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