return -EINVAL;
3742         }
3743
3744         if (dev->type != ARPHRD_ETHER) {
3745                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3746                 return -EINVAL;
3747         }
3748
3749         addr = nla_data(tb[NDA_LLADDR]);
3750
3751         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3752         if (err)
3753                 return err;
3754
3755         err = -EOPNOTSUPP;
3756
3757         /* Support fdb on master device the net/bridge default case */
3758         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3759             netif_is_bridge_port(dev)) {
3760                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3761                 const struct net_device_ops *ops = br_dev->netdev_ops;
3762
3763                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3764                                        nlh->nlmsg_flags, extack);
3765                 if (err)
3766                         goto out;
3767                 else
3768                         ndm->ndm_flags &= ~NTF_MASTER;
3769         }
3770
3771         /* Embedded bridge, macvlan, and any other device support */
3772         if ((ndm->ndm_flags & NTF_SELF)) {
3773                 if (dev->netdev_ops->ndo_fdb_add)
3774                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3775                                                            vid,
3776                                                            nlh->nlmsg_flags,
3777                                                            extack);
3778                 else
3779                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3780                                                nlh->nlmsg_flags);
3781
3782                 if (!err) {
3783                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3784                                         ndm->ndm_state);
3785                         ndm->ndm_flags &= ~NTF_SELF;
3786                 }
3787         }
3788 out:
3789         return err;
3790 }
3791
3792 /*
3793  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3794  */
3795 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3796                      struct nlattr *tb[],
3797                      struct net_device *dev,
3798                      const unsigned char *addr, u16 vid)
3799 {
3800         int err = -EINVAL;
3801
3802         /* If aging addresses are supported device will need to
3803          * implement its own handler for this.
3804          */
3805         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3806                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3807                 return err;
3808         }
3809
3810         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3811                 err = dev_uc_del(dev, addr);
3812         else if (is_multicast_ether_addr(addr))
3813                 err = dev_mc_del(dev, addr);
3814
3815         return err;
3816 }
3817 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3818
3819 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3820                         struct netlink_ext_ack *extack)
3821 {
3822         struct net *net = sock_net(skb->sk);
3823         struct ndmsg *ndm;
3824         struct nlattr *tb[NDA_MAX+1];
3825         struct net_device *dev;
3826         int err = -EINVAL;
3827         __u8 *addr;
3828         u16 vid;
3829
3830         if (!netlink_capable(skb, CAP_NET_ADMIN))
3831                 return -EPERM;
3832
3833         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3834                                      extack);
3835         if (err < 0)
3836                 return err;
3837
3838         ndm = nlmsg_data(nlh);
3839         if (ndm->ndm_ifindex == 0) {
3840                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3841                 return -EINVAL;
3842         }
3843
3844         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3845         if (dev == NULL) {
3846                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3847                 return -ENODEV;
3848         }
3849
3850         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3851                 NL_SET_ERR_MSG(extack, "invalid address");
3852                 return -EINVAL;
3853         }
3854
3855         if (dev->type != ARPHRD_ETHER) {
3856                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3857                 return -EINVAL;
3858         }
3859
3860         addr = nla_data(tb[NDA_LLADDR]);
3861
3862         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3863         if (err)
3864                 return err;
3865
3866         err = -EOPNOTSUPP;
3867
3868         /* Support fdb on master device the net/bridge default case */
3869         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3870             netif_is_bridge_port(dev)) {
3871                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3872                 const struct net_device_ops *ops = br_dev->netdev_ops;
3873
3874                 if (ops->ndo_fdb_del)
3875                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3876
3877                 if (err)
3878                         goto out;
3879                 else
3880                         ndm->ndm_flags &= ~NTF_MASTER;
3881         }
3882
3883         /* Embedded bridge, macvlan, and any other device support */
3884         if (ndm->ndm_flags & NTF_SELF) {
3885                 if (dev->netdev_ops->ndo_fdb_del)
3886                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3887                                                            vid);
3888                 else
3889                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3890
3891                 if (!err) {
3892                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3893                                         ndm->ndm_state);
3894                         ndm->ndm_flags &= ~NTF_SELF;
3895                 }
3896         }
3897 out:
3898         return err;
3899 }
3900
3901 static int nlmsg_populate_fdb(struct sk_buff *skb,
3902                               struct netlink_callback *cb,
3903                               struct net_device *dev,
3904                               int *idx,
3905                               struct netdev_hw_addr_list *list)
3906 {
3907         struct netdev_hw_addr *ha;
3908         int err;
3909         u32 portid, seq;
3910
3911         portid = NETLINK_CB(cb->skb).portid;
3912         seq = cb->nlh->nlmsg_seq;
3913
3914         list_for_each_entry(ha, &list->list, list) {
3915                 if (*idx < cb->args[2])
3916                         goto skip;
3917
3918                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3919                                               portid, seq,
3920                                               RTM_NEWNEIGH, NTF_SELF,
3921                                               NLM_F_MULTI, NUD_PERMANENT);
3922                 if (err < 0)
3923                         return err;
3924 skip:
3925                 *idx += 1;
3926         }
3927         return 0;
3928 }
3929
3930 /**
3931  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3932  * @skb: socket buffer to store message in
3933  * @cb: netlink callback
3934  * @dev: netdevice
3935  * @filter_dev: ignored
3936  * @idx: the number of FDB table entries dumped is added to *@idx
3937  *
3938  * Default netdevice operation to dump the existing unicast address list.
3939  * Returns number of addresses from list put in skb.
3940  */
3941 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3942                       struct netlink_callback *cb,
3943                       struct net_device *dev,
3944                       struct net_device *filter_dev,
3945                       int *idx)
3946 {
3947         int err;
3948
3949         if (dev->type != ARPHRD_ETHER)
3950                 return -EINVAL;
3951
3952         netif_addr_lock_bh(dev);
3953         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3954         if (err)
3955                 goto out;
3956         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3957 out:
3958         netif_addr_unlock_bh(dev);
3959         return err;
3960 }
3961 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3962
3963 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
3964                                  int *br_idx, int *brport_idx,
3965                                  struct netlink_ext_ack *extack)
3966 {
3967         struct nlattr *tb[NDA_MAX + 1];
3968         struct ndmsg *ndm;
3969         int err, i;
3970
3971         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
3972                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
3973                 return -EINVAL;
3974         }
3975
3976         ndm = nlmsg_data(nlh);
3977         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
3978             ndm->ndm_flags || ndm->ndm_type) {
3979                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
3980                 return -EINVAL;
3981         }
3982
3983         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
3984                                             NDA_MAX, NULL, extack);
3985         if (err < 0)
3986                 return err;
3987
3988         *brport_idx = ndm->ndm_ifindex;
3989         for (i = 0; i <= NDA_MAX; ++i) {
3990                 if (!tb[i])
3991                         continue;
3992
3993                 switch (i) {
3994                 case NDA_IFINDEX:
3995                         if (nla_len(tb[i]) != sizeof(u32)) {
3996                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
3997                                 return -EINVAL;
3998                         }
3999                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4000                         break;
4001                 case NDA_MASTER:
4002                         if (nla_len(tb[i]) != sizeof(u32)) {
4003                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4004                                 return -EINVAL;
4005                         }
4006                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4007                         break;
4008                 default:
4009                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4010                         return -EINVAL;
4011                 }
4012         }
4013
4014         return 0;
4015 }
4016
4017 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4018                                  int *br_idx, int *brport_idx,
4019                                  struct netlink_ext_ack *extack)
4020 {
4021         struct nlattr *tb[IFLA_MAX+1];
4022         int err;
4023
4024         /* A hack to preserve kernel<->userspace interface.
4025          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4026          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4027          * So, check for ndmsg with an optional u32 attribute (not used here).
4028          * Fortunately these sizes don't conflict with the size of ifinfomsg
4029          * with an optional attribute.
4030          */
4031         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4032             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4033              nla_attr_size(sizeof(u32)))) {
4034                 struct ifinfomsg *ifm;
4035
4036                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4037                                              tb, IFLA_MAX, ifla_policy,
4038                                              extack);
4039                 if (err < 0) {
4040                         return -EINVAL;
4041                 } else if (err == 0) {
4042                         if (tb[IFLA_MASTER])
4043                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4044                 }
4045
4046                 ifm = nlmsg_data(nlh);
4047                 *brport_idx = ifm->ifi_index;
4048         }
4049         return 0;
4050 }
4051
4052 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4053 {
4054         struct net_device *dev;
4055         struct net_device *br_dev = NULL;
4056         const struct net_device_ops *ops = NULL;
4057         const struct net_device_ops *cops = NULL;
4058         struct net *net = sock_net(skb->sk);
4059         struct hlist_head *head;
4060         int brport_idx = 0;
4061         int br_idx = 0;
4062         int h, s_h;
4063         int idx = 0, s_idx;
4064         int err = 0;
4065         int fidx = 0;
4066
4067         if (cb->strict_check)
4068                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4069                                             cb->extack);
4070         else
4071                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4072                                             cb->extack);
4073         if (err < 0)
4074                 return err;
4075
4076         if (br_idx) {
4077                 br_dev = __dev_get_by_index(net, br_idx);
4078                 if (!br_dev)
4079                         return -ENODEV;
4080
4081                 ops = br_dev->netdev_ops;
4082         }
4083
4084         s_h = cb->args[0];
4085         s_idx = cb->args[1];
4086
4087         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4088                 idx = 0;
4089                 head = &net->dev_index_head[h];
4090                 hlist_for_each_entry(dev, head, index_hlist) {
4091
4092                         if (brport_idx && (dev->ifindex != brport_idx))
4093                                 continue;
4094
4095                         if (!br_idx) { /* user did not specify a specific bridge */
4096                                 if (netif_is_bridge_port(dev)) {
4097                                         br_dev = netdev_master_upper_dev_get(dev);
4098                                         cops = br_dev->netdev_ops;
4099                                 }
4100                         } else {
4101                                 if (dev != br_dev &&
4102                                     !netif_is_bridge_port(dev))
4103                                         continue;
4104
4105                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4106                                     !(dev->priv_flags & IFF_EBRIDGE))
4107                                         continue;
4108                                 cops = ops;
4109                         }
4110
4111                         if (idx < s_idx)
4112                                 goto cont;
4113
4114                         if (netif_is_bridge_port(dev)) {
4115                                 if (cops && cops->ndo_fdb_dump) {
4116                                         err = cops->ndo_fdb_dump(skb, cb,
4117                                                                 br_dev, dev,
4118                                                                 &fidx);
4119                                         if (err == -EMSGSIZE)
4120                                                 goto out;
4121                                 }
4122                         }
4123
4124                         if (dev->netdev_ops->ndo_fdb_dump)
4125                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4126                                                                     dev, NULL,
4127                                                                     &fidx);
4128                         else
4129                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4130                                                         &fidx);
4131                         if (err == -EMSGSIZE)
4132                                 goto out;
4133
4134                         cops = NULL;
4135
4136                         /* reset fdb offset to 0 for rest of the interfaces */
4137                         cb->args[2] = 0;
4138                         fidx = 0;
4139 cont:
4140                         idx++;
4141                 }
4142         }
4143
4144 out:
4145         cb->args[0] = h;
4146         cb->args[1] = idx;
4147         cb->args[2] = fidx;
4148
4149         return skb->len;
4150 }
4151
4152 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4153                                 struct nlattr **tb, u8 *ndm_flags,
4154                                 int *br_idx, int *brport_idx, u8 **addr,
4155                                 u16 *vid, struct netlink_ext_ack *extack)
4156 {
4157         struct ndmsg *ndm;
4158         int err, i;
4159
4160         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4161                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4162                 return -EINVAL;
4163         }
4164
4165         ndm = nlmsg_data(nlh);
4166         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4167             ndm->ndm_type) {
4168                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4169                 return -EINVAL;
4170         }
4171
4172         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4173                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4174                 return -EINVAL;
4175         }
4176
4177         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4178                                             NDA_MAX, nda_policy, extack);
4179         if (err < 0)
4180                 return err;
4181
4182         *ndm_flags = ndm->ndm_flags;
4183         *brport_idx = ndm->ndm_ifindex;
4184         for (i = 0; i <= NDA_MAX; ++i) {
4185                 if (!tb[i])
4186                         continue;
4187
4188                 switch (i) {
4189                 case NDA_MASTER:
4190                         *br_idx = nla_get_u32(tb[i]);
4191                         break;
4192                 case NDA_LLADDR:
4193                         if (nla_len(tb[i]) != ETH_ALEN) {
4194                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4195                                 return -EINVAL;
4196                         }
4197                         *addr = nla_data(tb[i]);
4198                         break;
4199                 case NDA_VLAN:
4200                         err = fdb_vid_parse(tb[i], vid, extack);
4201                         if (err)
4202                                 return err;
4203                         break;
4204                 case NDA_VNI:
4205                         break;
4206                 default:
4207                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4208                         return -EINVAL;
4209                 }
4210         }
4211
4212         return 0;
4213 }
4214
4215 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4216                         struct netlink_ext_ack *extack)
4217 {
4218         struct net_device *dev = NULL, *br_dev = NULL;
4219         const struct net_device_ops *ops = NULL;
4220         struct net *net = sock_net(in_skb->sk);
4221         struct nlattr *tb[NDA_MAX + 1];
4222         struct sk_buff *skb;
4223         int brport_idx = 0;
4224         u8 ndm_flags = 0;
4225         int br_idx = 0;
4226         u8 *addr = NULL;
4227         u16 vid = 0;
4228         int err;
4229
4230         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4231                                    &brport_idx, &addr, &vid, extack);
4232         if (err < 0)
4233                 return err;
4234
4235         if (!addr) {
4236                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4237                 return -EINVAL;
4238         }
4239
4240         if (brport_idx) {
4241                 dev = __dev_get_by_index(net, brport_idx);
4242                 if (!dev) {
4243                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4244                         return -ENODEV;
4245                 }
4246         }
4247
4248         if (br_idx) {
4249                 if (dev) {
4250                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4251                         return -EINVAL;
4252                 }
4253
4254                 br_dev = __dev_get_by_index(net, br_idx);
4255                 if (!br_dev) {
4256                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4257                         return -EINVAL;
4258                 }
4259                 ops = br_dev->netdev_ops;
4260         }
4261
4262         if (dev) {
4263                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4264                         if (!netif_is_bridge_port(dev)) {
4265                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4266                                 return -EINVAL;
4267                         }
4268                         br_dev = netdev_master_upper_dev_get(dev);
4269                         if (!br_dev) {
4270                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4271                                 return -EINVAL;
4272                         }
4273                         ops = br_dev->netdev_ops;
4274                 } else {
4275                         if (!(ndm_flags & NTF_SELF)) {
4276                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4277                                 return -EINVAL;
4278                         }
4279                         ops = dev->netdev_ops;
4280                 }
4281         }
4282
4283         if (!br_dev && !dev) {
4284                 NL_SET_ERR_MSG(extack, "No device specified");
4285                 return -ENODEV;
4286         }
4287
4288         if (!ops || !ops->ndo_fdb_get) {
4289                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4290                 return -EOPNOTSUPP;
4291         }
4292
4293         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4294         if (!skb)
4295                 return -ENOBUFS;
4296
4297         if (br_dev)
4298                 dev = br_dev;
4299         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4300                                NETLINK_CB(in_skb).portid,
4301                                nlh->nlmsg_seq, extack);
4302         if (err)
4303                 goto out;
4304
4305         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4306 out:
4307         kfree_skb(skb);
4308         return err;
4309 }
4310
4311 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4312                                unsigned int attrnum, unsigned int flag)
4313 {
4314         if (mask & flag)
4315                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4316         return 0;
4317 }
4318
4319 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4320                             struct net_device *dev, u16 mode,
4321                             u32 flags, u32 mask, int nlflags,
4322                             u32 filter_mask,
4323                             int (*vlan_fill)(struct sk_buff *skb,
4324                                              struct net_device *dev,
4325                                              u32 filter_mask))
4326 {
4327         struct nlmsghdr *nlh;
4328         struct ifinfomsg *ifm;
4329         struct nlattr *br_afspec;
4330         struct nlattr *protinfo;
4331         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4332         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4333         int err = 0;
4334
4335         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4336         if (nlh == NULL)
4337                 return -EMSGSIZE;
4338
4339         ifm = nlmsg_data(nlh);
4340         ifm->ifi_family = AF_BRIDGE;
4341         ifm->__ifi_pad = 0;
4342         ifm->ifi_type = dev->type;
4343         ifm->ifi_index = dev->ifindex;
4344         ifm->ifi_flags = dev_get_flags(dev);
4345         ifm->ifi_change = 0;
4346
4347
4348         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4349             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4350             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4351             (br_dev &&
4352              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4353             (dev->addr_len &&
4354              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4355             (dev->ifindex != dev_get_iflink(dev) &&
4356              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4357                 goto nla_put_failure;
4358
4359         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4360         if (!br_afspec)
4361                 goto nla_put_failure;
4362
4363         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4364                 nla_nest_cancel(skb, br_afspec);
4365                 goto nla_put_failure;
4366         }
4367
4368         if (mode != BRIDGE_MODE_UNDEF) {
4369                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4370                         nla_nest_cancel(skb, br_afspec);
4371                         goto nla_put_failure;
4372                 }
4373         }
4374         if (vlan_fill) {
4375                 err = vlan_fill(skb, dev, filter_mask);
4376                 if (err) {
4377                         nla_nest_cancel(skb, br_afspec);
4378                         goto nla_put_failure;
4379                 }
4380         }
4381         nla_nest_end(skb, br_afspec);
4382
4383         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4384         if (!protinfo)
4385                 goto nla_put_failure;
4386
4387         if (brport_nla_put_flag(skb, flags, mask,
4388                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4389             brport_nla_put_flag(skb, flags, mask,
4390                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4391             brport_nla_put_flag(skb, flags, mask,
4392                                 IFLA_BRPORT_FAST_LEAVE,
4393                                 BR_MULTICAST_FAST_LEAVE) ||
4394             brport_nla_put_flag(skb, flags, mask,
4395                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4396             brport_nla_put_flag(skb, flags, mask,
4397                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4398             brport_nla_put_flag(skb, flags, mask,
4399                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4400             brport_nla_put_flag(skb, flags, mask,
4401                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4402             brport_nla_put_flag(skb, flags, mask,
4403                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
4404                 nla_nest_cancel(skb, protinfo);
4405                 goto nla_put_failure;
4406         }
4407
4408         nla_nest_end(skb, protinfo);
4409
4410         nlmsg_end(skb, nlh);
4411         return 0;
4412 nla_put_failure:
4413         nlmsg_cancel(skb, nlh);
4414         return err ? err : -EMSGSIZE;
4415 }
4416 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4417
4418 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4419                                     bool strict_check, u32 *filter_mask,
4420                                     struct netlink_ext_ack *extack)
4421 {
4422         struct nlattr *tb[IFLA_MAX+1];
4423         int err, i;
4424
4425         if (strict_check) {
4426                 struct ifinfomsg *ifm;
4427
4428                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4429                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4430                         return -EINVAL;
4431                 }
4432
4433                 ifm = nlmsg_data(nlh);
4434                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4435                     ifm->ifi_change || ifm->ifi_index) {
4436                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4437                         return -EINVAL;
4438                 }
4439
4440                 err = nlmsg_parse_deprecated_strict(nlh,
4441                                                     sizeof(struct ifinfomsg),
4442                                                     tb, IFLA_MAX, ifla_policy,
4443                                                     extack);
4444         } else {
4445                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4446                                              tb, IFLA_MAX, ifla_policy,
4447                                              extack);
4448         }
4449         if (err < 0)
4450                 return err;
4451
4452         /* new attributes should only be added with strict checking */
4453         for (i = 0; i <= IFLA_MAX; ++i) {
4454                 if (!tb[i])
4455                         continue;
4456
4457                 switch (i) {
4458                 case IFLA_EXT_MASK:
4459                         *filter_mask = nla_get_u32(tb[i]);
4460                         break;
4461                 default:
4462                         if (strict_check) {
4463                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4464                                 return -EINVAL;
4465                         }
4466                 }
4467         }
4468
4469         return 0;
4470 }
4471
4472 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4473 {
4474         const struct nlmsghdr *nlh = cb->nlh;
4475         struct net *net = sock_net(skb->sk);
4476         struct net_device *dev;
4477         int idx = 0;
4478         u32 portid = NETLINK_CB(cb->skb).portid;
4479         u32 seq = nlh->nlmsg_seq;
4480         u32 filter_mask = 0;
4481         int err;
4482
4483         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4484                                        cb->extack);
4485         if (err < 0 && cb->strict_check)
4486                 return err;
4487
4488         rcu_read_lock();
4489         for_each_netdev_rcu(net, dev) {
4490                 const struct net_device_ops *ops = dev->netdev_ops;
4491                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4492
4493                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4494                         if (idx >= cb->args[0]) {
4495                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4496                                                 skb, portid, seq, dev,
4497                                                 filter_mask, NLM_F_MULTI);
4498                                 if (err < 0 && err != -EOPNOTSUPP) {
4499                                         if (likely(skb->len))
4500                                                 break;
4501
4502                                         goto out_err;
4503                                 }
4504                         }
4505                         idx++;
4506                 }
4507
4508                 if (ops->ndo_bridge_getlink) {
4509                         if (idx >= cb->args[0]) {
4510                                 err = ops->ndo_bridge_getlink(skb, portid,
4511                                                               seq, dev,
4512                                                               filter_mask,
4513                                                               NLM_F_MULTI);
4514                                 if (err < 0 && err != -EOPNOTSUPP) {
4515                                         if (likely(skb->len))
4516                                                 break;
4517
4518                                         goto out_err;
4519                                 }
4520                         }
4521                         idx++;
4522                 }
4523         }
4524         err = skb->len;
4525 out_err:
4526         rcu_read_unlock();
4527         cb->args[0] = idx;
4528
4529         return err;
4530 }
4531
4532 static inline size_t bridge_nlmsg_size(void)
4533 {
4534         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4535                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4536                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4537                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4538                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4539                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4540                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4541                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4542                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4543                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4544                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4545 }
4546
4547 static int rtnl_bridge_notify(struct net_device *dev)
4548 {
4549         struct net *net = dev_net(dev);
4550         struct sk_buff *skb;
4551         int err = -EOPNOTSUPP;
4552
4553         if (!dev->netdev_ops->ndo_bridge_getlink)
4554                 return 0;
4555
4556         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4557         if (!skb) {
4558                 err = -ENOMEM;
4559                 goto errout;
4560         }
4561
4562         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4563         if (err < 0)
4564                 goto errout;
4565
4566         /* Notification info is only filled for bridge ports, not the bridge
4567          * device itself. Therefore, a zero notification length is valid and
4568          * should not result in an error.
4569          */
4570         if (!skb->len)
4571                 goto errout;
4572
4573         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4574         return 0;
4575 errout:
4576         WARN_ON(err == -EMSGSIZE);
4577         kfree_skb(skb);
4578         if (err)
4579                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4580         return err;
4581 }
4582
4583 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4584                                struct netlink_ext_ack *extack)
4585 {
4586         struct net *net = sock_net(skb->sk);
4587         struct ifinfomsg *ifm;
4588         struct net_device *dev;
4589         struct nlattr *br_spec, *attr = NULL;
4590         int rem, err = -EOPNOTSUPP;
4591         u16 flags = 0;
4592         bool have_flags = false;
4593
4594         if (nlmsg_len(nlh) < sizeof(*ifm))
4595                 return -EINVAL;
4596
4597         ifm = nlmsg_data(nlh);
4598         if (ifm->ifi_family != AF_BRIDGE)
4599                 return -EPFNOSUPPORT;
4600
4601         dev = __dev_get_by_index(net, ifm->ifi_index);
4602         if (!dev) {
4603                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4604                 return -ENODEV;
4605         }
4606
4607         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4608         if (br_spec) {
4609                 nla_for_each_nested(attr, br_spec, rem) {
4610                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
4611                                 if (nla_len(attr) < sizeof(flags))
4612                                         return -EINVAL;
4613
4614                                 have_flags = true;
4615                                 flags = nla_get_u16(attr);
4616                         }
4617
4618                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
4619                                 if (nla_len(attr) < sizeof(u16))
4620                                         return -EINVAL;
4621                         }
4622                 }
4623         }
4624
4625         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4626                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4627
4628                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4629                         err = -EOPNOTSUPP;
4630                         goto out;
4631                 }
4632
4633                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4634                                                              extack);
4635                 if (err)
4636                         goto out;
4637
4638                 flags &= ~BRIDGE_FLAGS_MASTER;
4639         }
4640
4641         if ((flags & BRIDGE_FLAGS_SELF)) {
4642                 if (!dev->netdev_ops->ndo_bridge_setlink)
4643                         err = -EOPNOTSUPP;
4644                 else
4645                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4646                                                                   flags,
4647                                                                   extack);
4648                 if (!err) {
4649                         flags &= ~BRIDGE_FLAGS_SELF;
4650
4651                         /* Generate event to notify upper layer of bridge
4652                          * change
4653                          */
4654                         err = rtnl_bridge_notify(dev);
4655                 }
4656         }
4657
4658         if (have_flags)
4659                 memcpy(nla_data(attr), &flags, sizeof(flags));
4660 out:
4661         return err;
4662 }
4663
4664 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4665                                struct netlink_ext_ack *extack)
4666 {
4667         struct net *net = sock_net(skb->sk);
4668         struct ifinfomsg *ifm;
4669         struct net_device *dev;
4670         struct nlattr *br_spec, *attr = NULL;
4671         int rem, err = -EOPNOTSUPP;
4672         u16 flags = 0;
4673         bool have_flags = false;
4674
4675         if (nlmsg_len(nlh) < sizeof(*ifm))
4676                 return -EINVAL;
4677
4678         ifm = nlmsg_data(nlh);
4679         if (ifm->ifi_family != AF_BRIDGE)
4680                 return -EPFNOSUPPORT;
4681
4682         dev = __dev_get_by_index(net, ifm->ifi_index);
4683         if (!dev) {
4684                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4685                 return -ENODEV;
4686         }
4687
4688         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4689         if (br_spec) {
4690                 nla_for_each_nested(attr, br_spec, rem) {
4691                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4692                                 if (nla_len(attr) < sizeof(flags))
4693                                         return -EINVAL;
4694
4695                                 have_flags = true;
4696                                 flags = nla_get_u16(attr);
4697                                 break;
4698                         }
4699                 }
4700         }
4701
4702         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4703                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4704
4705                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4706                         err = -EOPNOTSUPP;
4707                         goto out;
4708                 }
4709
4710                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4711                 if (err)
4712                         goto out;
4713
4714                 flags &= ~BRIDGE_FLAGS_MASTER;
4715         }
4716
4717         if ((flags & BRIDGE_FLAGS_SELF)) {
4718                 if (!dev->netdev_ops->ndo_bridge_dellink)
4719                         err = -EOPNOTSUPP;
4720                 else
4721                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4722                                                                   flags);
4723
4724                 if (!err) {
4725                         flags &= ~BRIDGE_FLAGS_SELF;
4726
4727                         /* Generate event to notify upper layer of bridge
4728                          * change
4729                          */
4730                         err = rtnl_bridge_notify(dev);
4731                 }
4732         }
4733
4734         if (have_flags)
4735                 memcpy(nla_data(attr), &flags, sizeof(flags));
4736 out:
4737         return err;
4738 }
4739
4740 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4741 {
4742         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4743                (!idxattr || idxattr == attrid);
4744 }
4745
4746 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4747 static int rtnl_get_offload_stats_attr_size(int attr_id)
4748 {
4749         switch (attr_id) {
4750         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4751                 return sizeof(struct rtnl_link_stats64);
4752         }
4753
4754         return 0;
4755 }
4756
4757 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4758                                   int *prividx)
4759 {
4760         struct nlattr *attr = NULL;
4761         int attr_id, size;
4762         void *attr_data;
4763         int err;
4764
4765         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4766               dev->netdev_ops->ndo_get_offload_stats))
4767                 return -ENODATA;
4768
4769         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4770              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4771                 if (attr_id < *prividx)
4772                         continue;
4773
4774                 size = rtnl_get_offload_stats_attr_size(attr_id);
4775                 if (!size)
4776                         continue;
4777
4778                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4779                         continue;
4780
4781                 attr = nla_reserve_64bit(skb, attr_id, size,
4782                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4783                 if (!attr)
4784                         goto nla_put_failure;
4785
4786                 attr_data = nla_data(attr);
4787                 memset(attr_data, 0, size);
4788                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4789                                                              attr_data);
4790                 if (err)
4791                         goto get_offload_stats_failure;
4792         }
4793
4794         if (!attr)
4795                 return -ENODATA;
4796
4797         *prividx = 0;
4798         return 0;
4799
4800 nla_put_failure:
4801         err = -EMSGSIZE;
4802 get_offload_stats_failure:
4803         *prividx = attr_id;
4804         return err;
4805 }
4806
4807 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4808 {
4809         int nla_size = 0;
4810         int attr_id;
4811         int size;
4812
4813         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4814               dev->netdev_ops->ndo_get_offload_stats))
4815                 return 0;
4816
4817         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4818              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4819                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4820                         continue;
4821                 size = rtnl_get_offload_stats_attr_size(attr_id);
4822                 nla_size += nla_total_size_64bit(size);
4823         }
4824
4825         if (nla_size != 0)
4826                 nla_size += nla_total_size(0);
4827
4828         return nla_size;
4829 }
4830
4831 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4832                                int type, u32 pid, u32 seq, u32 change,
4833                                unsigned int flags, unsigned int filter_mask,
4834                                int *idxattr, int *prividx)
4835 {
4836         struct if_stats_msg *ifsm;
4837         struct nlmsghdr *nlh;
4838         struct nlattr *attr;
4839         int s_prividx = *prividx;
4840         int err;
4841
4842         ASSERT_RTNL();
4843
4844         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4845         if (!nlh)
4846                 return -EMSGSIZE;
4847
4848         ifsm = nlmsg_data(nlh);
4849         ifsm->family = PF_UNSPEC;
4850         ifsm->pad1 = 0;
4851         ifsm->pad2 = 0;
4852         ifsm->ifindex = dev->ifindex;
4853         ifsm->filter_mask = filter_mask;
4854
4855         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4856                 struct rtnl_link_stats64 *sp;
4857
4858                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4859                                          sizeof(struct rtnl_link_stats64),
4860                                          IFLA_STATS_UNSPEC);
4861                 if (!attr)
4862                         goto nla_put_failure;
4863
4864                 sp = nla_data(attr);
4865                 dev_get_stats(dev, sp);
4866         }
4867
4868         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4869                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4870
4871                 if (ops && ops->fill_linkxstats) {
4872                         *idxattr = IFLA_STATS_LINK_XSTATS;
4873                         attr = nla_nest_start_noflag(skb,
4874                                                      IFLA_STATS_LINK_XSTATS);
4875                         if (!attr)
4876                                 goto nla_put_failure;
4877
4878                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4879                         nla_nest_end(skb, attr);
4880                         if (err)
4881                                 goto nla_put_failure;
4882                         *idxattr = 0;
4883                 }
4884         }
4885
4886         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4887                              *idxattr)) {
4888                 const struct rtnl_link_ops *ops = NULL;
4889                 const struct net_device *master;
4890
4891                 master = netdev_master_upper_dev_get(dev);
4892                 if (master)
4893                         ops = master->rtnl_link_ops;
4894                 if (ops && ops->fill_linkxstats) {
4895                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4896                         attr = nla_nest_start_noflag(skb,
4897                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
4898                         if (!attr)
4899                                 goto nla_put_failure;
4900
4901                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4902                         nla_nest_end(skb, attr);
4903                         if (err)
4904                                 goto nla_put_failure;
4905                         *idxattr = 0;
4906                 }
4907         }
4908
4909         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4910                              *idxattr)) {
4911                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4912                 attr = nla_nest_start_noflag(skb,
4913                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
4914                 if (!attr)
4915                         goto nla_put_failure;
4916
4917                 err = rtnl_get_offload_stats(skb, dev, prividx);
4918                 if (err == -ENODATA)
4919                         nla_nest_cancel(skb, attr);
4920                 else
4921                         nla_nest_end(skb, attr);
4922
4923                 if (err && err != -ENODATA)
4924                         goto nla_put_failure;
4925                 *idxattr = 0;
4926         }
4927
4928         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4929                 struct rtnl_af_ops *af_ops;
4930
4931                 *idxattr = IFLA_STATS_AF_SPEC;
4932                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
4933                 if (!attr)
4934                         goto nla_put_failure;
4935
4936                 rcu_read_lock();
4937                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4938                         if (af_ops->fill_stats_af) {
4939                                 struct nlattr *af;
4940                                 int err;
4941
4942                                 af = nla_nest_start_noflag(skb,
4943                                                            af_ops->family);
4944                                 if (!af) {
4945                                         rcu_read_unlock();
4946                                         goto nla_put_failure;
4947                                 }
4948                                 err = af_ops->fill_stats_af(skb, dev);
4949
4950                                 if (err == -ENODATA) {
4951                                         nla_nest_cancel(skb, af);
4952                                 } else if (err < 0) {
4953                                         rcu_read_unlock();
4954                                         goto nla_put_failure;
4955                                 }
4956
4957                                 nla_nest_end(skb, af);
4958                         }
4959                 }
4960                 rcu_read_unlock();
4961
4962                 nla_nest_end(skb, attr);
4963
4964                 *idxattr = 0;
4965         }
4966
4967         nlmsg_end(skb, nlh);
4968
4969         return 0;
4970
4971 nla_put_failure:
4972         /* not a multi message or no progress mean a real error */
4973         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4974                 nlmsg_cancel(skb, nlh);
4975         else
4976                 nlmsg_end(skb, nlh);
4977
4978         return -EMSGSIZE;
4979 }
4980
4981 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4982                                   u32 filter_mask)
4983 {
4984         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
4985
4986         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4987                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4988
4989         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4990                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4991                 int attr = IFLA_STATS_LINK_XSTATS;
4992
4993                 if (ops && ops->get_linkxstats_size) {
4994                         size += nla_total_size(ops->get_linkxstats_size(dev,
4995                                                                         attr));
4996                         /* for IFLA_STATS_LINK_XSTATS */
4997                         size += nla_total_size(0);
4998                 }
4999         }
5000
5001         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5002                 struct net_device *_dev = (struct net_device *)dev;
5003                 const struct rtnl_link_ops *ops = NULL;
5004                 const struct net_device *master;
5005
5006                 /* netdev_master_upper_dev_get can't take const */
5007                 master = netdev_master_upper_dev_get(_dev);
5008                 if (master)
5009                         ops = master->rtnl_link_ops;
5010                 if (ops && ops->get_linkxstats_size) {
5011                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5012
5013                         size += nla_total_size(ops->get_linkxstats_size(dev,
5014                                                                         attr));
5015                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5016                         size += nla_total_size(0);
5017                 }
5018         }
5019
5020         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5021                 size += rtnl_get_offload_stats_size(dev);
5022
5023         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5024                 struct rtnl_af_ops *af_ops;
5025
5026                 /* for IFLA_STATS_AF_SPEC */
5027                 size += nla_total_size(0);
5028
5029                 rcu_read_lock();
5030                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5031                         if (af_ops->get_stats_af_size) {
5032                                 size += nla_total_size(
5033                                         af_ops->get_stats_af_size(dev));
5034
5035                                 /* for AF_* */
5036                                 size += nla_total_size(0);
5037                         }
5038                 }
5039                 rcu_read_unlock();
5040         }
5041
5042         return size;
5043 }
5044
5045 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5046                                 bool is_dump, struct netlink_ext_ack *extack)
5047 {
5048         struct if_stats_msg *ifsm;
5049
5050         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5051                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5052                 return -EINVAL;
5053         }
5054
5055         if (!strict_check)
5056                 return 0;
5057
5058         ifsm = nlmsg_data(nlh);
5059
5060         /* only requests using strict checks can pass data to influence
5061          * the dump. The legacy exception is filter_mask.
5062          */
5063         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5064                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5065                 return -EINVAL;
5066         }
5067         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5068                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5069                 return -EINVAL;
5070         }
5071         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5072                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5073                 return -EINVAL;
5074         }
5075
5076         return 0;
5077 }
5078
5079 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5080                           struct netlink_ext_ack *extack)
5081 {
5082         struct net *net = sock_net(skb->sk);
5083         struct net_device *dev = NULL;
5084         int idxattr = 0, prividx = 0;
5085         struct if_stats_msg *ifsm;
5086         struct sk_buff *nskb;
5087         u32 filter_mask;
5088         int err;
5089
5090         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5091                                    false, extack);
5092         if (err)
5093                 return err;
5094
5095         ifsm = nlmsg_data(nlh);
5096         if (ifsm->ifindex > 0)
5097                 dev = __dev_get_by_index(net, ifsm->ifindex);
5098         else
5099                 return -EINVAL;
5100
5101         if (!dev)
5102                 return -ENODEV;
5103
5104         filter_mask = ifsm->filter_mask;
5105         if (!filter_mask)
5106                 return -EINVAL;
5107
5108         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5109         if (!nskb)
5110                 return -ENOBUFS;
5111
5112         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5113                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5114                                   0, filter_mask, &idxattr, &prividx);
5115         if (err < 0) {
5116                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5117                 WARN_ON(err == -EMSGSIZE);
5118                 kfree_skb(nskb);
5119         } else {
5120                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5121         }
5122
5123         return err;
5124 }
5125
5126 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5127 {
5128         struct netlink_ext_ack *extack = cb->extack;
5129         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5130         struct net *net = sock_net(skb->sk);
5131         unsigned int flags = NLM_F_MULTI;
5132         struct if_stats_msg *ifsm;
5133         struct hlist_head *head;
5134         struct net_device *dev;
5135         u32 filter_mask = 0;
5136         int idx = 0;
5137
5138         s_h = cb->args[0];
5139         s_idx = cb->args[1];
5140         s_idxattr = cb->args[2];
5141         s_prividx = cb->args[3];
5142
5143         cb->seq = net->dev_base_seq;
5144
5145         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5146         if (err)
5147                 return err;
5148
5149         ifsm = nlmsg_data(cb->nlh);
5150         filter_mask = ifsm->filter_mask;
5151         if (!filter_mask) {
5152                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5153                 return -EINVAL;
5154         }
5155
5156         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5157                 idx = 0;
5158                 head = &net->dev_index_head[h];
5159                 hlist_for_each_entry(dev, head, index_hlist) {
5160                         if (idx < s_idx)
5161                                 goto cont;
5162                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5163                                                   NETLINK_CB(cb->skb).portid,
5164                                                   cb->nlh->nlmsg_seq, 0,
5165                                                   flags, filter_mask,
5166                                                   &s_idxattr, &s_prividx);
5167                         /* If we ran out of room on the first message,
5168                          * we're in trouble
5169                          */
5170                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5171
5172                         if (err < 0)
5173                                 goto out;
5174                         s_prividx = 0;
5175                         s_idxattr = 0;
5176                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5177 cont:
5178                         idx++;
5179                 }
5180         }
5181 out:
5182         cb->args[3] = s_prividx;
5183         cb->args[2] = s_idxattr;
5184         cb->args[1] = idx;
5185         cb->args[0] = h;
5186
5187         return skb->len;
5188 }
5189
5190 /* Process one rtnetlink message. */
5191
5192 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5193                              struct netlink_ext_ack *extack)
5194 {
5195         struct net *net = sock_net(skb->sk);
5196         struct rtnl_link *link;
5197         struct module *owner;
5198         int err = -EOPNOTSUPP;
5199         rtnl_doit_func doit;
5200         unsigned int flags;
5201         int kind;
5202         int family;
5203         int type;
5204
5205         type = nlh->nlmsg_type;
5206         if (type > RTM_MAX)
5207                 return -EOPNOTSUPP;
5208
5209         type -= RTM_BASE;
5210
5211         /* All the messages must have at least 1 byte length */
5212         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5213                 return 0;
5214
5215         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5216         kind = type&3;
5217
5218         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5219                 return -EPERM;
5220
5221         rcu_read_lock();
5222         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5223                 struct sock *rtnl;
5224                 rtnl_dumpit_func dumpit;
5225                 u16 min_dump_alloc = 0;
5226
5227                 link = rtnl_get_link(family, type);
5228                 if (!link || !link->dumpit) {
5229                         family = PF_UNSPEC;
5230                         link = rtnl_get_link(family, type);
5231                         if (!link || !link->dumpit)
5232                                 goto err_unlock;
5233                 }
5234                 owner = link->owner;
5235                 dumpit = link->dumpit;
5236
5237                 if (type == RTM_GETLINK - RTM_BASE)
5238                         min_dump_alloc = rtnl_calcit(skb, nlh);
5239
5240                 err = 0;
5241                 /* need to do this before rcu_read_unlock() */
5242                 if (!try_module_get(owner))
5243                         err = -EPROTONOSUPPORT;
5244
5245                 rcu_read_unlock();
5246
5247                 rtnl = net->rtnl;
5248                 if (err == 0) {
5249                         struct netlink_dump_control c = {
5250                                 .dump           = dumpit,
5251                                 .min_dump_alloc = min_dump_alloc,
5252                                 .module         = owner,
5253                         };
5254                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5255                         /* netlink_dump_start() will keep a reference on
5256                          * module if dump is still in progress.
5257                          */
5258                         module_put(owner);
5259                 }
5260                 return err;
5261         }
5262
5263         link = rtnl_get_link(family, type);
5264         if (!link || !link->doit) {
5265                 family = PF_UNSPEC;
5266                 link = rtnl_get_link(PF_UNSPEC, type);
5267                 if (!link || !link->doit)
5268                         goto out_unlock;
5269         }
5270
5271         owner = link->owner;
5272         if (!try_module_get(owner)) {
5273                 err = -EPROTONOSUPPORT;
5274                 goto out_unlock;
5275         }
5276
5277         flags = link->flags;
5278         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5279                 doit = link->doit;
5280                 rcu_read_unlock();
5281                 if (doit)
5282                         err = doit(skb, nlh, extack);
5283                 module_put(owner);
5284                 return err;
5285         }
5286         rcu_read_unlock();
5287
5288         rtnl_lock();
5289         link = rtnl_get_link(family, type);
5290         if (link && link->doit)
5291                 err = link->doit(skb, nlh, extack);
5292         rtnl_unlock();
5293
5294         module_put(owner);
5295
5296         return err;
5297
5298 out_unlock:
5299         rcu_read_unlock();
5300         return err;
5301
5302 err_unlock:
5303         rcu_read_unlock();
5304         return -EOPNOTSUPP;
5305 }
5306
5307 static void rtnetlink_rcv(struct sk_buff *skb)
5308 {
5309         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5310 }
5311
5312 static int rtnetlink_bind(struct net *net, int group)
5313 {
5314         switch (group) {
5315         case RTNLGRP_IPV4_MROUTE_R:
5316         case RTNLGRP_IPV6_MROUTE_R:
5317                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5318                         return -EPERM;
5319                 break;
5320         }
5321         return 0;
5322 }
5323
5324 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5325 {
5326         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5327
5328         switch (event) {
5329         case NETDEV_REBOOT:
5330         case NETDEV_CHANGEMTU:
5331         case NETDEV_CHANGEADDR:
5332         case NETDEV_CHANGENAME:
5333         case NETDEV_FEAT_CHANGE:
5334         case NETDEV_BONDING_FAILOVER:
5335         case NETDEV_POST_TYPE_CHANGE:
5336         case NETDEV_NOTIFY_PEERS:
5337         case NETDEV_CHANGEUPPER:
5338         case NETDEV_RESEND_IGMP:
5339         case NETDEV_CHANGEINFODATA:
5340         case NETDEV_CHANGELOWERSTATE:
5341         case NETDEV_CHANGE_TX_QUEUE_LEN:
5342                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5343                                    GFP_KERNEL, NULL, 0);
5344                 break;
5345         default:
5346                 break;
5347         }
5348         return NOTIFY_DONE;
5349 }
5350
5351 static struct notifier_block rtnetlink_dev_notifier = {
5352         .notifier_call  = rtnetlink_event,
5353 };
5354
5355
5356 static int __net_init rtnetlink_net_init(struct net *net)
5357 {
5358         struct sock *sk;
5359         struct netlink_kernel_cfg cfg = {
5360                 .groups         = RTNLGRP_MAX,
5361                 .input          = rtnetlink_rcv,
5362                 .cb_mutex       = &rtnl_mutex,
5363                 .flags          = NL_CFG_F_NONROOT_RECV,
5364                 .bind           = rtnetlink_bind,
5365         };
5366
5367         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5368         if (!sk)
5369                 return -ENOMEM;
5370         net->rtnl = sk;
5371         return 0;
5372 }
5373
5374 static void __net_exit rtnetlink_net_exit(struct net *net)
5375 {
5376         netlink_kernel_release(net->rtnl);
5377         net->rtnl = NULL;
5378 }
5379
5380 static struct pernet_operations rtnetlink_net_ops = {
5381         .init = rtnetlink_net_init,
5382         .exit = rtnetlink_net_exit,
5383 };
5384
5385 void __init rtnetlink_init(void)
5386 {
5387         if (register_pernet_subsys(&rtnetlink_net_ops))
5388                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5389
5390         register_netdevice_notifier(&rtnetlink_dev_notifier);
5391
5392         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5393                       rtnl_dump_ifinfo, 0);
5394         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5395         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5396         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5397
5398         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5399         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5400         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5401
5402         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5403         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5404         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5405
5406         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5407         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5408         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5409
5410         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5411                       0);
5412 }