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 arithmetic 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     40
  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 *__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 __rcu **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 rcu_dereference_rtnl(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_dereference(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 modify 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 modify 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 __rcu **tab;
 290         struct rtnl_link *link;
 291         int msgindex;
 292
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294         msgindex = rtm_msgindex(msgtype);
 295
 296         rtnl_lock();
 297         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 298         if (!tab) {
 299                 rtnl_unlock();
 300                 return -ENOENT;
 301         }
 302
 303         link = rtnl_dereference(tab[msgindex]);
 304         rcu_assign_pointer(tab[msgindex], NULL);
 305         rtnl_unlock();
 306
 307         kfree_rcu(link, rcu);
 308
 309         return 0;
 310 }
 311 EXPORT_SYMBOL_GPL(rtnl_unregister);
 312
 313 /**
 314  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 315  * @protocol : Protocol family or PF_UNSPEC
 316  *
 317  * Identical to calling rtnl_unregster() for all registered message types
 318  * of a certain protocol family.
 319  */
 320 void rtnl_unregister_all(int protocol)
 321 {
 322         struct rtnl_link __rcu **tab;
 323         struct rtnl_link *link;
 324         int msgindex;
 325
 326         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 327
 328         rtnl_lock();
 329         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 330         if (!tab) {
 331                 rtnl_unlock();
 332                 return;
 333         }
 334         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 335         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 336                 link = rtnl_dereference(tab[msgindex]);
 337                 if (!link)
 338                         continue;
 339
 340                 rcu_assign_pointer(tab[msgindex], NULL);
 341                 kfree_rcu(link, rcu);
 342         }
 343         rtnl_unlock();
 344
 345         synchronize_net();
 346
 347         kfree(tab);
 348 }
 349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 350
 351 static LIST_HEAD(link_ops);
 352
 353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 354 {
 355         const struct rtnl_link_ops *ops;
 356
 357         list_for_each_entry(ops, &link_ops, list) {
 358                 if (!strcmp(ops->kind, kind))
 359                         return ops;
 360         }
 361         return NULL;
 362 }
 363
 364 /**
 365  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 366  * @ops: struct rtnl_link_ops * to register
 367  *
 368  * The caller must hold the rtnl_mutex. This function should be used
 369  * by drivers that create devices during module initialization. It
 370  * must be called before registering the devices.
 371  *
 372  * Returns 0 on success or a negative error code.
 373  */
 374 int __rtnl_link_register(struct rtnl_link_ops *ops)
 375 {
 376         if (rtnl_link_ops_get(ops->kind))
 377                 return -EEXIST;
 378
 379         /* The check for alloc/setup is here because if ops
 380          * does not have that filled up, it is not possible
 381          * to use the ops for creating device. So do not
 382          * fill up dellink as well. That disables rtnl_dellink.
 383          */
 384         if ((ops->alloc || ops->setup) && !ops->dellink)
 385                 ops->dellink = unregister_netdevice_queue;
 386
 387         list_add_tail(&ops->list, &link_ops);
 388         return 0;
 389 }
 390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 391
 392 /**
 393  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 394  * @ops: struct rtnl_link_ops * to register
 395  *
 396  * Returns 0 on success or a negative error code.
 397  */
 398 int rtnl_link_register(struct rtnl_link_ops *ops)
 399 {
 400         int err;
 401
 402         /* Sanity-check max sizes to avoid stack buffer overflow. */
 403         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 404                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 405                 return -EINVAL;
 406
 407         rtnl_lock();
 408         err = __rtnl_link_register(ops);
 409         rtnl_unlock();
 410         return err;
 411 }
 412 EXPORT_SYMBOL_GPL(rtnl_link_register);
 413
 414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 415 {
 416         struct net_device *dev;
 417         LIST_HEAD(list_kill);
 418
 419         for_each_netdev(net, dev) {
 420                 if (dev->rtnl_link_ops == ops)
 421                         ops->dellink(dev, &list_kill);
 422         }
 423         unregister_netdevice_many(&list_kill);
 424 }
 425
 426 /**
 427  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 428  * @ops: struct rtnl_link_ops * to unregister
 429  *
 430  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 431  * integrity (hold pernet_ops_rwsem for writing to close the race
 432  * with setup_net() and cleanup_net()).
 433  */
 434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 435 {
 436         struct net *net;
 437
 438         for_each_net(net) {
 439                 __rtnl_kill_links(net, ops);
 440         }
 441         list_del(&ops->list);
 442 }
 443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 444
 445 /* Return with the rtnl_lock held when there are no network
 446  * devices unregistering in any network namespace.
 447  */
 448 static void rtnl_lock_unregistering_all(void)
 449 {
 450         struct net *net;
 451         bool unregistering;
 452         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 453
 454         add_wait_queue(&netdev_unregistering_wq, &wait);
 455         for (;;) {
 456                 unregistering = false;
 457                 rtnl_lock();
 458                 /* We held write locked pernet_ops_rwsem, and parallel
 459                  * setup_net() and cleanup_net() are not possible.
 460                  */
 461                 for_each_net(net) {
 462                         if (net->dev_unreg_count > 0) {
 463                                 unregistering = true;
 464                                 break;
 465                         }
 466                 }
 467                 if (!unregistering)
 468                         break;
 469                 __rtnl_unlock();
 470
 471                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 472         }
 473         remove_wait_queue(&netdev_unregistering_wq, &wait);
 474 }
 475
 476 /**
 477  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 478  * @ops: struct rtnl_link_ops * to unregister
 479  */
 480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 481 {
 482         /* Close the race with setup_net() and cleanup_net() */
 483         down_write(&pernet_ops_rwsem);
 484         rtnl_lock_unregistering_all();
 485         __rtnl_link_unregister(ops);
 486         rtnl_unlock();
 487         up_write(&pernet_ops_rwsem);
 488 }
 489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 490
 491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 492 {
 493         struct net_device *master_dev;
 494         const struct rtnl_link_ops *ops;
 495         size_t size = 0;
 496
 497         rcu_read_lock();
 498
 499         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 500         if (!master_dev)
 501                 goto out;
 502
 503         ops = master_dev->rtnl_link_ops;
 504         if (!ops || !ops->get_slave_size)
 505                 goto out;
 506         /* IFLA_INFO_SLAVE_DATA + nested data */
 507         size = nla_total_size(sizeof(struct nlattr)) +
 508                ops->get_slave_size(master_dev, dev);
 509
 510 out:
 511         rcu_read_unlock();
 512         return size;
 513 }
 514
 515 static size_t rtnl_link_get_size(const struct net_device *dev)
 516 {
 517         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 518         size_t size;
 519
 520         if (!ops)
 521                 return 0;
 522
 523         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 524                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 525
 526         if (ops->get_size)
 527                 /* IFLA_INFO_DATA + nested data */
 528                 size += nla_total_size(sizeof(struct nlattr)) +
 529                         ops->get_size(dev);
 530
 531         if (ops->get_xstats_size)
 532                 /* IFLA_INFO_XSTATS */
 533                 size += nla_total_size(ops->get_xstats_size(dev));
 534
 535         size += rtnl_link_get_slave_info_data_size(dev);
 536
 537         return size;
 538 }
 539
 540 static LIST_HEAD(rtnl_af_ops);
 541
 542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 543 {
 544         const struct rtnl_af_ops *ops;
 545
 546         ASSERT_RTNL();
 547
 548         list_for_each_entry(ops, &rtnl_af_ops, list) {
 549                 if (ops->family == family)
 550                         return ops;
 551         }
 552
 553         return NULL;
 554 }
 555
 556 /**
 557  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 558  * @ops: struct rtnl_af_ops * to register
 559  *
 560  * Returns 0 on success or a negative error code.
 561  */
 562 void rtnl_af_register(struct rtnl_af_ops *ops)
 563 {
 564         rtnl_lock();
 565         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 566         rtnl_unlock();
 567 }
 568 EXPORT_SYMBOL_GPL(rtnl_af_register);
 569
 570 /**
 571  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 572  * @ops: struct rtnl_af_ops * to unregister
 573  */
 574 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 575 {
 576         rtnl_lock();
 577         list_del_rcu(&ops->list);
 578         rtnl_unlock();
 579
 580         synchronize_rcu();
 581 }
 582 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 583
 584 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 585                                     u32 ext_filter_mask)
 586 {
 587         struct rtnl_af_ops *af_ops;
 588         size_t size;
 589
 590         /* IFLA_AF_SPEC */
 591         size = nla_total_size(sizeof(struct nlattr));
 592
 593         rcu_read_lock();
 594         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 595                 if (af_ops->get_link_af_size) {
 596                         /* AF_* + nested data */
 597                         size += nla_total_size(sizeof(struct nlattr)) +
 598                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 599                 }
 600         }
 601         rcu_read_unlock();
 602
 603         return size;
 604 }
 605
 606 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 607 {
 608         struct net_device *master_dev;
 609         bool ret = false;
 610
 611         rcu_read_lock();
 612
 613         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 614         if (master_dev && master_dev->rtnl_link_ops)
 615                 ret = true;
 616         rcu_read_unlock();
 617         return ret;
 618 }
 619
 620 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 621                                      const struct net_device *dev)
 622 {
 623         struct net_device *master_dev;
 624         const struct rtnl_link_ops *ops;
 625         struct nlattr *slave_data;
 626         int err;
 627
 628         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 629         if (!master_dev)
 630                 return 0;
 631         ops = master_dev->rtnl_link_ops;
 632         if (!ops)
 633                 return 0;
 634         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 635                 return -EMSGSIZE;
 636         if (ops->fill_slave_info) {
 637                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 638                 if (!slave_data)
 639                         return -EMSGSIZE;
 640                 err = ops->fill_slave_info(skb, master_dev, dev);
 641                 if (err < 0)
 642                         goto err_cancel_slave_data;
 643                 nla_nest_end(skb, slave_data);
 644         }
 645         return 0;
 646
 647 err_cancel_slave_data:
 648         nla_nest_cancel(skb, slave_data);
 649         return err;
 650 }
 651
 652 static int rtnl_link_info_fill(struct sk_buff *skb,
 653                                const struct net_device *dev)
 654 {
 655         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 656         struct nlattr *data;
 657         int err;
 658
 659         if (!ops)
 660                 return 0;
 661         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 662                 return -EMSGSIZE;
 663         if (ops->fill_xstats) {
 664                 err = ops->fill_xstats(skb, dev);
 665                 if (err < 0)
 666                         return err;
 667         }
 668         if (ops->fill_info) {
 669                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 670                 if (data == NULL)
 671                         return -EMSGSIZE;
 672                 err = ops->fill_info(skb, dev);
 673                 if (err < 0)
 674                         goto err_cancel_data;
 675                 nla_nest_end(skb, data);
 676         }
 677         return 0;
 678
 679 err_cancel_data:
 680         nla_nest_cancel(skb, data);
 681         return err;
 682 }
 683
 684 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 685 {
 686         struct nlattr *linkinfo;
 687         int err = -EMSGSIZE;
 688
 689         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 690         if (linkinfo == NULL)
 691                 goto out;
 692
 693         err = rtnl_link_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         err = rtnl_link_slave_info_fill(skb, dev);
 698         if (err < 0)
 699                 goto err_cancel_link;
 700
 701         nla_nest_end(skb, linkinfo);
 702         return 0;
 703
 704 err_cancel_link:
 705         nla_nest_cancel(skb, linkinfo);
 706 out:
 707         return err;
 708 }
 709
 710 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 711 {
 712         struct sock *rtnl = net->rtnl;
 713
 714         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 715 }
 716
 717 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 718 {
 719         struct sock *rtnl = net->rtnl;
 720
 721         return nlmsg_unicast(rtnl, skb, pid);
 722 }
 723 EXPORT_SYMBOL(rtnl_unicast);
 724
 725 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 726                  struct nlmsghdr *nlh, gfp_t flags)
 727 {
 728         struct sock *rtnl = net->rtnl;
 729
 730         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 731 }
 732 EXPORT_SYMBOL(rtnl_notify);
 733
 734 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 735 {
 736         struct sock *rtnl = net->rtnl;
 737
 738         netlink_set_err(rtnl, 0, group, error);
 739 }
 740 EXPORT_SYMBOL(rtnl_set_sk_err);
 741
 742 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 743 {
 744         struct nlattr *mx;
 745         int i, valid = 0;
 746
 747         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 748         if (metrics == dst_default_metrics.metrics)
 749                 return 0;
 750
 751         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 752         if (mx == NULL)
 753                 return -ENOBUFS;
 754
 755         for (i = 0; i < RTAX_MAX; i++) {
 756                 if (metrics[i]) {
 757                         if (i == RTAX_CC_ALGO - 1) {
 758                                 char tmp[TCP_CA_NAME_MAX], *name;
 759
 760                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 761                                 if (!name)
 762                                         continue;
 763                                 if (nla_put_string(skb, i + 1, name))
 764                                         goto nla_put_failure;
 765                         } else if (i == RTAX_FEATURES - 1) {
 766                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 767
 768                                 if (!user_features)
 769                                         continue;
 770                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 771                                 if (nla_put_u32(skb, i + 1, user_features))
 772                                         goto nla_put_failure;
 773                         } else {
 774                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 775                                         goto nla_put_failure;
 776                         }
 777                         valid++;
 778                 }
 779         }
 780
 781         if (!valid) {
 782                 nla_nest_cancel(skb, mx);
 783                 return 0;
 784         }
 785
 786         return nla_nest_end(skb, mx);
 787
 788 nla_put_failure:
 789         nla_nest_cancel(skb, mx);
 790         return -EMSGSIZE;
 791 }
 792 EXPORT_SYMBOL(rtnetlink_put_metrics);
 793
 794 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 795                        long expires, u32 error)
 796 {
 797         struct rta_cacheinfo ci = {
 798                 .rta_error = error,
 799                 .rta_id =  id,
 800         };
 801
 802         if (dst) {
 803                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 804                 ci.rta_used = dst->__use;
 805                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 806         }
 807         if (expires) {
 808                 unsigned long clock;
 809
 810                 clock = jiffies_to_clock_t(abs(expires));
 811                 clock = min_t(unsigned long, clock, INT_MAX);
 812                 ci.rta_expires = (expires > 0) ? clock : -clock;
 813         }
 814         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 815 }
 816 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 817
 818 static void set_operstate(struct net_device *dev, unsigned char transition)
 819 {
 820         unsigned char operstate = dev->operstate;
 821
 822         switch (transition) {
 823         case IF_OPER_UP:
 824                 if ((operstate == IF_OPER_DORMANT ||
 825                      operstate == IF_OPER_TESTING ||
 826                      operstate == IF_OPER_UNKNOWN) &&
 827                     !netif_dormant(dev) && !netif_testing(dev))
 828                         operstate = IF_OPER_UP;
 829                 break;
 830
 831         case IF_OPER_TESTING:
 832                 if (operstate == IF_OPER_UP ||
 833                     operstate == IF_OPER_UNKNOWN)
 834                         operstate = IF_OPER_TESTING;
 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(&dev_base_lock);
 846                 dev->operstate = operstate;
 847                 write_unlock(&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 size_t rtnl_prop_list_size(const struct net_device *dev)
 987 {
 988         struct netdev_name_node *name_node;
 989         size_t size;
 990
 991         if (list_empty(&dev->name_node->list))
 992                 return 0;
 993         size = nla_total_size(0);
 994         list_for_each_entry(name_node, &dev->name_node->list, list)
 995                 size += nla_total_size(ALTIFNAMSIZ);
 996         return size;
 997 }
 998
 999 static size_t rtnl_proto_down_size(const struct net_device *dev)
1000 {
1001         size_t size = nla_total_size(1);
1002
1003         if (dev->proto_down_reason)
1004                 size += nla_total_size(0) + nla_total_size(4);
1005
1006         return size;
1007 }
1008
1009 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1010                                      u32 ext_filter_mask)
1011 {
1012         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1013                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1014                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1015                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1016                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1017                + nla_total_size(sizeof(struct rtnl_link_stats))
1018                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1019                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1020                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1021                + nla_total_size(4) /* IFLA_TXQLEN */
1022                + nla_total_size(4) /* IFLA_WEIGHT */
1023                + nla_total_size(4) /* IFLA_MTU */
1024                + nla_total_size(4) /* IFLA_LINK */
1025                + nla_total_size(4) /* IFLA_MASTER */
1026                + nla_total_size(1) /* IFLA_CARRIER */
1027                + nla_total_size(4) /* IFLA_PROMISCUITY */
1028                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1029                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1030                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1031                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1032                + nla_total_size(1) /* IFLA_OPERSTATE */
1033                + nla_total_size(1) /* IFLA_LINKMODE */
1034                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1035                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1036                + nla_total_size(4) /* IFLA_GROUP */
1037                + nla_total_size(ext_filter_mask
1038                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1039                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1040                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1041                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1042                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1043                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1044                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1045                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1046                + rtnl_xdp_size() /* IFLA_XDP */
1047                + nla_total_size(4)  /* IFLA_EVENT */
1048                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1049                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1050                + rtnl_proto_down_size(dev)  /* proto down */
1051                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1052                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1053                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1054                + nla_total_size(4)  /* IFLA_MIN_MTU */
1055                + nla_total_size(4)  /* IFLA_MAX_MTU */
1056                + rtnl_prop_list_size(dev)
1057                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1058                + 0;
1059 }
1060
1061 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1062 {
1063         struct nlattr *vf_ports;
1064         struct nlattr *vf_port;
1065         int vf;
1066         int err;
1067
1068         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1069         if (!vf_ports)
1070                 return -EMSGSIZE;
1071
1072         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1073                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1074                 if (!vf_port)
1075                         goto nla_put_failure;
1076                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1077                         goto nla_put_failure;
1078                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1079                 if (err == -EMSGSIZE)
1080                         goto nla_put_failure;
1081                 if (err) {
1082                         nla_nest_cancel(skb, vf_port);
1083                         continue;
1084                 }
1085                 nla_nest_end(skb, vf_port);
1086         }
1087
1088         nla_nest_end(skb, vf_ports);
1089
1090         return 0;
1091
1092 nla_put_failure:
1093         nla_nest_cancel(skb, vf_ports);
1094         return -EMSGSIZE;
1095 }
1096
1097 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1098 {
1099         struct nlattr *port_self;
1100         int err;
1101
1102         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1103         if (!port_self)
1104                 return -EMSGSIZE;
1105
1106         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1107         if (err) {
1108                 nla_nest_cancel(skb, port_self);
1109                 return (err == -EMSGSIZE) ? err : 0;
1110         }
1111
1112         nla_nest_end(skb, port_self);
1113
1114         return 0;
1115 }
1116
1117 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1118                           u32 ext_filter_mask)
1119 {
1120         int err;
1121
1122         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1123             !(ext_filter_mask & RTEXT_FILTER_VF))
1124                 return 0;
1125
1126         err = rtnl_port_self_fill(skb, dev);
1127         if (err)
1128                 return err;
1129
1130         if (dev_num_vf(dev->dev.parent)) {
1131                 err = rtnl_vf_ports_fill(skb, dev);
1132                 if (err)
1133                         return err;
1134         }
1135
1136         return 0;
1137 }
1138
1139 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1140 {
1141         int err;
1142         struct netdev_phys_item_id ppid;
1143
1144         err = dev_get_phys_port_id(dev, &ppid);
1145         if (err) {
1146                 if (err == -EOPNOTSUPP)
1147                         return 0;
1148                 return err;
1149         }
1150
1151         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1152                 return -EMSGSIZE;
1153
1154         return 0;
1155 }
1156
1157 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1158 {
1159         char name[IFNAMSIZ];
1160         int err;
1161
1162         err = dev_get_phys_port_name(dev, name, sizeof(name));
1163         if (err) {
1164                 if (err == -EOPNOTSUPP)
1165                         return 0;
1166                 return err;
1167         }
1168
1169         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1170                 return -EMSGSIZE;
1171
1172         return 0;
1173 }
1174
1175 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1176 {
1177         struct netdev_phys_item_id ppid = { };
1178         int err;
1179
1180         err = dev_get_port_parent_id(dev, &ppid, false);
1181         if (err) {
1182                 if (err == -EOPNOTSUPP)
1183                         return 0;
1184                 return err;
1185         }
1186
1187         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1188                 return -EMSGSIZE;
1189
1190         return 0;
1191 }
1192
1193 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1194                                               struct net_device *dev)
1195 {
1196         struct rtnl_link_stats64 *sp;
1197         struct nlattr *attr;
1198
1199         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1200                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1201         if (!attr)
1202                 return -EMSGSIZE;
1203
1204         sp = nla_data(attr);
1205         dev_get_stats(dev, sp);
1206
1207         attr = nla_reserve(skb, IFLA_STATS,
1208                            sizeof(struct rtnl_link_stats));
1209         if (!attr)
1210                 return -EMSGSIZE;
1211
1212         copy_rtnl_link_stats(nla_data(attr), sp);
1213
1214         return 0;
1215 }
1216
1217 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1218                                                struct net_device *dev,
1219                                                int vfs_num,
1220                                                struct nlattr *vfinfo,
1221                                                u32 ext_filter_mask)
1222 {
1223         struct ifla_vf_rss_query_en vf_rss_query_en;
1224         struct nlattr *vf, *vfstats, *vfvlanlist;
1225         struct ifla_vf_link_state vf_linkstate;
1226         struct ifla_vf_vlan_info vf_vlan_info;
1227         struct ifla_vf_spoofchk vf_spoofchk;
1228         struct ifla_vf_tx_rate vf_tx_rate;
1229         struct ifla_vf_stats vf_stats;
1230         struct ifla_vf_trust vf_trust;
1231         struct ifla_vf_vlan vf_vlan;
1232         struct ifla_vf_rate vf_rate;
1233         struct ifla_vf_mac vf_mac;
1234         struct ifla_vf_broadcast vf_broadcast;
1235         struct ifla_vf_info ivi;
1236         struct ifla_vf_guid node_guid;
1237         struct ifla_vf_guid port_guid;
1238
1239         memset(&ivi, 0, sizeof(ivi));
1240
1241         /* Not all SR-IOV capable drivers support the
1242          * spoofcheck and "RSS query enable" query.  Preset to
1243          * -1 so the user space tool can detect that the driver
1244          * didn't report anything.
1245          */
1246         ivi.spoofchk = -1;
1247         ivi.rss_query_en = -1;
1248         ivi.trusted = -1;
1249         /* The default value for VF link state is "auto"
1250          * IFLA_VF_LINK_STATE_AUTO which equals zero
1251          */
1252         ivi.linkstate = 0;
1253         /* VLAN Protocol by default is 802.1Q */
1254         ivi.vlan_proto = htons(ETH_P_8021Q);
1255         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1256                 return 0;
1257
1258         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1259         memset(&node_guid, 0, sizeof(node_guid));
1260         memset(&port_guid, 0, sizeof(port_guid));
1261
1262         vf_mac.vf =
1263                 vf_vlan.vf =
1264                 vf_vlan_info.vf =
1265                 vf_rate.vf =
1266                 vf_tx_rate.vf =
1267                 vf_spoofchk.vf =
1268                 vf_linkstate.vf =
1269                 vf_rss_query_en.vf =
1270                 vf_trust.vf =
1271                 node_guid.vf =
1272                 port_guid.vf = ivi.vf;
1273
1274         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1275         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1276         vf_vlan.vlan = ivi.vlan;
1277         vf_vlan.qos = ivi.qos;
1278         vf_vlan_info.vlan = ivi.vlan;
1279         vf_vlan_info.qos = ivi.qos;
1280         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1281         vf_tx_rate.rate = ivi.max_tx_rate;
1282         vf_rate.min_tx_rate = ivi.min_tx_rate;
1283         vf_rate.max_tx_rate = ivi.max_tx_rate;
1284         vf_spoofchk.setting = ivi.spoofchk;
1285         vf_linkstate.link_state = ivi.linkstate;
1286         vf_rss_query_en.setting = ivi.rss_query_en;
1287         vf_trust.setting = ivi.trusted;
1288         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1289         if (!vf)
1290                 goto nla_put_vfinfo_failure;
1291         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1292             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1293             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1294             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1295                     &vf_rate) ||
1296             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1297                     &vf_tx_rate) ||
1298             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1299                     &vf_spoofchk) ||
1300             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1301                     &vf_linkstate) ||
1302             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1303                     sizeof(vf_rss_query_en),
1304                     &vf_rss_query_en) ||
1305             nla_put(skb, IFLA_VF_TRUST,
1306                     sizeof(vf_trust), &vf_trust))
1307                 goto nla_put_vf_failure;
1308
1309         if (dev->netdev_ops->ndo_get_vf_guid &&
1310             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1311                                               &port_guid)) {
1312                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1313                             &node_guid) ||
1314                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1315                             &port_guid))
1316                         goto nla_put_vf_failure;
1317         }
1318         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1319         if (!vfvlanlist)
1320                 goto nla_put_vf_failure;
1321         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1322                     &vf_vlan_info)) {
1323                 nla_nest_cancel(skb, vfvlanlist);
1324                 goto nla_put_vf_failure;
1325         }
1326         nla_nest_end(skb, vfvlanlist);
1327         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1328                 memset(&vf_stats, 0, sizeof(vf_stats));
1329                 if (dev->netdev_ops->ndo_get_vf_stats)
1330                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1331                                                           &vf_stats);
1332                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1333                 if (!vfstats)
1334                         goto nla_put_vf_failure;
1335                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1336                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1337                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1338                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1339                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1340                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1341                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1342                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1343                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1344                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1345                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1346                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1347                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1348                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1349                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1350                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1351                         nla_nest_cancel(skb, vfstats);
1352                         goto nla_put_vf_failure;
1353                 }
1354                 nla_nest_end(skb, vfstats);
1355         }
1356         nla_nest_end(skb, vf);
1357         return 0;
1358
1359 nla_put_vf_failure:
1360         nla_nest_cancel(skb, vf);
1361 nla_put_vfinfo_failure:
1362         nla_nest_cancel(skb, vfinfo);
1363         return -EMSGSIZE;
1364 }
1365
1366 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1367                                            struct net_device *dev,
1368                                            u32 ext_filter_mask)
1369 {
1370         struct nlattr *vfinfo;
1371         int i, num_vfs;
1372
1373         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1374                 return 0;
1375
1376         num_vfs = dev_num_vf(dev->dev.parent);
1377         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1378                 return -EMSGSIZE;
1379
1380         if (!dev->netdev_ops->ndo_get_vf_config)
1381                 return 0;
1382
1383         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1384         if (!vfinfo)
1385                 return -EMSGSIZE;
1386
1387         for (i = 0; i < num_vfs; i++) {
1388                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo, ext_filter_mask))
1389                         return -EMSGSIZE;
1390         }
1391
1392         nla_nest_end(skb, vfinfo);
1393         return 0;
1394 }
1395
1396 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1397 {
1398         struct rtnl_link_ifmap map;
1399
1400         memset(&map, 0, sizeof(map));
1401         map.mem_start   = dev->mem_start;
1402         map.mem_end     = dev->mem_end;
1403         map.base_addr   = dev->base_addr;
1404         map.irq         = dev->irq;
1405         map.dma         = dev->dma;
1406         map.port        = dev->if_port;
1407
1408         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1409                 return -EMSGSIZE;
1410
1411         return 0;
1412 }
1413
1414 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1415 {
1416         const struct bpf_prog *generic_xdp_prog;
1417
1418         ASSERT_RTNL();
1419
1420         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1421         if (!generic_xdp_prog)
1422                 return 0;
1423         return generic_xdp_prog->aux->id;
1424 }
1425
1426 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1427 {
1428         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1429 }
1430
1431 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1432 {
1433         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1434 }
1435
1436 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1437                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1438                                u32 (*get_prog_id)(struct net_device *dev))
1439 {
1440         u32 curr_id;
1441         int err;
1442
1443         curr_id = get_prog_id(dev);
1444         if (!curr_id)
1445                 return 0;
1446
1447         *prog_id = curr_id;
1448         err = nla_put_u32(skb, attr, curr_id);
1449         if (err)
1450                 return err;
1451
1452         if (*mode != XDP_ATTACHED_NONE)
1453                 *mode = XDP_ATTACHED_MULTI;
1454         else
1455                 *mode = tgt_mode;
1456
1457         return 0;
1458 }
1459
1460 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1461 {
1462         struct nlattr *xdp;
1463         u32 prog_id;
1464         int err;
1465         u8 mode;
1466
1467         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1468         if (!xdp)
1469                 return -EMSGSIZE;
1470
1471         prog_id = 0;
1472         mode = XDP_ATTACHED_NONE;
1473         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1474                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1475         if (err)
1476                 goto err_cancel;
1477         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1478                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1479         if (err)
1480                 goto err_cancel;
1481         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1482                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1483         if (err)
1484                 goto err_cancel;
1485
1486         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1487         if (err)
1488                 goto err_cancel;
1489
1490         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1491                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1492                 if (err)
1493                         goto err_cancel;
1494         }
1495
1496         nla_nest_end(skb, xdp);
1497         return 0;
1498
1499 err_cancel:
1500         nla_nest_cancel(skb, xdp);
1501         return err;
1502 }
1503
1504 static u32 rtnl_get_event(unsigned long event)
1505 {
1506         u32 rtnl_event_type = IFLA_EVENT_NONE;
1507
1508         switch (event) {
1509         case NETDEV_REBOOT:
1510                 rtnl_event_type = IFLA_EVENT_REBOOT;
1511                 break;
1512         case NETDEV_FEAT_CHANGE:
1513                 rtnl_event_type = IFLA_EVENT_FEATURES;
1514                 break;
1515         case NETDEV_BONDING_FAILOVER:
1516                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1517                 break;
1518         case NETDEV_NOTIFY_PEERS:
1519                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1520                 break;
1521         case NETDEV_RESEND_IGMP:
1522                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1523                 break;
1524         case NETDEV_CHANGEINFODATA:
1525                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1526                 break;
1527         default:
1528                 break;
1529         }
1530
1531         return rtnl_event_type;
1532 }
1533
1534 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1535 {
1536         const struct net_device *upper_dev;
1537         int ret = 0;
1538
1539         rcu_read_lock();
1540
1541         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1542         if (upper_dev)
1543                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1544
1545         rcu_read_unlock();
1546         return ret;
1547 }
1548
1549 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1550                           bool force)
1551 {
1552         int ifindex = dev_get_iflink(dev);
1553
1554         if (force || dev->ifindex != ifindex)
1555                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1556
1557         return 0;
1558 }
1559
1560 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1561                                               struct net_device *dev)
1562 {
1563         char buf[IFALIASZ];
1564         int ret;
1565
1566         ret = dev_get_alias(dev, buf, sizeof(buf));
1567         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1568 }
1569
1570 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1571                                   const struct net_device *dev,
1572                                   struct net *src_net, gfp_t gfp)
1573 {
1574         bool put_iflink = false;
1575
1576         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1577                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1578
1579                 if (!net_eq(dev_net(dev), link_net)) {
1580                         int id = peernet2id_alloc(src_net, link_net, gfp);
1581
1582                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1583                                 return -EMSGSIZE;
1584
1585                         put_iflink = true;
1586                 }
1587         }
1588
1589         return nla_put_iflink(skb, dev, put_iflink);
1590 }
1591
1592 static int rtnl_fill_link_af(struct sk_buff *skb,
1593                              const struct net_device *dev,
1594                              u32 ext_filter_mask)
1595 {
1596         const struct rtnl_af_ops *af_ops;
1597         struct nlattr *af_spec;
1598
1599         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1600         if (!af_spec)
1601                 return -EMSGSIZE;
1602
1603         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1604                 struct nlattr *af;
1605                 int err;
1606
1607                 if (!af_ops->fill_link_af)
1608                         continue;
1609
1610                 af = nla_nest_start_noflag(skb, af_ops->family);
1611                 if (!af)
1612                         return -EMSGSIZE;
1613
1614                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1615                 /*
1616                  * Caller may return ENODATA to indicate that there
1617                  * was no data to be dumped. This is not an error, it
1618                  * means we should trim the attribute header and
1619                  * continue.
1620                  */
1621                 if (err == -ENODATA)
1622                         nla_nest_cancel(skb, af);
1623                 else if (err < 0)
1624                         return -EMSGSIZE;
1625
1626                 nla_nest_end(skb, af);
1627         }
1628
1629         nla_nest_end(skb, af_spec);
1630         return 0;
1631 }
1632
1633 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1634                                  const struct net_device *dev)
1635 {
1636         struct netdev_name_node *name_node;
1637         int count = 0;
1638
1639         list_for_each_entry(name_node, &dev->name_node->list, list) {
1640                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1641                         return -EMSGSIZE;
1642                 count++;
1643         }
1644         return count;
1645 }
1646
1647 static int rtnl_fill_prop_list(struct sk_buff *skb,
1648                                const struct net_device *dev)
1649 {
1650         struct nlattr *prop_list;
1651         int ret;
1652
1653         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1654         if (!prop_list)
1655                 return -EMSGSIZE;
1656
1657         ret = rtnl_fill_alt_ifnames(skb, dev);
1658         if (ret <= 0)
1659                 goto nest_cancel;
1660
1661         nla_nest_end(skb, prop_list);
1662         return 0;
1663
1664 nest_cancel:
1665         nla_nest_cancel(skb, prop_list);
1666         return ret;
1667 }
1668
1669 static int rtnl_fill_proto_down(struct sk_buff *skb,
1670                                 const struct net_device *dev)
1671 {
1672         struct nlattr *pr;
1673         u32 preason;
1674
1675         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1676                 goto nla_put_failure;
1677
1678         preason = dev->proto_down_reason;
1679         if (!preason)
1680                 return 0;
1681
1682         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1683         if (!pr)
1684                 return -EMSGSIZE;
1685
1686         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1687                 nla_nest_cancel(skb, pr);
1688                 goto nla_put_failure;
1689         }
1690
1691         nla_nest_end(skb, pr);
1692         return 0;
1693
1694 nla_put_failure:
1695         return -EMSGSIZE;
1696 }
1697
1698 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1699                             struct net_device *dev, struct net *src_net,
1700                             int type, u32 pid, u32 seq, u32 change,
1701                             unsigned int flags, u32 ext_filter_mask,
1702                             u32 event, int *new_nsid, int new_ifindex,
1703                             int tgt_netnsid, gfp_t gfp)
1704 {
1705         struct ifinfomsg *ifm;
1706         struct nlmsghdr *nlh;
1707         struct Qdisc *qdisc;
1708
1709         ASSERT_RTNL();
1710         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1711         if (nlh == NULL)
1712                 return -EMSGSIZE;
1713
1714         ifm = nlmsg_data(nlh);
1715         ifm->ifi_family = AF_UNSPEC;
1716         ifm->__ifi_pad = 0;
1717         ifm->ifi_type = dev->type;
1718         ifm->ifi_index = dev->ifindex;
1719         ifm->ifi_flags = dev_get_flags(dev);
1720         ifm->ifi_change = change;
1721
1722         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1723                 goto nla_put_failure;
1724
1725         qdisc = rtnl_dereference(dev->qdisc);
1726         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1727             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1728             nla_put_u8(skb, IFLA_OPERSTATE,
1729                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1730             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1731             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1732             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1733             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1734             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1735             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1736             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1737             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1738             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1739 #ifdef CONFIG_RPS
1740             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1741 #endif
1742             put_master_ifindex(skb, dev) ||
1743             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1744             (qdisc &&
1745              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1746             nla_put_ifalias(skb, dev) ||
1747             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1748                         atomic_read(&dev->carrier_up_count) +
1749                         atomic_read(&dev->carrier_down_count)) ||
1750             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1751                         atomic_read(&dev->carrier_up_count)) ||
1752             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1753                         atomic_read(&dev->carrier_down_count)))
1754                 goto nla_put_failure;
1755
1756         if (rtnl_fill_proto_down(skb, dev))
1757                 goto nla_put_failure;
1758
1759         if (event != IFLA_EVENT_NONE) {
1760                 if (nla_put_u32(skb, IFLA_EVENT, event))
1761                         goto nla_put_failure;
1762         }
1763
1764         if (rtnl_fill_link_ifmap(skb, dev))
1765                 goto nla_put_failure;
1766
1767         if (dev->addr_len) {
1768                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1769                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1770                         goto nla_put_failure;
1771         }
1772
1773         if (rtnl_phys_port_id_fill(skb, dev))
1774                 goto nla_put_failure;
1775
1776         if (rtnl_phys_port_name_fill(skb, dev))
1777                 goto nla_put_failure;
1778
1779         if (rtnl_phys_switch_id_fill(skb, dev))
1780                 goto nla_put_failure;
1781
1782         if (rtnl_fill_stats(skb, dev))
1783                 goto nla_put_failure;
1784
1785         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1786                 goto nla_put_failure;
1787
1788         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1789                 goto nla_put_failure;
1790
1791         if (rtnl_xdp_fill(skb, dev))
1792                 goto nla_put_failure;
1793
1794         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1795                 if (rtnl_link_fill(skb, dev) < 0)
1796                         goto nla_put_failure;
1797         }
1798
1799         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1800                 goto nla_put_failure;
1801
1802         if (new_nsid &&
1803             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1804                 goto nla_put_failure;
1805         if (new_ifindex &&
1806             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1807                 goto nla_put_failure;
1808
1809         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1810             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1811                 goto nla_put_failure;
1812
1813         rcu_read_lock();
1814         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1815                 goto nla_put_failure_rcu;
1816         rcu_read_unlock();
1817
1818         if (rtnl_fill_prop_list(skb, dev))
1819                 goto nla_put_failure;
1820
1821         if (dev->dev.parent &&
1822             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1823                            dev_name(dev->dev.parent)))
1824                 goto nla_put_failure;
1825
1826         if (dev->dev.parent && dev->dev.parent->bus &&
1827             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1828                            dev->dev.parent->bus->name))
1829                 goto nla_put_failure;
1830
1831         nlmsg_end(skb, nlh);
1832         return 0;
1833
1834 nla_put_failure_rcu:
1835         rcu_read_unlock();
1836 nla_put_failure:
1837         nlmsg_cancel(skb, nlh);
1838         return -EMSGSIZE;
1839 }
1840
1841 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1842         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1843         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1844         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1845         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1846         [IFLA_MTU]              = { .type = NLA_U32 },
1847         [IFLA_LINK]             = { .type = NLA_U32 },
1848         [IFLA_MASTER]           = { .type = NLA_U32 },
1849         [IFLA_CARRIER]          = { .type = NLA_U8 },
1850         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1851         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1852         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1853         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1854         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1855         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1856         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1857         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1858          * allow 0-length string (needed to remove an alias).
1859          */
1860         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1861         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1862         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1863         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1864         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1865         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1866         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1867         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1868         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1869         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1870         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1871         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1872         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1873         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1874         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1875         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1876         [IFLA_XDP]              = { .type = NLA_NESTED },
1877         [IFLA_EVENT]            = { .type = NLA_U32 },
1878         [IFLA_GROUP]            = { .type = NLA_U32 },
1879         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1880         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1881         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1882         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1883         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1884         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1885         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1886                                     .len = ALTIFNAMSIZ - 1 },
1887         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1888         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1889         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1890         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1891 };
1892
1893 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1894         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1895         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1896         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1897         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1898 };
1899
1900 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1901         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1902         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1903         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1904         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1905         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1906         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1907         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1908         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1909         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1910         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1911         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1912         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1913         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1914 };
1915
1916 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1917         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1918         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1919                                     .len = PORT_PROFILE_MAX },
1920         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1921                                       .len = PORT_UUID_MAX },
1922         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1923                                     .len = PORT_UUID_MAX },
1924         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1925         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1926
1927         /* Unused, but we need to keep it here since user space could
1928          * fill it. It's also broken with regard to NLA_BINARY use in
1929          * combination with structs.
1930          */
1931         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1932                                     .len = sizeof(struct ifla_port_vsi) },
1933 };
1934
1935 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1936         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1937         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1938         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1939         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1940         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1941         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1942 };
1943
1944 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1945 {
1946         const struct rtnl_link_ops *ops = NULL;
1947         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1948
1949         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1950                 return NULL;
1951
1952         if (linfo[IFLA_INFO_KIND]) {
1953                 char kind[MODULE_NAME_LEN];
1954
1955                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1956                 ops = rtnl_link_ops_get(kind);
1957         }
1958
1959         return ops;
1960 }
1961
1962 static bool link_master_filtered(struct net_device *dev, int master_idx)
1963 {
1964         struct net_device *master;
1965
1966         if (!master_idx)
1967                 return false;
1968
1969         master = netdev_master_upper_dev_get(dev);
1970
1971         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
1972          * another invalid value for ifindex to denote "no master".
1973          */
1974         if (master_idx == -1)
1975                 return !!master;
1976
1977         if (!master || master->ifindex != master_idx)
1978                 return true;
1979
1980         return false;
1981 }
1982
1983 static bool link_kind_filtered(const struct net_device *dev,
1984                                const struct rtnl_link_ops *kind_ops)
1985 {
1986         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1987                 return true;
1988
1989         return false;
1990 }
1991
1992 static bool link_dump_filtered(struct net_device *dev,
1993                                int master_idx,
1994                                const struct rtnl_link_ops *kind_ops)
1995 {
1996         if (link_master_filtered(dev, master_idx) ||
1997             link_kind_filtered(dev, kind_ops))
1998                 return true;
1999
2000         return false;
2001 }
2002
2003 /**
2004  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2005  * @sk: netlink socket
2006  * @netnsid: network namespace identifier
2007  *
2008  * Returns the network namespace identified by netnsid on success or an error
2009  * pointer on failure.
2010  */
2011 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2012 {
2013         struct net *net;
2014
2015         net = get_net_ns_by_id(sock_net(sk), netnsid);
2016         if (!net)
2017                 return ERR_PTR(-EINVAL);
2018
2019         /* For now, the caller is required to have CAP_NET_ADMIN in
2020          * the user namespace owning the target net ns.
2021          */
2022         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2023                 put_net(net);
2024                 return ERR_PTR(-EACCES);
2025         }
2026         return net;
2027 }
2028 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2029
2030 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2031                                       bool strict_check, struct nlattr **tb,
2032                                       struct netlink_ext_ack *extack)
2033 {
2034         int hdrlen;
2035
2036         if (strict_check) {
2037                 struct ifinfomsg *ifm;
2038
2039                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2040                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2041                         return -EINVAL;
2042                 }
2043
2044                 ifm = nlmsg_data(nlh);
2045                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2046                     ifm->ifi_change) {
2047                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2048                         return -EINVAL;
2049                 }
2050                 if (ifm->ifi_index) {
2051                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2052                         return -EINVAL;
2053                 }
2054
2055                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2056                                                      IFLA_MAX, ifla_policy,
2057                                                      extack);
2058         }
2059
2060         /* A hack to preserve kernel<->userspace interface.
2061          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2062          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2063          * what iproute2 < v3.9.0 used.
2064          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2065          * attribute, its netlink message is shorter than struct ifinfomsg.
2066          */
2067         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2068                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2069
2070         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2071                                       extack);
2072 }
2073
2074 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2075 {
2076         struct netlink_ext_ack *extack = cb->extack;
2077         const struct nlmsghdr *nlh = cb->nlh;
2078         struct net *net = sock_net(skb->sk);
2079         struct net *tgt_net = net;
2080         int h, s_h;
2081         int idx = 0, s_idx;
2082         struct net_device *dev;
2083         struct hlist_head *head;
2084         struct nlattr *tb[IFLA_MAX+1];
2085         u32 ext_filter_mask = 0;
2086         const struct rtnl_link_ops *kind_ops = NULL;
2087         unsigned int flags = NLM_F_MULTI;
2088         int master_idx = 0;
2089         int netnsid = -1;
2090         int err, i;
2091
2092         s_h = cb->args[0];
2093         s_idx = cb->args[1];
2094
2095         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2096         if (err < 0) {
2097                 if (cb->strict_check)
2098                         return err;
2099
2100                 goto walk_entries;
2101         }
2102
2103         for (i = 0; i <= IFLA_MAX; ++i) {
2104                 if (!tb[i])
2105                         continue;
2106
2107                 /* new attributes should only be added with strict checking */
2108                 switch (i) {
2109                 case IFLA_TARGET_NETNSID:
2110                         netnsid = nla_get_s32(tb[i]);
2111                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2112                         if (IS_ERR(tgt_net)) {
2113                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2114                                 return PTR_ERR(tgt_net);
2115                         }
2116                         break;
2117                 case IFLA_EXT_MASK:
2118                         ext_filter_mask = nla_get_u32(tb[i]);
2119                         break;
2120                 case IFLA_MASTER:
2121                         master_idx = nla_get_u32(tb[i]);
2122                         break;
2123                 case IFLA_LINKINFO:
2124                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2125                         break;
2126                 default:
2127                         if (cb->strict_check) {
2128                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2129                                 return -EINVAL;
2130                         }
2131                 }
2132         }
2133
2134         if (master_idx || kind_ops)
2135                 flags |= NLM_F_DUMP_FILTERED;
2136
2137 walk_entries:
2138         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2139                 idx = 0;
2140                 head = &tgt_net->dev_index_head[h];
2141                 hlist_for_each_entry(dev, head, index_hlist) {
2142                         if (link_dump_filtered(dev, master_idx, kind_ops))
2143                                 goto cont;
2144                         if (idx < s_idx)
2145                                 goto cont;
2146                         err = rtnl_fill_ifinfo(skb, dev, net,
2147                                                RTM_NEWLINK,
2148                                                NETLINK_CB(cb->skb).portid,
2149                                                nlh->nlmsg_seq, 0, flags,
2150                                                ext_filter_mask, 0, NULL, 0,
2151                                                netnsid, GFP_KERNEL);
2152
2153                         if (err < 0) {
2154                                 if (likely(skb->len))
2155                                         goto out;
2156
2157                                 goto out_err;
2158                         }
2159 cont:
2160                         idx++;
2161                 }
2162         }
2163 out:
2164         err = skb->len;
2165 out_err:
2166         cb->args[1] = idx;
2167         cb->args[0] = h;
2168         cb->seq = tgt_net->dev_base_seq;
2169         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2170         if (netnsid >= 0)
2171                 put_net(tgt_net);
2172
2173         return err;
2174 }
2175
2176 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2177                              struct netlink_ext_ack *exterr)
2178 {
2179         const struct ifinfomsg *ifmp;
2180         const struct nlattr *attrs;
2181         size_t len;
2182
2183         ifmp = nla_data(nla_peer);
2184         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2185         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2186
2187         if (ifmp->ifi_index < 0) {
2188                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2189                                     "ifindex can't be negative");
2190                 return -EINVAL;
2191         }
2192
2193         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2194                                     exterr);
2195 }
2196 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2197
2198 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2199 {
2200         struct net *net;
2201         /* Examine the link attributes and figure out which
2202          * network namespace we are talking about.
2203          */
2204         if (tb[IFLA_NET_NS_PID])
2205                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2206         else if (tb[IFLA_NET_NS_FD])
2207                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2208         else
2209                 net = get_net(src_net);
2210         return net;
2211 }
2212 EXPORT_SYMBOL(rtnl_link_get_net);
2213
2214 /* Figure out which network namespace we are talking about by
2215  * examining the link attributes in the following order:
2216  *
2217  * 1. IFLA_NET_NS_PID
2218  * 2. IFLA_NET_NS_FD
2219  * 3. IFLA_TARGET_NETNSID
2220  */
2221 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2222                                                struct nlattr *tb[])
2223 {
2224         struct net *net;
2225
2226         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2227                 return rtnl_link_get_net(src_net, tb);
2228
2229         if (!tb[IFLA_TARGET_NETNSID])
2230                 return get_net(src_net);
2231
2232         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2233         if (!net)
2234                 return ERR_PTR(-EINVAL);
2235
2236         return net;
2237 }
2238
2239 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2240                                              struct net *src_net,
2241                                              struct nlattr *tb[], int cap)
2242 {
2243         struct net *net;
2244
2245         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2246         if (IS_ERR(net))
2247                 return net;
2248
2249         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2250                 put_net(net);
2251                 return ERR_PTR(-EPERM);
2252         }
2253
2254         return net;
2255 }
2256
2257 /* Verify that rtnetlink requests do not pass additional properties
2258  * potentially referring to different network namespaces.
2259  */
2260 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2261                                     struct netlink_ext_ack *extack,
2262                                     bool netns_id_only)
2263 {
2264
2265         if (netns_id_only) {
2266                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2267                         return 0;
2268
2269                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2270                 return -EOPNOTSUPP;
2271         }
2272
2273         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2274                 goto invalid_attr;
2275
2276         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2277                 goto invalid_attr;
2278
2279         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2280                 goto invalid_attr;
2281
2282         return 0;
2283
2284 invalid_attr:
2285         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2286         return -EINVAL;
2287 }
2288
2289 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2290                             struct netlink_ext_ack *extack)
2291 {
2292         if (dev) {
2293                 if (tb[IFLA_ADDRESS] &&
2294                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2295                         return -EINVAL;
2296
2297                 if (tb[IFLA_BROADCAST] &&
2298                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2299                         return -EINVAL;
2300         }
2301
2302         if (tb[IFLA_AF_SPEC]) {
2303                 struct nlattr *af;
2304                 int rem, err;
2305
2306                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2307                         const struct rtnl_af_ops *af_ops;
2308
2309                         af_ops = rtnl_af_lookup(nla_type(af));
2310                         if (!af_ops)
2311                                 return -EAFNOSUPPORT;
2312
2313                         if (!af_ops->set_link_af)
2314                                 return -EOPNOTSUPP;
2315
2316                         if (af_ops->validate_link_af) {
2317                                 err = af_ops->validate_link_af(dev, af, extack);
2318                                 if (err < 0)
2319                                         return err;
2320                         }
2321                 }
2322         }
2323
2324         return 0;
2325 }
2326
2327 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2328                                   int guid_type)
2329 {
2330         const struct net_device_ops *ops = dev->netdev_ops;
2331
2332         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2333 }
2334
2335 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2336 {
2337         if (dev->type != ARPHRD_INFINIBAND)
2338                 return -EOPNOTSUPP;
2339
2340         return handle_infiniband_guid(dev, ivt, guid_type);
2341 }
2342
2343 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2344 {
2345         const struct net_device_ops *ops = dev->netdev_ops;
2346         int err = -EINVAL;
2347
2348         if (tb[IFLA_VF_MAC]) {
2349                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2350
2351                 if (ivm->vf >= INT_MAX)
2352                         return -EINVAL;
2353                 err = -EOPNOTSUPP;
2354                 if (ops->ndo_set_vf_mac)
2355                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2356                                                   ivm->mac);
2357                 if (err < 0)
2358                         return err;
2359         }
2360
2361         if (tb[IFLA_VF_VLAN]) {
2362                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2363
2364                 if (ivv->vf >= INT_MAX)
2365                         return -EINVAL;
2366                 err = -EOPNOTSUPP;
2367                 if (ops->ndo_set_vf_vlan)
2368                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2369                                                    ivv->qos,
2370                                                    htons(ETH_P_8021Q));
2371                 if (err < 0)
2372                         return err;
2373         }
2374
2375         if (tb[IFLA_VF_VLAN_LIST]) {
2376                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2377                 struct nlattr *attr;
2378                 int rem, len = 0;
2379
2380                 err = -EOPNOTSUPP;
2381                 if (!ops->ndo_set_vf_vlan)
2382                         return err;
2383
2384                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2385                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2386                             nla_len(attr) < NLA_HDRLEN) {
2387                                 return -EINVAL;
2388                         }
2389                         if (len >= MAX_VLAN_LIST_LEN)
2390                                 return -EOPNOTSUPP;
2391                         ivvl[len] = nla_data(attr);
2392
2393                         len++;
2394                 }
2395                 if (len == 0)
2396                         return -EINVAL;
2397
2398                 if (ivvl[0]->vf >= INT_MAX)
2399                         return -EINVAL;
2400                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2401                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2402                 if (err < 0)
2403                         return err;
2404         }
2405
2406         if (tb[IFLA_VF_TX_RATE]) {
2407                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2408                 struct ifla_vf_info ivf;
2409
2410                 if (ivt->vf >= INT_MAX)
2411                         return -EINVAL;
2412                 err = -EOPNOTSUPP;
2413                 if (ops->ndo_get_vf_config)
2414                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2415                 if (err < 0)
2416                         return err;
2417
2418                 err = -EOPNOTSUPP;
2419                 if (ops->ndo_set_vf_rate)
2420                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2421                                                    ivf.min_tx_rate,
2422                                                    ivt->rate);
2423                 if (err < 0)
2424                         return err;
2425         }
2426
2427         if (tb[IFLA_VF_RATE]) {
2428                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2429
2430                 if (ivt->vf >= INT_MAX)
2431                         return -EINVAL;
2432                 err = -EOPNOTSUPP;
2433                 if (ops->ndo_set_vf_rate)
2434                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2435                                                    ivt->min_tx_rate,
2436                                                    ivt->max_tx_rate);
2437                 if (err < 0)
2438                         return err;
2439         }
2440
2441         if (tb[IFLA_VF_SPOOFCHK]) {
2442                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2443
2444                 if (ivs->vf >= INT_MAX)
2445                         return -EINVAL;
2446                 err = -EOPNOTSUPP;
2447                 if (ops->ndo_set_vf_spoofchk)
2448                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2449                                                        ivs->setting);
2450                 if (err < 0)
2451                         return err;
2452         }
2453
2454         if (tb[IFLA_VF_LINK_STATE]) {
2455                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2456
2457                 if (ivl->vf >= INT_MAX)
2458                         return -EINVAL;
2459                 err = -EOPNOTSUPP;
2460                 if (ops->ndo_set_vf_link_state)
2461                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2462                                                          ivl->link_state);
2463                 if (err < 0)
2464                         return err;
2465         }
2466
2467         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2468                 struct ifla_vf_rss_query_en *ivrssq_en;
2469
2470                 err = -EOPNOTSUPP;
2471                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2472                 if (ivrssq_en->vf >= INT_MAX)
2473                         return -EINVAL;
2474                 if (ops->ndo_set_vf_rss_query_en)
2475                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2476                                                            ivrssq_en->setting);
2477                 if (err < 0)
2478                         return err;
2479         }
2480
2481         if (tb[IFLA_VF_TRUST]) {
2482                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2483
2484                 if (ivt->vf >= INT_MAX)
2485                         return -EINVAL;
2486                 err = -EOPNOTSUPP;
2487                 if (ops->ndo_set_vf_trust)
2488                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2489                 if (err < 0)
2490                         return err;
2491         }
2492
2493         if (tb[IFLA_VF_IB_NODE_GUID]) {
2494                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2495
2496                 if (ivt->vf >= INT_MAX)
2497                         return -EINVAL;
2498                 if (!ops->ndo_set_vf_guid)
2499                         return -EOPNOTSUPP;
2500                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2501         }
2502
2503         if (tb[IFLA_VF_IB_PORT_GUID]) {
2504                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2505
2506                 if (ivt->vf >= INT_MAX)
2507                         return -EINVAL;
2508                 if (!ops->ndo_set_vf_guid)
2509                         return -EOPNOTSUPP;
2510
2511                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2512         }
2513
2514         return err;
2515 }
2516
2517 static int do_set_master(struct net_device *dev, int ifindex,
2518                          struct netlink_ext_ack *extack)
2519 {
2520         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2521         const struct net_device_ops *ops;
2522         int err;
2523
2524         if (upper_dev) {
2525                 if (upper_dev->ifindex == ifindex)
2526                         return 0;
2527                 ops = upper_dev->netdev_ops;
2528                 if (ops->ndo_del_slave) {
2529                         err = ops->ndo_del_slave(upper_dev, dev);
2530                         if (err)
2531                                 return err;
2532                 } else {
2533                         return -EOPNOTSUPP;
2534                 }
2535         }
2536
2537         if (ifindex) {
2538                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2539                 if (!upper_dev)
2540                         return -EINVAL;
2541                 ops = upper_dev->netdev_ops;
2542                 if (ops->ndo_add_slave) {
2543                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2544                         if (err)
2545                                 return err;
2546                 } else {
2547                         return -EOPNOTSUPP;
2548                 }
2549         }
2550         return 0;
2551 }
2552
2553 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2554         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2555         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2556 };
2557
2558 static int do_set_proto_down(struct net_device *dev,
2559                              struct nlattr *nl_proto_down,
2560                              struct nlattr *nl_proto_down_reason,
2561                              struct netlink_ext_ack *extack)
2562 {
2563         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2564         const struct net_device_ops *ops = dev->netdev_ops;
2565         unsigned long mask = 0;
2566         u32 value;
2567         bool proto_down;
2568         int err;
2569
2570         if (!ops->ndo_change_proto_down) {
2571                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2572                 return -EOPNOTSUPP;
2573         }
2574
2575         if (nl_proto_down_reason) {
2576                 err = nla_parse_nested_deprecated(pdreason,
2577                                                   IFLA_PROTO_DOWN_REASON_MAX,
2578                                                   nl_proto_down_reason,
2579                                                   ifla_proto_down_reason_policy,
2580                                                   NULL);
2581                 if (err < 0)
2582                         return err;
2583
2584                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2585                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2586                         return -EINVAL;
2587                 }
2588
2589                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2590
2591                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2592                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2593
2594                 dev_change_proto_down_reason(dev, mask, value);
2595         }
2596
2597         if (nl_proto_down) {
2598                 proto_down = nla_get_u8(nl_proto_down);
2599
2600                 /* Don't turn off protodown if there are active reasons */
2601                 if (!proto_down && dev->proto_down_reason) {
2602                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2603                         return -EBUSY;
2604                 }
2605                 err = dev_change_proto_down(dev,
2606                                             proto_down);
2607                 if (err)
2608                         return err;
2609         }
2610
2611         return 0;
2612 }
2613
2614 #define DO_SETLINK_MODIFIED     0x01
2615 /* notify flag means notify + modified. */
2616 #define DO_SETLINK_NOTIFY       0x03
2617 static int do_setlink(const struct sk_buff *skb,
2618                       struct net_device *dev, struct ifinfomsg *ifm,
2619                       struct netlink_ext_ack *extack,
2620                       struct nlattr **tb, char *ifname, int status)
2621 {
2622         const struct net_device_ops *ops = dev->netdev_ops;
2623         int err;
2624
2625         err = validate_linkmsg(dev, tb, extack);
2626         if (err < 0)
2627                 return err;
2628
2629         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2630                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2631                 struct net *net;
2632                 int new_ifindex;
2633
2634                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2635                                                 tb, CAP_NET_ADMIN);
2636                 if (IS_ERR(net)) {
2637                         err = PTR_ERR(net);
2638                         goto errout;
2639                 }
2640
2641                 if (tb[IFLA_NEW_IFINDEX])
2642                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2643                 else
2644                         new_ifindex = 0;
2645
2646                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2647                 put_net(net);
2648                 if (err)
2649                         goto errout;
2650                 status |= DO_SETLINK_MODIFIED;
2651         }
2652
2653         if (tb[IFLA_MAP]) {
2654                 struct rtnl_link_ifmap *u_map;
2655                 struct ifmap k_map;
2656
2657                 if (!ops->ndo_set_config) {
2658                         err = -EOPNOTSUPP;
2659                         goto errout;
2660                 }
2661
2662                 if (!netif_device_present(dev)) {
2663                         err = -ENODEV;
2664                         goto errout;
2665                 }
2666
2667                 u_map = nla_data(tb[IFLA_MAP]);
2668                 k_map.mem_start = (unsigned long) u_map->mem_start;
2669                 k_map.mem_end = (unsigned long) u_map->mem_end;
2670                 k_map.base_addr = (unsigned short) u_map->base_addr;
2671                 k_map.irq = (unsigned char) u_map->irq;
2672                 k_map.dma = (unsigned char) u_map->dma;
2673                 k_map.port = (unsigned char) u_map->port;
2674
2675                 err = ops->ndo_set_config(dev, &k_map);
2676                 if (err < 0)
2677                         goto errout;
2678
2679                 status |= DO_SETLINK_NOTIFY;
2680         }
2681
2682         if (tb[IFLA_ADDRESS]) {
2683                 struct sockaddr *sa;
2684                 int len;
2685
2686                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2687                                                   sizeof(*sa));
2688                 sa = kmalloc(len, GFP_KERNEL);
2689                 if (!sa) {
2690                         err = -ENOMEM;
2691                         goto errout;
2692                 }
2693                 sa->sa_family = dev->type;
2694                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2695                        dev->addr_len);
2696                 err = dev_set_mac_address_user(dev, sa, extack);
2697                 kfree(sa);
2698                 if (err)
2699                         goto errout;
2700                 status |= DO_SETLINK_MODIFIED;
2701         }
2702
2703         if (tb[IFLA_MTU]) {
2704                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2705                 if (err < 0)
2706                         goto errout;
2707                 status |= DO_SETLINK_MODIFIED;
2708         }
2709
2710         if (tb[IFLA_GROUP]) {
2711                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2712                 status |= DO_SETLINK_NOTIFY;
2713         }
2714
2715         /*
2716          * Interface selected by interface index but interface
2717          * name provided implies that a name change has been
2718          * requested.
2719          */
2720         if (ifm->ifi_index > 0 && ifname[0]) {
2721                 err = dev_change_name(dev, ifname);
2722                 if (err < 0)
2723                         goto errout;
2724                 status |= DO_SETLINK_MODIFIED;
2725         }
2726
2727         if (tb[IFLA_IFALIAS]) {
2728                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2729                                     nla_len(tb[IFLA_IFALIAS]));
2730                 if (err < 0)
2731                         goto errout;
2732                 status |= DO_SETLINK_NOTIFY;
2733         }
2734
2735         if (tb[IFLA_BROADCAST]) {
2736                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2737                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2738         }
2739
2740         if (ifm->ifi_flags || ifm->ifi_change) {
2741                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2742                                        extack);
2743                 if (err < 0)
2744                         goto errout;
2745         }
2746
2747         if (tb[IFLA_MASTER]) {
2748                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2749                 if (err)
2750                         goto errout;
2751                 status |= DO_SETLINK_MODIFIED;
2752         }
2753
2754         if (tb[IFLA_CARRIER]) {
2755                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2756                 if (err)
2757                         goto errout;
2758                 status |= DO_SETLINK_MODIFIED;
2759         }
2760
2761         if (tb[IFLA_TXQLEN]) {
2762                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2763
2764                 err = dev_change_tx_queue_len(dev, value);
2765                 if (err)
2766                         goto errout;
2767                 status |= DO_SETLINK_MODIFIED;
2768         }
2769
2770         if (tb[IFLA_GSO_MAX_SIZE]) {
2771                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2772
2773                 if (max_size > GSO_MAX_SIZE) {
2774                         err = -EINVAL;
2775                         goto errout;
2776                 }
2777
2778                 if (dev->gso_max_size ^ max_size) {
2779                         netif_set_gso_max_size(dev, max_size);
2780                         status |= DO_SETLINK_MODIFIED;
2781                 }
2782         }
2783
2784         if (tb[IFLA_GSO_MAX_SEGS]) {
2785                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2786
2787                 if (max_segs > GSO_MAX_SEGS) {
2788                         err = -EINVAL;
2789                         goto errout;
2790                 }
2791
2792                 if (dev->gso_max_segs ^ max_segs) {
2793                         dev->gso_max_segs = max_segs;
2794                         status |= DO_SETLINK_MODIFIED;
2795                 }
2796         }
2797
2798         if (tb[IFLA_OPERSTATE])
2799                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2800
2801         if (tb[IFLA_LINKMODE]) {
2802                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2803
2804                 write_lock(&dev_base_lock);
2805                 if (dev->link_mode ^ value)
2806                         status |= DO_SETLINK_NOTIFY;
2807                 dev->link_mode = value;
2808                 write_unlock(&dev_base_lock);
2809         }
2810
2811         if (tb[IFLA_VFINFO_LIST]) {
2812                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2813                 struct nlattr *attr;
2814                 int rem;
2815
2816                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2817                         if (nla_type(attr) != IFLA_VF_INFO ||
2818                             nla_len(attr) < NLA_HDRLEN) {
2819                                 err = -EINVAL;
2820                                 goto errout;
2821                         }
2822                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2823                                                           attr,
2824                                                           ifla_vf_policy,
2825                                                           NULL);
2826                         if (err < 0)
2827                                 goto errout;
2828                         err = do_setvfinfo(dev, vfinfo);
2829                         if (err < 0)
2830                                 goto errout;
2831                         status |= DO_SETLINK_NOTIFY;
2832                 }
2833         }
2834         err = 0;
2835
2836         if (tb[IFLA_VF_PORTS]) {
2837                 struct nlattr *port[IFLA_PORT_MAX+1];
2838                 struct nlattr *attr;
2839                 int vf;
2840                 int rem;
2841
2842                 err = -EOPNOTSUPP;
2843                 if (!ops->ndo_set_vf_port)
2844                         goto errout;
2845
2846                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2847                         if (nla_type(attr) != IFLA_VF_PORT ||
2848                             nla_len(attr) < NLA_HDRLEN) {
2849                                 err = -EINVAL;
2850                                 goto errout;
2851                         }
2852                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2853                                                           attr,
2854                                                           ifla_port_policy,
2855                                                           NULL);
2856                         if (err < 0)
2857                                 goto errout;
2858                         if (!port[IFLA_PORT_VF]) {
2859                                 err = -EOPNOTSUPP;
2860                                 goto errout;
2861                         }
2862                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2863                         err = ops->ndo_set_vf_port(dev, vf, port);
2864                         if (err < 0)
2865                                 goto errout;
2866                         status |= DO_SETLINK_NOTIFY;
2867                 }
2868         }
2869         err = 0;
2870
2871         if (tb[IFLA_PORT_SELF]) {
2872                 struct nlattr *port[IFLA_PORT_MAX+1];
2873
2874                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2875                                                   tb[IFLA_PORT_SELF],
2876                                                   ifla_port_policy, NULL);
2877                 if (err < 0)
2878                         goto errout;
2879
2880                 err = -EOPNOTSUPP;
2881                 if (ops->ndo_set_vf_port)
2882                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2883                 if (err < 0)
2884                         goto errout;
2885                 status |= DO_SETLINK_NOTIFY;
2886         }
2887
2888         if (tb[IFLA_AF_SPEC]) {
2889                 struct nlattr *af;
2890                 int rem;
2891
2892                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2893                         const struct rtnl_af_ops *af_ops;
2894
2895                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2896
2897                         err = af_ops->set_link_af(dev, af, extack);
2898                         if (err < 0)
2899                                 goto errout;
2900
2901                         status |= DO_SETLINK_NOTIFY;
2902                 }
2903         }
2904         err = 0;
2905
2906         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2907                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2908                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2909                 if (err)
2910                         goto errout;
2911                 status |= DO_SETLINK_NOTIFY;
2912         }
2913
2914         if (tb[IFLA_XDP]) {
2915                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2916                 u32 xdp_flags = 0;
2917
2918                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2919                                                   tb[IFLA_XDP],
2920                                                   ifla_xdp_policy, NULL);
2921                 if (err < 0)
2922                         goto errout;
2923
2924                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2925                         err = -EINVAL;
2926                         goto errout;
2927                 }
2928
2929                 if (xdp[IFLA_XDP_FLAGS]) {
2930                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2931                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2932                                 err = -EINVAL;
2933                                 goto errout;
2934                         }
2935                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2936                                 err = -EINVAL;
2937                                 goto errout;
2938                         }
2939                 }
2940
2941                 if (xdp[IFLA_XDP_FD]) {
2942                         int expected_fd = -1;
2943
2944                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2945                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2946                                         err = -EINVAL;
2947                                         goto errout;
2948                                 }
2949                                 expected_fd =
2950                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2951                         }
2952
2953                         err = dev_change_xdp_fd(dev, extack,
2954                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2955                                                 expected_fd,
2956                                                 xdp_flags);
2957                         if (err)
2958                                 goto errout;
2959                         status |= DO_SETLINK_NOTIFY;
2960                 }
2961         }
2962
2963 errout:
2964         if (status & DO_SETLINK_MODIFIED) {
2965                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2966                         netdev_state_change(dev);
2967
2968                 if (err < 0)
2969                         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",
2970                                              dev->name);
2971         }
2972
2973         return err;
2974 }
2975
2976 static struct net_device *rtnl_dev_get(struct net *net,
2977                                        struct nlattr *ifname_attr,
2978                                        struct nlattr *altifname_attr,
2979                                        char *ifname)
2980 {
2981         char buffer[ALTIFNAMSIZ];
2982
2983         if (!ifname) {
2984                 ifname = buffer;
2985                 if (ifname_attr)
2986                         nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2987                 else if (altifname_attr)
2988                         nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2989                 else
2990                         return NULL;
2991         }
2992
2993         return __dev_get_by_name(net, ifname);
2994 }
2995
2996 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2997                         struct netlink_ext_ack *extack)
2998 {
2999         struct net *net = sock_net(skb->sk);
3000         struct ifinfomsg *ifm;
3001         struct net_device *dev;
3002         int err;
3003         struct nlattr *tb[IFLA_MAX+1];
3004         char ifname[IFNAMSIZ];
3005
3006         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3007                                      ifla_policy, extack);
3008         if (err < 0)
3009                 goto errout;
3010
3011         err = rtnl_ensure_unique_netns(tb, extack, false);
3012         if (err < 0)
3013                 goto errout;
3014
3015         if (tb[IFLA_IFNAME])
3016                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3017         else
3018                 ifname[0] = '\0';
3019
3020         err = -EINVAL;
3021         ifm = nlmsg_data(nlh);
3022         if (ifm->ifi_index > 0)
3023                 dev = __dev_get_by_index(net, ifm->ifi_index);
3024         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3025                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3026         else
3027                 goto errout;
3028
3029         if (dev == NULL) {
3030                 err = -ENODEV;
3031                 goto errout;
3032         }
3033
3034         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3035 errout:
3036         return err;
3037 }
3038
3039 static int rtnl_group_dellink(const struct net *net, int group)
3040 {
3041         struct net_device *dev, *aux;
3042         LIST_HEAD(list_kill);
3043         bool found = false;
3044
3045         if (!group)
3046                 return -EPERM;
3047
3048         for_each_netdev(net, dev) {
3049                 if (dev->group == group) {
3050                         const struct rtnl_link_ops *ops;
3051
3052                         found = true;
3053                         ops = dev->rtnl_link_ops;
3054                         if (!ops || !ops->dellink)
3055                                 return -EOPNOTSUPP;
3056                 }
3057         }
3058
3059         if (!found)
3060                 return -ENODEV;
3061
3062         for_each_netdev_safe(net, dev, aux) {
3063                 if (dev->group == group) {
3064                         const struct rtnl_link_ops *ops;
3065
3066                         ops = dev->rtnl_link_ops;
3067                         ops->dellink(dev, &list_kill);
3068                 }
3069         }
3070         unregister_netdevice_many(&list_kill);
3071
3072         return 0;
3073 }
3074
3075 int rtnl_delete_link(struct net_device *dev)
3076 {
3077         const struct rtnl_link_ops *ops;
3078         LIST_HEAD(list_kill);
3079
3080         ops = dev->rtnl_link_ops;
3081         if (!ops || !ops->dellink)
3082                 return -EOPNOTSUPP;
3083
3084         ops->dellink(dev, &list_kill);
3085         unregister_netdevice_many(&list_kill);
3086
3087         return 0;
3088 }
3089 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3090
3091 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3092                         struct netlink_ext_ack *extack)
3093 {
3094         struct net *net = sock_net(skb->sk);
3095         struct net *tgt_net = net;
3096         struct net_device *dev = NULL;
3097         struct ifinfomsg *ifm;
3098         struct nlattr *tb[IFLA_MAX+1];
3099         int err;
3100         int netnsid = -1;
3101
3102         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3103                                      ifla_policy, extack);
3104         if (err < 0)
3105                 return err;
3106
3107         err = rtnl_ensure_unique_netns(tb, extack, true);
3108         if (err < 0)
3109                 return err;
3110
3111         if (tb[IFLA_TARGET_NETNSID]) {
3112                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3113                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3114                 if (IS_ERR(tgt_net))
3115                         return PTR_ERR(tgt_net);
3116         }
3117
3118         err = -EINVAL;
3119         ifm = nlmsg_data(nlh);
3120         if (ifm->ifi_index > 0)
3121                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3122         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3123                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3124                                    tb[IFLA_ALT_IFNAME], NULL);
3125         else if (tb[IFLA_GROUP])
3126                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3127         else
3128                 goto out;
3129
3130         if (!dev) {
3131                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3132                         err = -ENODEV;
3133
3134                 goto out;
3135         }
3136
3137         err = rtnl_delete_link(dev);
3138
3139 out:
3140         if (netnsid >= 0)
3141                 put_net(tgt_net);
3142
3143         return err;
3144 }
3145
3146 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3147 {
3148         unsigned int old_flags;
3149         int err;
3150
3151         old_flags = dev->flags;
3152         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3153                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3154                                          NULL);
3155                 if (err < 0)
3156                         return err;
3157         }
3158
3159         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3160                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3161         } else {
3162                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3163                 __dev_notify_flags(dev, old_flags, ~0U);
3164         }
3165         return 0;
3166 }
3167 EXPORT_SYMBOL(rtnl_configure_link);
3168
3169 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3170                                     unsigned char name_assign_type,
3171                                     const struct rtnl_link_ops *ops,
3172                                     struct nlattr *tb[],
3173                                     struct netlink_ext_ack *extack)
3174 {
3175         struct net_device *dev;
3176         unsigned int num_tx_queues = 1;
3177         unsigned int num_rx_queues = 1;
3178         int err;
3179
3180         if (tb[IFLA_NUM_TX_QUEUES])
3181                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3182         else if (ops->get_num_tx_queues)
3183                 num_tx_queues = ops->get_num_tx_queues();
3184
3185         if (tb[IFLA_NUM_RX_QUEUES])
3186                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3187         else if (ops->get_num_rx_queues)
3188                 num_rx_queues = ops->get_num_rx_queues();
3189
3190         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3191                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3192                 return ERR_PTR(-EINVAL);
3193         }
3194
3195         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3196                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3197                 return ERR_PTR(-EINVAL);
3198         }
3199
3200         if (ops->alloc) {
3201                 dev = ops->alloc(tb, ifname, name_assign_type,
3202                                  num_tx_queues, num_rx_queues);
3203                 if (IS_ERR(dev))
3204                         return dev;
3205         } else {
3206                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3207                                        name_assign_type, ops->setup,
3208                                        num_tx_queues, num_rx_queues);
3209         }
3210
3211         if (!dev)
3212                 return ERR_PTR(-ENOMEM);
3213
3214         err = validate_linkmsg(dev, tb, extack);
3215         if (err < 0) {
3216                 free_netdev(dev);
3217                 return ERR_PTR(err);
3218         }
3219
3220         dev_net_set(dev, net);
3221         dev->rtnl_link_ops = ops;
3222         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3223
3224         if (tb[IFLA_MTU]) {
3225                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3226
3227                 err = dev_validate_mtu(dev, mtu, extack);
3228                 if (err) {
3229                         free_netdev(dev);
3230                         return ERR_PTR(err);
3231                 }
3232                 dev->mtu = mtu;
3233         }
3234         if (tb[IFLA_ADDRESS]) {
3235                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3236                                 nla_len(tb[IFLA_ADDRESS]));
3237                 dev->addr_assign_type = NET_ADDR_SET;
3238         }
3239         if (tb[IFLA_BROADCAST])
3240                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3241                                 nla_len(tb[IFLA_BROADCAST]));
3242         if (tb[IFLA_TXQLEN])
3243                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3244         if (tb[IFLA_OPERSTATE])
3245                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3246         if (tb[IFLA_LINKMODE])
3247                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3248         if (tb[IFLA_GROUP])
3249                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3250         if (tb[IFLA_GSO_MAX_SIZE])
3251                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3252         if (tb[IFLA_GSO_MAX_SEGS])
3253                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3254
3255         return dev;
3256 }
3257 EXPORT_SYMBOL(rtnl_create_link);
3258
3259 static int rtnl_group_changelink(const struct sk_buff *skb,
3260                 struct net *net, int group,
3261                 struct ifinfomsg *ifm,
3262                 struct netlink_ext_ack *extack,
3263                 struct nlattr **tb)
3264 {
3265         struct net_device *dev, *aux;
3266         int err;
3267
3268         for_each_netdev_safe(net, dev, aux) {
3269                 if (dev->group == group) {
3270                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3271                         if (err < 0)
3272                                 return err;
3273                 }
3274         }
3275
3276         return 0;
3277 }
3278
3279 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3280                           struct nlattr **attr, struct netlink_ext_ack *extack)
3281 {
3282         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3283         unsigned char name_assign_type = NET_NAME_USER;
3284         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3285         const struct rtnl_link_ops *m_ops;
3286         struct net_device *master_dev;
3287         struct net *net = sock_net(skb->sk);
3288         const struct rtnl_link_ops *ops;
3289         struct nlattr *tb[IFLA_MAX + 1];
3290         struct net *dest_net, *link_net;
3291         struct nlattr **slave_data;
3292         char kind[MODULE_NAME_LEN];
3293         struct net_device *dev;
3294         struct ifinfomsg *ifm;
3295         char ifname[IFNAMSIZ];
3296         struct nlattr **data;
3297         bool link_specified;
3298         int err;
3299
3300 #ifdef CONFIG_MODULES
3301 replay:
3302 #endif
3303         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3304                                      ifla_policy, extack);
3305         if (err < 0)
3306                 return err;
3307
3308         err = rtnl_ensure_unique_netns(tb, extack, false);
3309         if (err < 0)
3310                 return err;
3311
3312         if (tb[IFLA_IFNAME])
3313                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3314         else
3315                 ifname[0] = '\0';
3316
3317         ifm = nlmsg_data(nlh);
3318         if (ifm->ifi_index > 0) {
3319                 link_specified = true;
3320                 dev = __dev_get_by_index(net, ifm->ifi_index);
3321         } else if (ifm->ifi_index < 0) {
3322                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3323                 return -EINVAL;
3324         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3325                 link_specified = true;
3326                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3327         } else {
3328                 link_specified = false;
3329                 dev = NULL;
3330         }
3331
3332         master_dev = NULL;
3333         m_ops = NULL;
3334         if (dev) {
3335                 master_dev = netdev_master_upper_dev_get(dev);
3336                 if (master_dev)
3337                         m_ops = master_dev->rtnl_link_ops;
3338         }
3339
3340         err = validate_linkmsg(dev, tb, extack);
3341         if (err < 0)
3342                 return err;
3343
3344         if (tb[IFLA_LINKINFO]) {
3345                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3346                                                   tb[IFLA_LINKINFO],
3347                                                   ifla_info_policy, NULL);
3348                 if (err < 0)
3349                         return err;
3350         } else
3351                 memset(linkinfo, 0, sizeof(linkinfo));
3352
3353         if (linkinfo[IFLA_INFO_KIND]) {
3354                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3355                 ops = rtnl_link_ops_get(kind);
3356         } else {
3357                 kind[0] = '\0';
3358                 ops = NULL;
3359         }
3360
3361         data = NULL;
3362         if (ops) {
3363                 if (ops->maxtype > RTNL_MAX_TYPE)
3364                         return -EINVAL;
3365
3366                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3367                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3368                                                           linkinfo[IFLA_INFO_DATA],
3369                                                           ops->policy, extack);
3370                         if (err < 0)
3371                                 return err;
3372                         data = attr;
3373                 }
3374                 if (ops->validate) {
3375                         err = ops->validate(tb, data, extack);
3376                         if (err < 0)
3377                                 return err;
3378                 }
3379         }
3380
3381         slave_data = NULL;
3382         if (m_ops) {
3383                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3384                         return -EINVAL;
3385
3386                 if (m_ops->slave_maxtype &&
3387                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3388                         err = nla_parse_nested_deprecated(slave_attr,
3389                                                           m_ops->slave_maxtype,
3390                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3391                                                           m_ops->slave_policy,
3392                                                           extack);
3393                         if (err < 0)
3394                                 return err;
3395                         slave_data = slave_attr;
3396                 }
3397         }
3398
3399         if (dev) {
3400                 int status = 0;
3401
3402                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3403                         return -EEXIST;
3404                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3405                         return -EOPNOTSUPP;
3406
3407                 if (linkinfo[IFLA_INFO_DATA]) {
3408                         if (!ops || ops != dev->rtnl_link_ops ||
3409                             !ops->changelink)
3410                                 return -EOPNOTSUPP;
3411
3412                         err = ops->changelink(dev, tb, data, extack);
3413                         if (err < 0)
3414                                 return err;
3415                         status |= DO_SETLINK_NOTIFY;
3416                 }
3417
3418                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3419                         if (!m_ops || !m_ops->slave_changelink)
3420                                 return -EOPNOTSUPP;
3421
3422                         err = m_ops->slave_changelink(master_dev, dev, tb,
3423                                                       slave_data, extack);
3424                         if (err < 0)
3425                                 return err;
3426                         status |= DO_SETLINK_NOTIFY;
3427                 }
3428
3429                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3430         }
3431
3432         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3433                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3434                  * or it's for a group
3435                 */
3436                 if (link_specified)
3437                         return -ENODEV;
3438                 if (tb[IFLA_GROUP])
3439                         return rtnl_group_changelink(skb, net,
3440                                                 nla_get_u32(tb[IFLA_GROUP]),
3441                                                 ifm, extack, tb);
3442                 return -ENODEV;
3443         }
3444
3445         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3446                 return -EOPNOTSUPP;
3447
3448         if (!ops) {
3449 #ifdef CONFIG_MODULES
3450                 if (kind[0]) {
3451                         __rtnl_unlock();
3452                         request_module("rtnl-link-%s", kind);
3453                         rtnl_lock();
3454                         ops = rtnl_link_ops_get(kind);
3455                         if (ops)
3456                                 goto replay;
3457                 }
3458 #endif
3459                 NL_SET_ERR_MSG(extack, "Unknown device type");
3460                 return -EOPNOTSUPP;
3461         }
3462
3463         if (!ops->alloc && !ops->setup)
3464                 return -EOPNOTSUPP;
3465
3466         if (!ifname[0]) {
3467                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3468                 name_assign_type = NET_NAME_ENUM;
3469         }
3470
3471         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3472         if (IS_ERR(dest_net))
3473                 return PTR_ERR(dest_net);
3474
3475         if (tb[IFLA_LINK_NETNSID]) {
3476                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3477
3478                 link_net = get_net_ns_by_id(dest_net, id);
3479                 if (!link_net) {
3480                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3481                         err =  -EINVAL;
3482                         goto out;
3483                 }
3484                 err = -EPERM;
3485                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3486                         goto out;
3487         } else {
3488                 link_net = NULL;
3489         }
3490
3491         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3492                                name_assign_type, ops, tb, extack);
3493         if (IS_ERR(dev)) {
3494                 err = PTR_ERR(dev);
3495                 goto out;
3496         }
3497
3498         dev->ifindex = ifm->ifi_index;
3499
3500         if (ops->newlink)
3501                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3502         else
3503                 err = register_netdevice(dev);
3504         if (err < 0) {
3505                 free_netdev(dev);
3506                 goto out;
3507         }
3508
3509         err = rtnl_configure_link(dev, ifm);
3510         if (err < 0)
3511                 goto out_unregister;
3512         if (link_net) {
3513                 err = dev_change_net_namespace(dev, dest_net, ifname);
3514                 if (err < 0)
3515                         goto out_unregister;
3516         }
3517         if (tb[IFLA_MASTER]) {
3518                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3519                 if (err)
3520                         goto out_unregister;
3521         }
3522 out:
3523         if (link_net)
3524                 put_net(link_net);
3525         put_net(dest_net);
3526         return err;
3527 out_unregister:
3528         if (ops->newlink) {
3529                 LIST_HEAD(list_kill);
3530
3531                 ops->dellink(dev, &list_kill);
3532                 unregister_netdevice_many(&list_kill);
3533         } else {
3534                 unregister_netdevice(dev);
3535         }
3536         goto out;
3537 }
3538
3539 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3540                         struct netlink_ext_ack *extack)
3541 {
3542         struct nlattr **attr;
3543         int ret;
3544
3545         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3546         if (!attr)
3547                 return -ENOMEM;
3548
3549         ret = __rtnl_newlink(skb, nlh, attr, extack);
3550         kfree(attr);
3551         return ret;
3552 }
3553
3554 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3555                                   const struct nlmsghdr *nlh,
3556                                   struct nlattr **tb,
3557                                   struct netlink_ext_ack *extack)
3558 {
3559         struct ifinfomsg *ifm;
3560         int i, err;
3561
3562         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3563                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3564                 return -EINVAL;
3565         }
3566
3567         if (!netlink_strict_get_check(skb))
3568                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3569                                               ifla_policy, extack);
3570
3571         ifm = nlmsg_data(nlh);
3572         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3573             ifm->ifi_change) {
3574                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3575                 return -EINVAL;
3576         }
3577
3578         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3579                                             ifla_policy, extack);
3580         if (err)
3581                 return err;
3582
3583         for (i = 0; i <= IFLA_MAX; i++) {
3584                 if (!tb[i])
3585                         continue;
3586
3587                 switch (i) {
3588                 case IFLA_IFNAME:
3589                 case IFLA_ALT_IFNAME:
3590                 case IFLA_EXT_MASK:
3591                 case IFLA_TARGET_NETNSID:
3592                         break;
3593                 default:
3594                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3595                         return -EINVAL;
3596                 }
3597         }
3598
3599         return 0;
3600 }
3601
3602 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3603                         struct netlink_ext_ack *extack)
3604 {
3605         struct net *net = sock_net(skb->sk);
3606         struct net *tgt_net = net;
3607         struct ifinfomsg *ifm;
3608         struct nlattr *tb[IFLA_MAX+1];
3609         struct net_device *dev = NULL;
3610         struct sk_buff *nskb;
3611         int netnsid = -1;
3612         int err;
3613         u32 ext_filter_mask = 0;
3614
3615         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3616         if (err < 0)
3617                 return err;
3618
3619         err = rtnl_ensure_unique_netns(tb, extack, true);
3620         if (err < 0)
3621                 return err;
3622
3623         if (tb[IFLA_TARGET_NETNSID]) {
3624                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3625                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3626                 if (IS_ERR(tgt_net))
3627                         return PTR_ERR(tgt_net);
3628         }
3629
3630         if (tb[IFLA_EXT_MASK])
3631                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3632
3633         err = -EINVAL;
3634         ifm = nlmsg_data(nlh);
3635         if (ifm->ifi_index > 0)
3636                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3637         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3638                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3639                                    tb[IFLA_ALT_IFNAME], NULL);
3640         else
3641                 goto out;
3642
3643         err = -ENODEV;
3644         if (dev == NULL)
3645                 goto out;
3646
3647         err = -ENOBUFS;
3648         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3649         if (nskb == NULL)
3650                 goto out;
3651
3652         err = rtnl_fill_ifinfo(nskb, dev, net,
3653                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3654                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3655                                0, NULL, 0, netnsid, GFP_KERNEL);
3656         if (err < 0) {
3657                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3658                 WARN_ON(err == -EMSGSIZE);
3659                 kfree_skb(nskb);
3660         } else
3661                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3662 out:
3663         if (netnsid >= 0)
3664                 put_net(tgt_net);
3665
3666         return err;
3667 }
3668
3669 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3670                            bool *changed, struct netlink_ext_ack *extack)
3671 {
3672         char *alt_ifname;
3673         size_t size;
3674         int err;
3675
3676         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3677         if (err)
3678                 return err;
3679
3680         if (cmd == RTM_NEWLINKPROP) {
3681                 size = rtnl_prop_list_size(dev);
3682                 size += nla_total_size(ALTIFNAMSIZ);
3683                 if (size >= U16_MAX) {
3684                         NL_SET_ERR_MSG(extack,
3685                                        "effective property list too long");
3686                         return -EINVAL;
3687                 }
3688         }
3689
3690         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3691         if (!alt_ifname)
3692                 return -ENOMEM;
3693
3694         if (cmd == RTM_NEWLINKPROP) {
3695                 err = netdev_name_node_alt_create(dev, alt_ifname);
3696                 if (!err)
3697                         alt_ifname = NULL;
3698         } else if (cmd == RTM_DELLINKPROP) {
3699                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3700         } else {
3701                 WARN_ON_ONCE(1);
3702                 err = -EINVAL;
3703         }
3704
3705         kfree(alt_ifname);
3706         if (!err)
3707                 *changed = true;
3708         return err;
3709 }
3710
3711 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3712                          struct netlink_ext_ack *extack)
3713 {
3714         struct net *net = sock_net(skb->sk);
3715         struct nlattr *tb[IFLA_MAX + 1];
3716         struct net_device *dev;
3717         struct ifinfomsg *ifm;
3718         bool changed = false;
3719         struct nlattr *attr;
3720         int err, rem;
3721
3722         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3723         if (err)
3724                 return err;
3725
3726         err = rtnl_ensure_unique_netns(tb, extack, true);
3727         if (err)
3728                 return err;
3729
3730         ifm = nlmsg_data(nlh);
3731         if (ifm->ifi_index > 0)
3732                 dev = __dev_get_by_index(net, ifm->ifi_index);
3733         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3734                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3735                                    tb[IFLA_ALT_IFNAME], NULL);
3736         else
3737                 return -EINVAL;
3738
3739         if (!dev)
3740                 return -ENODEV;
3741
3742         if (!tb[IFLA_PROP_LIST])
3743                 return 0;
3744
3745         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3746                 switch (nla_type(attr)) {
3747                 case IFLA_ALT_IFNAME:
3748                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3749                         if (err)
3750                                 return err;
3751                         break;
3752                 }
3753         }
3754
3755         if (changed)
3756                 netdev_state_change(dev);
3757         return 0;
3758 }
3759
3760 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3761                             struct netlink_ext_ack *extack)
3762 {
3763         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3764 }
3765
3766 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3767                             struct netlink_ext_ack *extack)
3768 {
3769         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3770 }
3771
3772 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3773 {
3774         struct net *net = sock_net(skb->sk);
3775         size_t min_ifinfo_dump_size = 0;
3776         struct nlattr *tb[IFLA_MAX+1];
3777         u32 ext_filter_mask = 0;
3778         struct net_device *dev;
3779         int hdrlen;
3780
3781         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3782         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3783                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3784
3785         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3786                 if (tb[IFLA_EXT_MASK])
3787                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3788         }
3789
3790         if (!ext_filter_mask)
3791                 return NLMSG_GOODSIZE;
3792         /*
3793          * traverse the list of net devices and compute the minimum
3794          * buffer size based upon the filter mask.
3795          */
3796         rcu_read_lock();
3797         for_each_netdev_rcu(net, dev) {
3798                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3799                                            if_nlmsg_size(dev, ext_filter_mask));
3800         }
3801         rcu_read_unlock();
3802
3803         return nlmsg_total_size(min_ifinfo_dump_size);
3804 }
3805
3806 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3807 {
3808         int idx;
3809         int s_idx = cb->family;
3810         int type = cb->nlh->nlmsg_type - RTM_BASE;
3811         int ret = 0;
3812
3813         if (s_idx == 0)
3814                 s_idx = 1;
3815
3816         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3817                 struct rtnl_link __rcu **tab;
3818                 struct rtnl_link *link;
3819                 rtnl_dumpit_func dumpit;
3820
3821                 if (idx < s_idx || idx == PF_PACKET)
3822                         continue;
3823
3824                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3825                         continue;
3826
3827                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3828                 if (!tab)
3829                         continue;
3830
3831                 link = rcu_dereference_rtnl(tab[type]);
3832                 if (!link)
3833                         continue;
3834
3835                 dumpit = link->dumpit;
3836                 if (!dumpit)
3837                         continue;
3838
3839                 if (idx > s_idx) {
3840                         memset(&cb->args[0], 0, sizeof(cb->args));
3841                         cb->prev_seq = 0;
3842                         cb->seq = 0;
3843                 }
3844                 ret = dumpit(skb, cb);
3845                 if (ret)
3846                         break;
3847         }
3848         cb->family = idx;
3849
3850         return skb->len ? : ret;
3851 }
3852
3853 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3854                                        unsigned int change,
3855                                        u32 event, gfp_t flags, int *new_nsid,
3856                                        int new_ifindex)
3857 {
3858         struct net *net = dev_net(dev);
3859         struct sk_buff *skb;
3860         int err = -ENOBUFS;
3861         size_t if_info_size;
3862
3863         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3864         if (skb == NULL)
3865                 goto errout;
3866
3867         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3868                                type, 0, 0, change, 0, 0, event,
3869                                new_nsid, new_ifindex, -1, flags);
3870         if (err < 0) {
3871                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3872                 WARN_ON(err == -EMSGSIZE);
3873                 kfree_skb(skb);
3874                 goto errout;
3875         }
3876         return skb;
3877 errout:
3878         if (err < 0)
3879                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3880         return NULL;
3881 }
3882
3883 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3884 {
3885         struct net *net = dev_net(dev);
3886
3887         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3888 }
3889
3890 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3891                                unsigned int change, u32 event,
3892                                gfp_t flags, int *new_nsid, int new_ifindex)
3893 {
3894         struct sk_buff *skb;
3895
3896         if (dev->reg_state != NETREG_REGISTERED)
3897                 return;
3898
3899         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3900                                      new_ifindex);
3901         if (skb)
3902                 rtmsg_ifinfo_send(skb, dev, flags);
3903 }
3904
3905 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3906                   gfp_t flags)
3907 {
3908         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3909                            NULL, 0);
3910 }
3911
3912 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3913                          gfp_t flags, int *new_nsid, int new_ifindex)
3914 {
3915         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3916                            new_nsid, new_ifindex);
3917 }
3918
3919 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3920                                    struct net_device *dev,
3921                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3922                                    int type, unsigned int flags,
3923                                    int nlflags, u16 ndm_state)
3924 {
3925         struct nlmsghdr *nlh;
3926         struct ndmsg *ndm;
3927
3928         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3929         if (!nlh)
3930                 return -EMSGSIZE;
3931
3932         ndm = nlmsg_data(nlh);
3933         ndm->ndm_family  = AF_BRIDGE;
3934         ndm->ndm_pad1    = 0;
3935         ndm->ndm_pad2    = 0;
3936         ndm->ndm_flags   = flags;
3937         ndm->ndm_type    = 0;
3938         ndm->ndm_ifindex = dev->ifindex;
3939         ndm->ndm_state   = ndm_state;
3940
3941         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
3942                 goto nla_put_failure;
3943         if (vid)
3944                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3945                         goto nla_put_failure;
3946
3947         nlmsg_end(skb, nlh);
3948         return 0;
3949
3950 nla_put_failure:
3951         nlmsg_cancel(skb, nlh);
3952         return -EMSGSIZE;
3953 }
3954
3955 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
3956 {
3957         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3958                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
3959                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3960                0;
3961 }
3962
3963 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3964                             u16 ndm_state)
3965 {
3966         struct net *net = dev_net(dev);
3967         struct sk_buff *skb;
3968         int err = -ENOBUFS;
3969
3970         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
3971         if (!skb)
3972                 goto errout;
3973
3974         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3975                                       0, 0, type, NTF_SELF, 0, ndm_state);
3976         if (err < 0) {
3977                 kfree_skb(skb);
3978                 goto errout;
3979         }
3980
3981         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3982         return;
3983 errout:
3984         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3985 }
3986
3987 /*
3988  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3989  */
3990 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3991                      struct nlattr *tb[],
3992                      struct net_device *dev,
3993                      const unsigned char *addr, u16 vid,
3994                      u16 flags)
3995 {
3996         int err = -EINVAL;
3997
3998         /* If aging addresses are supported device will need to
3999          * implement its own handler for this.
4000          */
4001         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4002                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4003                 return err;
4004         }
4005
4006         if (vid) {
4007                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4008                 return err;
4009         }
4010
4011         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4012                 err = dev_uc_add_excl(dev, addr);
4013         else if (is_multicast_ether_addr(addr))
4014                 err = dev_mc_add_excl(dev, addr);
4015
4016         /* Only return duplicate errors if NLM_F_EXCL is set */
4017         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4018                 err = 0;
4019
4020         return err;
4021 }
4022 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4023
4024 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4025                          struct netlink_ext_ack *extack)
4026 {
4027         u16 vid = 0;
4028
4029         if (vlan_attr) {
4030                 if (nla_len(vlan_attr) != sizeof(u16)) {
4031                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4032                         return -EINVAL;
4033                 }
4034
4035                 vid = nla_get_u16(vlan_attr);
4036
4037                 if (!vid || vid >= VLAN_VID_MASK) {
4038                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4039                         return -EINVAL;
4040                 }
4041         }
4042         *p_vid = vid;
4043         return 0;
4044 }
4045
4046 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4047                         struct netlink_ext_ack *extack)
4048 {
4049         struct net *net = sock_net(skb->sk);
4050         struct ndmsg *ndm;
4051         struct nlattr *tb[NDA_MAX+1];
4052         struct net_device *dev;
4053         u8 *addr;
4054         u16 vid;
4055         int err;
4056
4057         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4058                                      extack);
4059         if (err < 0)
4060                 return err;
4061
4062         ndm = nlmsg_data(nlh);
4063         if (ndm->ndm_ifindex == 0) {
4064                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4065                 return -EINVAL;
4066         }
4067
4068         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4069         if (dev == NULL) {
4070                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4071                 return -ENODEV;
4072         }
4073
4074         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4075                 NL_SET_ERR_MSG(extack, "invalid address");
4076                 return -EINVAL;
4077         }
4078
4079         if (dev->type != ARPHRD_ETHER) {
4080                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4081                 return -EINVAL;
4082         }
4083
4084         addr = nla_data(tb[NDA_LLADDR]);
4085
4086         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4087         if (err)
4088                 return err;
4089
4090         err = -EOPNOTSUPP;
4091
4092         /* Support fdb on master device the net/bridge default case */
4093         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4094             netif_is_bridge_port(dev)) {
4095                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4096                 const struct net_device_ops *ops = br_dev->netdev_ops;
4097
4098                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4099                                        nlh->nlmsg_flags, extack);
4100                 if (err)
4101                         goto out;
4102                 else
4103                         ndm->ndm_flags &= ~NTF_MASTER;
4104         }
4105
4106         /* Embedded bridge, macvlan, and any other device support */
4107         if ((ndm->ndm_flags & NTF_SELF)) {
4108                 if (dev->netdev_ops->ndo_fdb_add)
4109                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4110                                                            vid,
4111                                                            nlh->nlmsg_flags,
4112                                                            extack);
4113                 else
4114                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4115                                                nlh->nlmsg_flags);
4116
4117                 if (!err) {
4118                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4119                                         ndm->ndm_state);
4120                         ndm->ndm_flags &= ~NTF_SELF;
4121                 }
4122         }
4123 out:
4124         return err;
4125 }
4126
4127 /*
4128  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4129  */
4130 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4131                      struct nlattr *tb[],
4132                      struct net_device *dev,
4133                      const unsigned char *addr, u16 vid)
4134 {
4135         int err = -EINVAL;
4136
4137         /* If aging addresses are supported device will need to
4138          * implement its own handler for this.
4139          */
4140         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4141                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4142                 return err;
4143         }
4144
4145         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4146                 err = dev_uc_del(dev, addr);
4147         else if (is_multicast_ether_addr(addr))
4148                 err = dev_mc_del(dev, addr);
4149
4150         return err;
4151 }
4152 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4153
4154 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4155                         struct netlink_ext_ack *extack)
4156 {
4157         struct net *net = sock_net(skb->sk);
4158         struct ndmsg *ndm;
4159         struct nlattr *tb[NDA_MAX+1];
4160         struct net_device *dev;
4161         __u8 *addr;
4162         int err;
4163         u16 vid;
4164
4165         if (!netlink_capable(skb, CAP_NET_ADMIN))
4166                 return -EPERM;
4167
4168         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4169                                      extack);
4170         if (err < 0)
4171                 return err;
4172
4173         ndm = nlmsg_data(nlh);
4174         if (ndm->ndm_ifindex == 0) {
4175                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4176                 return -EINVAL;
4177         }
4178
4179         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4180         if (dev == NULL) {
4181                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4182                 return -ENODEV;
4183         }
4184
4185         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4186                 NL_SET_ERR_MSG(extack, "invalid address");
4187                 return -EINVAL;
4188         }
4189
4190         if (dev->type != ARPHRD_ETHER) {
4191                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4192                 return -EINVAL;
4193         }
4194
4195         addr = nla_data(tb[NDA_LLADDR]);
4196
4197         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4198         if (err)
4199                 return err;
4200
4201         err = -EOPNOTSUPP;
4202
4203         /* Support fdb on master device the net/bridge default case */
4204         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4205             netif_is_bridge_port(dev)) {
4206                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4207                 const struct net_device_ops *ops = br_dev->netdev_ops;
4208
4209                 if (ops->ndo_fdb_del)
4210                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4211
4212                 if (err)
4213                         goto out;
4214                 else
4215                         ndm->ndm_flags &= ~NTF_MASTER;
4216         }
4217
4218         /* Embedded bridge, macvlan, and any other device support */
4219         if (ndm->ndm_flags & NTF_SELF) {
4220                 if (dev->netdev_ops->ndo_fdb_del)
4221                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4222                                                            vid);
4223                 else
4224                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4225
4226                 if (!err) {
4227                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4228                                         ndm->ndm_state);
4229                         ndm->ndm_flags &= ~NTF_SELF;
4230                 }
4231         }
4232 out:
4233         return err;
4234 }
4235
4236 static int nlmsg_populate_fdb(struct sk_buff *skb,
4237                               struct netlink_callback *cb,
4238                               struct net_device *dev,
4239                               int *idx,
4240                               struct netdev_hw_addr_list *list)
4241 {
4242         struct netdev_hw_addr *ha;
4243         int err;
4244         u32 portid, seq;
4245
4246         portid = NETLINK_CB(cb->skb).portid;
4247         seq = cb->nlh->nlmsg_seq;
4248
4249         list_for_each_entry(ha, &list->list, list) {
4250                 if (*idx < cb->args[2])
4251                         goto skip;
4252
4253                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4254                                               portid, seq,
4255                                               RTM_NEWNEIGH, NTF_SELF,
4256                                               NLM_F_MULTI, NUD_PERMANENT);
4257                 if (err < 0)
4258                         return err;
4259 skip:
4260                 *idx += 1;
4261         }
4262         return 0;
4263 }
4264
4265 /**
4266  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4267  * @skb: socket buffer to store message in
4268  * @cb: netlink callback
4269  * @dev: netdevice
4270  * @filter_dev: ignored
4271  * @idx: the number of FDB table entries dumped is added to *@idx
4272  *
4273  * Default netdevice operation to dump the existing unicast address list.
4274  * Returns number of addresses from list put in skb.
4275  */
4276 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4277                       struct netlink_callback *cb,
4278                       struct net_device *dev,
4279                       struct net_device *filter_dev,
4280                       int *idx)
4281 {
4282         int err;
4283
4284         if (dev->type != ARPHRD_ETHER)
4285                 return -EINVAL;
4286
4287         netif_addr_lock_bh(dev);
4288         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4289         if (err)
4290                 goto out;
4291         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4292 out:
4293         netif_addr_unlock_bh(dev);
4294         return err;
4295 }
4296 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4297
4298 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4299                                  int *br_idx, int *brport_idx,
4300                                  struct netlink_ext_ack *extack)
4301 {
4302         struct nlattr *tb[NDA_MAX + 1];
4303         struct ndmsg *ndm;
4304         int err, i;
4305
4306         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4307                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4308                 return -EINVAL;
4309         }
4310
4311         ndm = nlmsg_data(nlh);
4312         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4313             ndm->ndm_flags || ndm->ndm_type) {
4314                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4315                 return -EINVAL;
4316         }
4317
4318         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4319                                             NDA_MAX, NULL, extack);
4320         if (err < 0)
4321                 return err;
4322
4323         *brport_idx = ndm->ndm_ifindex;
4324         for (i = 0; i <= NDA_MAX; ++i) {
4325                 if (!tb[i])
4326                         continue;
4327
4328                 switch (i) {
4329                 case NDA_IFINDEX:
4330                         if (nla_len(tb[i]) != sizeof(u32)) {
4331                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4332                                 return -EINVAL;
4333                         }
4334                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4335                         break;
4336                 case NDA_MASTER:
4337                         if (nla_len(tb[i]) != sizeof(u32)) {
4338                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4339                                 return -EINVAL;
4340                         }
4341                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4342                         break;
4343                 default:
4344                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4345                         return -EINVAL;
4346                 }
4347         }
4348
4349         return 0;
4350 }
4351
4352 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4353                                  int *br_idx, int *brport_idx,
4354                                  struct netlink_ext_ack *extack)
4355 {
4356         struct nlattr *tb[IFLA_MAX+1];
4357         int err;
4358
4359         /* A hack to preserve kernel<->userspace interface.
4360          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4361          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4362          * So, check for ndmsg with an optional u32 attribute (not used here).
4363          * Fortunately these sizes don't conflict with the size of ifinfomsg
4364          * with an optional attribute.
4365          */
4366         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4367             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4368              nla_attr_size(sizeof(u32)))) {
4369                 struct ifinfomsg *ifm;
4370
4371                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4372                                              tb, IFLA_MAX, ifla_policy,
4373                                              extack);
4374                 if (err < 0) {
4375                         return -EINVAL;
4376                 } else if (err == 0) {
4377                         if (tb[IFLA_MASTER])
4378                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4379                 }
4380
4381                 ifm = nlmsg_data(nlh);
4382                 *brport_idx = ifm->ifi_index;
4383         }
4384         return 0;
4385 }
4386
4387 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4388 {
4389         struct net_device *dev;
4390         struct net_device *br_dev = NULL;
4391         const struct net_device_ops *ops = NULL;
4392         const struct net_device_ops *cops = NULL;
4393         struct net *net = sock_net(skb->sk);
4394         struct hlist_head *head;
4395         int brport_idx = 0;
4396         int br_idx = 0;
4397         int h, s_h;
4398         int idx = 0, s_idx;
4399         int err = 0;
4400         int fidx = 0;
4401
4402         if (cb->strict_check)
4403                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4404                                             cb->extack);
4405         else
4406                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4407                                             cb->extack);
4408         if (err < 0)
4409                 return err;
4410
4411         if (br_idx) {
4412                 br_dev = __dev_get_by_index(net, br_idx);
4413                 if (!br_dev)
4414                         return -ENODEV;
4415
4416                 ops = br_dev->netdev_ops;
4417         }
4418
4419         s_h = cb->args[0];
4420         s_idx = cb->args[1];
4421
4422         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4423                 idx = 0;
4424                 head = &net->dev_index_head[h];
4425                 hlist_for_each_entry(dev, head, index_hlist) {
4426
4427                         if (brport_idx && (dev->ifindex != brport_idx))
4428                                 continue;
4429
4430                         if (!br_idx) { /* user did not specify a specific bridge */
4431                                 if (netif_is_bridge_port(dev)) {
4432                                         br_dev = netdev_master_upper_dev_get(dev);
4433                                         cops = br_dev->netdev_ops;
4434                                 }
4435                         } else {
4436                                 if (dev != br_dev &&
4437                                     !netif_is_bridge_port(dev))
4438                                         continue;
4439
4440                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4441                                     !(dev->priv_flags & IFF_EBRIDGE))
4442                                         continue;
4443                                 cops = ops;
4444                         }
4445
4446                         if (idx < s_idx)
4447                                 goto cont;
4448
4449                         if (netif_is_bridge_port(dev)) {
4450                                 if (cops && cops->ndo_fdb_dump) {
4451                                         err = cops->ndo_fdb_dump(skb, cb,
4452                                                                 br_dev, dev,
4453                                                                 &fidx);
4454                                         if (err == -EMSGSIZE)
4455                                                 goto out;
4456                                 }
4457                         }
4458
4459                         if (dev->netdev_ops->ndo_fdb_dump)
4460                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4461                                                                     dev, NULL,
4462                                                                     &fidx);
4463                         else
4464                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4465                                                         &fidx);
4466                         if (err == -EMSGSIZE)
4467                                 goto out;
4468
4469                         cops = NULL;
4470
4471                         /* reset fdb offset to 0 for rest of the interfaces */
4472                         cb->args[2] = 0;
4473                         fidx = 0;
4474 cont:
4475                         idx++;
4476                 }
4477         }
4478
4479 out:
4480         cb->args[0] = h;
4481         cb->args[1] = idx;
4482         cb->args[2] = fidx;
4483
4484         return skb->len;
4485 }
4486
4487 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4488                                 struct nlattr **tb, u8 *ndm_flags,
4489                                 int *br_idx, int *brport_idx, u8 **addr,
4490                                 u16 *vid, struct netlink_ext_ack *extack)
4491 {
4492         struct ndmsg *ndm;
4493         int err, i;
4494
4495         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4496                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4497                 return -EINVAL;
4498         }
4499
4500         ndm = nlmsg_data(nlh);
4501         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4502             ndm->ndm_type) {
4503                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4504                 return -EINVAL;
4505         }
4506
4507         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4508                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4509                 return -EINVAL;
4510         }
4511
4512         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4513                                             NDA_MAX, nda_policy, extack);
4514         if (err < 0)
4515                 return err;
4516
4517         *ndm_flags = ndm->ndm_flags;
4518         *brport_idx = ndm->ndm_ifindex;
4519         for (i = 0; i <= NDA_MAX; ++i) {
4520                 if (!tb[i])
4521                         continue;
4522
4523                 switch (i) {
4524                 case NDA_MASTER:
4525                         *br_idx = nla_get_u32(tb[i]);
4526                         break;
4527                 case NDA_LLADDR:
4528                         if (nla_len(tb[i]) != ETH_ALEN) {
4529                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4530                                 return -EINVAL;
4531                         }
4532                         *addr = nla_data(tb[i]);
4533                         break;
4534                 case NDA_VLAN:
4535                         err = fdb_vid_parse(tb[i], vid, extack);
4536                         if (err)
4537                                 return err;
4538                         break;
4539                 case NDA_VNI:
4540                         break;
4541                 default:
4542                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4543                         return -EINVAL;
4544                 }
4545         }
4546
4547         return 0;
4548 }
4549
4550 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4551                         struct netlink_ext_ack *extack)
4552 {
4553         struct net_device *dev = NULL, *br_dev = NULL;
4554         const struct net_device_ops *ops = NULL;
4555         struct net *net = sock_net(in_skb->sk);
4556         struct nlattr *tb[NDA_MAX + 1];
4557         struct sk_buff *skb;
4558         int brport_idx = 0;
4559         u8 ndm_flags = 0;
4560         int br_idx = 0;
4561         u8 *addr = NULL;
4562         u16 vid = 0;
4563         int err;
4564
4565         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4566                                    &brport_idx, &addr, &vid, extack);
4567         if (err < 0)
4568                 return err;
4569
4570         if (!addr) {
4571                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4572                 return -EINVAL;
4573         }
4574
4575         if (brport_idx) {
4576                 dev = __dev_get_by_index(net, brport_idx);
4577                 if (!dev) {
4578                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4579                         return -ENODEV;
4580                 }
4581         }
4582
4583         if (br_idx) {
4584                 if (dev) {
4585                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4586                         return -EINVAL;
4587                 }
4588
4589                 br_dev = __dev_get_by_index(net, br_idx);
4590                 if (!br_dev) {
4591                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4592                         return -EINVAL;
4593                 }
4594                 ops = br_dev->netdev_ops;
4595         }
4596
4597         if (dev) {
4598                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4599                         if (!netif_is_bridge_port(dev)) {
4600                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4601                                 return -EINVAL;
4602                         }
4603                         br_dev = netdev_master_upper_dev_get(dev);
4604                         if (!br_dev) {
4605                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4606                                 return -EINVAL;
4607                         }
4608                         ops = br_dev->netdev_ops;
4609                 } else {
4610                         if (!(ndm_flags & NTF_SELF)) {
4611                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4612                                 return -EINVAL;
4613                         }
4614                         ops = dev->netdev_ops;
4615                 }
4616         }
4617
4618         if (!br_dev && !dev) {
4619                 NL_SET_ERR_MSG(extack, "No device specified");
4620                 return -ENODEV;
4621         }
4622
4623         if (!ops || !ops->ndo_fdb_get) {
4624                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4625                 return -EOPNOTSUPP;
4626         }
4627
4628         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4629         if (!skb)
4630                 return -ENOBUFS;
4631
4632         if (br_dev)
4633                 dev = br_dev;
4634         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4635                                NETLINK_CB(in_skb).portid,
4636                                nlh->nlmsg_seq, extack);
4637         if (err)
4638                 goto out;
4639
4640         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4641 out:
4642         kfree_skb(skb);
4643         return err;
4644 }
4645
4646 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4647                                unsigned int attrnum, unsigned int flag)
4648 {
4649         if (mask & flag)
4650                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4651         return 0;
4652 }
4653
4654 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4655                             struct net_device *dev, u16 mode,
4656                             u32 flags, u32 mask, int nlflags,
4657                             u32 filter_mask,
4658                             int (*vlan_fill)(struct sk_buff *skb,
4659                                              struct net_device *dev,
4660                                              u32 filter_mask))
4661 {
4662         struct nlmsghdr *nlh;
4663         struct ifinfomsg *ifm;
4664         struct nlattr *br_afspec;
4665         struct nlattr *protinfo;
4666         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4667         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4668         int err = 0;
4669
4670         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4671         if (nlh == NULL)
4672                 return -EMSGSIZE;
4673
4674         ifm = nlmsg_data(nlh);
4675         ifm->ifi_family = AF_BRIDGE;
4676         ifm->__ifi_pad = 0;
4677         ifm->ifi_type = dev->type;
4678         ifm->ifi_index = dev->ifindex;
4679         ifm->ifi_flags = dev_get_flags(dev);
4680         ifm->ifi_change = 0;
4681
4682
4683         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4684             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4685             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4686             (br_dev &&
4687              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4688             (dev->addr_len &&
4689              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4690             (dev->ifindex != dev_get_iflink(dev) &&
4691              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4692                 goto nla_put_failure;
4693
4694         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4695         if (!br_afspec)
4696                 goto nla_put_failure;
4697
4698         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4699                 nla_nest_cancel(skb, br_afspec);
4700                 goto nla_put_failure;
4701         }
4702
4703         if (mode != BRIDGE_MODE_UNDEF) {
4704                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4705                         nla_nest_cancel(skb, br_afspec);
4706                         goto nla_put_failure;
4707                 }
4708         }
4709         if (vlan_fill) {
4710                 err = vlan_fill(skb, dev, filter_mask);
4711                 if (err) {
4712                         nla_nest_cancel(skb, br_afspec);
4713                         goto nla_put_failure;
4714                 }
4715         }
4716         nla_nest_end(skb, br_afspec);
4717
4718         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4719         if (!protinfo)
4720                 goto nla_put_failure;
4721
4722         if (brport_nla_put_flag(skb, flags, mask,
4723                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4724             brport_nla_put_flag(skb, flags, mask,
4725                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4726             brport_nla_put_flag(skb, flags, mask,
4727                                 IFLA_BRPORT_FAST_LEAVE,
4728                                 BR_MULTICAST_FAST_LEAVE) ||
4729             brport_nla_put_flag(skb, flags, mask,
4730                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4731             brport_nla_put_flag(skb, flags, mask,
4732                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4733             brport_nla_put_flag(skb, flags, mask,
4734                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4735             brport_nla_put_flag(skb, flags, mask,
4736                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4737             brport_nla_put_flag(skb, flags, mask,
4738                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4739             brport_nla_put_flag(skb, flags, mask,
4740                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4741             brport_nla_put_flag(skb, flags, mask,
4742                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4743                 nla_nest_cancel(skb, protinfo);
4744                 goto nla_put_failure;
4745         }
4746
4747         nla_nest_end(skb, protinfo);
4748
4749         nlmsg_end(skb, nlh);
4750         return 0;
4751 nla_put_failure:
4752         nlmsg_cancel(skb, nlh);
4753         return err ? err : -EMSGSIZE;
4754 }
4755 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4756
4757 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4758                                     bool strict_check, u32 *filter_mask,
4759                                     struct netlink_ext_ack *extack)
4760 {
4761         struct nlattr *tb[IFLA_MAX+1];
4762         int err, i;
4763
4764         if (strict_check) {
4765                 struct ifinfomsg *ifm;
4766
4767                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4768                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4769                         return -EINVAL;
4770                 }
4771
4772                 ifm = nlmsg_data(nlh);
4773                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4774                     ifm->ifi_change || ifm->ifi_index) {
4775                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4776                         return -EINVAL;
4777                 }
4778
4779                 err = nlmsg_parse_deprecated_strict(nlh,
4780                                                     sizeof(struct ifinfomsg),
4781                                                     tb, IFLA_MAX, ifla_policy,
4782                                                     extack);
4783         } else {
4784                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4785                                              tb, IFLA_MAX, ifla_policy,
4786                                              extack);
4787         }
4788         if (err < 0)
4789                 return err;
4790
4791         /* new attributes should only be added with strict checking */
4792         for (i = 0; i <= IFLA_MAX; ++i) {
4793                 if (!tb[i])
4794                         continue;
4795
4796                 switch (i) {
4797                 case IFLA_EXT_MASK:
4798                         *filter_mask = nla_get_u32(tb[i]);
4799                         break;
4800                 default:
4801                         if (strict_check) {
4802                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4803                                 return -EINVAL;
4804                         }
4805                 }
4806         }
4807
4808         return 0;
4809 }
4810
4811 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4812 {
4813         const struct nlmsghdr *nlh = cb->nlh;
4814         struct net *net = sock_net(skb->sk);
4815         struct net_device *dev;
4816         int idx = 0;
4817         u32 portid = NETLINK_CB(cb->skb).portid;
4818         u32 seq = nlh->nlmsg_seq;
4819         u32 filter_mask = 0;
4820         int err;
4821
4822         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4823                                        cb->extack);
4824         if (err < 0 && cb->strict_check)
4825                 return err;
4826
4827         rcu_read_lock();
4828         for_each_netdev_rcu(net, dev) {
4829                 const struct net_device_ops *ops = dev->netdev_ops;
4830                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4831
4832                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4833                         if (idx >= cb->args[0]) {
4834                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4835                                                 skb, portid, seq, dev,
4836                                                 filter_mask, NLM_F_MULTI);
4837                                 if (err < 0 && err != -EOPNOTSUPP) {
4838                                         if (likely(skb->len))
4839                                                 break;
4840
4841                                         goto out_err;
4842                                 }
4843                         }
4844                         idx++;
4845                 }
4846
4847                 if (ops->ndo_bridge_getlink) {
4848                         if (idx >= cb->args[0]) {
4849                                 err = ops->ndo_bridge_getlink(skb, portid,
4850                                                               seq, dev,
4851                                                               filter_mask,
4852                                                               NLM_F_MULTI);
4853                                 if (err < 0 && err != -EOPNOTSUPP) {
4854                                         if (likely(skb->len))
4855                                                 break;
4856
4857                                         goto out_err;
4858                                 }
4859                         }
4860                         idx++;
4861                 }
4862         }
4863         err = skb->len;
4864 out_err:
4865         rcu_read_unlock();
4866         cb->args[0] = idx;
4867
4868         return err;
4869 }
4870
4871 static inline size_t bridge_nlmsg_size(void)
4872 {
4873         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4874                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4875                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4876                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4877                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4878                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4879                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4880                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4881                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4882                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4883                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4884 }
4885
4886 static int rtnl_bridge_notify(struct net_device *dev)
4887 {
4888         struct net *net = dev_net(dev);
4889         struct sk_buff *skb;
4890         int err = -EOPNOTSUPP;
4891
4892         if (!dev->netdev_ops->ndo_bridge_getlink)
4893                 return 0;
4894
4895         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4896         if (!skb) {
4897                 err = -ENOMEM;
4898                 goto errout;
4899         }
4900
4901         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4902         if (err < 0)
4903                 goto errout;
4904
4905         /* Notification info is only filled for bridge ports, not the bridge
4906          * device itself. Therefore, a zero notification length is valid and
4907          * should not result in an error.
4908          */
4909         if (!skb->len)
4910                 goto errout;
4911
4912         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4913         return 0;
4914 errout:
4915         WARN_ON(err == -EMSGSIZE);
4916         kfree_skb(skb);
4917         if (err)
4918                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4919         return err;
4920 }
4921
4922 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4923                                struct netlink_ext_ack *extack)
4924 {
4925         struct net *net = sock_net(skb->sk);
4926         struct ifinfomsg *ifm;
4927         struct net_device *dev;
4928         struct nlattr *br_spec, *attr = NULL;
4929         int rem, err = -EOPNOTSUPP;
4930         u16 flags = 0;
4931         bool have_flags = false;
4932
4933         if (nlmsg_len(nlh) < sizeof(*ifm))
4934                 return -EINVAL;
4935
4936         ifm = nlmsg_data(nlh);
4937         if (ifm->ifi_family != AF_BRIDGE)
4938                 return -EPFNOSUPPORT;
4939
4940         dev = __dev_get_by_index(net, ifm->ifi_index);
4941         if (!dev) {
4942                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4943                 return -ENODEV;
4944         }
4945
4946         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4947         if (br_spec) {
4948                 nla_for_each_nested(attr, br_spec, rem) {
4949                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
4950                                 if (nla_len(attr) < sizeof(flags))
4951                                         return -EINVAL;
4952
4953                                 have_flags = true;
4954                                 flags = nla_get_u16(attr);
4955                         }
4956
4957                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
4958                                 if (nla_len(attr) < sizeof(u16))
4959                                         return -EINVAL;
4960                         }
4961                 }
4962         }
4963
4964         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4965                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4966
4967                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4968                         err = -EOPNOTSUPP;
4969                         goto out;
4970                 }
4971
4972                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4973                                                              extack);
4974                 if (err)
4975                         goto out;
4976
4977                 flags &= ~BRIDGE_FLAGS_MASTER;
4978         }
4979
4980         if ((flags & BRIDGE_FLAGS_SELF)) {
4981                 if (!dev->netdev_ops->ndo_bridge_setlink)
4982                         err = -EOPNOTSUPP;
4983                 else
4984                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4985                                                                   flags,
4986                                                                   extack);
4987                 if (!err) {
4988                         flags &= ~BRIDGE_FLAGS_SELF;
4989
4990                         /* Generate event to notify upper layer of bridge
4991                          * change
4992                          */
4993                         err = rtnl_bridge_notify(dev);
4994                 }
4995         }
4996
4997         if (have_flags)
4998                 memcpy(nla_data(attr), &flags, sizeof(flags));
4999 out:
5000         return err;
5001 }
5002
5003 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5004                                struct netlink_ext_ack *extack)
5005 {
5006         struct net *net = sock_net(skb->sk);
5007         struct ifinfomsg *ifm;
5008         struct net_device *dev;
5009         struct nlattr *br_spec, *attr = NULL;
5010         int rem, err = -EOPNOTSUPP;
5011         u16 flags = 0;
5012         bool have_flags = false;
5013
5014         if (nlmsg_len(nlh) < sizeof(*ifm))
5015                 return -EINVAL;
5016
5017         ifm = nlmsg_data(nlh);
5018         if (ifm->ifi_family != AF_BRIDGE)
5019                 return -EPFNOSUPPORT;
5020
5021         dev = __dev_get_by_index(net, ifm->ifi_index);
5022         if (!dev) {
5023                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5024                 return -ENODEV;
5025         }
5026
5027         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5028         if (br_spec) {
5029                 nla_for_each_nested(attr, br_spec, rem) {
5030                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5031                                 if (nla_len(attr) < sizeof(flags))
5032                                         return -EINVAL;
5033
5034                                 have_flags = true;
5035                                 flags = nla_get_u16(attr);
5036                                 break;
5037                         }
5038                 }
5039         }
5040
5041         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5042                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5043
5044                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5045                         err = -EOPNOTSUPP;
5046                         goto out;
5047                 }
5048
5049                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5050                 if (err)
5051                         goto out;
5052
5053                 flags &= ~BRIDGE_FLAGS_MASTER;
5054         }
5055
5056         if ((flags & BRIDGE_FLAGS_SELF)) {
5057                 if (!dev->netdev_ops->ndo_bridge_dellink)
5058                         err = -EOPNOTSUPP;
5059                 else
5060                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5061                                                                   flags);
5062
5063                 if (!err) {
5064                         flags &= ~BRIDGE_FLAGS_SELF;
5065
5066                         /* Generate event to notify upper layer of bridge
5067                          * change
5068                          */
5069                         err = rtnl_bridge_notify(dev);
5070                 }
5071         }
5072
5073         if (have_flags)
5074                 memcpy(nla_data(attr), &flags, sizeof(flags));
5075 out:
5076         return err;
5077 }
5078
5079 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5080 {
5081         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5082                (!idxattr || idxattr == attrid);
5083 }
5084
5085 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5086 static int rtnl_get_offload_stats_attr_size(int attr_id)
5087 {
5088         switch (attr_id) {
5089         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5090                 return sizeof(struct rtnl_link_stats64);
5091         }
5092
5093         return 0;
5094 }
5095
5096 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5097                                   int *prividx)
5098 {
5099         struct nlattr *attr = NULL;
5100         int attr_id, size;
5101         void *attr_data;
5102         int err;
5103
5104         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5105               dev->netdev_ops->ndo_get_offload_stats))
5106                 return -ENODATA;
5107
5108         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5109              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5110                 if (attr_id < *prividx)
5111                         continue;
5112
5113                 size = rtnl_get_offload_stats_attr_size(attr_id);
5114                 if (!size)
5115                         continue;
5116
5117                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5118                         continue;
5119
5120                 attr = nla_reserve_64bit(skb, attr_id, size,
5121                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5122                 if (!attr)
5123                         goto nla_put_failure;
5124
5125                 attr_data = nla_data(attr);
5126                 memset(attr_data, 0, size);
5127                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5128                                                              attr_data);
5129                 if (err)
5130                         goto get_offload_stats_failure;
5131         }
5132
5133         if (!attr)
5134                 return -ENODATA;
5135
5136         *prividx = 0;
5137         return 0;
5138
5139 nla_put_failure:
5140         err = -EMSGSIZE;
5141 get_offload_stats_failure:
5142         *prividx = attr_id;
5143         return err;
5144 }
5145
5146 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5147 {
5148         int nla_size = 0;
5149         int attr_id;
5150         int size;
5151
5152         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5153               dev->netdev_ops->ndo_get_offload_stats))
5154                 return 0;
5155
5156         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5157              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5158                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5159                         continue;
5160                 size = rtnl_get_offload_stats_attr_size(attr_id);
5161                 nla_size += nla_total_size_64bit(size);
5162         }
5163
5164         if (nla_size != 0)
5165                 nla_size += nla_total_size(0);
5166
5167         return nla_size;
5168 }
5169
5170 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5171                                int type, u32 pid, u32 seq, u32 change,
5172                                unsigned int flags, unsigned int filter_mask,
5173                                int *idxattr, int *prividx)
5174 {
5175         struct if_stats_msg *ifsm;
5176         struct nlmsghdr *nlh;
5177         struct nlattr *attr;
5178         int s_prividx = *prividx;
5179         int err;
5180
5181         ASSERT_RTNL();
5182
5183         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5184         if (!nlh)
5185                 return -EMSGSIZE;
5186
5187         ifsm = nlmsg_data(nlh);
5188         ifsm->family = PF_UNSPEC;
5189         ifsm->pad1 = 0;
5190         ifsm->pad2 = 0;
5191         ifsm->ifindex = dev->ifindex;
5192         ifsm->filter_mask = filter_mask;
5193
5194         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5195                 struct rtnl_link_stats64 *sp;
5196
5197                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5198                                          sizeof(struct rtnl_link_stats64),
5199                                          IFLA_STATS_UNSPEC);
5200                 if (!attr)
5201                         goto nla_put_failure;
5202
5203                 sp = nla_data(attr);
5204                 dev_get_stats(dev, sp);
5205         }
5206
5207         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5208                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5209
5210                 if (ops && ops->fill_linkxstats) {
5211                         *idxattr = IFLA_STATS_LINK_XSTATS;
5212                         attr = nla_nest_start_noflag(skb,
5213                                                      IFLA_STATS_LINK_XSTATS);
5214                         if (!attr)
5215                                 goto nla_put_failure;
5216
5217                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5218                         nla_nest_end(skb, attr);
5219                         if (err)
5220                                 goto nla_put_failure;
5221                         *idxattr = 0;
5222                 }
5223         }
5224
5225         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5226                              *idxattr)) {
5227                 const struct rtnl_link_ops *ops = NULL;
5228                 const struct net_device *master;
5229
5230                 master = netdev_master_upper_dev_get(dev);
5231                 if (master)
5232                         ops = master->rtnl_link_ops;
5233                 if (ops && ops->fill_linkxstats) {
5234                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5235                         attr = nla_nest_start_noflag(skb,
5236                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5237                         if (!attr)
5238                                 goto nla_put_failure;
5239
5240                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5241                         nla_nest_end(skb, attr);
5242                         if (err)
5243                                 goto nla_put_failure;
5244                         *idxattr = 0;
5245                 }
5246         }
5247
5248         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5249                              *idxattr)) {
5250                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5251                 attr = nla_nest_start_noflag(skb,
5252                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5253                 if (!attr)
5254                         goto nla_put_failure;
5255
5256                 err = rtnl_get_offload_stats(skb, dev, prividx);
5257                 if (err == -ENODATA)
5258                         nla_nest_cancel(skb, attr);
5259                 else
5260                         nla_nest_end(skb, attr);
5261
5262                 if (err && err != -ENODATA)
5263                         goto nla_put_failure;
5264                 *idxattr = 0;
5265         }
5266
5267         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5268                 struct rtnl_af_ops *af_ops;
5269
5270                 *idxattr = IFLA_STATS_AF_SPEC;
5271                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5272                 if (!attr)
5273                         goto nla_put_failure;
5274
5275                 rcu_read_lock();
5276                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5277                         if (af_ops->fill_stats_af) {
5278                                 struct nlattr *af;
5279                                 int err;
5280
5281                                 af = nla_nest_start_noflag(skb,
5282                                                            af_ops->family);
5283                                 if (!af) {
5284                                         rcu_read_unlock();
5285                                         goto nla_put_failure;
5286                                 }
5287                                 err = af_ops->fill_stats_af(skb, dev);
5288
5289                                 if (err == -ENODATA) {
5290                                         nla_nest_cancel(skb, af);
5291                                 } else if (err < 0) {
5292                                         rcu_read_unlock();
5293                                         goto nla_put_failure;
5294                                 }
5295
5296                                 nla_nest_end(skb, af);
5297                         }
5298                 }
5299                 rcu_read_unlock();
5300
5301                 nla_nest_end(skb, attr);
5302
5303                 *idxattr = 0;
5304         }
5305
5306         nlmsg_end(skb, nlh);
5307
5308         return 0;
5309
5310 nla_put_failure:
5311         /* not a multi message or no progress mean a real error */
5312         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5313                 nlmsg_cancel(skb, nlh);
5314         else
5315                 nlmsg_end(skb, nlh);
5316
5317         return -EMSGSIZE;
5318 }
5319
5320 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5321                                   u32 filter_mask)
5322 {
5323         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5324
5325         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5326                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5327
5328         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5329                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5330                 int attr = IFLA_STATS_LINK_XSTATS;
5331
5332                 if (ops && ops->get_linkxstats_size) {
5333                         size += nla_total_size(ops->get_linkxstats_size(dev,
5334                                                                         attr));
5335                         /* for IFLA_STATS_LINK_XSTATS */
5336                         size += nla_total_size(0);
5337                 }
5338         }
5339
5340         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5341                 struct net_device *_dev = (struct net_device *)dev;
5342                 const struct rtnl_link_ops *ops = NULL;
5343                 const struct net_device *master;
5344
5345                 /* netdev_master_upper_dev_get can't take const */
5346                 master = netdev_master_upper_dev_get(_dev);
5347                 if (master)
5348                         ops = master->rtnl_link_ops;
5349                 if (ops && ops->get_linkxstats_size) {
5350                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5351
5352                         size += nla_total_size(ops->get_linkxstats_size(dev,
5353                                                                         attr));
5354                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5355                         size += nla_total_size(0);
5356                 }
5357         }
5358
5359         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5360                 size += rtnl_get_offload_stats_size(dev);
5361
5362         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5363                 struct rtnl_af_ops *af_ops;
5364
5365                 /* for IFLA_STATS_AF_SPEC */
5366                 size += nla_total_size(0);
5367
5368                 rcu_read_lock();
5369                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5370                         if (af_ops->get_stats_af_size) {
5371                                 size += nla_total_size(
5372                                         af_ops->get_stats_af_size(dev));
5373
5374                                 /* for AF_* */
5375                                 size += nla_total_size(0);
5376                         }
5377                 }
5378                 rcu_read_unlock();
5379         }
5380
5381         return size;
5382 }
5383
5384 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5385                                 bool is_dump, struct netlink_ext_ack *extack)
5386 {
5387         struct if_stats_msg *ifsm;
5388
5389         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5390                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5391                 return -EINVAL;
5392         }
5393
5394         if (!strict_check)
5395                 return 0;
5396
5397         ifsm = nlmsg_data(nlh);
5398
5399         /* only requests using strict checks can pass data to influence
5400          * the dump. The legacy exception is filter_mask.
5401          */
5402         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5403                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5404                 return -EINVAL;
5405         }
5406         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5407                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5408                 return -EINVAL;
5409         }
5410         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5411                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5412                 return -EINVAL;
5413         }
5414
5415         return 0;
5416 }
5417
5418 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5419                           struct netlink_ext_ack *extack)
5420 {
5421         struct net *net = sock_net(skb->sk);
5422         struct net_device *dev = NULL;
5423         int idxattr = 0, prividx = 0;
5424         struct if_stats_msg *ifsm;
5425         struct sk_buff *nskb;
5426         u32 filter_mask;
5427         int err;
5428
5429         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5430                                    false, extack);
5431         if (err)
5432                 return err;
5433
5434         ifsm = nlmsg_data(nlh);
5435         if (ifsm->ifindex > 0)
5436                 dev = __dev_get_by_index(net, ifsm->ifindex);
5437         else
5438                 return -EINVAL;
5439
5440         if (!dev)
5441                 return -ENODEV;
5442
5443         filter_mask = ifsm->filter_mask;
5444         if (!filter_mask)
5445                 return -EINVAL;
5446
5447         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5448         if (!nskb)
5449                 return -ENOBUFS;
5450
5451         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5452                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5453                                   0, filter_mask, &idxattr, &prividx);
5454         if (err < 0) {
5455                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5456                 WARN_ON(err == -EMSGSIZE);
5457                 kfree_skb(nskb);
5458         } else {
5459                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5460         }
5461
5462         return err;
5463 }
5464
5465 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5466 {
5467         struct netlink_ext_ack *extack = cb->extack;
5468         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5469         struct net *net = sock_net(skb->sk);
5470         unsigned int flags = NLM_F_MULTI;
5471         struct if_stats_msg *ifsm;
5472         struct hlist_head *head;
5473         struct net_device *dev;
5474         u32 filter_mask = 0;
5475         int idx = 0;
5476
5477         s_h = cb->args[0];
5478         s_idx = cb->args[1];
5479         s_idxattr = cb->args[2];
5480         s_prividx = cb->args[3];
5481
5482         cb->seq = net->dev_base_seq;
5483
5484         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5485         if (err)
5486                 return err;
5487
5488         ifsm = nlmsg_data(cb->nlh);
5489         filter_mask = ifsm->filter_mask;
5490         if (!filter_mask) {
5491                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5492                 return -EINVAL;
5493         }
5494
5495         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5496                 idx = 0;
5497                 head = &net->dev_index_head[h];
5498                 hlist_for_each_entry(dev, head, index_hlist) {
5499                         if (idx < s_idx)
5500                                 goto cont;
5501                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5502                                                   NETLINK_CB(cb->skb).portid,
5503                                                   cb->nlh->nlmsg_seq, 0,
5504                                                   flags, filter_mask,
5505                                                   &s_idxattr, &s_prividx);
5506                         /* If we ran out of room on the first message,
5507                          * we're in trouble
5508                          */
5509                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5510
5511                         if (err < 0)
5512                                 goto out;
5513                         s_prividx = 0;
5514                         s_idxattr = 0;
5515                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5516 cont:
5517                         idx++;
5518                 }
5519         }
5520 out:
5521         cb->args[3] = s_prividx;
5522         cb->args[2] = s_idxattr;
5523         cb->args[1] = idx;
5524         cb->args[0] = h;
5525
5526         return skb->len;
5527 }
5528
5529 /* Process one rtnetlink message. */
5530
5531 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5532                              struct netlink_ext_ack *extack)
5533 {
5534         struct net *net = sock_net(skb->sk);
5535         struct rtnl_link *link;
5536         struct module *owner;
5537         int err = -EOPNOTSUPP;
5538         rtnl_doit_func doit;
5539         unsigned int flags;
5540         int kind;
5541         int family;
5542         int type;
5543
5544         type = nlh->nlmsg_type;
5545         if (type > RTM_MAX)
5546                 return -EOPNOTSUPP;
5547
5548         type -= RTM_BASE;
5549
5550         /* All the messages must have at least 1 byte length */
5551         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5552                 return 0;
5553
5554         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5555         kind = type&3;
5556
5557         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5558                 return -EPERM;
5559
5560         rcu_read_lock();
5561         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5562                 struct sock *rtnl;
5563                 rtnl_dumpit_func dumpit;
5564                 u32 min_dump_alloc = 0;
5565
5566                 link = rtnl_get_link(family, type);
5567                 if (!link || !link->dumpit) {
5568                         family = PF_UNSPEC;
5569                         link = rtnl_get_link(family, type);
5570                         if (!link || !link->dumpit)
5571                                 goto err_unlock;
5572                 }
5573                 owner = link->owner;
5574                 dumpit = link->dumpit;
5575
5576                 if (type == RTM_GETLINK - RTM_BASE)
5577                         min_dump_alloc = rtnl_calcit(skb, nlh);
5578
5579                 err = 0;
5580                 /* need to do this before rcu_read_unlock() */
5581                 if (!try_module_get(owner))
5582                         err = -EPROTONOSUPPORT;
5583
5584                 rcu_read_unlock();
5585
5586                 rtnl = net->rtnl;
5587                 if (err == 0) {
5588                         struct netlink_dump_control c = {
5589                                 .dump           = dumpit,
5590                                 .min_dump_alloc = min_dump_alloc,
5591                                 .module         = owner,
5592                         };
5593                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5594                         /* netlink_dump_start() will keep a reference on
5595                          * module if dump is still in progress.
5596                          */
5597                         module_put(owner);
5598                 }
5599                 return err;
5600         }
5601
5602         link = rtnl_get_link(family, type);
5603         if (!link || !link->doit) {
5604                 family = PF_UNSPEC;
5605                 link = rtnl_get_link(PF_UNSPEC, type);
5606                 if (!link || !link->doit)
5607                         goto out_unlock;
5608         }
5609
5610         owner = link->owner;
5611         if (!try_module_get(owner)) {
5612                 err = -EPROTONOSUPPORT;
5613                 goto out_unlock;
5614         }
5615
5616         flags = link->flags;
5617         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5618                 doit = link->doit;
5619                 rcu_read_unlock();
5620                 if (doit)
5621                         err = doit(skb, nlh, extack);
5622                 module_put(owner);
5623                 return err;
5624         }
5625         rcu_read_unlock();
5626
5627         rtnl_lock();
5628         link = rtnl_get_link(family, type);
5629         if (link && link->doit)
5630                 err = link->doit(skb, nlh, extack);
5631         rtnl_unlock();
5632
5633         module_put(owner);
5634
5635         return err;
5636
5637 out_unlock:
5638         rcu_read_unlock();
5639         return err;
5640
5641 err_unlock:
5642         rcu_read_unlock();
5643         return -EOPNOTSUPP;
5644 }
5645
5646 static void rtnetlink_rcv(struct sk_buff *skb)
5647 {
5648         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5649 }
5650
5651 static int rtnetlink_bind(struct net *net, int group)
5652 {
5653         switch (group) {
5654         case RTNLGRP_IPV4_MROUTE_R:
5655         case RTNLGRP_IPV6_MROUTE_R:
5656                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5657                         return -EPERM;
5658                 break;
5659         }
5660         return 0;
5661 }
5662
5663 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5664 {
5665         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5666
5667         switch (event) {
5668         case NETDEV_REBOOT:
5669         case NETDEV_CHANGEMTU:
5670         case NETDEV_CHANGEADDR:
5671         case NETDEV_CHANGENAME:
5672         case NETDEV_FEAT_CHANGE:
5673         case NETDEV_BONDING_FAILOVER:
5674         case NETDEV_POST_TYPE_CHANGE:
5675         case NETDEV_NOTIFY_PEERS:
5676         case NETDEV_CHANGEUPPER:
5677         case NETDEV_RESEND_IGMP:
5678         case NETDEV_CHANGEINFODATA:
5679         case NETDEV_CHANGELOWERSTATE:
5680         case NETDEV_CHANGE_TX_QUEUE_LEN:
5681                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5682                                    GFP_KERNEL, NULL, 0);
5683                 break;
5684         default:
5685                 break;
5686         }
5687         return NOTIFY_DONE;
5688 }
5689
5690 static struct notifier_block rtnetlink_dev_notifier = {
5691         .notifier_call  = rtnetlink_event,
5692 };
5693
5694
5695 static int __net_init rtnetlink_net_init(struct net *net)
5696 {
5697         struct sock *sk;
5698         struct netlink_kernel_cfg cfg = {
5699                 .groups         = RTNLGRP_MAX,
5700                 .input          = rtnetlink_rcv,
5701                 .cb_mutex       = &rtnl_mutex,
5702                 .flags          = NL_CFG_F_NONROOT_RECV,
5703                 .bind           = rtnetlink_bind,
5704         };
5705
5706         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5707         if (!sk)
5708                 return -ENOMEM;
5709         net->rtnl = sk;
5710         return 0;
5711 }
5712
5713 static void __net_exit rtnetlink_net_exit(struct net *net)
5714 {
5715         netlink_kernel_release(net->rtnl);
5716         net->rtnl = NULL;
5717 }
5718
5719 static struct pernet_operations rtnetlink_net_ops = {
5720         .init = rtnetlink_net_init,
5721         .exit = rtnetlink_net_exit,
5722 };
5723
5724 void __init rtnetlink_init(void)
5725 {
5726         if (register_pernet_subsys(&rtnetlink_net_ops))
5727                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5728
5729         register_netdevice_notifier(&rtnetlink_dev_notifier);
5730
5731         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5732                       rtnl_dump_ifinfo, 0);
5733         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5734         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5735         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5736
5737         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5738         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5739         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5740
5741         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5742         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5743
5744         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5745         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5746         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5747
5748         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5749         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5750         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5751
5752         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5753                       0);
5754 }