net/core/rtnetlink.c

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