net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 143
 144 static inline int rtm_msgindex(int msgtype)
 145 {
 146         int msgindex = msgtype - RTM_BASE;
 147
 148         /*
 149          * msgindex < 0 implies someone tried to register a netlink
 150          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 151          * the message type has not been added to linux/rtnetlink.h
 152          */
 153         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 154
 155         return msgindex;
 156 }
 157
 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 159 {
 160         struct rtnl_link **tab;
 161
 162         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 163                 protocol = PF_UNSPEC;
 164
 165         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 166         if (!tab)
 167                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 168
 169         return tab[msgtype];
 170 }
 171
 172 static int rtnl_register_internal(struct module *owner,
 173                                   int protocol, int msgtype,
 174                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 175                                   unsigned int flags)
 176 {
 177         struct rtnl_link *link, *old;
 178         struct rtnl_link __rcu **tab;
 179         int msgindex;
 180         int ret = -ENOBUFS;
 181
 182         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 183         msgindex = rtm_msgindex(msgtype);
 184
 185         rtnl_lock();
 186         tab = rtnl_msg_handlers[protocol];
 187         if (tab == NULL) {
 188                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 189                 if (!tab)
 190                         goto unlock;
 191
 192                 /* ensures we see the 0 stores */
 193                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 194         }
 195
 196         old = rtnl_dereference(tab[msgindex]);
 197         if (old) {
 198                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         } else {
 202                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 203                 if (!link)
 204                         goto unlock;
 205         }
 206
 207         WARN_ON(link->owner && link->owner != owner);
 208         link->owner = owner;
 209
 210         WARN_ON(doit && link->doit && link->doit != doit);
 211         if (doit)
 212                 link->doit = doit;
 213         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 214         if (dumpit)
 215                 link->dumpit = dumpit;
 216
 217         link->flags |= flags;
 218
 219         /* publish protocol:msgtype */
 220         rcu_assign_pointer(tab[msgindex], link);
 221         ret = 0;
 222         if (old)
 223                 kfree_rcu(old, rcu);
 224 unlock:
 225         rtnl_unlock();
 226         return ret;
 227 }
 228
 229 /**
 230  * rtnl_register_module - Register a rtnetlink message type
 231  *
 232  * @owner: module registering the hook (THIS_MODULE)
 233  * @protocol: Protocol family or PF_UNSPEC
 234  * @msgtype: rtnetlink message type
 235  * @doit: Function pointer called for each request message
 236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link **tab, *link;
 290         int msgindex;
 291
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293         msgindex = rtm_msgindex(msgtype);
 294
 295         rtnl_lock();
 296         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 297         if (!tab) {
 298                 rtnl_unlock();
 299                 return -ENOENT;
 300         }
 301
 302         link = tab[msgindex];
 303         rcu_assign_pointer(tab[msgindex], NULL);
 304         rtnl_unlock();
 305
 306         kfree_rcu(link, rcu);
 307
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_unregister);
 311
 312 /**
 313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 314  * @protocol : Protocol family or PF_UNSPEC
 315  *
 316  * Identical to calling rtnl_unregster() for all registered message types
 317  * of a certain protocol family.
 318  */
 319 void rtnl_unregister_all(int protocol)
 320 {
 321         struct rtnl_link **tab, *link;
 322         int msgindex;
 323
 324         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 325
 326         rtnl_lock();
 327         tab = rtnl_msg_handlers[protocol];
 328         if (!tab) {
 329                 rtnl_unlock();
 330                 return;
 331         }
 332         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 333         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 334                 link = tab[msgindex];
 335                 if (!link)
 336                         continue;
 337
 338                 rcu_assign_pointer(tab[msgindex], NULL);
 339                 kfree_rcu(link, rcu);
 340         }
 341         rtnl_unlock();
 342
 343         synchronize_net();
 344
 345         kfree(tab);
 346 }
 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 348
 349 static LIST_HEAD(link_ops);
 350
 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 352 {
 353         const struct rtnl_link_ops *ops;
 354
 355         list_for_each_entry(ops, &link_ops, list) {
 356                 if (!strcmp(ops->kind, kind))
 357                         return ops;
 358         }
 359         return NULL;
 360 }
 361
 362 /**
 363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 364  * @ops: struct rtnl_link_ops * to register
 365  *
 366  * The caller must hold the rtnl_mutex. This function should be used
 367  * by drivers that create devices during module initialization. It
 368  * must be called before registering the devices.
 369  *
 370  * Returns 0 on success or a negative error code.
 371  */
 372 int __rtnl_link_register(struct rtnl_link_ops *ops)
 373 {
 374         if (rtnl_link_ops_get(ops->kind))
 375                 return -EEXIST;
 376
 377         /* The check for setup is here because if ops
 378          * does not have that filled up, it is not possible
 379          * to use the ops for creating device. So do not
 380          * fill up dellink as well. That disables rtnl_dellink.
 381          */
 382         if (ops->setup && !ops->dellink)
 383                 ops->dellink = unregister_netdevice_queue;
 384
 385         list_add_tail(&ops->list, &link_ops);
 386         return 0;
 387 }
 388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 389
 390 /**
 391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 392  * @ops: struct rtnl_link_ops * to register
 393  *
 394  * Returns 0 on success or a negative error code.
 395  */
 396 int rtnl_link_register(struct rtnl_link_ops *ops)
 397 {
 398         int err;
 399
 400         /* Sanity-check max sizes to avoid stack buffer overflow. */
 401         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 402                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 403                 return -EINVAL;
 404
 405         rtnl_lock();
 406         err = __rtnl_link_register(ops);
 407         rtnl_unlock();
 408         return err;
 409 }
 410 EXPORT_SYMBOL_GPL(rtnl_link_register);
 411
 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 413 {
 414         struct net_device *dev;
 415         LIST_HEAD(list_kill);
 416
 417         for_each_netdev(net, dev) {
 418                 if (dev->rtnl_link_ops == ops)
 419                         ops->dellink(dev, &list_kill);
 420         }
 421         unregister_netdevice_many(&list_kill);
 422 }
 423
 424 /**
 425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 426  * @ops: struct rtnl_link_ops * to unregister
 427  *
 428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 429  * integrity (hold pernet_ops_rwsem for writing to close the race
 430  * with setup_net() and cleanup_net()).
 431  */
 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 433 {
 434         struct net *net;
 435
 436         for_each_net(net) {
 437                 __rtnl_kill_links(net, ops);
 438         }
 439         list_del(&ops->list);
 440 }
 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 442
 443 /* Return with the rtnl_lock held when there are no network
 444  * devices unregistering in any network namespace.
 445  */
 446 static void rtnl_lock_unregistering_all(void)
 447 {
 448         struct net *net;
 449         bool unregistering;
 450         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 451
 452         add_wait_queue(&netdev_unregistering_wq, &wait);
 453         for (;;) {
 454                 unregistering = false;
 455                 rtnl_lock();
 456                 /* We held write locked pernet_ops_rwsem, and parallel
 457                  * setup_net() and cleanup_net() are not possible.
 458                  */
 459                 for_each_net(net) {
 460                         if (net->dev_unreg_count > 0) {
 461                                 unregistering = true;
 462                                 break;
 463                         }
 464                 }
 465                 if (!unregistering)
 466                         break;
 467                 __rtnl_unlock();
 468
 469                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 470         }
 471         remove_wait_queue(&netdev_unregistering_wq, &wait);
 472 }
 473
 474 /**
 475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 476  * @ops: struct rtnl_link_ops * to unregister
 477  */
 478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 479 {
 480         /* Close the race with setup_net() and cleanup_net() */
 481         down_write(&pernet_ops_rwsem);
 482         rtnl_lock_unregistering_all();
 483         __rtnl_link_unregister(ops);
 484         rtnl_unlock();
 485         up_write(&pernet_ops_rwsem);
 486 }
 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 488
 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 490 {
 491         struct net_device *master_dev;
 492         const struct rtnl_link_ops *ops;
 493         size_t size = 0;
 494
 495         rcu_read_lock();
 496
 497         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 498         if (!master_dev)
 499                 goto out;
 500
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops || !ops->get_slave_size)
 503                 goto out;
 504         /* IFLA_INFO_SLAVE_DATA + nested data */
 505         size = nla_total_size(sizeof(struct nlattr)) +
 506                ops->get_slave_size(master_dev, dev);
 507
 508 out:
 509         rcu_read_unlock();
 510         return size;
 511 }
 512
 513 static size_t rtnl_link_get_size(const struct net_device *dev)
 514 {
 515         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 516         size_t size;
 517
 518         if (!ops)
 519                 return 0;
 520
 521         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 522                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 523
 524         if (ops->get_size)
 525                 /* IFLA_INFO_DATA + nested data */
 526                 size += nla_total_size(sizeof(struct nlattr)) +
 527                         ops->get_size(dev);
 528
 529         if (ops->get_xstats_size)
 530                 /* IFLA_INFO_XSTATS */
 531                 size += nla_total_size(ops->get_xstats_size(dev));
 532
 533         size += rtnl_link_get_slave_info_data_size(dev);
 534
 535         return size;
 536 }
 537
 538 static LIST_HEAD(rtnl_af_ops);
 539
 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 541 {
 542         const struct rtnl_af_ops *ops;
 543
 544         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 545                 if (ops->family == family)
 546                         return ops;
 547         }
 548
 549         return NULL;
 550 }
 551
 552 /**
 553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 554  * @ops: struct rtnl_af_ops * to register
 555  *
 556  * Returns 0 on success or a negative error code.
 557  */
 558 void rtnl_af_register(struct rtnl_af_ops *ops)
 559 {
 560         rtnl_lock();
 561         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 562         rtnl_unlock();
 563 }
 564 EXPORT_SYMBOL_GPL(rtnl_af_register);
 565
 566 /**
 567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 568  * @ops: struct rtnl_af_ops * to unregister
 569  */
 570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 571 {
 572         rtnl_lock();
 573         list_del_rcu(&ops->list);
 574         rtnl_unlock();
 575
 576         synchronize_rcu();
 577 }
 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 579
 580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 581                                     u32 ext_filter_mask)
 582 {
 583         struct rtnl_af_ops *af_ops;
 584         size_t size;
 585
 586         /* IFLA_AF_SPEC */
 587         size = nla_total_size(sizeof(struct nlattr));
 588
 589         rcu_read_lock();
 590         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 591                 if (af_ops->get_link_af_size) {
 592                         /* AF_* + nested data */
 593                         size += nla_total_size(sizeof(struct nlattr)) +
 594                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 595                 }
 596         }
 597         rcu_read_unlock();
 598
 599         return size;
 600 }
 601
 602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 603 {
 604         struct net_device *master_dev;
 605         bool ret = false;
 606
 607         rcu_read_lock();
 608
 609         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 610         if (master_dev && master_dev->rtnl_link_ops)
 611                 ret = true;
 612         rcu_read_unlock();
 613         return ret;
 614 }
 615
 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 617                                      const struct net_device *dev)
 618 {
 619         struct net_device *master_dev;
 620         const struct rtnl_link_ops *ops;
 621         struct nlattr *slave_data;
 622         int err;
 623
 624         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 625         if (!master_dev)
 626                 return 0;
 627         ops = master_dev->rtnl_link_ops;
 628         if (!ops)
 629                 return 0;
 630         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 631                 return -EMSGSIZE;
 632         if (ops->fill_slave_info) {
 633                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 634                 if (!slave_data)
 635                         return -EMSGSIZE;
 636                 err = ops->fill_slave_info(skb, master_dev, dev);
 637                 if (err < 0)
 638                         goto err_cancel_slave_data;
 639                 nla_nest_end(skb, slave_data);
 640         }
 641         return 0;
 642
 643 err_cancel_slave_data:
 644         nla_nest_cancel(skb, slave_data);
 645         return err;
 646 }
 647
 648 static int rtnl_link_info_fill(struct sk_buff *skb,
 649                                const struct net_device *dev)
 650 {
 651         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 652         struct nlattr *data;
 653         int err;
 654
 655         if (!ops)
 656                 return 0;
 657         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 658                 return -EMSGSIZE;
 659         if (ops->fill_xstats) {
 660                 err = ops->fill_xstats(skb, dev);
 661                 if (err < 0)
 662                         return err;
 663         }
 664         if (ops->fill_info) {
 665                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 666                 if (data == NULL)
 667                         return -EMSGSIZE;
 668                 err = ops->fill_info(skb, dev);
 669                 if (err < 0)
 670                         goto err_cancel_data;
 671                 nla_nest_end(skb, data);
 672         }
 673         return 0;
 674
 675 err_cancel_data:
 676         nla_nest_cancel(skb, data);
 677         return err;
 678 }
 679
 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 681 {
 682         struct nlattr *linkinfo;
 683         int err = -EMSGSIZE;
 684
 685         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 686         if (linkinfo == NULL)
 687                 goto out;
 688
 689         err = rtnl_link_info_fill(skb, dev);
 690         if (err < 0)
 691                 goto err_cancel_link;
 692
 693         err = rtnl_link_slave_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         nla_nest_end(skb, linkinfo);
 698         return 0;
 699
 700 err_cancel_link:
 701         nla_nest_cancel(skb, linkinfo);
 702 out:
 703         return err;
 704 }
 705
 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 707 {
 708         struct sock *rtnl = net->rtnl;
 709         int err = 0;
 710
 711         NETLINK_CB(skb).dst_group = group;
 712         if (echo)
 713                 refcount_inc(&skb->users);
 714         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 715         if (echo)
 716                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 717         return err;
 718 }
 719
 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 721 {
 722         struct sock *rtnl = net->rtnl;
 723
 724         return nlmsg_unicast(rtnl, skb, pid);
 725 }
 726 EXPORT_SYMBOL(rtnl_unicast);
 727
 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 729                  struct nlmsghdr *nlh, gfp_t flags)
 730 {
 731         struct sock *rtnl = net->rtnl;
 732         int report = 0;
 733
 734         if (nlh)
 735                 report = nlmsg_report(nlh);
 736
 737         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 738 }
 739 EXPORT_SYMBOL(rtnl_notify);
 740
 741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 742 {
 743         struct sock *rtnl = net->rtnl;
 744
 745         netlink_set_err(rtnl, 0, group, error);
 746 }
 747 EXPORT_SYMBOL(rtnl_set_sk_err);
 748
 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 750 {
 751         struct nlattr *mx;
 752         int i, valid = 0;
 753
 754         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 755         if (metrics == dst_default_metrics.metrics)
 756                 return 0;
 757
 758         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 759         if (mx == NULL)
 760                 return -ENOBUFS;
 761
 762         for (i = 0; i < RTAX_MAX; i++) {
 763                 if (metrics[i]) {
 764                         if (i == RTAX_CC_ALGO - 1) {
 765                                 char tmp[TCP_CA_NAME_MAX], *name;
 766
 767                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 768                                 if (!name)
 769                                         continue;
 770                                 if (nla_put_string(skb, i + 1, name))
 771                                         goto nla_put_failure;
 772                         } else if (i == RTAX_FEATURES - 1) {
 773                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 774
 775                                 if (!user_features)
 776                                         continue;
 777                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 778                                 if (nla_put_u32(skb, i + 1, user_features))
 779                                         goto nla_put_failure;
 780                         } else {
 781                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 782                                         goto nla_put_failure;
 783                         }
 784                         valid++;
 785                 }
 786         }
 787
 788         if (!valid) {
 789                 nla_nest_cancel(skb, mx);
 790                 return 0;
 791         }
 792
 793         return nla_nest_end(skb, mx);
 794
 795 nla_put_failure:
 796         nla_nest_cancel(skb, mx);
 797         return -EMSGSIZE;
 798 }
 799 EXPORT_SYMBOL(rtnetlink_put_metrics);
 800
 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 802                        long expires, u32 error)
 803 {
 804         struct rta_cacheinfo ci = {
 805                 .rta_error = error,
 806                 .rta_id =  id,
 807         };
 808
 809         if (dst) {
 810                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 811                 ci.rta_used = dst->__use;
 812                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 813         }
 814         if (expires) {
 815                 unsigned long clock;
 816
 817                 clock = jiffies_to_clock_t(abs(expires));
 818                 clock = min_t(unsigned long, clock, INT_MAX);
 819                 ci.rta_expires = (expires > 0) ? clock : -clock;
 820         }
 821         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 822 }
 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 824
 825 static void set_operstate(struct net_device *dev, unsigned char transition)
 826 {
 827         unsigned char operstate = dev->operstate;
 828
 829         switch (transition) {
 830         case IF_OPER_UP:
 831                 if ((operstate == IF_OPER_DORMANT ||
 832                      operstate == IF_OPER_TESTING ||
 833                      operstate == IF_OPER_UNKNOWN) &&
 834                     !netif_dormant(dev) && !netif_testing(dev))
 835                         operstate = IF_OPER_UP;
 836                 break;
 837
 838         case IF_OPER_TESTING:
 839                 if (operstate == IF_OPER_UP ||
 840                     operstate == IF_OPER_UNKNOWN)
 841                         operstate = IF_OPER_TESTING;
 842                 break;
 843
 844         case IF_OPER_DORMANT:
 845                 if (operstate == IF_OPER_UP ||
 846                     operstate == IF_OPER_UNKNOWN)
 847                         operstate = IF_OPER_DORMANT;
 848                 break;
 849         }
 850
 851         if (dev->operstate != operstate) {
 852                 write_lock_bh(&dev_base_lock);
 853                 dev->operstate = operstate;
 854                 write_unlock_bh(&dev_base_lock);
 855                 netdev_state_change(dev);
 856         }
 857 }
 858
 859 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 860 {
 861         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 862                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 863 }
 864
 865 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 866                                            const struct ifinfomsg *ifm)
 867 {
 868         unsigned int flags = ifm->ifi_flags;
 869
 870         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 871         if (ifm->ifi_change)
 872                 flags = (flags & ifm->ifi_change) |
 873                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 874
 875         return flags;
 876 }
 877
 878 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 879                                  const struct rtnl_link_stats64 *b)
 880 {
 881         a->rx_packets = b->rx_packets;
 882         a->tx_packets = b->tx_packets;
 883         a->rx_bytes = b->rx_bytes;
 884         a->tx_bytes = b->tx_bytes;
 885         a->rx_errors = b->rx_errors;
 886         a->tx_errors = b->tx_errors;
 887         a->rx_dropped = b->rx_dropped;
 888         a->tx_dropped = b->tx_dropped;
 889
 890         a->multicast = b->multicast;
 891         a->collisions = b->collisions;
 892
 893         a->rx_length_errors = b->rx_length_errors;
 894         a->rx_over_errors = b->rx_over_errors;
 895         a->rx_crc_errors = b->rx_crc_errors;
 896         a->rx_frame_errors = b->rx_frame_errors;
 897         a->rx_fifo_errors = b->rx_fifo_errors;
 898         a->rx_missed_errors = b->rx_missed_errors;
 899
 900         a->tx_aborted_errors = b->tx_aborted_errors;
 901         a->tx_carrier_errors = b->tx_carrier_errors;
 902         a->tx_fifo_errors = b->tx_fifo_errors;
 903         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 904         a->tx_window_errors = b->tx_window_errors;
 905
 906         a->rx_compressed = b->rx_compressed;
 907         a->tx_compressed = b->tx_compressed;
 908
 909         a->rx_nohandler = b->rx_nohandler;
 910 }
 911
 912 /* All VF info */
 913 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 914                                    u32 ext_filter_mask)
 915 {
 916         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 917                 int num_vfs = dev_num_vf(dev->dev.parent);
 918                 size_t size = nla_total_size(0);
 919                 size += num_vfs *
 920                         (nla_total_size(0) +
 921                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 922                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 923                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 924                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 925                          nla_total_size(MAX_VLAN_LIST_LEN *
 926                                         sizeof(struct ifla_vf_vlan_info)) +
 927                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 928                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 929                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 930                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 931                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 932                          nla_total_size(sizeof(struct ifla_vf_trust)));
 933                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 934                         size += num_vfs *
 935                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 936                                  /* IFLA_VF_STATS_RX_PACKETS */
 937                                  nla_total_size_64bit(sizeof(__u64)) +
 938                                  /* IFLA_VF_STATS_TX_PACKETS */
 939                                  nla_total_size_64bit(sizeof(__u64)) +
 940                                  /* IFLA_VF_STATS_RX_BYTES */
 941                                  nla_total_size_64bit(sizeof(__u64)) +
 942                                  /* IFLA_VF_STATS_TX_BYTES */
 943                                  nla_total_size_64bit(sizeof(__u64)) +
 944                                  /* IFLA_VF_STATS_BROADCAST */
 945                                  nla_total_size_64bit(sizeof(__u64)) +
 946                                  /* IFLA_VF_STATS_MULTICAST */
 947                                  nla_total_size_64bit(sizeof(__u64)) +
 948                                  /* IFLA_VF_STATS_RX_DROPPED */
 949                                  nla_total_size_64bit(sizeof(__u64)) +
 950                                  /* IFLA_VF_STATS_TX_DROPPED */
 951                                  nla_total_size_64bit(sizeof(__u64)));
 952                 }
 953                 return size;
 954         } else
 955                 return 0;
 956 }
 957
 958 static size_t rtnl_port_size(const struct net_device *dev,
 959                              u32 ext_filter_mask)
 960 {
 961         size_t port_size = nla_total_size(4)            /* PORT_VF */
 962                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 963                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 964                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 965                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 966                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 967         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 968         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 969                 + port_size;
 970         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 971                 + port_size;
 972
 973         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 974             !(ext_filter_mask & RTEXT_FILTER_VF))
 975                 return 0;
 976         if (dev_num_vf(dev->dev.parent))
 977                 return port_self_size + vf_ports_size +
 978                         vf_port_size * dev_num_vf(dev->dev.parent);
 979         else
 980                 return port_self_size;
 981 }
 982
 983 static size_t rtnl_xdp_size(void)
 984 {
 985         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 986                           nla_total_size(1) +   /* XDP_ATTACHED */
 987                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 988                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 989
 990         return xdp_size;
 991 }
 992
 993 static size_t rtnl_prop_list_size(const struct net_device *dev)
 994 {
 995         struct netdev_name_node *name_node;
 996         size_t size;
 997
 998         if (list_empty(&dev->name_node->list))
 999                 return 0;
1000         size = nla_total_size(0);
1001         list_for_each_entry(name_node, &dev->name_node->list, list)
1002                 size += nla_total_size(ALTIFNAMSIZ);
1003         return size;
1004 }
1005
1006 static size_t rtnl_proto_down_size(const struct net_device *dev)
1007 {
1008         size_t size = nla_total_size(1);
1009
1010         if (dev->proto_down_reason)
1011                 size += nla_total_size(0) + nla_total_size(4);
1012
1013         return size;
1014 }
1015
1016 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1017                                      u32 ext_filter_mask)
1018 {
1019         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1020                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1021                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1022                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1023                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1024                + nla_total_size(sizeof(struct rtnl_link_stats))
1025                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1026                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1027                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1028                + nla_total_size(4) /* IFLA_TXQLEN */
1029                + nla_total_size(4) /* IFLA_WEIGHT */
1030                + nla_total_size(4) /* IFLA_MTU */
1031                + nla_total_size(4) /* IFLA_LINK */
1032                + nla_total_size(4) /* IFLA_MASTER */
1033                + nla_total_size(1) /* IFLA_CARRIER */
1034                + nla_total_size(4) /* IFLA_PROMISCUITY */
1035                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1036                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1037                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1038                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1039                + nla_total_size(1) /* IFLA_OPERSTATE */
1040                + nla_total_size(1) /* IFLA_LINKMODE */
1041                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1042                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1043                + nla_total_size(4) /* IFLA_GROUP */
1044                + nla_total_size(ext_filter_mask
1045                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1046                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1047                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1048                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1049                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1050                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1051                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1052                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1053                + rtnl_xdp_size() /* IFLA_XDP */
1054                + nla_total_size(4)  /* IFLA_EVENT */
1055                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1056                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1057                + rtnl_proto_down_size(dev)  /* proto down */
1058                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1059                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1060                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1061                + nla_total_size(4)  /* IFLA_MIN_MTU */
1062                + nla_total_size(4)  /* IFLA_MAX_MTU */
1063                + rtnl_prop_list_size(dev)
1064                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1065                + 0;
1066 }
1067
1068 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1069 {
1070         struct nlattr *vf_ports;
1071         struct nlattr *vf_port;
1072         int vf;
1073         int err;
1074
1075         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1076         if (!vf_ports)
1077                 return -EMSGSIZE;
1078
1079         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1080                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1081                 if (!vf_port)
1082                         goto nla_put_failure;
1083                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1084                         goto nla_put_failure;
1085                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1086                 if (err == -EMSGSIZE)
1087                         goto nla_put_failure;
1088                 if (err) {
1089                         nla_nest_cancel(skb, vf_port);
1090                         continue;
1091                 }
1092                 nla_nest_end(skb, vf_port);
1093         }
1094
1095         nla_nest_end(skb, vf_ports);
1096
1097         return 0;
1098
1099 nla_put_failure:
1100         nla_nest_cancel(skb, vf_ports);
1101         return -EMSGSIZE;
1102 }
1103
1104 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1105 {
1106         struct nlattr *port_self;
1107         int err;
1108
1109         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1110         if (!port_self)
1111                 return -EMSGSIZE;
1112
1113         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1114         if (err) {
1115                 nla_nest_cancel(skb, port_self);
1116                 return (err == -EMSGSIZE) ? err : 0;
1117         }
1118
1119         nla_nest_end(skb, port_self);
1120
1121         return 0;
1122 }
1123
1124 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1125                           u32 ext_filter_mask)
1126 {
1127         int err;
1128
1129         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1130             !(ext_filter_mask & RTEXT_FILTER_VF))
1131                 return 0;
1132
1133         err = rtnl_port_self_fill(skb, dev);
1134         if (err)
1135                 return err;
1136
1137         if (dev_num_vf(dev->dev.parent)) {
1138                 err = rtnl_vf_ports_fill(skb, dev);
1139                 if (err)
1140                         return err;
1141         }
1142
1143         return 0;
1144 }
1145
1146 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1147 {
1148         int err;
1149         struct netdev_phys_item_id ppid;
1150
1151         err = dev_get_phys_port_id(dev, &ppid);
1152         if (err) {
1153                 if (err == -EOPNOTSUPP)
1154                         return 0;
1155                 return err;
1156         }
1157
1158         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1159                 return -EMSGSIZE;
1160
1161         return 0;
1162 }
1163
1164 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1165 {
1166         char name[IFNAMSIZ];
1167         int err;
1168
1169         err = dev_get_phys_port_name(dev, name, sizeof(name));
1170         if (err) {
1171                 if (err == -EOPNOTSUPP)
1172                         return 0;
1173                 return err;
1174         }
1175
1176         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1177                 return -EMSGSIZE;
1178
1179         return 0;
1180 }
1181
1182 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1183 {
1184         struct netdev_phys_item_id ppid = { };
1185         int err;
1186
1187         err = dev_get_port_parent_id(dev, &ppid, false);
1188         if (err) {
1189                 if (err == -EOPNOTSUPP)
1190                         return 0;
1191                 return err;
1192         }
1193
1194         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1195                 return -EMSGSIZE;
1196
1197         return 0;
1198 }
1199
1200 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1201                                               struct net_device *dev)
1202 {
1203         struct rtnl_link_stats64 *sp;
1204         struct nlattr *attr;
1205
1206         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1207                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1208         if (!attr)
1209                 return -EMSGSIZE;
1210
1211         sp = nla_data(attr);
1212         dev_get_stats(dev, sp);
1213
1214         attr = nla_reserve(skb, IFLA_STATS,
1215                            sizeof(struct rtnl_link_stats));
1216         if (!attr)
1217                 return -EMSGSIZE;
1218
1219         copy_rtnl_link_stats(nla_data(attr), sp);
1220
1221         return 0;
1222 }
1223
1224 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1225                                                struct net_device *dev,
1226                                                int vfs_num,
1227                                                struct nlattr *vfinfo,
1228                                                u32 ext_filter_mask)
1229 {
1230         struct ifla_vf_rss_query_en vf_rss_query_en;
1231         struct nlattr *vf, *vfstats, *vfvlanlist;
1232         struct ifla_vf_link_state vf_linkstate;
1233         struct ifla_vf_vlan_info vf_vlan_info;
1234         struct ifla_vf_spoofchk vf_spoofchk;
1235         struct ifla_vf_tx_rate vf_tx_rate;
1236         struct ifla_vf_stats vf_stats;
1237         struct ifla_vf_trust vf_trust;
1238         struct ifla_vf_vlan vf_vlan;
1239         struct ifla_vf_rate vf_rate;
1240         struct ifla_vf_mac vf_mac;
1241         struct ifla_vf_broadcast vf_broadcast;
1242         struct ifla_vf_info ivi;
1243         struct ifla_vf_guid node_guid;
1244         struct ifla_vf_guid port_guid;
1245
1246         memset(&ivi, 0, sizeof(ivi));
1247
1248         /* Not all SR-IOV capable drivers support the
1249          * spoofcheck and "RSS query enable" query.  Preset to
1250          * -1 so the user space tool can detect that the driver
1251          * didn't report anything.
1252          */
1253         ivi.spoofchk = -1;
1254         ivi.rss_query_en = -1;
1255         ivi.trusted = -1;
1256         /* The default value for VF link state is "auto"
1257          * IFLA_VF_LINK_STATE_AUTO which equals zero
1258          */
1259         ivi.linkstate = 0;
1260         /* VLAN Protocol by default is 802.1Q */
1261         ivi.vlan_proto = htons(ETH_P_8021Q);
1262         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1263                 return 0;
1264
1265         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1266         memset(&node_guid, 0, sizeof(node_guid));
1267         memset(&port_guid, 0, sizeof(port_guid));
1268
1269         vf_mac.vf =
1270                 vf_vlan.vf =
1271                 vf_vlan_info.vf =
1272                 vf_rate.vf =
1273                 vf_tx_rate.vf =
1274                 vf_spoofchk.vf =
1275                 vf_linkstate.vf =
1276                 vf_rss_query_en.vf =
1277                 vf_trust.vf =
1278                 node_guid.vf =
1279                 port_guid.vf = ivi.vf;
1280
1281         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1282         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1283         vf_vlan.vlan = ivi.vlan;
1284         vf_vlan.qos = ivi.qos;
1285         vf_vlan_info.vlan = ivi.vlan;
1286         vf_vlan_info.qos = ivi.qos;
1287         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1288         vf_tx_rate.rate = ivi.max_tx_rate;
1289         vf_rate.min_tx_rate = ivi.min_tx_rate;
1290         vf_rate.max_tx_rate = ivi.max_tx_rate;
1291         vf_spoofchk.setting = ivi.spoofchk;
1292         vf_linkstate.link_state = ivi.linkstate;
1293         vf_rss_query_en.setting = ivi.rss_query_en;
1294         vf_trust.setting = ivi.trusted;
1295         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1296         if (!vf)
1297                 goto nla_put_vfinfo_failure;
1298         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1299             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1300             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1301             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1302                     &vf_rate) ||
1303             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1304                     &vf_tx_rate) ||
1305             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1306                     &vf_spoofchk) ||
1307             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1308                     &vf_linkstate) ||
1309             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1310                     sizeof(vf_rss_query_en),
1311                     &vf_rss_query_en) ||
1312             nla_put(skb, IFLA_VF_TRUST,
1313                     sizeof(vf_trust), &vf_trust))
1314                 goto nla_put_vf_failure;
1315
1316         if (dev->netdev_ops->ndo_get_vf_guid &&
1317             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1318                                               &port_guid)) {
1319                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1320                             &node_guid) ||
1321                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1322                             &port_guid))
1323                         goto nla_put_vf_failure;
1324         }
1325         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1326         if (!vfvlanlist)
1327                 goto nla_put_vf_failure;
1328         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1329                     &vf_vlan_info)) {
1330                 nla_nest_cancel(skb, vfvlanlist);
1331                 goto nla_put_vf_failure;
1332         }
1333         nla_nest_end(skb, vfvlanlist);
1334         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1335                 memset(&vf_stats, 0, sizeof(vf_stats));
1336                 if (dev->netdev_ops->ndo_get_vf_stats)
1337                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1338                                                           &vf_stats);
1339                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1340                 if (!vfstats)
1341                         goto nla_put_vf_failure;
1342                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1343                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1344                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1345                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1346                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1347                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1348                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1349                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1350                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1351                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1352                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1353                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1354                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1355                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1356                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1357                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1358                         nla_nest_cancel(skb, vfstats);
1359                         goto nla_put_vf_failure;
1360                 }
1361                 nla_nest_end(skb, vfstats);
1362         }
1363         nla_nest_end(skb, vf);
1364         return 0;
1365
1366 nla_put_vf_failure:
1367         nla_nest_cancel(skb, vf);
1368 nla_put_vfinfo_failure:
1369         nla_nest_cancel(skb, vfinfo);
1370         return -EMSGSIZE;
1371 }
1372
1373 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1374                                            struct net_device *dev,
1375                                            u32 ext_filter_mask)
1376 {
1377         struct nlattr *vfinfo;
1378         int i, num_vfs;
1379
1380         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1381                 return 0;
1382
1383         num_vfs = dev_num_vf(dev->dev.parent);
1384         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1385                 return -EMSGSIZE;
1386
1387         if (!dev->netdev_ops->ndo_get_vf_config)
1388                 return 0;
1389
1390         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1391         if (!vfinfo)
1392                 return -EMSGSIZE;
1393
1394         for (i = 0; i < num_vfs; i++) {
1395                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo, ext_filter_mask))
1396                         return -EMSGSIZE;
1397         }
1398
1399         nla_nest_end(skb, vfinfo);
1400         return 0;
1401 }
1402
1403 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1404 {
1405         struct rtnl_link_ifmap map;
1406
1407         memset(&map, 0, sizeof(map));
1408         map.mem_start   = dev->mem_start;
1409         map.mem_end     = dev->mem_end;
1410         map.base_addr   = dev->base_addr;
1411         map.irq         = dev->irq;
1412         map.dma         = dev->dma;
1413         map.port        = dev->if_port;
1414
1415         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1416                 return -EMSGSIZE;
1417
1418         return 0;
1419 }
1420
1421 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1422 {
1423         const struct bpf_prog *generic_xdp_prog;
1424
1425         ASSERT_RTNL();
1426
1427         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1428         if (!generic_xdp_prog)
1429                 return 0;
1430         return generic_xdp_prog->aux->id;
1431 }
1432
1433 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1434 {
1435         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1436 }
1437
1438 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1439 {
1440         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1441 }
1442
1443 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1444                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1445                                u32 (*get_prog_id)(struct net_device *dev))
1446 {
1447         u32 curr_id;
1448         int err;
1449
1450         curr_id = get_prog_id(dev);
1451         if (!curr_id)
1452                 return 0;
1453
1454         *prog_id = curr_id;
1455         err = nla_put_u32(skb, attr, curr_id);
1456         if (err)
1457                 return err;
1458
1459         if (*mode != XDP_ATTACHED_NONE)
1460                 *mode = XDP_ATTACHED_MULTI;
1461         else
1462                 *mode = tgt_mode;
1463
1464         return 0;
1465 }
1466
1467 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1468 {
1469         struct nlattr *xdp;
1470         u32 prog_id;
1471         int err;
1472         u8 mode;
1473
1474         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1475         if (!xdp)
1476                 return -EMSGSIZE;
1477
1478         prog_id = 0;
1479         mode = XDP_ATTACHED_NONE;
1480         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1481                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1482         if (err)
1483                 goto err_cancel;
1484         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1485                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1486         if (err)
1487                 goto err_cancel;
1488         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1489                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1490         if (err)
1491                 goto err_cancel;
1492
1493         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1494         if (err)
1495                 goto err_cancel;
1496
1497         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1498                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1499                 if (err)
1500                         goto err_cancel;
1501         }
1502
1503         nla_nest_end(skb, xdp);
1504         return 0;
1505
1506 err_cancel:
1507         nla_nest_cancel(skb, xdp);
1508         return err;
1509 }
1510
1511 static u32 rtnl_get_event(unsigned long event)
1512 {
1513         u32 rtnl_event_type = IFLA_EVENT_NONE;
1514
1515         switch (event) {
1516         case NETDEV_REBOOT:
1517                 rtnl_event_type = IFLA_EVENT_REBOOT;
1518                 break;
1519         case NETDEV_FEAT_CHANGE:
1520                 rtnl_event_type = IFLA_EVENT_FEATURES;
1521                 break;
1522         case NETDEV_BONDING_FAILOVER:
1523                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1524                 break;
1525         case NETDEV_NOTIFY_PEERS:
1526                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1527                 break;
1528         case NETDEV_RESEND_IGMP:
1529                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1530                 break;
1531         case NETDEV_CHANGEINFODATA:
1532                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1533                 break;
1534         default:
1535                 break;
1536         }
1537
1538         return rtnl_event_type;
1539 }
1540
1541 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1542 {
1543         const struct net_device *upper_dev;
1544         int ret = 0;
1545
1546         rcu_read_lock();
1547
1548         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1549         if (upper_dev)
1550                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1551
1552         rcu_read_unlock();
1553         return ret;
1554 }
1555
1556 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1557                           bool force)
1558 {
1559         int ifindex = dev_get_iflink(dev);
1560
1561         if (force || dev->ifindex != ifindex)
1562                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1563
1564         return 0;
1565 }
1566
1567 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1568                                               struct net_device *dev)
1569 {
1570         char buf[IFALIASZ];
1571         int ret;
1572
1573         ret = dev_get_alias(dev, buf, sizeof(buf));
1574         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1575 }
1576
1577 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1578                                   const struct net_device *dev,
1579                                   struct net *src_net, gfp_t gfp)
1580 {
1581         bool put_iflink = false;
1582
1583         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1584                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1585
1586                 if (!net_eq(dev_net(dev), link_net)) {
1587                         int id = peernet2id_alloc(src_net, link_net, gfp);
1588
1589                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1590                                 return -EMSGSIZE;
1591
1592                         put_iflink = true;
1593                 }
1594         }
1595
1596         return nla_put_iflink(skb, dev, put_iflink);
1597 }
1598
1599 static int rtnl_fill_link_af(struct sk_buff *skb,
1600                              const struct net_device *dev,
1601                              u32 ext_filter_mask)
1602 {
1603         const struct rtnl_af_ops *af_ops;
1604         struct nlattr *af_spec;
1605
1606         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1607         if (!af_spec)
1608                 return -EMSGSIZE;
1609
1610         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1611                 struct nlattr *af;
1612                 int err;
1613
1614                 if (!af_ops->fill_link_af)
1615                         continue;
1616
1617                 af = nla_nest_start_noflag(skb, af_ops->family);
1618                 if (!af)
1619                         return -EMSGSIZE;
1620
1621                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1622                 /*
1623                  * Caller may return ENODATA to indicate that there
1624                  * was no data to be dumped. This is not an error, it
1625                  * means we should trim the attribute header and
1626                  * continue.
1627                  */
1628                 if (err == -ENODATA)
1629                         nla_nest_cancel(skb, af);
1630                 else if (err < 0)
1631                         return -EMSGSIZE;
1632
1633                 nla_nest_end(skb, af);
1634         }
1635
1636         nla_nest_end(skb, af_spec);
1637         return 0;
1638 }
1639
1640 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1641                                  const struct net_device *dev)
1642 {
1643         struct netdev_name_node *name_node;
1644         int count = 0;
1645
1646         list_for_each_entry(name_node, &dev->name_node->list, list) {
1647                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1648                         return -EMSGSIZE;
1649                 count++;
1650         }
1651         return count;
1652 }
1653
1654 static int rtnl_fill_prop_list(struct sk_buff *skb,
1655                                const struct net_device *dev)
1656 {
1657         struct nlattr *prop_list;
1658         int ret;
1659
1660         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1661         if (!prop_list)
1662                 return -EMSGSIZE;
1663
1664         ret = rtnl_fill_alt_ifnames(skb, dev);
1665         if (ret <= 0)
1666                 goto nest_cancel;
1667
1668         nla_nest_end(skb, prop_list);
1669         return 0;
1670
1671 nest_cancel:
1672         nla_nest_cancel(skb, prop_list);
1673         return ret;
1674 }
1675
1676 static int rtnl_fill_proto_down(struct sk_buff *skb,
1677                                 const struct net_device *dev)
1678 {
1679         struct nlattr *pr;
1680         u32 preason;
1681
1682         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1683                 goto nla_put_failure;
1684
1685         preason = dev->proto_down_reason;
1686         if (!preason)
1687                 return 0;
1688
1689         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1690         if (!pr)
1691                 return -EMSGSIZE;
1692
1693         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1694                 nla_nest_cancel(skb, pr);
1695                 goto nla_put_failure;
1696         }
1697
1698         nla_nest_end(skb, pr);
1699         return 0;
1700
1701 nla_put_failure:
1702         return -EMSGSIZE;
1703 }
1704
1705 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1706                             struct net_device *dev, struct net *src_net,
1707                             int type, u32 pid, u32 seq, u32 change,
1708                             unsigned int flags, u32 ext_filter_mask,
1709                             u32 event, int *new_nsid, int new_ifindex,
1710                             int tgt_netnsid, gfp_t gfp)
1711 {
1712         struct ifinfomsg *ifm;
1713         struct nlmsghdr *nlh;
1714         struct Qdisc *qdisc;
1715
1716         ASSERT_RTNL();
1717         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1718         if (nlh == NULL)
1719                 return -EMSGSIZE;
1720
1721         ifm = nlmsg_data(nlh);
1722         ifm->ifi_family = AF_UNSPEC;
1723         ifm->__ifi_pad = 0;
1724         ifm->ifi_type = dev->type;
1725         ifm->ifi_index = dev->ifindex;
1726         ifm->ifi_flags = dev_get_flags(dev);
1727         ifm->ifi_change = change;
1728
1729         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1730                 goto nla_put_failure;
1731
1732         qdisc = rtnl_dereference(dev->qdisc);
1733         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1734             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1735             nla_put_u8(skb, IFLA_OPERSTATE,
1736                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1737             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1738             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1739             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1740             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1741             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1742             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1743             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1744             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1745             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1746 #ifdef CONFIG_RPS
1747             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1748 #endif
1749             put_master_ifindex(skb, dev) ||
1750             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1751             (qdisc &&
1752              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1753             nla_put_ifalias(skb, dev) ||
1754             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1755                         atomic_read(&dev->carrier_up_count) +
1756                         atomic_read(&dev->carrier_down_count)) ||
1757             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1758                         atomic_read(&dev->carrier_up_count)) ||
1759             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1760                         atomic_read(&dev->carrier_down_count)))
1761                 goto nla_put_failure;
1762
1763         if (rtnl_fill_proto_down(skb, dev))
1764                 goto nla_put_failure;
1765
1766         if (event != IFLA_EVENT_NONE) {
1767                 if (nla_put_u32(skb, IFLA_EVENT, event))
1768                         goto nla_put_failure;
1769         }
1770
1771         if (rtnl_fill_link_ifmap(skb, dev))
1772                 goto nla_put_failure;
1773
1774         if (dev->addr_len) {
1775                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1776                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1777                         goto nla_put_failure;
1778         }
1779
1780         if (rtnl_phys_port_id_fill(skb, dev))
1781                 goto nla_put_failure;
1782
1783         if (rtnl_phys_port_name_fill(skb, dev))
1784                 goto nla_put_failure;
1785
1786         if (rtnl_phys_switch_id_fill(skb, dev))
1787                 goto nla_put_failure;
1788
1789         if (rtnl_fill_stats(skb, dev))
1790                 goto nla_put_failure;
1791
1792         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1793                 goto nla_put_failure;
1794
1795         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1796                 goto nla_put_failure;
1797
1798         if (rtnl_xdp_fill(skb, dev))
1799                 goto nla_put_failure;
1800
1801         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1802                 if (rtnl_link_fill(skb, dev) < 0)
1803                         goto nla_put_failure;
1804         }
1805
1806         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1807                 goto nla_put_failure;
1808
1809         if (new_nsid &&
1810             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1811                 goto nla_put_failure;
1812         if (new_ifindex &&
1813             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1814                 goto nla_put_failure;
1815
1816         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1817             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1818                 goto nla_put_failure;
1819
1820         rcu_read_lock();
1821         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1822                 goto nla_put_failure_rcu;
1823         rcu_read_unlock();
1824
1825         if (rtnl_fill_prop_list(skb, dev))
1826                 goto nla_put_failure;
1827
1828         nlmsg_end(skb, nlh);
1829         return 0;
1830
1831 nla_put_failure_rcu:
1832         rcu_read_unlock();
1833 nla_put_failure:
1834         nlmsg_cancel(skb, nlh);
1835         return -EMSGSIZE;
1836 }
1837
1838 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1839         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1840         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1841         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1842         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1843         [IFLA_MTU]              = { .type = NLA_U32 },
1844         [IFLA_LINK]             = { .type = NLA_U32 },
1845         [IFLA_MASTER]           = { .type = NLA_U32 },
1846         [IFLA_CARRIER]          = { .type = NLA_U8 },
1847         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1848         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1849         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1850         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1851         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1852         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1853         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1854         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1855          * allow 0-length string (needed to remove an alias).
1856          */
1857         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1858         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1859         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1860         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1861         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1862         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1863         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1864         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1865         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1866         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1867         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1868         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1869         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1870         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1871         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1872         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1873         [IFLA_XDP]              = { .type = NLA_NESTED },
1874         [IFLA_EVENT]            = { .type = NLA_U32 },
1875         [IFLA_GROUP]            = { .type = NLA_U32 },
1876         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1877         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1878         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1879         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1880         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1881         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1882         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1883                                     .len = ALTIFNAMSIZ - 1 },
1884         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1885         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1886 };
1887
1888 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1889         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1890         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1891         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1892         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1893 };
1894
1895 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1896         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1897         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1898         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1899         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1900         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1901         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1902         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1903         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1904         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1905         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1906         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1907         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1908         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1909 };
1910
1911 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1912         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1913         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1914                                     .len = PORT_PROFILE_MAX },
1915         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1916                                       .len = PORT_UUID_MAX },
1917         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1918                                     .len = PORT_UUID_MAX },
1919         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1920         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1921
1922         /* Unused, but we need to keep it here since user space could
1923          * fill it. It's also broken with regard to NLA_BINARY use in
1924          * combination with structs.
1925          */
1926         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1927                                     .len = sizeof(struct ifla_port_vsi) },
1928 };
1929
1930 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1931         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1932         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1933         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1934         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1935         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1936         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1937 };
1938
1939 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1940 {
1941         const struct rtnl_link_ops *ops = NULL;
1942         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1943
1944         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1945                 return NULL;
1946
1947         if (linfo[IFLA_INFO_KIND]) {
1948                 char kind[MODULE_NAME_LEN];
1949
1950                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1951                 ops = rtnl_link_ops_get(kind);
1952         }
1953
1954         return ops;
1955 }
1956
1957 static bool link_master_filtered(struct net_device *dev, int master_idx)
1958 {
1959         struct net_device *master;
1960
1961         if (!master_idx)
1962                 return false;
1963
1964         master = netdev_master_upper_dev_get(dev);
1965         if (!master || master->ifindex != master_idx)
1966                 return true;
1967
1968         return false;
1969 }
1970
1971 static bool link_kind_filtered(const struct net_device *dev,
1972                                const struct rtnl_link_ops *kind_ops)
1973 {
1974         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1975                 return true;
1976
1977         return false;
1978 }
1979
1980 static bool link_dump_filtered(struct net_device *dev,
1981                                int master_idx,
1982                                const struct rtnl_link_ops *kind_ops)
1983 {
1984         if (link_master_filtered(dev, master_idx) ||
1985             link_kind_filtered(dev, kind_ops))
1986                 return true;
1987
1988         return false;
1989 }
1990
1991 /**
1992  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1993  * @sk: netlink socket
1994  * @netnsid: network namespace identifier
1995  *
1996  * Returns the network namespace identified by netnsid on success or an error
1997  * pointer on failure.
1998  */
1999 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2000 {
2001         struct net *net;
2002
2003         net = get_net_ns_by_id(sock_net(sk), netnsid);
2004         if (!net)
2005                 return ERR_PTR(-EINVAL);
2006
2007         /* For now, the caller is required to have CAP_NET_ADMIN in
2008          * the user namespace owning the target net ns.
2009          */
2010         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2011                 put_net(net);
2012                 return ERR_PTR(-EACCES);
2013         }
2014         return net;
2015 }
2016 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2017
2018 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2019                                       bool strict_check, struct nlattr **tb,
2020                                       struct netlink_ext_ack *extack)
2021 {
2022         int hdrlen;
2023
2024         if (strict_check) {
2025                 struct ifinfomsg *ifm;
2026
2027                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2028                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2029                         return -EINVAL;
2030                 }
2031
2032                 ifm = nlmsg_data(nlh);
2033                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2034                     ifm->ifi_change) {
2035                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2036                         return -EINVAL;
2037                 }
2038                 if (ifm->ifi_index) {
2039                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2040                         return -EINVAL;
2041                 }
2042
2043                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2044                                                      IFLA_MAX, ifla_policy,
2045                                                      extack);
2046         }
2047
2048         /* A hack to preserve kernel<->userspace interface.
2049          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2050          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2051          * what iproute2 < v3.9.0 used.
2052          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2053          * attribute, its netlink message is shorter than struct ifinfomsg.
2054          */
2055         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2056                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2057
2058         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2059                                       extack);
2060 }
2061
2062 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2063 {
2064         struct netlink_ext_ack *extack = cb->extack;
2065         const struct nlmsghdr *nlh = cb->nlh;
2066         struct net *net = sock_net(skb->sk);
2067         struct net *tgt_net = net;
2068         int h, s_h;
2069         int idx = 0, s_idx;
2070         struct net_device *dev;
2071         struct hlist_head *head;
2072         struct nlattr *tb[IFLA_MAX+1];
2073         u32 ext_filter_mask = 0;
2074         const struct rtnl_link_ops *kind_ops = NULL;
2075         unsigned int flags = NLM_F_MULTI;
2076         int master_idx = 0;
2077         int netnsid = -1;
2078         int err, i;
2079
2080         s_h = cb->args[0];
2081         s_idx = cb->args[1];
2082
2083         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2084         if (err < 0) {
2085                 if (cb->strict_check)
2086                         return err;
2087
2088                 goto walk_entries;
2089         }
2090
2091         for (i = 0; i <= IFLA_MAX; ++i) {
2092                 if (!tb[i])
2093                         continue;
2094
2095                 /* new attributes should only be added with strict checking */
2096                 switch (i) {
2097                 case IFLA_TARGET_NETNSID:
2098                         netnsid = nla_get_s32(tb[i]);
2099                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2100                         if (IS_ERR(tgt_net)) {
2101                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2102                                 return PTR_ERR(tgt_net);
2103                         }
2104                         break;
2105                 case IFLA_EXT_MASK:
2106                         ext_filter_mask = nla_get_u32(tb[i]);
2107                         break;
2108                 case IFLA_MASTER:
2109                         master_idx = nla_get_u32(tb[i]);
2110                         break;
2111                 case IFLA_LINKINFO:
2112                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2113                         break;
2114                 default:
2115                         if (cb->strict_check) {
2116                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2117                                 return -EINVAL;
2118                         }
2119                 }
2120         }
2121
2122         if (master_idx || kind_ops)
2123                 flags |= NLM_F_DUMP_FILTERED;
2124
2125 walk_entries:
2126         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2127                 idx = 0;
2128                 head = &tgt_net->dev_index_head[h];
2129                 hlist_for_each_entry(dev, head, index_hlist) {
2130                         if (link_dump_filtered(dev, master_idx, kind_ops))
2131                                 goto cont;
2132                         if (idx < s_idx)
2133                                 goto cont;
2134                         err = rtnl_fill_ifinfo(skb, dev, net,
2135                                                RTM_NEWLINK,
2136                                                NETLINK_CB(cb->skb).portid,
2137                                                nlh->nlmsg_seq, 0, flags,
2138                                                ext_filter_mask, 0, NULL, 0,
2139                                                netnsid, GFP_KERNEL);
2140
2141                         if (err < 0) {
2142                                 if (likely(skb->len))
2143                                         goto out;
2144
2145                                 goto out_err;
2146                         }
2147 cont:
2148                         idx++;
2149                 }
2150         }
2151 out:
2152         err = skb->len;
2153 out_err:
2154         cb->args[1] = idx;
2155         cb->args[0] = h;
2156         cb->seq = net->dev_base_seq;
2157         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2158         if (netnsid >= 0)
2159                 put_net(tgt_net);
2160
2161         return err;
2162 }
2163
2164 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2165                              struct netlink_ext_ack *exterr)
2166 {
2167         const struct ifinfomsg *ifmp;
2168         const struct nlattr *attrs;
2169         size_t len;
2170
2171         ifmp = nla_data(nla_peer);
2172         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2173         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2174
2175         if (ifmp->ifi_index < 0) {
2176                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2177                                     "ifindex can't be negative");
2178                 return -EINVAL;
2179         }
2180
2181         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2182                                     exterr);
2183 }
2184 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2185
2186 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2187 {
2188         struct net *net;
2189         /* Examine the link attributes and figure out which
2190          * network namespace we are talking about.
2191          */
2192         if (tb[IFLA_NET_NS_PID])
2193                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2194         else if (tb[IFLA_NET_NS_FD])
2195                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2196         else
2197                 net = get_net(src_net);
2198         return net;
2199 }
2200 EXPORT_SYMBOL(rtnl_link_get_net);
2201
2202 /* Figure out which network namespace we are talking about by
2203  * examining the link attributes in the following order:
2204  *
2205  * 1. IFLA_NET_NS_PID
2206  * 2. IFLA_NET_NS_FD
2207  * 3. IFLA_TARGET_NETNSID
2208  */
2209 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2210                                                struct nlattr *tb[])
2211 {
2212         struct net *net;
2213
2214         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2215                 return rtnl_link_get_net(src_net, tb);
2216
2217         if (!tb[IFLA_TARGET_NETNSID])
2218                 return get_net(src_net);
2219
2220         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2221         if (!net)
2222                 return ERR_PTR(-EINVAL);
2223
2224         return net;
2225 }
2226
2227 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2228                                              struct net *src_net,
2229                                              struct nlattr *tb[], int cap)
2230 {
2231         struct net *net;
2232
2233         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2234         if (IS_ERR(net))
2235                 return net;
2236
2237         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2238                 put_net(net);
2239                 return ERR_PTR(-EPERM);
2240         }
2241
2242         return net;
2243 }
2244
2245 /* Verify that rtnetlink requests do not pass additional properties
2246  * potentially referring to different network namespaces.
2247  */
2248 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2249                                     struct netlink_ext_ack *extack,
2250                                     bool netns_id_only)
2251 {
2252
2253         if (netns_id_only) {
2254                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2255                         return 0;
2256
2257                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2258                 return -EOPNOTSUPP;
2259         }
2260
2261         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2262                 goto invalid_attr;
2263
2264         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2265                 goto invalid_attr;
2266
2267         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2268                 goto invalid_attr;
2269
2270         return 0;
2271
2272 invalid_attr:
2273         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2274         return -EINVAL;
2275 }
2276
2277 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2278 {
2279         if (dev) {
2280                 if (tb[IFLA_ADDRESS] &&
2281                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2282                         return -EINVAL;
2283
2284                 if (tb[IFLA_BROADCAST] &&
2285                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2286                         return -EINVAL;
2287         }
2288
2289         if (tb[IFLA_AF_SPEC]) {
2290                 struct nlattr *af;
2291                 int rem, err;
2292
2293                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2294                         const struct rtnl_af_ops *af_ops;
2295
2296                         rcu_read_lock();
2297                         af_ops = rtnl_af_lookup(nla_type(af));
2298                         if (!af_ops) {
2299                                 rcu_read_unlock();
2300                                 return -EAFNOSUPPORT;
2301                         }
2302
2303                         if (!af_ops->set_link_af) {
2304                                 rcu_read_unlock();
2305                                 return -EOPNOTSUPP;
2306                         }
2307
2308                         if (af_ops->validate_link_af) {
2309                                 err = af_ops->validate_link_af(dev, af);
2310                                 if (err < 0) {
2311                                         rcu_read_unlock();
2312                                         return err;
2313                                 }
2314                         }
2315
2316                         rcu_read_unlock();
2317                 }
2318         }
2319
2320         return 0;
2321 }
2322
2323 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2324                                   int guid_type)
2325 {
2326         const struct net_device_ops *ops = dev->netdev_ops;
2327
2328         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2329 }
2330
2331 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2332 {
2333         if (dev->type != ARPHRD_INFINIBAND)
2334                 return -EOPNOTSUPP;
2335
2336         return handle_infiniband_guid(dev, ivt, guid_type);
2337 }
2338
2339 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2340 {
2341         const struct net_device_ops *ops = dev->netdev_ops;
2342         int err = -EINVAL;
2343
2344         if (tb[IFLA_VF_MAC]) {
2345                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2346
2347                 if (ivm->vf >= INT_MAX)
2348                         return -EINVAL;
2349                 err = -EOPNOTSUPP;
2350                 if (ops->ndo_set_vf_mac)
2351                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2352                                                   ivm->mac);
2353                 if (err < 0)
2354                         return err;
2355         }
2356
2357         if (tb[IFLA_VF_VLAN]) {
2358                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2359
2360                 if (ivv->vf >= INT_MAX)
2361                         return -EINVAL;
2362                 err = -EOPNOTSUPP;
2363                 if (ops->ndo_set_vf_vlan)
2364                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2365                                                    ivv->qos,
2366                                                    htons(ETH_P_8021Q));
2367                 if (err < 0)
2368                         return err;
2369         }
2370
2371         if (tb[IFLA_VF_VLAN_LIST]) {
2372                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2373                 struct nlattr *attr;
2374                 int rem, len = 0;
2375
2376                 err = -EOPNOTSUPP;
2377                 if (!ops->ndo_set_vf_vlan)
2378                         return err;
2379
2380                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2381                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2382                             nla_len(attr) < sizeof(struct ifla_vf_vlan_info)) {
2383                                 return -EINVAL;
2384                         }
2385                         if (len >= MAX_VLAN_LIST_LEN)
2386                                 return -EOPNOTSUPP;
2387                         ivvl[len] = nla_data(attr);
2388
2389                         len++;
2390                 }
2391                 if (len == 0)
2392                         return -EINVAL;
2393
2394                 if (ivvl[0]->vf >= INT_MAX)
2395                         return -EINVAL;
2396                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2397                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2398                 if (err < 0)
2399                         return err;
2400         }
2401
2402         if (tb[IFLA_VF_TX_RATE]) {
2403                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2404                 struct ifla_vf_info ivf;
2405
2406                 if (ivt->vf >= INT_MAX)
2407                         return -EINVAL;
2408                 err = -EOPNOTSUPP;
2409                 if (ops->ndo_get_vf_config)
2410                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2411                 if (err < 0)
2412                         return err;
2413
2414                 err = -EOPNOTSUPP;
2415                 if (ops->ndo_set_vf_rate)
2416                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2417                                                    ivf.min_tx_rate,
2418                                                    ivt->rate);
2419                 if (err < 0)
2420                         return err;
2421         }
2422
2423         if (tb[IFLA_VF_RATE]) {
2424                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2425
2426                 if (ivt->vf >= INT_MAX)
2427                         return -EINVAL;
2428                 err = -EOPNOTSUPP;
2429                 if (ops->ndo_set_vf_rate)
2430                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2431                                                    ivt->min_tx_rate,
2432                                                    ivt->max_tx_rate);
2433                 if (err < 0)
2434                         return err;
2435         }
2436
2437         if (tb[IFLA_VF_SPOOFCHK]) {
2438                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2439
2440                 if (ivs->vf >= INT_MAX)
2441                         return -EINVAL;
2442                 err = -EOPNOTSUPP;
2443                 if (ops->ndo_set_vf_spoofchk)
2444                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2445                                                        ivs->setting);
2446                 if (err < 0)
2447                         return err;
2448         }
2449
2450         if (tb[IFLA_VF_LINK_STATE]) {
2451                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2452
2453                 if (ivl->vf >= INT_MAX)
2454                         return -EINVAL;
2455                 err = -EOPNOTSUPP;
2456                 if (ops->ndo_set_vf_link_state)
2457                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2458                                                          ivl->link_state);
2459                 if (err < 0)
2460                         return err;
2461         }
2462
2463         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2464                 struct ifla_vf_rss_query_en *ivrssq_en;
2465
2466                 err = -EOPNOTSUPP;
2467                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2468                 if (ivrssq_en->vf >= INT_MAX)
2469                         return -EINVAL;
2470                 if (ops->ndo_set_vf_rss_query_en)
2471                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2472                                                            ivrssq_en->setting);
2473                 if (err < 0)
2474                         return err;
2475         }
2476
2477         if (tb[IFLA_VF_TRUST]) {
2478                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2479
2480                 if (ivt->vf >= INT_MAX)
2481                         return -EINVAL;
2482                 err = -EOPNOTSUPP;
2483                 if (ops->ndo_set_vf_trust)
2484                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2485                 if (err < 0)
2486                         return err;
2487         }
2488
2489         if (tb[IFLA_VF_IB_NODE_GUID]) {
2490                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2491
2492                 if (ivt->vf >= INT_MAX)
2493                         return -EINVAL;
2494                 if (!ops->ndo_set_vf_guid)
2495                         return -EOPNOTSUPP;
2496                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2497         }
2498
2499         if (tb[IFLA_VF_IB_PORT_GUID]) {
2500                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2501
2502                 if (ivt->vf >= INT_MAX)
2503                         return -EINVAL;
2504                 if (!ops->ndo_set_vf_guid)
2505                         return -EOPNOTSUPP;
2506
2507                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2508         }
2509
2510         return err;
2511 }
2512
2513 static int do_set_master(struct net_device *dev, int ifindex,
2514                          struct netlink_ext_ack *extack)
2515 {
2516         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2517         const struct net_device_ops *ops;
2518         int err;
2519
2520         if (upper_dev) {
2521                 if (upper_dev->ifindex == ifindex)
2522                         return 0;
2523                 ops = upper_dev->netdev_ops;
2524                 if (ops->ndo_del_slave) {
2525                         err = ops->ndo_del_slave(upper_dev, dev);
2526                         if (err)
2527                                 return err;
2528                 } else {
2529                         return -EOPNOTSUPP;
2530                 }
2531         }
2532
2533         if (ifindex) {
2534                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2535                 if (!upper_dev)
2536                         return -EINVAL;
2537                 ops = upper_dev->netdev_ops;
2538                 if (ops->ndo_add_slave) {
2539                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2540                         if (err)
2541                                 return err;
2542                 } else {
2543                         return -EOPNOTSUPP;
2544                 }
2545         }
2546         return 0;
2547 }
2548
2549 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2550         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2551         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2552 };
2553
2554 static int do_set_proto_down(struct net_device *dev,
2555                              struct nlattr *nl_proto_down,
2556                              struct nlattr *nl_proto_down_reason,
2557                              struct netlink_ext_ack *extack)
2558 {
2559         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2560         const struct net_device_ops *ops = dev->netdev_ops;
2561         unsigned long mask = 0;
2562         u32 value;
2563         bool proto_down;
2564         int err;
2565
2566         if (!ops->ndo_change_proto_down) {
2567                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2568                 return -EOPNOTSUPP;
2569         }
2570
2571         if (nl_proto_down_reason) {
2572                 err = nla_parse_nested_deprecated(pdreason,
2573                                                   IFLA_PROTO_DOWN_REASON_MAX,
2574                                                   nl_proto_down_reason,
2575                                                   ifla_proto_down_reason_policy,
2576                                                   NULL);
2577                 if (err < 0)
2578                         return err;
2579
2580                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2581                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2582                         return -EINVAL;
2583                 }
2584
2585                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2586
2587                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2588                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2589
2590                 dev_change_proto_down_reason(dev, mask, value);
2591         }
2592
2593         if (nl_proto_down) {
2594                 proto_down = nla_get_u8(nl_proto_down);
2595
2596                 /* Dont turn off protodown if there are active reasons */
2597                 if (!proto_down && dev->proto_down_reason) {
2598                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2599                         return -EBUSY;
2600                 }
2601                 err = dev_change_proto_down(dev,
2602                                             proto_down);
2603                 if (err)
2604                         return err;
2605         }
2606
2607         return 0;
2608 }
2609
2610 #define DO_SETLINK_MODIFIED     0x01
2611 /* notify flag means notify + modified. */
2612 #define DO_SETLINK_NOTIFY       0x03
2613 static int do_setlink(const struct sk_buff *skb,
2614                       struct net_device *dev, struct ifinfomsg *ifm,
2615                       struct netlink_ext_ack *extack,
2616                       struct nlattr **tb, char *ifname, int status)
2617 {
2618         const struct net_device_ops *ops = dev->netdev_ops;
2619         int err;
2620
2621         err = validate_linkmsg(dev, tb);
2622         if (err < 0)
2623                 return err;
2624
2625         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2626                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2627                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2628                                                             tb, CAP_NET_ADMIN);
2629                 if (IS_ERR(net)) {
2630                         err = PTR_ERR(net);
2631                         goto errout;
2632                 }
2633
2634                 err = dev_change_net_namespace(dev, net, pat);
2635                 put_net(net);
2636                 if (err)
2637                         goto errout;
2638                 status |= DO_SETLINK_MODIFIED;
2639         }
2640
2641         if (tb[IFLA_MAP]) {
2642                 struct rtnl_link_ifmap *u_map;
2643                 struct ifmap k_map;
2644
2645                 if (!ops->ndo_set_config) {
2646                         err = -EOPNOTSUPP;
2647                         goto errout;
2648                 }
2649
2650                 if (!netif_device_present(dev)) {
2651                         err = -ENODEV;
2652                         goto errout;
2653                 }
2654
2655                 u_map = nla_data(tb[IFLA_MAP]);
2656                 k_map.mem_start = (unsigned long) u_map->mem_start;
2657                 k_map.mem_end = (unsigned long) u_map->mem_end;
2658                 k_map.base_addr = (unsigned short) u_map->base_addr;
2659                 k_map.irq = (unsigned char) u_map->irq;
2660                 k_map.dma = (unsigned char) u_map->dma;
2661                 k_map.port = (unsigned char) u_map->port;
2662
2663                 err = ops->ndo_set_config(dev, &k_map);
2664                 if (err < 0)
2665                         goto errout;
2666
2667                 status |= DO_SETLINK_NOTIFY;
2668         }
2669
2670         if (tb[IFLA_ADDRESS]) {
2671                 struct sockaddr *sa;
2672                 int len;
2673
2674                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2675                                                   sizeof(*sa));
2676                 sa = kmalloc(len, GFP_KERNEL);
2677                 if (!sa) {
2678                         err = -ENOMEM;
2679                         goto errout;
2680                 }
2681                 sa->sa_family = dev->type;
2682                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2683                        dev->addr_len);
2684                 err = dev_set_mac_address_user(dev, sa, extack);
2685                 kfree(sa);
2686                 if (err)
2687                         goto errout;
2688                 status |= DO_SETLINK_MODIFIED;
2689         }
2690
2691         if (tb[IFLA_MTU]) {
2692                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2693                 if (err < 0)
2694                         goto errout;
2695                 status |= DO_SETLINK_MODIFIED;
2696         }
2697
2698         if (tb[IFLA_GROUP]) {
2699                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2700                 status |= DO_SETLINK_NOTIFY;
2701         }
2702
2703         /*
2704          * Interface selected by interface index but interface
2705          * name provided implies that a name change has been
2706          * requested.
2707          */
2708         if (ifm->ifi_index > 0 && ifname[0]) {
2709                 err = dev_change_name(dev, ifname);
2710                 if (err < 0)
2711                         goto errout;
2712                 status |= DO_SETLINK_MODIFIED;
2713         }
2714
2715         if (tb[IFLA_IFALIAS]) {
2716                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2717                                     nla_len(tb[IFLA_IFALIAS]));
2718                 if (err < 0)
2719                         goto errout;
2720                 status |= DO_SETLINK_NOTIFY;
2721         }
2722
2723         if (tb[IFLA_BROADCAST]) {
2724                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2725                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2726         }
2727
2728         if (ifm->ifi_flags || ifm->ifi_change) {
2729                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2730                                        extack);
2731                 if (err < 0)
2732                         goto errout;
2733         }
2734
2735         if (tb[IFLA_MASTER]) {
2736                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2737                 if (err)
2738                         goto errout;
2739                 status |= DO_SETLINK_MODIFIED;
2740         }
2741
2742         if (tb[IFLA_CARRIER]) {
2743                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2744                 if (err)
2745                         goto errout;
2746                 status |= DO_SETLINK_MODIFIED;
2747         }
2748
2749         if (tb[IFLA_TXQLEN]) {
2750                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2751
2752                 err = dev_change_tx_queue_len(dev, value);
2753                 if (err)
2754                         goto errout;
2755                 status |= DO_SETLINK_MODIFIED;
2756         }
2757
2758         if (tb[IFLA_GSO_MAX_SIZE]) {
2759                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2760
2761                 if (max_size > GSO_MAX_SIZE) {
2762                         err = -EINVAL;
2763                         goto errout;
2764                 }
2765
2766                 if (dev->gso_max_size ^ max_size) {
2767                         netif_set_gso_max_size(dev, max_size);
2768                         status |= DO_SETLINK_MODIFIED;
2769                 }
2770         }
2771
2772         if (tb[IFLA_GSO_MAX_SEGS]) {
2773                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2774
2775                 if (max_segs > GSO_MAX_SEGS) {
2776                         err = -EINVAL;
2777                         goto errout;
2778                 }
2779
2780                 if (dev->gso_max_segs ^ max_segs) {
2781                         dev->gso_max_segs = max_segs;
2782                         status |= DO_SETLINK_MODIFIED;
2783                 }
2784         }
2785
2786         if (tb[IFLA_OPERSTATE])
2787                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2788
2789         if (tb[IFLA_LINKMODE]) {
2790                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2791
2792                 write_lock_bh(&dev_base_lock);
2793                 if (dev->link_mode ^ value)
2794                         status |= DO_SETLINK_NOTIFY;
2795                 dev->link_mode = value;
2796                 write_unlock_bh(&dev_base_lock);
2797         }
2798
2799         if (tb[IFLA_VFINFO_LIST]) {
2800                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2801                 struct nlattr *attr;
2802                 int rem;
2803
2804                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2805                         if (nla_type(attr) != IFLA_VF_INFO ||
2806                             nla_len(attr) < NLA_HDRLEN) {
2807                                 err = -EINVAL;
2808                                 goto errout;
2809                         }
2810                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2811                                                           attr,
2812                                                           ifla_vf_policy,
2813                                                           NULL);
2814                         if (err < 0)
2815                                 goto errout;
2816                         err = do_setvfinfo(dev, vfinfo);
2817                         if (err < 0)
2818                                 goto errout;
2819                         status |= DO_SETLINK_NOTIFY;
2820                 }
2821         }
2822         err = 0;
2823
2824         if (tb[IFLA_VF_PORTS]) {
2825                 struct nlattr *port[IFLA_PORT_MAX+1];
2826                 struct nlattr *attr;
2827                 int vf;
2828                 int rem;
2829
2830                 err = -EOPNOTSUPP;
2831                 if (!ops->ndo_set_vf_port)
2832                         goto errout;
2833
2834                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2835                         if (nla_type(attr) != IFLA_VF_PORT ||
2836                             nla_len(attr) < NLA_HDRLEN) {
2837                                 err = -EINVAL;
2838                                 goto errout;
2839                         }
2840                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2841                                                           attr,
2842                                                           ifla_port_policy,
2843                                                           NULL);
2844                         if (err < 0)
2845                                 goto errout;
2846                         if (!port[IFLA_PORT_VF]) {
2847                                 err = -EOPNOTSUPP;
2848                                 goto errout;
2849                         }
2850                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2851                         err = ops->ndo_set_vf_port(dev, vf, port);
2852                         if (err < 0)
2853                                 goto errout;
2854                         status |= DO_SETLINK_NOTIFY;
2855                 }
2856         }
2857         err = 0;
2858
2859         if (tb[IFLA_PORT_SELF]) {
2860                 struct nlattr *port[IFLA_PORT_MAX+1];
2861
2862                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2863                                                   tb[IFLA_PORT_SELF],
2864                                                   ifla_port_policy, NULL);
2865                 if (err < 0)
2866                         goto errout;
2867
2868                 err = -EOPNOTSUPP;
2869                 if (ops->ndo_set_vf_port)
2870                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2871                 if (err < 0)
2872                         goto errout;
2873                 status |= DO_SETLINK_NOTIFY;
2874         }
2875
2876         if (tb[IFLA_AF_SPEC]) {
2877                 struct nlattr *af;
2878                 int rem;
2879
2880                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2881                         const struct rtnl_af_ops *af_ops;
2882
2883                         rcu_read_lock();
2884
2885                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2886
2887                         err = af_ops->set_link_af(dev, af);
2888                         if (err < 0) {
2889                                 rcu_read_unlock();
2890                                 goto errout;
2891                         }
2892
2893                         rcu_read_unlock();
2894                         status |= DO_SETLINK_NOTIFY;
2895                 }
2896         }
2897         err = 0;
2898
2899         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2900                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2901                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2902                 if (err)
2903                         goto errout;
2904                 status |= DO_SETLINK_NOTIFY;
2905         }
2906
2907         if (tb[IFLA_XDP]) {
2908                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2909                 u32 xdp_flags = 0;
2910
2911                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2912                                                   tb[IFLA_XDP],
2913                                                   ifla_xdp_policy, NULL);
2914                 if (err < 0)
2915                         goto errout;
2916
2917                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2918                         err = -EINVAL;
2919                         goto errout;
2920                 }
2921
2922                 if (xdp[IFLA_XDP_FLAGS]) {
2923                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2924                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2925                                 err = -EINVAL;
2926                                 goto errout;
2927                         }
2928                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2929                                 err = -EINVAL;
2930                                 goto errout;
2931                         }
2932                 }
2933
2934                 if (xdp[IFLA_XDP_FD]) {
2935                         int expected_fd = -1;
2936
2937                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2938                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2939                                         err = -EINVAL;
2940                                         goto errout;
2941                                 }
2942                                 expected_fd =
2943                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2944                         }
2945
2946                         err = dev_change_xdp_fd(dev, extack,
2947                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2948                                                 expected_fd,
2949                                                 xdp_flags);
2950                         if (err)
2951                                 goto errout;
2952                         status |= DO_SETLINK_NOTIFY;
2953                 }
2954         }
2955
2956 errout:
2957         if (status & DO_SETLINK_MODIFIED) {
2958                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2959                         netdev_state_change(dev);
2960
2961                 if (err < 0)
2962                         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",
2963                                              dev->name);
2964         }
2965
2966         return err;
2967 }
2968
2969 static struct net_device *rtnl_dev_get(struct net *net,
2970                                        struct nlattr *ifname_attr,
2971                                        struct nlattr *altifname_attr,
2972                                        char *ifname)
2973 {
2974         char buffer[ALTIFNAMSIZ];
2975
2976         if (!ifname) {
2977                 ifname = buffer;
2978                 if (ifname_attr)
2979                         nla_strlcpy(ifname, ifname_attr, IFNAMSIZ);
2980                 else if (altifname_attr)
2981                         nla_strlcpy(ifname, altifname_attr, ALTIFNAMSIZ);
2982                 else
2983                         return NULL;
2984         }
2985
2986         return __dev_get_by_name(net, ifname);
2987 }
2988
2989 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2990                         struct netlink_ext_ack *extack)
2991 {
2992         struct net *net = sock_net(skb->sk);
2993         struct ifinfomsg *ifm;
2994         struct net_device *dev;
2995         int err;
2996         struct nlattr *tb[IFLA_MAX+1];
2997         char ifname[IFNAMSIZ];
2998
2999         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3000                                      ifla_policy, extack);
3001         if (err < 0)
3002                 goto errout;
3003
3004         err = rtnl_ensure_unique_netns(tb, extack, false);
3005         if (err < 0)
3006                 goto errout;
3007
3008         if (tb[IFLA_IFNAME])
3009                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3010         else
3011                 ifname[0] = '\0';
3012
3013         err = -EINVAL;
3014         ifm = nlmsg_data(nlh);
3015         if (ifm->ifi_index > 0)
3016                 dev = __dev_get_by_index(net, ifm->ifi_index);
3017         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3018                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3019         else
3020                 goto errout;
3021
3022         if (dev == NULL) {
3023                 err = -ENODEV;
3024                 goto errout;
3025         }
3026
3027         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3028 errout:
3029         return err;
3030 }
3031
3032 static int rtnl_group_dellink(const struct net *net, int group)
3033 {
3034         struct net_device *dev, *aux;
3035         LIST_HEAD(list_kill);
3036         bool found = false;
3037
3038         if (!group)
3039                 return -EPERM;
3040
3041         for_each_netdev(net, dev) {
3042                 if (dev->group == group) {
3043                         const struct rtnl_link_ops *ops;
3044
3045                         found = true;
3046                         ops = dev->rtnl_link_ops;
3047                         if (!ops || !ops->dellink)
3048                                 return -EOPNOTSUPP;
3049                 }
3050         }
3051
3052         if (!found)
3053                 return -ENODEV;
3054
3055         for_each_netdev_safe(net, dev, aux) {
3056                 if (dev->group == group) {
3057                         const struct rtnl_link_ops *ops;
3058
3059                         ops = dev->rtnl_link_ops;
3060                         ops->dellink(dev, &list_kill);
3061                 }
3062         }
3063         unregister_netdevice_many(&list_kill);
3064
3065         return 0;
3066 }
3067
3068 int rtnl_delete_link(struct net_device *dev)
3069 {
3070         const struct rtnl_link_ops *ops;
3071         LIST_HEAD(list_kill);
3072
3073         ops = dev->rtnl_link_ops;
3074         if (!ops || !ops->dellink)
3075                 return -EOPNOTSUPP;
3076
3077         ops->dellink(dev, &list_kill);
3078         unregister_netdevice_many(&list_kill);
3079
3080         return 0;
3081 }
3082 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3083
3084 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3085                         struct netlink_ext_ack *extack)
3086 {
3087         struct net *net = sock_net(skb->sk);
3088         struct net *tgt_net = net;
3089         struct net_device *dev = NULL;
3090         struct ifinfomsg *ifm;
3091         struct nlattr *tb[IFLA_MAX+1];
3092         int err;
3093         int netnsid = -1;
3094
3095         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3096                                      ifla_policy, extack);
3097         if (err < 0)
3098                 return err;
3099
3100         err = rtnl_ensure_unique_netns(tb, extack, true);
3101         if (err < 0)
3102                 return err;
3103
3104         if (tb[IFLA_TARGET_NETNSID]) {
3105                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3106                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3107                 if (IS_ERR(tgt_net))
3108                         return PTR_ERR(tgt_net);
3109         }
3110
3111         err = -EINVAL;
3112         ifm = nlmsg_data(nlh);
3113         if (ifm->ifi_index > 0)
3114                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3115         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3116                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3117                                    tb[IFLA_ALT_IFNAME], NULL);
3118         else if (tb[IFLA_GROUP])
3119                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3120         else
3121                 goto out;
3122
3123         if (!dev) {
3124                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3125                         err = -ENODEV;
3126
3127                 goto out;
3128         }
3129
3130         err = rtnl_delete_link(dev);
3131
3132 out:
3133         if (netnsid >= 0)
3134                 put_net(tgt_net);
3135
3136         return err;
3137 }
3138
3139 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3140 {
3141         unsigned int old_flags;
3142         int err;
3143
3144         old_flags = dev->flags;
3145         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3146                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3147                                          NULL);
3148                 if (err < 0)
3149                         return err;
3150         }
3151
3152         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3153                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3154         } else {
3155                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3156                 __dev_notify_flags(dev, old_flags, ~0U);
3157         }
3158         return 0;
3159 }
3160 EXPORT_SYMBOL(rtnl_configure_link);
3161
3162 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3163                                     unsigned char name_assign_type,
3164                                     const struct rtnl_link_ops *ops,
3165                                     struct nlattr *tb[],
3166                                     struct netlink_ext_ack *extack)
3167 {
3168         struct net_device *dev;
3169         unsigned int num_tx_queues = 1;
3170         unsigned int num_rx_queues = 1;
3171
3172         if (tb[IFLA_NUM_TX_QUEUES])
3173                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3174         else if (ops->get_num_tx_queues)
3175                 num_tx_queues = ops->get_num_tx_queues();
3176
3177         if (tb[IFLA_NUM_RX_QUEUES])
3178                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3179         else if (ops->get_num_rx_queues)
3180                 num_rx_queues = ops->get_num_rx_queues();
3181
3182         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3183                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3184                 return ERR_PTR(-EINVAL);
3185         }
3186
3187         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3188                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3189                 return ERR_PTR(-EINVAL);
3190         }
3191
3192         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3193                                ops->setup, num_tx_queues, num_rx_queues);
3194         if (!dev)
3195                 return ERR_PTR(-ENOMEM);
3196
3197         dev_net_set(dev, net);
3198         dev->rtnl_link_ops = ops;
3199         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3200
3201         if (tb[IFLA_MTU]) {
3202                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3203                 int err;
3204
3205                 err = dev_validate_mtu(dev, mtu, extack);
3206                 if (err) {
3207                         free_netdev(dev);
3208                         return ERR_PTR(err);
3209                 }
3210                 dev->mtu = mtu;
3211         }
3212         if (tb[IFLA_ADDRESS]) {
3213                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3214                                 nla_len(tb[IFLA_ADDRESS]));
3215                 dev->addr_assign_type = NET_ADDR_SET;
3216         }
3217         if (tb[IFLA_BROADCAST])
3218                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3219                                 nla_len(tb[IFLA_BROADCAST]));
3220         if (tb[IFLA_TXQLEN])
3221                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3222         if (tb[IFLA_OPERSTATE])
3223                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3224         if (tb[IFLA_LINKMODE])
3225                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3226         if (tb[IFLA_GROUP])
3227                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3228         if (tb[IFLA_GSO_MAX_SIZE])
3229                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3230         if (tb[IFLA_GSO_MAX_SEGS])
3231                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3232
3233         return dev;
3234 }
3235 EXPORT_SYMBOL(rtnl_create_link);
3236
3237 static int rtnl_group_changelink(const struct sk_buff *skb,
3238                 struct net *net, int group,
3239                 struct ifinfomsg *ifm,
3240                 struct netlink_ext_ack *extack,
3241                 struct nlattr **tb)
3242 {
3243         struct net_device *dev, *aux;
3244         int err;
3245
3246         for_each_netdev_safe(net, dev, aux) {
3247                 if (dev->group == group) {
3248                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3249                         if (err < 0)
3250                                 return err;
3251                 }
3252         }
3253
3254         return 0;
3255 }
3256
3257 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3258                           struct nlattr **attr, struct netlink_ext_ack *extack)
3259 {
3260         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3261         unsigned char name_assign_type = NET_NAME_USER;
3262         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3263         const struct rtnl_link_ops *m_ops;
3264         struct net_device *master_dev;
3265         struct net *net = sock_net(skb->sk);
3266         const struct rtnl_link_ops *ops;
3267         struct nlattr *tb[IFLA_MAX + 1];
3268         struct net *dest_net, *link_net;
3269         struct nlattr **slave_data;
3270         char kind[MODULE_NAME_LEN];
3271         struct net_device *dev;
3272         struct ifinfomsg *ifm;
3273         char ifname[IFNAMSIZ];
3274         struct nlattr **data;
3275         bool link_specified;
3276         int err;
3277
3278 #ifdef CONFIG_MODULES
3279 replay:
3280 #endif
3281         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3282                                      ifla_policy, extack);
3283         if (err < 0)
3284                 return err;
3285
3286         err = rtnl_ensure_unique_netns(tb, extack, false);
3287         if (err < 0)
3288                 return err;
3289
3290         if (tb[IFLA_IFNAME])
3291                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3292         else
3293                 ifname[0] = '\0';
3294
3295         ifm = nlmsg_data(nlh);
3296         if (ifm->ifi_index > 0) {
3297                 link_specified = true;
3298                 dev = __dev_get_by_index(net, ifm->ifi_index);
3299         } else if (ifm->ifi_index < 0) {
3300                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3301                 return -EINVAL;
3302         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3303                 link_specified = true;
3304                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3305         } else {
3306                 link_specified = false;
3307                 dev = NULL;
3308         }
3309
3310         master_dev = NULL;
3311         m_ops = NULL;
3312         if (dev) {
3313                 master_dev = netdev_master_upper_dev_get(dev);
3314                 if (master_dev)
3315                         m_ops = master_dev->rtnl_link_ops;
3316         }
3317
3318         err = validate_linkmsg(dev, tb);
3319         if (err < 0)
3320                 return err;
3321
3322         if (tb[IFLA_LINKINFO]) {
3323                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3324                                                   tb[IFLA_LINKINFO],
3325                                                   ifla_info_policy, NULL);
3326                 if (err < 0)
3327                         return err;
3328         } else
3329                 memset(linkinfo, 0, sizeof(linkinfo));
3330
3331         if (linkinfo[IFLA_INFO_KIND]) {
3332                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3333                 ops = rtnl_link_ops_get(kind);
3334         } else {
3335                 kind[0] = '\0';
3336                 ops = NULL;
3337         }
3338
3339         data = NULL;
3340         if (ops) {
3341                 if (ops->maxtype > RTNL_MAX_TYPE)
3342                         return -EINVAL;
3343
3344                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3345                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3346                                                           linkinfo[IFLA_INFO_DATA],
3347                                                           ops->policy, extack);
3348                         if (err < 0)
3349                                 return err;
3350                         data = attr;
3351                 }
3352                 if (ops->validate) {
3353                         err = ops->validate(tb, data, extack);
3354                         if (err < 0)
3355                                 return err;
3356                 }
3357         }
3358
3359         slave_data = NULL;
3360         if (m_ops) {
3361                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3362                         return -EINVAL;
3363
3364                 if (m_ops->slave_maxtype &&
3365                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3366                         err = nla_parse_nested_deprecated(slave_attr,
3367                                                           m_ops->slave_maxtype,
3368                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3369                                                           m_ops->slave_policy,
3370                                                           extack);
3371                         if (err < 0)
3372                                 return err;
3373                         slave_data = slave_attr;
3374                 }
3375         }
3376
3377         if (dev) {
3378                 int status = 0;
3379
3380                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3381                         return -EEXIST;
3382                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3383                         return -EOPNOTSUPP;
3384
3385                 if (linkinfo[IFLA_INFO_DATA]) {
3386                         if (!ops || ops != dev->rtnl_link_ops ||
3387                             !ops->changelink)
3388                                 return -EOPNOTSUPP;
3389
3390                         err = ops->changelink(dev, tb, data, extack);
3391                         if (err < 0)
3392                                 return err;
3393                         status |= DO_SETLINK_NOTIFY;
3394                 }
3395
3396                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3397                         if (!m_ops || !m_ops->slave_changelink)
3398                                 return -EOPNOTSUPP;
3399
3400                         err = m_ops->slave_changelink(master_dev, dev, tb,
3401                                                       slave_data, extack);
3402                         if (err < 0)
3403                                 return err;
3404                         status |= DO_SETLINK_NOTIFY;
3405                 }
3406
3407                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3408         }
3409
3410         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3411                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3412                  * or it's for a group
3413                 */
3414                 if (link_specified)
3415                         return -ENODEV;
3416                 if (tb[IFLA_GROUP])
3417                         return rtnl_group_changelink(skb, net,
3418                                                 nla_get_u32(tb[IFLA_GROUP]),
3419                                                 ifm, extack, tb);
3420                 return -ENODEV;
3421         }
3422
3423         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3424                 return -EOPNOTSUPP;
3425
3426         if (!ops) {
3427 #ifdef CONFIG_MODULES
3428                 if (kind[0]) {
3429                         __rtnl_unlock();
3430                         request_module("rtnl-link-%s", kind);
3431                         rtnl_lock();
3432                         ops = rtnl_link_ops_get(kind);
3433                         if (ops)
3434                                 goto replay;
3435                 }
3436 #endif
3437                 NL_SET_ERR_MSG(extack, "Unknown device type");
3438                 return -EOPNOTSUPP;
3439         }
3440
3441         if (!ops->setup)
3442                 return -EOPNOTSUPP;
3443
3444         if (!ifname[0]) {
3445                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3446                 name_assign_type = NET_NAME_ENUM;
3447         }
3448
3449         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3450         if (IS_ERR(dest_net))
3451                 return PTR_ERR(dest_net);
3452
3453         if (tb[IFLA_LINK_NETNSID]) {
3454                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3455
3456                 link_net = get_net_ns_by_id(dest_net, id);
3457                 if (!link_net) {
3458                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3459                         err =  -EINVAL;
3460                         goto out;
3461                 }
3462                 err = -EPERM;
3463                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3464                         goto out;
3465         } else {
3466                 link_net = NULL;
3467         }
3468
3469         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3470                                name_assign_type, ops, tb, extack);
3471         if (IS_ERR(dev)) {
3472                 err = PTR_ERR(dev);
3473                 goto out;
3474         }
3475
3476         dev->ifindex = ifm->ifi_index;
3477
3478         if (ops->newlink)
3479                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3480         else
3481                 err = register_netdevice(dev);
3482         if (err < 0) {
3483                 free_netdev(dev);
3484                 goto out;
3485         }
3486
3487         err = rtnl_configure_link(dev, ifm);
3488         if (err < 0)
3489                 goto out_unregister;
3490         if (link_net) {
3491                 err = dev_change_net_namespace(dev, dest_net, ifname);
3492                 if (err < 0)
3493                         goto out_unregister;
3494         }
3495         if (tb[IFLA_MASTER]) {
3496                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3497                 if (err)
3498                         goto out_unregister;
3499         }
3500 out:
3501         if (link_net)
3502                 put_net(link_net);
3503         put_net(dest_net);
3504         return err;
3505 out_unregister:
3506         if (ops->newlink) {
3507                 LIST_HEAD(list_kill);
3508
3509                 ops->dellink(dev, &list_kill);
3510                 unregister_netdevice_many(&list_kill);
3511         } else {
3512                 unregister_netdevice(dev);
3513         }
3514         goto out;
3515 }
3516
3517 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3518                         struct netlink_ext_ack *extack)
3519 {
3520         struct nlattr **attr;
3521         int ret;
3522
3523         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3524         if (!attr)
3525                 return -ENOMEM;
3526
3527         ret = __rtnl_newlink(skb, nlh, attr, extack);
3528         kfree(attr);
3529         return ret;
3530 }
3531
3532 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3533                                   const struct nlmsghdr *nlh,
3534                                   struct nlattr **tb,
3535                                   struct netlink_ext_ack *extack)
3536 {
3537         struct ifinfomsg *ifm;
3538         int i, err;
3539
3540         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3541                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3542                 return -EINVAL;
3543         }
3544
3545         if (!netlink_strict_get_check(skb))
3546                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3547                                               ifla_policy, extack);
3548
3549         ifm = nlmsg_data(nlh);
3550         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3551             ifm->ifi_change) {
3552                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3553                 return -EINVAL;
3554         }
3555
3556         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3557                                             ifla_policy, extack);
3558         if (err)
3559                 return err;
3560
3561         for (i = 0; i <= IFLA_MAX; i++) {
3562                 if (!tb[i])
3563                         continue;
3564
3565                 switch (i) {
3566                 case IFLA_IFNAME:
3567                 case IFLA_ALT_IFNAME:
3568                 case IFLA_EXT_MASK:
3569                 case IFLA_TARGET_NETNSID:
3570                         break;
3571                 default:
3572                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3573                         return -EINVAL;
3574                 }
3575         }
3576
3577         return 0;
3578 }
3579
3580 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3581                         struct netlink_ext_ack *extack)
3582 {
3583         struct net *net = sock_net(skb->sk);
3584         struct net *tgt_net = net;
3585         struct ifinfomsg *ifm;
3586         struct nlattr *tb[IFLA_MAX+1];
3587         struct net_device *dev = NULL;
3588         struct sk_buff *nskb;
3589         int netnsid = -1;
3590         int err;
3591         u32 ext_filter_mask = 0;
3592
3593         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3594         if (err < 0)
3595                 return err;
3596
3597         err = rtnl_ensure_unique_netns(tb, extack, true);
3598         if (err < 0)
3599                 return err;
3600
3601         if (tb[IFLA_TARGET_NETNSID]) {
3602                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3603                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3604                 if (IS_ERR(tgt_net))
3605                         return PTR_ERR(tgt_net);
3606         }
3607
3608         if (tb[IFLA_EXT_MASK])
3609                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3610
3611         err = -EINVAL;
3612         ifm = nlmsg_data(nlh);
3613         if (ifm->ifi_index > 0)
3614                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3615         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3616                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3617                                    tb[IFLA_ALT_IFNAME], NULL);
3618         else
3619                 goto out;
3620
3621         err = -ENODEV;
3622         if (dev == NULL)
3623                 goto out;
3624
3625         err = -ENOBUFS;
3626         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3627         if (nskb == NULL)
3628                 goto out;
3629
3630         err = rtnl_fill_ifinfo(nskb, dev, net,
3631                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3632                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3633                                0, NULL, 0, netnsid, GFP_KERNEL);
3634         if (err < 0) {
3635                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3636                 WARN_ON(err == -EMSGSIZE);
3637                 kfree_skb(nskb);
3638         } else
3639                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3640 out:
3641         if (netnsid >= 0)
3642                 put_net(tgt_net);
3643
3644         return err;
3645 }
3646
3647 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3648                            bool *changed, struct netlink_ext_ack *extack)
3649 {
3650         char *alt_ifname;
3651         size_t size;
3652         int err;
3653
3654         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3655         if (err)
3656                 return err;
3657
3658         if (cmd == RTM_NEWLINKPROP) {
3659                 size = rtnl_prop_list_size(dev);
3660                 size += nla_total_size(ALTIFNAMSIZ);
3661                 if (size >= U16_MAX) {
3662                         NL_SET_ERR_MSG(extack,
3663                                        "effective property list too long");
3664                         return -EINVAL;
3665                 }
3666         }
3667
3668         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3669         if (!alt_ifname)
3670                 return -ENOMEM;
3671
3672         if (cmd == RTM_NEWLINKPROP) {
3673                 err = netdev_name_node_alt_create(dev, alt_ifname);
3674                 if (!err)
3675                         alt_ifname = NULL;
3676         } else if (cmd == RTM_DELLINKPROP) {
3677                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3678         } else {
3679                 WARN_ON_ONCE(1);
3680                 err = -EINVAL;
3681         }
3682
3683         kfree(alt_ifname);
3684         if (!err)
3685                 *changed = true;
3686         return err;
3687 }
3688
3689 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3690                          struct netlink_ext_ack *extack)
3691 {
3692         struct net *net = sock_net(skb->sk);
3693         struct nlattr *tb[IFLA_MAX + 1];
3694         struct net_device *dev;
3695         struct ifinfomsg *ifm;
3696         bool changed = false;
3697         struct nlattr *attr;
3698         int err, rem;
3699
3700         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3701         if (err)
3702                 return err;
3703
3704         err = rtnl_ensure_unique_netns(tb, extack, true);
3705         if (err)
3706                 return err;
3707
3708         ifm = nlmsg_data(nlh);
3709         if (ifm->ifi_index > 0)
3710                 dev = __dev_get_by_index(net, ifm->ifi_index);
3711         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3712                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3713                                    tb[IFLA_ALT_IFNAME], NULL);
3714         else
3715                 return -EINVAL;
3716
3717         if (!dev)
3718                 return -ENODEV;
3719
3720         if (!tb[IFLA_PROP_LIST])
3721                 return 0;
3722
3723         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3724                 switch (nla_type(attr)) {
3725                 case IFLA_ALT_IFNAME:
3726                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3727                         if (err)
3728                                 return err;
3729                         break;
3730                 }
3731         }
3732
3733         if (changed)
3734                 netdev_state_change(dev);
3735         return 0;
3736 }
3737
3738 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3739                             struct netlink_ext_ack *extack)
3740 {
3741         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3742 }
3743
3744 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3745                             struct netlink_ext_ack *extack)
3746 {
3747         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3748 }
3749
3750 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3751 {
3752         struct net *net = sock_net(skb->sk);
3753         size_t min_ifinfo_dump_size = 0;
3754         struct nlattr *tb[IFLA_MAX+1];
3755         u32 ext_filter_mask = 0;
3756         struct net_device *dev;
3757         int hdrlen;
3758
3759         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3760         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3761                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3762
3763         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3764                 if (tb[IFLA_EXT_MASK])
3765                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3766         }
3767
3768         if (!ext_filter_mask)
3769                 return NLMSG_GOODSIZE;
3770         /*
3771          * traverse the list of net devices and compute the minimum
3772          * buffer size based upon the filter mask.
3773          */
3774         rcu_read_lock();
3775         for_each_netdev_rcu(net, dev) {
3776                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3777                                            if_nlmsg_size(dev, ext_filter_mask));
3778         }
3779         rcu_read_unlock();
3780
3781         return nlmsg_total_size(min_ifinfo_dump_size);
3782 }
3783
3784 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3785 {
3786         int idx;
3787         int s_idx = cb->family;
3788         int type = cb->nlh->nlmsg_type - RTM_BASE;
3789         int ret = 0;
3790
3791         if (s_idx == 0)
3792                 s_idx = 1;
3793
3794         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3795                 struct rtnl_link **tab;
3796                 struct rtnl_link *link;
3797                 rtnl_dumpit_func dumpit;
3798
3799                 if (idx < s_idx || idx == PF_PACKET)
3800                         continue;
3801
3802                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3803                         continue;
3804
3805                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3806                 if (!tab)
3807                         continue;
3808
3809                 link = tab[type];
3810                 if (!link)
3811                         continue;
3812
3813                 dumpit = link->dumpit;
3814                 if (!dumpit)
3815                         continue;
3816
3817                 if (idx > s_idx) {
3818                         memset(&cb->args[0], 0, sizeof(cb->args));
3819                         cb->prev_seq = 0;
3820                         cb->seq = 0;
3821                 }
3822                 ret = dumpit(skb, cb);
3823                 if (ret)
3824                         break;
3825         }
3826         cb->family = idx;
3827
3828         return skb->len ? : ret;
3829 }
3830
3831 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3832                                        unsigned int change,
3833                                        u32 event, gfp_t flags, int *new_nsid,
3834                                        int new_ifindex)
3835 {
3836         struct net *net = dev_net(dev);
3837         struct sk_buff *skb;
3838         int err = -ENOBUFS;
3839         size_t if_info_size;
3840
3841         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3842         if (skb == NULL)
3843                 goto errout;
3844
3845         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3846                                type, 0, 0, change, 0, 0, event,
3847                                new_nsid, new_ifindex, -1, flags);
3848         if (err < 0) {
3849                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3850                 WARN_ON(err == -EMSGSIZE);
3851                 kfree_skb(skb);
3852                 goto errout;
3853         }
3854         return skb;
3855 errout:
3856         if (err < 0)
3857                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3858         return NULL;
3859 }
3860
3861 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3862 {
3863         struct net *net = dev_net(dev);
3864
3865         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3866 }
3867
3868 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3869                                unsigned int change, u32 event,
3870                                gfp_t flags, int *new_nsid, int new_ifindex)
3871 {
3872         struct sk_buff *skb;
3873
3874         if (dev->reg_state != NETREG_REGISTERED)
3875                 return;
3876
3877         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3878                                      new_ifindex);
3879         if (skb)
3880                 rtmsg_ifinfo_send(skb, dev, flags);
3881 }
3882
3883 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3884                   gfp_t flags)
3885 {
3886         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3887                            NULL, 0);
3888 }
3889
3890 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3891                          gfp_t flags, int *new_nsid, int new_ifindex)
3892 {
3893         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3894                            new_nsid, new_ifindex);
3895 }
3896
3897 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3898                                    struct net_device *dev,
3899                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3900                                    int type, unsigned int flags,
3901                                    int nlflags, u16 ndm_state)
3902 {
3903         struct nlmsghdr *nlh;
3904         struct ndmsg *ndm;
3905
3906         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3907         if (!nlh)
3908                 return -EMSGSIZE;
3909
3910         ndm = nlmsg_data(nlh);
3911         ndm->ndm_family  = AF_BRIDGE;
3912         ndm->ndm_pad1    = 0;
3913         ndm->ndm_pad2    = 0;
3914         ndm->ndm_flags   = flags;
3915         ndm->ndm_type    = 0;
3916         ndm->ndm_ifindex = dev->ifindex;
3917         ndm->ndm_state   = ndm_state;
3918
3919         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
3920                 goto nla_put_failure;
3921         if (vid)
3922                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3923                         goto nla_put_failure;
3924
3925         nlmsg_end(skb, nlh);
3926         return 0;
3927
3928 nla_put_failure:
3929         nlmsg_cancel(skb, nlh);
3930         return -EMSGSIZE;
3931 }
3932
3933 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
3934 {
3935         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3936                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
3937                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3938                0;
3939 }
3940
3941 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3942                             u16 ndm_state)
3943 {
3944         struct net *net = dev_net(dev);
3945         struct sk_buff *skb;
3946         int err = -ENOBUFS;
3947
3948         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
3949         if (!skb)
3950                 goto errout;
3951
3952         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3953                                       0, 0, type, NTF_SELF, 0, ndm_state);
3954         if (err < 0) {
3955                 kfree_skb(skb);
3956                 goto errout;
3957         }
3958
3959         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3960         return;
3961 errout:
3962         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3963 }
3964
3965 /*
3966  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3967  */
3968 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3969                      struct nlattr *tb[],
3970                      struct net_device *dev,
3971                      const unsigned char *addr, u16 vid,
3972                      u16 flags)
3973 {
3974         int err = -EINVAL;
3975
3976         /* If aging addresses are supported device will need to
3977          * implement its own handler for this.
3978          */
3979         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3980                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3981                 return err;
3982         }
3983
3984         if (vid) {
3985                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3986                 return err;
3987         }
3988
3989         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3990                 err = dev_uc_add_excl(dev, addr);
3991         else if (is_multicast_ether_addr(addr))
3992                 err = dev_mc_add_excl(dev, addr);
3993
3994         /* Only return duplicate errors if NLM_F_EXCL is set */
3995         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3996                 err = 0;
3997
3998         return err;
3999 }
4000 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4001
4002 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4003                          struct netlink_ext_ack *extack)
4004 {
4005         u16 vid = 0;
4006
4007         if (vlan_attr) {
4008                 if (nla_len(vlan_attr) != sizeof(u16)) {
4009                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4010                         return -EINVAL;
4011                 }
4012
4013                 vid = nla_get_u16(vlan_attr);
4014
4015                 if (!vid || vid >= VLAN_VID_MASK) {
4016                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4017                         return -EINVAL;
4018                 }
4019         }
4020         *p_vid = vid;
4021         return 0;
4022 }
4023
4024 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4025                         struct netlink_ext_ack *extack)
4026 {
4027         struct net *net = sock_net(skb->sk);
4028         struct ndmsg *ndm;
4029         struct nlattr *tb[NDA_MAX+1];
4030         struct net_device *dev;
4031         u8 *addr;
4032         u16 vid;
4033         int err;
4034
4035         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4036                                      extack);
4037         if (err < 0)
4038                 return err;
4039
4040         ndm = nlmsg_data(nlh);
4041         if (ndm->ndm_ifindex == 0) {
4042                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4043                 return -EINVAL;
4044         }
4045
4046         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4047         if (dev == NULL) {
4048                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4049                 return -ENODEV;
4050         }
4051
4052         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4053                 NL_SET_ERR_MSG(extack, "invalid address");
4054                 return -EINVAL;
4055         }
4056
4057         if (dev->type != ARPHRD_ETHER) {
4058                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4059                 return -EINVAL;
4060         }
4061
4062         addr = nla_data(tb[NDA_LLADDR]);
4063
4064         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4065         if (err)
4066                 return err;
4067
4068         err = -EOPNOTSUPP;
4069
4070         /* Support fdb on master device the net/bridge default case */
4071         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4072             netif_is_bridge_port(dev)) {
4073                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4074                 const struct net_device_ops *ops = br_dev->netdev_ops;
4075
4076                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4077                                        nlh->nlmsg_flags, extack);
4078                 if (err)
4079                         goto out;
4080                 else
4081                         ndm->ndm_flags &= ~NTF_MASTER;
4082         }
4083
4084         /* Embedded bridge, macvlan, and any other device support */
4085         if ((ndm->ndm_flags & NTF_SELF)) {
4086                 if (dev->netdev_ops->ndo_fdb_add)
4087                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4088                                                            vid,
4089                                                            nlh->nlmsg_flags,
4090                                                            extack);
4091                 else
4092                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4093                                                nlh->nlmsg_flags);
4094
4095                 if (!err) {
4096                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4097                                         ndm->ndm_state);
4098                         ndm->ndm_flags &= ~NTF_SELF;
4099                 }
4100         }
4101 out:
4102         return err;
4103 }
4104
4105 /*
4106  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4107  */
4108 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4109                      struct nlattr *tb[],
4110                      struct net_device *dev,
4111                      const unsigned char *addr, u16 vid)
4112 {
4113         int err = -EINVAL;
4114
4115         /* If aging addresses are supported device will need to
4116          * implement its own handler for this.
4117          */
4118         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4119                 pr_info("%s: FDB only supports static addresses\n", dev->name);
4120                 return err;
4121         }
4122
4123         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4124                 err = dev_uc_del(dev, addr);
4125         else if (is_multicast_ether_addr(addr))
4126                 err = dev_mc_del(dev, addr);
4127
4128         return err;
4129 }
4130 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4131
4132 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4133                         struct netlink_ext_ack *extack)
4134 {
4135         struct net *net = sock_net(skb->sk);
4136         struct ndmsg *ndm;
4137         struct nlattr *tb[NDA_MAX+1];
4138         struct net_device *dev;
4139         __u8 *addr;
4140         int err;
4141         u16 vid;
4142
4143         if (!netlink_capable(skb, CAP_NET_ADMIN))
4144                 return -EPERM;
4145
4146         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4147                                      extack);
4148         if (err < 0)
4149                 return err;
4150
4151         ndm = nlmsg_data(nlh);
4152         if (ndm->ndm_ifindex == 0) {
4153                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4154                 return -EINVAL;
4155         }
4156
4157         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4158         if (dev == NULL) {
4159                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4160                 return -ENODEV;
4161         }
4162
4163         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4164                 NL_SET_ERR_MSG(extack, "invalid address");
4165                 return -EINVAL;
4166         }
4167
4168         if (dev->type != ARPHRD_ETHER) {
4169                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4170                 return -EINVAL;
4171         }
4172
4173         addr = nla_data(tb[NDA_LLADDR]);
4174
4175         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4176         if (err)
4177                 return err;
4178
4179         err = -EOPNOTSUPP;
4180
4181         /* Support fdb on master device the net/bridge default case */
4182         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4183             netif_is_bridge_port(dev)) {
4184                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4185                 const struct net_device_ops *ops = br_dev->netdev_ops;
4186
4187                 if (ops->ndo_fdb_del)
4188                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4189
4190                 if (err)
4191                         goto out;
4192                 else
4193                         ndm->ndm_flags &= ~NTF_MASTER;
4194         }
4195
4196         /* Embedded bridge, macvlan, and any other device support */
4197         if (ndm->ndm_flags & NTF_SELF) {
4198                 if (dev->netdev_ops->ndo_fdb_del)
4199                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4200                                                            vid);
4201                 else
4202                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4203
4204                 if (!err) {
4205                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4206                                         ndm->ndm_state);
4207                         ndm->ndm_flags &= ~NTF_SELF;
4208                 }
4209         }
4210 out:
4211         return err;
4212 }
4213
4214 static int nlmsg_populate_fdb(struct sk_buff *skb,
4215                               struct netlink_callback *cb,
4216                               struct net_device *dev,
4217                               int *idx,
4218                               struct netdev_hw_addr_list *list)
4219 {
4220         struct netdev_hw_addr *ha;
4221         int err;
4222         u32 portid, seq;
4223
4224         portid = NETLINK_CB(cb->skb).portid;
4225         seq = cb->nlh->nlmsg_seq;
4226
4227         list_for_each_entry(ha, &list->list, list) {
4228                 if (*idx < cb->args[2])
4229                         goto skip;
4230
4231                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4232                                               portid, seq,
4233                                               RTM_NEWNEIGH, NTF_SELF,
4234                                               NLM_F_MULTI, NUD_PERMANENT);
4235                 if (err < 0)
4236                         return err;
4237 skip:
4238                 *idx += 1;
4239         }
4240         return 0;
4241 }
4242
4243 /**
4244  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4245  * @skb: socket buffer to store message in
4246  * @cb: netlink callback
4247  * @dev: netdevice
4248  * @filter_dev: ignored
4249  * @idx: the number of FDB table entries dumped is added to *@idx
4250  *
4251  * Default netdevice operation to dump the existing unicast address list.
4252  * Returns number of addresses from list put in skb.
4253  */
4254 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4255                       struct netlink_callback *cb,
4256                       struct net_device *dev,
4257                       struct net_device *filter_dev,
4258                       int *idx)
4259 {
4260         int err;
4261
4262         if (dev->type != ARPHRD_ETHER)
4263                 return -EINVAL;
4264
4265         netif_addr_lock_bh(dev);
4266         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4267         if (err)
4268                 goto out;
4269         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4270 out:
4271         netif_addr_unlock_bh(dev);
4272         return err;
4273 }
4274 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4275
4276 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4277                                  int *br_idx, int *brport_idx,
4278                                  struct netlink_ext_ack *extack)
4279 {
4280         struct nlattr *tb[NDA_MAX + 1];
4281         struct ndmsg *ndm;
4282         int err, i;
4283
4284         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4285                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4286                 return -EINVAL;
4287         }
4288
4289         ndm = nlmsg_data(nlh);
4290         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4291             ndm->ndm_flags || ndm->ndm_type) {
4292                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4293                 return -EINVAL;
4294         }
4295
4296         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4297                                             NDA_MAX, NULL, extack);
4298         if (err < 0)
4299                 return err;
4300
4301         *brport_idx = ndm->ndm_ifindex;
4302         for (i = 0; i <= NDA_MAX; ++i) {
4303                 if (!tb[i])
4304                         continue;
4305
4306                 switch (i) {
4307                 case NDA_IFINDEX:
4308                         if (nla_len(tb[i]) != sizeof(u32)) {
4309                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4310                                 return -EINVAL;
4311                         }
4312                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4313                         break;
4314                 case NDA_MASTER:
4315                         if (nla_len(tb[i]) != sizeof(u32)) {
4316                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4317                                 return -EINVAL;
4318                         }
4319                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4320                         break;
4321                 default:
4322                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4323                         return -EINVAL;
4324                 }
4325         }
4326
4327         return 0;
4328 }
4329
4330 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4331                                  int *br_idx, int *brport_idx,
4332                                  struct netlink_ext_ack *extack)
4333 {
4334         struct nlattr *tb[IFLA_MAX+1];
4335         int err;
4336
4337         /* A hack to preserve kernel<->userspace interface.
4338          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4339          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4340          * So, check for ndmsg with an optional u32 attribute (not used here).
4341          * Fortunately these sizes don't conflict with the size of ifinfomsg
4342          * with an optional attribute.
4343          */
4344         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4345             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4346              nla_attr_size(sizeof(u32)))) {
4347                 struct ifinfomsg *ifm;
4348
4349                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4350                                              tb, IFLA_MAX, ifla_policy,
4351                                              extack);
4352                 if (err < 0) {
4353                         return -EINVAL;
4354                 } else if (err == 0) {
4355                         if (tb[IFLA_MASTER])
4356                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4357                 }
4358
4359                 ifm = nlmsg_data(nlh);
4360                 *brport_idx = ifm->ifi_index;
4361         }
4362         return 0;
4363 }
4364
4365 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4366 {
4367         struct net_device *dev;
4368         struct net_device *br_dev = NULL;
4369         const struct net_device_ops *ops = NULL;
4370         const struct net_device_ops *cops = NULL;
4371         struct net *net = sock_net(skb->sk);
4372         struct hlist_head *head;
4373         int brport_idx = 0;
4374         int br_idx = 0;
4375         int h, s_h;
4376         int idx = 0, s_idx;
4377         int err = 0;
4378         int fidx = 0;
4379
4380         if (cb->strict_check)
4381                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4382                                             cb->extack);
4383         else
4384                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4385                                             cb->extack);
4386         if (err < 0)
4387                 return err;
4388
4389         if (br_idx) {
4390                 br_dev = __dev_get_by_index(net, br_idx);
4391                 if (!br_dev)
4392                         return -ENODEV;
4393
4394                 ops = br_dev->netdev_ops;
4395         }
4396
4397         s_h = cb->args[0];
4398         s_idx = cb->args[1];
4399
4400         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4401                 idx = 0;
4402                 head = &net->dev_index_head[h];
4403                 hlist_for_each_entry(dev, head, index_hlist) {
4404
4405                         if (brport_idx && (dev->ifindex != brport_idx))
4406                                 continue;
4407
4408                         if (!br_idx) { /* user did not specify a specific bridge */
4409                                 if (netif_is_bridge_port(dev)) {
4410                                         br_dev = netdev_master_upper_dev_get(dev);
4411                                         cops = br_dev->netdev_ops;
4412                                 }
4413                         } else {
4414                                 if (dev != br_dev &&
4415                                     !netif_is_bridge_port(dev))
4416                                         continue;
4417
4418                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4419                                     !(dev->priv_flags & IFF_EBRIDGE))
4420                                         continue;
4421                                 cops = ops;
4422                         }
4423
4424                         if (idx < s_idx)
4425                                 goto cont;
4426
4427                         if (netif_is_bridge_port(dev)) {
4428                                 if (cops && cops->ndo_fdb_dump) {
4429                                         err = cops->ndo_fdb_dump(skb, cb,
4430                                                                 br_dev, dev,
4431                                                                 &fidx);
4432                                         if (err == -EMSGSIZE)
4433                                                 goto out;
4434                                 }
4435                         }
4436
4437                         if (dev->netdev_ops->ndo_fdb_dump)
4438                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4439                                                                     dev, NULL,
4440                                                                     &fidx);
4441                         else
4442                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4443                                                         &fidx);
4444                         if (err == -EMSGSIZE)
4445                                 goto out;
4446
4447                         cops = NULL;
4448
4449                         /* reset fdb offset to 0 for rest of the interfaces */
4450                         cb->args[2] = 0;
4451                         fidx = 0;
4452 cont:
4453                         idx++;
4454                 }
4455         }
4456
4457 out:
4458         cb->args[0] = h;
4459         cb->args[1] = idx;
4460         cb->args[2] = fidx;
4461
4462         return skb->len;
4463 }
4464
4465 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4466                                 struct nlattr **tb, u8 *ndm_flags,
4467                                 int *br_idx, int *brport_idx, u8 **addr,
4468                                 u16 *vid, struct netlink_ext_ack *extack)
4469 {
4470         struct ndmsg *ndm;
4471         int err, i;
4472
4473         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4474                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4475                 return -EINVAL;
4476         }
4477
4478         ndm = nlmsg_data(nlh);
4479         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4480             ndm->ndm_type) {
4481                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4482                 return -EINVAL;
4483         }
4484
4485         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4486                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4487                 return -EINVAL;
4488         }
4489
4490         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4491                                             NDA_MAX, nda_policy, extack);
4492         if (err < 0)
4493                 return err;
4494
4495         *ndm_flags = ndm->ndm_flags;
4496         *brport_idx = ndm->ndm_ifindex;
4497         for (i = 0; i <= NDA_MAX; ++i) {
4498                 if (!tb[i])
4499                         continue;
4500
4501                 switch (i) {
4502                 case NDA_MASTER:
4503                         *br_idx = nla_get_u32(tb[i]);
4504                         break;
4505                 case NDA_LLADDR:
4506                         if (nla_len(tb[i]) != ETH_ALEN) {
4507                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4508                                 return -EINVAL;
4509                         }
4510                         *addr = nla_data(tb[i]);
4511                         break;
4512                 case NDA_VLAN:
4513                         err = fdb_vid_parse(tb[i], vid, extack);
4514                         if (err)
4515                                 return err;
4516                         break;
4517                 case NDA_VNI:
4518                         break;
4519                 default:
4520                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4521                         return -EINVAL;
4522                 }
4523         }
4524
4525         return 0;
4526 }
4527
4528 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4529                         struct netlink_ext_ack *extack)
4530 {
4531         struct net_device *dev = NULL, *br_dev = NULL;
4532         const struct net_device_ops *ops = NULL;
4533         struct net *net = sock_net(in_skb->sk);
4534         struct nlattr *tb[NDA_MAX + 1];
4535         struct sk_buff *skb;
4536         int brport_idx = 0;
4537         u8 ndm_flags = 0;
4538         int br_idx = 0;
4539         u8 *addr = NULL;
4540         u16 vid = 0;
4541         int err;
4542
4543         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4544                                    &brport_idx, &addr, &vid, extack);
4545         if (err < 0)
4546                 return err;
4547
4548         if (!addr) {
4549                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4550                 return -EINVAL;
4551         }
4552
4553         if (brport_idx) {
4554                 dev = __dev_get_by_index(net, brport_idx);
4555                 if (!dev) {
4556                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4557                         return -ENODEV;
4558                 }
4559         }
4560
4561         if (br_idx) {
4562                 if (dev) {
4563                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4564                         return -EINVAL;
4565                 }
4566
4567                 br_dev = __dev_get_by_index(net, br_idx);
4568                 if (!br_dev) {
4569                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4570                         return -EINVAL;
4571                 }
4572                 ops = br_dev->netdev_ops;
4573         }
4574
4575         if (dev) {
4576                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4577                         if (!netif_is_bridge_port(dev)) {
4578                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4579                                 return -EINVAL;
4580                         }
4581                         br_dev = netdev_master_upper_dev_get(dev);
4582                         if (!br_dev) {
4583                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4584                                 return -EINVAL;
4585                         }
4586                         ops = br_dev->netdev_ops;
4587                 } else {
4588                         if (!(ndm_flags & NTF_SELF)) {
4589                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4590                                 return -EINVAL;
4591                         }
4592                         ops = dev->netdev_ops;
4593                 }
4594         }
4595
4596         if (!br_dev && !dev) {
4597                 NL_SET_ERR_MSG(extack, "No device specified");
4598                 return -ENODEV;
4599         }
4600
4601         if (!ops || !ops->ndo_fdb_get) {
4602                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4603                 return -EOPNOTSUPP;
4604         }
4605
4606         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4607         if (!skb)
4608                 return -ENOBUFS;
4609
4610         if (br_dev)
4611                 dev = br_dev;
4612         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4613                                NETLINK_CB(in_skb).portid,
4614                                nlh->nlmsg_seq, extack);
4615         if (err)
4616                 goto out;
4617
4618         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4619 out:
4620         kfree_skb(skb);
4621         return err;
4622 }
4623
4624 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4625                                unsigned int attrnum, unsigned int flag)
4626 {
4627         if (mask & flag)
4628                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4629         return 0;
4630 }
4631
4632 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4633                             struct net_device *dev, u16 mode,
4634                             u32 flags, u32 mask, int nlflags,
4635                             u32 filter_mask,
4636                             int (*vlan_fill)(struct sk_buff *skb,
4637                                              struct net_device *dev,
4638                                              u32 filter_mask))
4639 {
4640         struct nlmsghdr *nlh;
4641         struct ifinfomsg *ifm;
4642         struct nlattr *br_afspec;
4643         struct nlattr *protinfo;
4644         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4645         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4646         int err = 0;
4647
4648         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4649         if (nlh == NULL)
4650                 return -EMSGSIZE;
4651
4652         ifm = nlmsg_data(nlh);
4653         ifm->ifi_family = AF_BRIDGE;
4654         ifm->__ifi_pad = 0;
4655         ifm->ifi_type = dev->type;
4656         ifm->ifi_index = dev->ifindex;
4657         ifm->ifi_flags = dev_get_flags(dev);
4658         ifm->ifi_change = 0;
4659
4660
4661         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4662             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4663             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4664             (br_dev &&
4665              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4666             (dev->addr_len &&
4667              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4668             (dev->ifindex != dev_get_iflink(dev) &&
4669              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4670                 goto nla_put_failure;
4671
4672         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4673         if (!br_afspec)
4674                 goto nla_put_failure;
4675
4676         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4677                 nla_nest_cancel(skb, br_afspec);
4678                 goto nla_put_failure;
4679         }
4680
4681         if (mode != BRIDGE_MODE_UNDEF) {
4682                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4683                         nla_nest_cancel(skb, br_afspec);
4684                         goto nla_put_failure;
4685                 }
4686         }
4687         if (vlan_fill) {
4688                 err = vlan_fill(skb, dev, filter_mask);
4689                 if (err) {
4690                         nla_nest_cancel(skb, br_afspec);
4691                         goto nla_put_failure;
4692                 }
4693         }
4694         nla_nest_end(skb, br_afspec);
4695
4696         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4697         if (!protinfo)
4698                 goto nla_put_failure;
4699
4700         if (brport_nla_put_flag(skb, flags, mask,
4701                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4702             brport_nla_put_flag(skb, flags, mask,
4703                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4704             brport_nla_put_flag(skb, flags, mask,
4705                                 IFLA_BRPORT_FAST_LEAVE,
4706                                 BR_MULTICAST_FAST_LEAVE) ||
4707             brport_nla_put_flag(skb, flags, mask,
4708                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4709             brport_nla_put_flag(skb, flags, mask,
4710                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4711             brport_nla_put_flag(skb, flags, mask,
4712                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4713             brport_nla_put_flag(skb, flags, mask,
4714                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4715             brport_nla_put_flag(skb, flags, mask,
4716                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4717             brport_nla_put_flag(skb, flags, mask,
4718                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4719             brport_nla_put_flag(skb, flags, mask,
4720                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4721                 nla_nest_cancel(skb, protinfo);
4722                 goto nla_put_failure;
4723         }
4724
4725         nla_nest_end(skb, protinfo);
4726
4727         nlmsg_end(skb, nlh);
4728         return 0;
4729 nla_put_failure:
4730         nlmsg_cancel(skb, nlh);
4731         return err ? err : -EMSGSIZE;
4732 }
4733 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4734
4735 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4736                                     bool strict_check, u32 *filter_mask,
4737                                     struct netlink_ext_ack *extack)
4738 {
4739         struct nlattr *tb[IFLA_MAX+1];
4740         int err, i;
4741
4742         if (strict_check) {
4743                 struct ifinfomsg *ifm;
4744
4745                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4746                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4747                         return -EINVAL;
4748                 }
4749
4750                 ifm = nlmsg_data(nlh);
4751                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4752                     ifm->ifi_change || ifm->ifi_index) {
4753                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4754                         return -EINVAL;
4755                 }
4756
4757                 err = nlmsg_parse_deprecated_strict(nlh,
4758                                                     sizeof(struct ifinfomsg),
4759                                                     tb, IFLA_MAX, ifla_policy,
4760                                                     extack);
4761         } else {
4762                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4763                                              tb, IFLA_MAX, ifla_policy,
4764                                              extack);
4765         }
4766         if (err < 0)
4767                 return err;
4768
4769         /* new attributes should only be added with strict checking */
4770         for (i = 0; i <= IFLA_MAX; ++i) {
4771                 if (!tb[i])
4772                         continue;
4773
4774                 switch (i) {
4775                 case IFLA_EXT_MASK:
4776                         *filter_mask = nla_get_u32(tb[i]);
4777                         break;
4778                 default:
4779                         if (strict_check) {
4780                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4781                                 return -EINVAL;
4782                         }
4783                 }
4784         }
4785
4786         return 0;
4787 }
4788
4789 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4790 {
4791         const struct nlmsghdr *nlh = cb->nlh;
4792         struct net *net = sock_net(skb->sk);
4793         struct net_device *dev;
4794         int idx = 0;
4795         u32 portid = NETLINK_CB(cb->skb).portid;
4796         u32 seq = nlh->nlmsg_seq;
4797         u32 filter_mask = 0;
4798         int err;
4799
4800         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4801                                        cb->extack);
4802         if (err < 0 && cb->strict_check)
4803                 return err;
4804
4805         rcu_read_lock();
4806         for_each_netdev_rcu(net, dev) {
4807                 const struct net_device_ops *ops = dev->netdev_ops;
4808                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4809
4810                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4811                         if (idx >= cb->args[0]) {
4812                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4813                                                 skb, portid, seq, dev,
4814                                                 filter_mask, NLM_F_MULTI);
4815                                 if (err < 0 && err != -EOPNOTSUPP) {
4816                                         if (likely(skb->len))
4817                                                 break;
4818
4819                                         goto out_err;
4820                                 }
4821                         }
4822                         idx++;
4823                 }
4824
4825                 if (ops->ndo_bridge_getlink) {
4826                         if (idx >= cb->args[0]) {
4827                                 err = ops->ndo_bridge_getlink(skb, portid,
4828                                                               seq, dev,
4829                                                               filter_mask,
4830                                                               NLM_F_MULTI);
4831                                 if (err < 0 && err != -EOPNOTSUPP) {
4832                                         if (likely(skb->len))
4833                                                 break;
4834
4835                                         goto out_err;
4836                                 }
4837                         }
4838                         idx++;
4839                 }
4840         }
4841         err = skb->len;
4842 out_err:
4843         rcu_read_unlock();
4844         cb->args[0] = idx;
4845
4846         return err;
4847 }
4848
4849 static inline size_t bridge_nlmsg_size(void)
4850 {
4851         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4852                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4853                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4854                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4855                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4856                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4857                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4858                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4859                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4860                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4861                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4862 }
4863
4864 static int rtnl_bridge_notify(struct net_device *dev)
4865 {
4866         struct net *net = dev_net(dev);
4867         struct sk_buff *skb;
4868         int err = -EOPNOTSUPP;
4869
4870         if (!dev->netdev_ops->ndo_bridge_getlink)
4871                 return 0;
4872
4873         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4874         if (!skb) {
4875                 err = -ENOMEM;
4876                 goto errout;
4877         }
4878
4879         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4880         if (err < 0)
4881                 goto errout;
4882
4883         /* Notification info is only filled for bridge ports, not the bridge
4884          * device itself. Therefore, a zero notification length is valid and
4885          * should not result in an error.
4886          */
4887         if (!skb->len)
4888                 goto errout;
4889
4890         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4891         return 0;
4892 errout:
4893         WARN_ON(err == -EMSGSIZE);
4894         kfree_skb(skb);
4895         if (err)
4896                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4897         return err;
4898 }
4899
4900 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4901                                struct netlink_ext_ack *extack)
4902 {
4903         struct net *net = sock_net(skb->sk);
4904         struct ifinfomsg *ifm;
4905         struct net_device *dev;
4906         struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
4907         int rem, err = -EOPNOTSUPP;
4908         u16 flags = 0;
4909
4910         if (nlmsg_len(nlh) < sizeof(*ifm))
4911                 return -EINVAL;
4912
4913         ifm = nlmsg_data(nlh);
4914         if (ifm->ifi_family != AF_BRIDGE)
4915                 return -EPFNOSUPPORT;
4916
4917         dev = __dev_get_by_index(net, ifm->ifi_index);
4918         if (!dev) {
4919                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4920                 return -ENODEV;
4921         }
4922
4923         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4924         if (br_spec) {
4925                 nla_for_each_nested(attr, br_spec, rem) {
4926                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
4927                                 if (nla_len(attr) < sizeof(flags))
4928                                         return -EINVAL;
4929
4930                                 br_flags_attr = attr;
4931                                 flags = nla_get_u16(attr);
4932                         }
4933
4934                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
4935                                 if (nla_len(attr) < sizeof(u16))
4936                                         return -EINVAL;
4937                         }
4938                 }
4939         }
4940
4941         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4942                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4943
4944                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4945                         err = -EOPNOTSUPP;
4946                         goto out;
4947                 }
4948
4949                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4950                                                              extack);
4951                 if (err)
4952                         goto out;
4953
4954                 flags &= ~BRIDGE_FLAGS_MASTER;
4955         }
4956
4957         if ((flags & BRIDGE_FLAGS_SELF)) {
4958                 if (!dev->netdev_ops->ndo_bridge_setlink)
4959                         err = -EOPNOTSUPP;
4960                 else
4961                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4962                                                                   flags,
4963                                                                   extack);
4964                 if (!err) {
4965                         flags &= ~BRIDGE_FLAGS_SELF;
4966
4967                         /* Generate event to notify upper layer of bridge
4968                          * change
4969                          */
4970                         err = rtnl_bridge_notify(dev);
4971                 }
4972         }
4973
4974         if (br_flags_attr)
4975                 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
4976 out:
4977         return err;
4978 }
4979
4980 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4981                                struct netlink_ext_ack *extack)
4982 {
4983         struct net *net = sock_net(skb->sk);
4984         struct ifinfomsg *ifm;
4985         struct net_device *dev;
4986         struct nlattr *br_spec, *attr = NULL;
4987         int rem, err = -EOPNOTSUPP;
4988         u16 flags = 0;
4989         bool have_flags = false;
4990
4991         if (nlmsg_len(nlh) < sizeof(*ifm))
4992                 return -EINVAL;
4993
4994         ifm = nlmsg_data(nlh);
4995         if (ifm->ifi_family != AF_BRIDGE)
4996                 return -EPFNOSUPPORT;
4997
4998         dev = __dev_get_by_index(net, ifm->ifi_index);
4999         if (!dev) {
5000                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5001                 return -ENODEV;
5002         }
5003
5004         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5005         if (br_spec) {
5006                 nla_for_each_nested(attr, br_spec, rem) {
5007                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5008                                 if (nla_len(attr) < sizeof(flags))
5009                                         return -EINVAL;
5010
5011                                 have_flags = true;
5012                                 flags = nla_get_u16(attr);
5013                                 break;
5014                         }
5015                 }
5016         }
5017
5018         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5019                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5020
5021                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5022                         err = -EOPNOTSUPP;
5023                         goto out;
5024                 }
5025
5026                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5027                 if (err)
5028                         goto out;
5029
5030                 flags &= ~BRIDGE_FLAGS_MASTER;
5031         }
5032
5033         if ((flags & BRIDGE_FLAGS_SELF)) {
5034                 if (!dev->netdev_ops->ndo_bridge_dellink)
5035                         err = -EOPNOTSUPP;
5036                 else
5037                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5038                                                                   flags);
5039
5040                 if (!err) {
5041                         flags &= ~BRIDGE_FLAGS_SELF;
5042
5043                         /* Generate event to notify upper layer of bridge
5044                          * change
5045                          */
5046                         err = rtnl_bridge_notify(dev);
5047                 }
5048         }
5049
5050         if (have_flags)
5051                 memcpy(nla_data(attr), &flags, sizeof(flags));
5052 out:
5053         return err;
5054 }
5055
5056 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5057 {
5058         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5059                (!idxattr || idxattr == attrid);
5060 }
5061
5062 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5063 static int rtnl_get_offload_stats_attr_size(int attr_id)
5064 {
5065         switch (attr_id) {
5066         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5067                 return sizeof(struct rtnl_link_stats64);
5068         }
5069
5070         return 0;
5071 }
5072
5073 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5074                                   int *prividx)
5075 {
5076         struct nlattr *attr = NULL;
5077         int attr_id, size;
5078         void *attr_data;
5079         int err;
5080
5081         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5082               dev->netdev_ops->ndo_get_offload_stats))
5083                 return -ENODATA;
5084
5085         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5086              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5087                 if (attr_id < *prividx)
5088                         continue;
5089
5090                 size = rtnl_get_offload_stats_attr_size(attr_id);
5091                 if (!size)
5092                         continue;
5093
5094                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5095                         continue;
5096
5097                 attr = nla_reserve_64bit(skb, attr_id, size,
5098                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5099                 if (!attr)
5100                         goto nla_put_failure;
5101
5102                 attr_data = nla_data(attr);
5103                 memset(attr_data, 0, size);
5104                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5105                                                              attr_data);
5106                 if (err)
5107                         goto get_offload_stats_failure;
5108         }
5109
5110         if (!attr)
5111                 return -ENODATA;
5112
5113         *prividx = 0;
5114         return 0;
5115
5116 nla_put_failure:
5117         err = -EMSGSIZE;
5118 get_offload_stats_failure:
5119         *prividx = attr_id;
5120         return err;
5121 }
5122
5123 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5124 {
5125         int nla_size = 0;
5126         int attr_id;
5127         int size;
5128
5129         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5130               dev->netdev_ops->ndo_get_offload_stats))
5131                 return 0;
5132
5133         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5134              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5135                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5136                         continue;
5137                 size = rtnl_get_offload_stats_attr_size(attr_id);
5138                 nla_size += nla_total_size_64bit(size);
5139         }
5140
5141         if (nla_size != 0)
5142                 nla_size += nla_total_size(0);
5143
5144         return nla_size;
5145 }
5146
5147 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5148                                int type, u32 pid, u32 seq, u32 change,
5149                                unsigned int flags, unsigned int filter_mask,
5150                                int *idxattr, int *prividx)
5151 {
5152         struct if_stats_msg *ifsm;
5153         struct nlmsghdr *nlh;
5154         struct nlattr *attr;
5155         int s_prividx = *prividx;
5156         int err;
5157
5158         ASSERT_RTNL();
5159
5160         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5161         if (!nlh)
5162                 return -EMSGSIZE;
5163
5164         ifsm = nlmsg_data(nlh);
5165         ifsm->family = PF_UNSPEC;
5166         ifsm->pad1 = 0;
5167         ifsm->pad2 = 0;
5168         ifsm->ifindex = dev->ifindex;
5169         ifsm->filter_mask = filter_mask;
5170
5171         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5172                 struct rtnl_link_stats64 *sp;
5173
5174                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5175                                          sizeof(struct rtnl_link_stats64),
5176                                          IFLA_STATS_UNSPEC);
5177                 if (!attr)
5178                         goto nla_put_failure;
5179
5180                 sp = nla_data(attr);
5181                 dev_get_stats(dev, sp);
5182         }
5183
5184         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5185                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5186
5187                 if (ops && ops->fill_linkxstats) {
5188                         *idxattr = IFLA_STATS_LINK_XSTATS;
5189                         attr = nla_nest_start_noflag(skb,
5190                                                      IFLA_STATS_LINK_XSTATS);
5191                         if (!attr)
5192                                 goto nla_put_failure;
5193
5194                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5195                         nla_nest_end(skb, attr);
5196                         if (err)
5197                                 goto nla_put_failure;
5198                         *idxattr = 0;
5199                 }
5200         }
5201
5202         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5203                              *idxattr)) {
5204                 const struct rtnl_link_ops *ops = NULL;
5205                 const struct net_device *master;
5206
5207                 master = netdev_master_upper_dev_get(dev);
5208                 if (master)
5209                         ops = master->rtnl_link_ops;
5210                 if (ops && ops->fill_linkxstats) {
5211                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5212                         attr = nla_nest_start_noflag(skb,
5213                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
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_OFFLOAD_XSTATS,
5226                              *idxattr)) {
5227                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5228                 attr = nla_nest_start_noflag(skb,
5229                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5230                 if (!attr)
5231                         goto nla_put_failure;
5232
5233                 err = rtnl_get_offload_stats(skb, dev, prividx);
5234                 if (err == -ENODATA)
5235                         nla_nest_cancel(skb, attr);
5236                 else
5237                         nla_nest_end(skb, attr);
5238
5239                 if (err && err != -ENODATA)
5240                         goto nla_put_failure;
5241                 *idxattr = 0;
5242         }
5243
5244         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5245                 struct rtnl_af_ops *af_ops;
5246
5247                 *idxattr = IFLA_STATS_AF_SPEC;
5248                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5249                 if (!attr)
5250                         goto nla_put_failure;
5251
5252                 rcu_read_lock();
5253                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5254                         if (af_ops->fill_stats_af) {
5255                                 struct nlattr *af;
5256                                 int err;
5257
5258                                 af = nla_nest_start_noflag(skb,
5259                                                            af_ops->family);
5260                                 if (!af) {
5261                                         rcu_read_unlock();
5262                                         goto nla_put_failure;
5263                                 }
5264                                 err = af_ops->fill_stats_af(skb, dev);
5265
5266                                 if (err == -ENODATA) {
5267                                         nla_nest_cancel(skb, af);
5268                                 } else if (err < 0) {
5269                                         rcu_read_unlock();
5270                                         goto nla_put_failure;
5271                                 }
5272
5273                                 nla_nest_end(skb, af);
5274                         }
5275                 }
5276                 rcu_read_unlock();
5277
5278                 nla_nest_end(skb, attr);
5279
5280                 *idxattr = 0;
5281         }
5282
5283         nlmsg_end(skb, nlh);
5284
5285         return 0;
5286
5287 nla_put_failure:
5288         /* not a multi message or no progress mean a real error */
5289         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5290                 nlmsg_cancel(skb, nlh);
5291         else
5292                 nlmsg_end(skb, nlh);
5293
5294         return -EMSGSIZE;
5295 }
5296
5297 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5298                                   u32 filter_mask)
5299 {
5300         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5301
5302         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5303                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5304
5305         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5306                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5307                 int attr = IFLA_STATS_LINK_XSTATS;
5308
5309                 if (ops && ops->get_linkxstats_size) {
5310                         size += nla_total_size(ops->get_linkxstats_size(dev,
5311                                                                         attr));
5312                         /* for IFLA_STATS_LINK_XSTATS */
5313                         size += nla_total_size(0);
5314                 }
5315         }
5316
5317         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5318                 struct net_device *_dev = (struct net_device *)dev;
5319                 const struct rtnl_link_ops *ops = NULL;
5320                 const struct net_device *master;
5321
5322                 /* netdev_master_upper_dev_get can't take const */
5323                 master = netdev_master_upper_dev_get(_dev);
5324                 if (master)
5325                         ops = master->rtnl_link_ops;
5326                 if (ops && ops->get_linkxstats_size) {
5327                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5328
5329                         size += nla_total_size(ops->get_linkxstats_size(dev,
5330                                                                         attr));
5331                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5332                         size += nla_total_size(0);
5333                 }
5334         }
5335
5336         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5337                 size += rtnl_get_offload_stats_size(dev);
5338
5339         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5340                 struct rtnl_af_ops *af_ops;
5341
5342                 /* for IFLA_STATS_AF_SPEC */
5343                 size += nla_total_size(0);
5344
5345                 rcu_read_lock();
5346                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5347                         if (af_ops->get_stats_af_size) {
5348                                 size += nla_total_size(
5349                                         af_ops->get_stats_af_size(dev));
5350
5351                                 /* for AF_* */
5352                                 size += nla_total_size(0);
5353                         }
5354                 }
5355                 rcu_read_unlock();
5356         }
5357
5358         return size;
5359 }
5360
5361 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5362                                 bool is_dump, struct netlink_ext_ack *extack)
5363 {
5364         struct if_stats_msg *ifsm;
5365
5366         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5367                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5368                 return -EINVAL;
5369         }
5370
5371         if (!strict_check)
5372                 return 0;
5373
5374         ifsm = nlmsg_data(nlh);
5375
5376         /* only requests using strict checks can pass data to influence
5377          * the dump. The legacy exception is filter_mask.
5378          */
5379         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5380                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5381                 return -EINVAL;
5382         }
5383         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5384                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5385                 return -EINVAL;
5386         }
5387         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5388                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5389                 return -EINVAL;
5390         }
5391
5392         return 0;
5393 }
5394
5395 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5396                           struct netlink_ext_ack *extack)
5397 {
5398         struct net *net = sock_net(skb->sk);
5399         struct net_device *dev = NULL;
5400         int idxattr = 0, prividx = 0;
5401         struct if_stats_msg *ifsm;
5402         struct sk_buff *nskb;
5403         u32 filter_mask;
5404         int err;
5405
5406         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5407                                    false, extack);
5408         if (err)
5409                 return err;
5410
5411         ifsm = nlmsg_data(nlh);
5412         if (ifsm->ifindex > 0)
5413                 dev = __dev_get_by_index(net, ifsm->ifindex);
5414         else
5415                 return -EINVAL;
5416
5417         if (!dev)
5418                 return -ENODEV;
5419
5420         filter_mask = ifsm->filter_mask;
5421         if (!filter_mask)
5422                 return -EINVAL;
5423
5424         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5425         if (!nskb)
5426                 return -ENOBUFS;
5427
5428         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5429                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5430                                   0, filter_mask, &idxattr, &prividx);
5431         if (err < 0) {
5432                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5433                 WARN_ON(err == -EMSGSIZE);
5434                 kfree_skb(nskb);
5435         } else {
5436                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5437         }
5438
5439         return err;
5440 }
5441
5442 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5443 {
5444         struct netlink_ext_ack *extack = cb->extack;
5445         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5446         struct net *net = sock_net(skb->sk);
5447         unsigned int flags = NLM_F_MULTI;
5448         struct if_stats_msg *ifsm;
5449         struct hlist_head *head;
5450         struct net_device *dev;
5451         u32 filter_mask = 0;
5452         int idx = 0;
5453
5454         s_h = cb->args[0];
5455         s_idx = cb->args[1];
5456         s_idxattr = cb->args[2];
5457         s_prividx = cb->args[3];
5458
5459         cb->seq = net->dev_base_seq;
5460
5461         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5462         if (err)
5463                 return err;
5464
5465         ifsm = nlmsg_data(cb->nlh);
5466         filter_mask = ifsm->filter_mask;
5467         if (!filter_mask) {
5468                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5469                 return -EINVAL;
5470         }
5471
5472         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5473                 idx = 0;
5474                 head = &net->dev_index_head[h];
5475                 hlist_for_each_entry(dev, head, index_hlist) {
5476                         if (idx < s_idx)
5477                                 goto cont;
5478                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5479                                                   NETLINK_CB(cb->skb).portid,
5480                                                   cb->nlh->nlmsg_seq, 0,
5481                                                   flags, filter_mask,
5482                                                   &s_idxattr, &s_prividx);
5483                         /* If we ran out of room on the first message,
5484                          * we're in trouble
5485                          */
5486                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5487
5488                         if (err < 0)
5489                                 goto out;
5490                         s_prividx = 0;
5491                         s_idxattr = 0;
5492                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5493 cont:
5494                         idx++;
5495                 }
5496         }
5497 out:
5498         cb->args[3] = s_prividx;
5499         cb->args[2] = s_idxattr;
5500         cb->args[1] = idx;
5501         cb->args[0] = h;
5502
5503         return skb->len;
5504 }
5505
5506 /* Process one rtnetlink message. */
5507
5508 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5509                              struct netlink_ext_ack *extack)
5510 {
5511         struct net *net = sock_net(skb->sk);
5512         struct rtnl_link *link;
5513         struct module *owner;
5514         int err = -EOPNOTSUPP;
5515         rtnl_doit_func doit;
5516         unsigned int flags;
5517         int kind;
5518         int family;
5519         int type;
5520
5521         type = nlh->nlmsg_type;
5522         if (type > RTM_MAX)
5523                 return -EOPNOTSUPP;
5524
5525         type -= RTM_BASE;
5526
5527         /* All the messages must have at least 1 byte length */
5528         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5529                 return 0;
5530
5531         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5532         kind = type&3;
5533
5534         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5535                 return -EPERM;
5536
5537         rcu_read_lock();
5538         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5539                 struct sock *rtnl;
5540                 rtnl_dumpit_func dumpit;
5541                 u32 min_dump_alloc = 0;
5542
5543                 link = rtnl_get_link(family, type);
5544                 if (!link || !link->dumpit) {
5545                         family = PF_UNSPEC;
5546                         link = rtnl_get_link(family, type);
5547                         if (!link || !link->dumpit)
5548                                 goto err_unlock;
5549                 }
5550                 owner = link->owner;
5551                 dumpit = link->dumpit;
5552
5553                 if (type == RTM_GETLINK - RTM_BASE)
5554                         min_dump_alloc = rtnl_calcit(skb, nlh);
5555
5556                 err = 0;
5557                 /* need to do this before rcu_read_unlock() */
5558                 if (!try_module_get(owner))
5559                         err = -EPROTONOSUPPORT;
5560
5561                 rcu_read_unlock();
5562
5563                 rtnl = net->rtnl;
5564                 if (err == 0) {
5565                         struct netlink_dump_control c = {
5566                                 .dump           = dumpit,
5567                                 .min_dump_alloc = min_dump_alloc,
5568                                 .module         = owner,
5569                         };
5570                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5571                         /* netlink_dump_start() will keep a reference on
5572                          * module if dump is still in progress.
5573                          */
5574                         module_put(owner);
5575                 }
5576                 return err;
5577         }
5578
5579         link = rtnl_get_link(family, type);
5580         if (!link || !link->doit) {
5581                 family = PF_UNSPEC;
5582                 link = rtnl_get_link(PF_UNSPEC, type);
5583                 if (!link || !link->doit)
5584                         goto out_unlock;
5585         }
5586
5587         owner = link->owner;
5588         if (!try_module_get(owner)) {
5589                 err = -EPROTONOSUPPORT;
5590                 goto out_unlock;
5591         }
5592
5593         flags = link->flags;
5594         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5595                 doit = link->doit;
5596                 rcu_read_unlock();
5597                 if (doit)
5598                         err = doit(skb, nlh, extack);
5599                 module_put(owner);
5600                 return err;
5601         }
5602         rcu_read_unlock();
5603
5604         rtnl_lock();
5605         link = rtnl_get_link(family, type);
5606         if (link && link->doit)
5607                 err = link->doit(skb, nlh, extack);
5608         rtnl_unlock();
5609
5610         module_put(owner);
5611
5612         return err;
5613
5614 out_unlock:
5615         rcu_read_unlock();
5616         return err;
5617
5618 err_unlock:
5619         rcu_read_unlock();
5620         return -EOPNOTSUPP;
5621 }
5622
5623 static void rtnetlink_rcv(struct sk_buff *skb)
5624 {
5625         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5626 }
5627
5628 static int rtnetlink_bind(struct net *net, int group)
5629 {
5630         switch (group) {
5631         case RTNLGRP_IPV4_MROUTE_R:
5632         case RTNLGRP_IPV6_MROUTE_R:
5633                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5634                         return -EPERM;
5635                 break;
5636         }
5637         return 0;
5638 }
5639
5640 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5641 {
5642         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5643
5644         switch (event) {
5645         case NETDEV_REBOOT:
5646         case NETDEV_CHANGEMTU:
5647         case NETDEV_CHANGEADDR:
5648         case NETDEV_CHANGENAME:
5649         case NETDEV_FEAT_CHANGE:
5650         case NETDEV_BONDING_FAILOVER:
5651         case NETDEV_POST_TYPE_CHANGE:
5652         case NETDEV_NOTIFY_PEERS:
5653         case NETDEV_CHANGEUPPER:
5654         case NETDEV_RESEND_IGMP:
5655         case NETDEV_CHANGEINFODATA:
5656         case NETDEV_CHANGELOWERSTATE:
5657         case NETDEV_CHANGE_TX_QUEUE_LEN:
5658                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5659                                    GFP_KERNEL, NULL, 0);
5660                 break;
5661         default:
5662                 break;
5663         }
5664         return NOTIFY_DONE;
5665 }
5666
5667 static struct notifier_block rtnetlink_dev_notifier = {
5668         .notifier_call  = rtnetlink_event,
5669 };
5670
5671
5672 static int __net_init rtnetlink_net_init(struct net *net)
5673 {
5674         struct sock *sk;
5675         struct netlink_kernel_cfg cfg = {
5676                 .groups         = RTNLGRP_MAX,
5677                 .input          = rtnetlink_rcv,
5678                 .cb_mutex       = &rtnl_mutex,
5679                 .flags          = NL_CFG_F_NONROOT_RECV,
5680                 .bind           = rtnetlink_bind,
5681         };
5682
5683         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5684         if (!sk)
5685                 return -ENOMEM;
5686         net->rtnl = sk;
5687         return 0;
5688 }
5689
5690 static void __net_exit rtnetlink_net_exit(struct net *net)
5691 {
5692         netlink_kernel_release(net->rtnl);
5693         net->rtnl = NULL;
5694 }
5695
5696 static struct pernet_operations rtnetlink_net_ops = {
5697         .init = rtnetlink_net_init,
5698         .exit = rtnetlink_net_exit,
5699 };
5700
5701 void __init rtnetlink_init(void)
5702 {
5703         if (register_pernet_subsys(&rtnetlink_net_ops))
5704                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5705
5706         register_netdevice_notifier(&rtnetlink_dev_notifier);
5707
5708         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5709                       rtnl_dump_ifinfo, 0);
5710         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5711         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5712         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5713
5714         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5715         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5716         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5717
5718         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5719         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5720
5721         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5722         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5723         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5724
5725         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5726         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5727         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5728
5729         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5730                       0);
5731 }