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