net/core/rtnetlink.c

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