net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetic was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56 #include <net/devlink.h>
  57 #if IS_ENABLED(CONFIG_IPV6)
  58 #include <net/addrconf.h>
  59 #endif
  60
  61 #include "dev.h"
  62
  63 #define RTNL_MAX_TYPE           50
  64 #define RTNL_SLAVE_MAX_TYPE     43
  65
  66 struct rtnl_link {
  67         rtnl_doit_func          doit;
  68         rtnl_dumpit_func        dumpit;
  69         struct module           *owner;
  70         unsigned int            flags;
  71         struct rcu_head         rcu;
  72 };
  73
  74 static DEFINE_MUTEX(rtnl_mutex);
  75
  76 void rtnl_lock(void)
  77 {
  78         mutex_lock(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock);
  81
  82 int rtnl_lock_killable(void)
  83 {
  84         return mutex_lock_killable(&rtnl_mutex);
  85 }
  86 EXPORT_SYMBOL(rtnl_lock_killable);
  87
  88 static struct sk_buff *defer_kfree_skb_list;
  89 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  90 {
  91         if (head && tail) {
  92                 tail->next = defer_kfree_skb_list;
  93                 defer_kfree_skb_list = head;
  94         }
  95 }
  96 EXPORT_SYMBOL(rtnl_kfree_skbs);
  97
  98 void __rtnl_unlock(void)
  99 {
 100         struct sk_buff *head = defer_kfree_skb_list;
 101
 102         defer_kfree_skb_list = NULL;
 103
 104         /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
 105          * is used. In some places, e.g. in cfg80211, we have code that will do
 106          * something like
 107          *   rtnl_lock()
 108          *   wiphy_lock()
 109          *   ...
 110          *   rtnl_unlock()
 111          *
 112          * and because netdev_run_todo() acquires the RTNL for items on the list
 113          * we could cause a situation such as this:
 114          * Thread 1                     Thread 2
 115          *                                rtnl_lock()
 116          *                                unregister_netdevice()
 117          *                                __rtnl_unlock()
 118          * rtnl_lock()
 119          * wiphy_lock()
 120          * rtnl_unlock()
 121          *   netdev_run_todo()
 122          *     __rtnl_unlock()
 123          *
 124          *     // list not empty now
 125          *     // because of thread 2
 126          *                                rtnl_lock()
 127          *     while (!list_empty(...))
 128          *       rtnl_lock()
 129          *                                wiphy_lock()
 130          * **** DEADLOCK ****
 131          *
 132          * However, usage of __rtnl_unlock() is rare, and so we can ensure that
 133          * it's not used in cases where something is added to do the list.
 134          */
 135         WARN_ON(!list_empty(&net_todo_list));
 136
 137         mutex_unlock(&rtnl_mutex);
 138
 139         while (head) {
 140                 struct sk_buff *next = head->next;
 141
 142                 kfree_skb(head);
 143                 cond_resched();
 144                 head = next;
 145         }
 146 }
 147
 148 void rtnl_unlock(void)
 149 {
 150         /* This fellow will unlock it for us. */
 151         netdev_run_todo();
 152 }
 153 EXPORT_SYMBOL(rtnl_unlock);
 154
 155 int rtnl_trylock(void)
 156 {
 157         return mutex_trylock(&rtnl_mutex);
 158 }
 159 EXPORT_SYMBOL(rtnl_trylock);
 160
 161 int rtnl_is_locked(void)
 162 {
 163         return mutex_is_locked(&rtnl_mutex);
 164 }
 165 EXPORT_SYMBOL(rtnl_is_locked);
 166
 167 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 168 {
 169         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 170 }
 171 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 172
 173 #ifdef CONFIG_PROVE_LOCKING
 174 bool lockdep_rtnl_is_held(void)
 175 {
 176         return lockdep_is_held(&rtnl_mutex);
 177 }
 178 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 179 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 180
 181 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 182
 183 static inline int rtm_msgindex(int msgtype)
 184 {
 185         int msgindex = msgtype - RTM_BASE;
 186
 187         /*
 188          * msgindex < 0 implies someone tried to register a netlink
 189          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 190          * the message type has not been added to linux/rtnetlink.h
 191          */
 192         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 193
 194         return msgindex;
 195 }
 196
 197 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 198 {
 199         struct rtnl_link __rcu **tab;
 200
 201         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 202                 protocol = PF_UNSPEC;
 203
 204         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 205         if (!tab)
 206                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 207
 208         return rcu_dereference_rtnl(tab[msgtype]);
 209 }
 210
 211 static int rtnl_register_internal(struct module *owner,
 212                                   int protocol, int msgtype,
 213                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 214                                   unsigned int flags)
 215 {
 216         struct rtnl_link *link, *old;
 217         struct rtnl_link __rcu **tab;
 218         int msgindex;
 219         int ret = -ENOBUFS;
 220
 221         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 222         msgindex = rtm_msgindex(msgtype);
 223
 224         rtnl_lock();
 225         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 226         if (tab == NULL) {
 227                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 228                 if (!tab)
 229                         goto unlock;
 230
 231                 /* ensures we see the 0 stores */
 232                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 233         }
 234
 235         old = rtnl_dereference(tab[msgindex]);
 236         if (old) {
 237                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 238                 if (!link)
 239                         goto unlock;
 240         } else {
 241                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 242                 if (!link)
 243                         goto unlock;
 244         }
 245
 246         WARN_ON(link->owner && link->owner != owner);
 247         link->owner = owner;
 248
 249         WARN_ON(doit && link->doit && link->doit != doit);
 250         if (doit)
 251                 link->doit = doit;
 252         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 253         if (dumpit)
 254                 link->dumpit = dumpit;
 255
 256         WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
 257                 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
 258         link->flags |= flags;
 259
 260         /* publish protocol:msgtype */
 261         rcu_assign_pointer(tab[msgindex], link);
 262         ret = 0;
 263         if (old)
 264                 kfree_rcu(old, rcu);
 265 unlock:
 266         rtnl_unlock();
 267         return ret;
 268 }
 269
 270 /**
 271  * rtnl_register_module - Register a rtnetlink message type
 272  *
 273  * @owner: module registering the hook (THIS_MODULE)
 274  * @protocol: Protocol family or PF_UNSPEC
 275  * @msgtype: rtnetlink message type
 276  * @doit: Function pointer called for each request message
 277  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 278  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 279  *
 280  * Like rtnl_register, but for use by removable modules.
 281  */
 282 int rtnl_register_module(struct module *owner,
 283                          int protocol, int msgtype,
 284                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 285                          unsigned int flags)
 286 {
 287         return rtnl_register_internal(owner, protocol, msgtype,
 288                                       doit, dumpit, flags);
 289 }
 290 EXPORT_SYMBOL_GPL(rtnl_register_module);
 291
 292 /**
 293  * rtnl_register - Register a rtnetlink message type
 294  * @protocol: Protocol family or PF_UNSPEC
 295  * @msgtype: rtnetlink message type
 296  * @doit: Function pointer called for each request message
 297  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 298  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 299  *
 300  * Registers the specified function pointers (at least one of them has
 301  * to be non-NULL) to be called whenever a request message for the
 302  * specified protocol family and message type is received.
 303  *
 304  * The special protocol family PF_UNSPEC may be used to define fallback
 305  * function pointers for the case when no entry for the specific protocol
 306  * family exists.
 307  */
 308 void rtnl_register(int protocol, int msgtype,
 309                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 310                    unsigned int flags)
 311 {
 312         int err;
 313
 314         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 315                                      flags);
 316         if (err)
 317                 pr_err("Unable to register rtnetlink message handler, "
 318                        "protocol = %d, message type = %d\n", protocol, msgtype);
 319 }
 320
 321 /**
 322  * rtnl_unregister - Unregister a rtnetlink message type
 323  * @protocol: Protocol family or PF_UNSPEC
 324  * @msgtype: rtnetlink message type
 325  *
 326  * Returns 0 on success or a negative error code.
 327  */
 328 int rtnl_unregister(int protocol, int msgtype)
 329 {
 330         struct rtnl_link __rcu **tab;
 331         struct rtnl_link *link;
 332         int msgindex;
 333
 334         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 335         msgindex = rtm_msgindex(msgtype);
 336
 337         rtnl_lock();
 338         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 339         if (!tab) {
 340                 rtnl_unlock();
 341                 return -ENOENT;
 342         }
 343
 344         link = rtnl_dereference(tab[msgindex]);
 345         RCU_INIT_POINTER(tab[msgindex], NULL);
 346         rtnl_unlock();
 347
 348         kfree_rcu(link, rcu);
 349
 350         return 0;
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_unregister);
 353
 354 /**
 355  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 356  * @protocol : Protocol family or PF_UNSPEC
 357  *
 358  * Identical to calling rtnl_unregster() for all registered message types
 359  * of a certain protocol family.
 360  */
 361 void rtnl_unregister_all(int protocol)
 362 {
 363         struct rtnl_link __rcu **tab;
 364         struct rtnl_link *link;
 365         int msgindex;
 366
 367         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 368
 369         rtnl_lock();
 370         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 371         if (!tab) {
 372                 rtnl_unlock();
 373                 return;
 374         }
 375         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 376         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 377                 link = rtnl_dereference(tab[msgindex]);
 378                 if (!link)
 379                         continue;
 380
 381                 RCU_INIT_POINTER(tab[msgindex], NULL);
 382                 kfree_rcu(link, rcu);
 383         }
 384         rtnl_unlock();
 385
 386         synchronize_net();
 387
 388         kfree(tab);
 389 }
 390 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 391
 392 static LIST_HEAD(link_ops);
 393
 394 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 395 {
 396         const struct rtnl_link_ops *ops;
 397
 398         list_for_each_entry(ops, &link_ops, list) {
 399                 if (!strcmp(ops->kind, kind))
 400                         return ops;
 401         }
 402         return NULL;
 403 }
 404
 405 /**
 406  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 407  * @ops: struct rtnl_link_ops * to register
 408  *
 409  * The caller must hold the rtnl_mutex. This function should be used
 410  * by drivers that create devices during module initialization. It
 411  * must be called before registering the devices.
 412  *
 413  * Returns 0 on success or a negative error code.
 414  */
 415 int __rtnl_link_register(struct rtnl_link_ops *ops)
 416 {
 417         if (rtnl_link_ops_get(ops->kind))
 418                 return -EEXIST;
 419
 420         /* The check for alloc/setup is here because if ops
 421          * does not have that filled up, it is not possible
 422          * to use the ops for creating device. So do not
 423          * fill up dellink as well. That disables rtnl_dellink.
 424          */
 425         if ((ops->alloc || ops->setup) && !ops->dellink)
 426                 ops->dellink = unregister_netdevice_queue;
 427
 428         list_add_tail(&ops->list, &link_ops);
 429         return 0;
 430 }
 431 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 432
 433 /**
 434  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 435  * @ops: struct rtnl_link_ops * to register
 436  *
 437  * Returns 0 on success or a negative error code.
 438  */
 439 int rtnl_link_register(struct rtnl_link_ops *ops)
 440 {
 441         int err;
 442
 443         /* Sanity-check max sizes to avoid stack buffer overflow. */
 444         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 445                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 446                 return -EINVAL;
 447
 448         rtnl_lock();
 449         err = __rtnl_link_register(ops);
 450         rtnl_unlock();
 451         return err;
 452 }
 453 EXPORT_SYMBOL_GPL(rtnl_link_register);
 454
 455 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 456 {
 457         struct net_device *dev;
 458         LIST_HEAD(list_kill);
 459
 460         for_each_netdev(net, dev) {
 461                 if (dev->rtnl_link_ops == ops)
 462                         ops->dellink(dev, &list_kill);
 463         }
 464         unregister_netdevice_many(&list_kill);
 465 }
 466
 467 /**
 468  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 469  * @ops: struct rtnl_link_ops * to unregister
 470  *
 471  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 472  * integrity (hold pernet_ops_rwsem for writing to close the race
 473  * with setup_net() and cleanup_net()).
 474  */
 475 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 476 {
 477         struct net *net;
 478
 479         for_each_net(net) {
 480                 __rtnl_kill_links(net, ops);
 481         }
 482         list_del(&ops->list);
 483 }
 484 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 485
 486 /* Return with the rtnl_lock held when there are no network
 487  * devices unregistering in any network namespace.
 488  */
 489 static void rtnl_lock_unregistering_all(void)
 490 {
 491         struct net *net;
 492         bool unregistering;
 493         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 494
 495         add_wait_queue(&netdev_unregistering_wq, &wait);
 496         for (;;) {
 497                 unregistering = false;
 498                 rtnl_lock();
 499                 /* We held write locked pernet_ops_rwsem, and parallel
 500                  * setup_net() and cleanup_net() are not possible.
 501                  */
 502                 for_each_net(net) {
 503                         if (atomic_read(&net->dev_unreg_count) > 0) {
 504                                 unregistering = true;
 505                                 break;
 506                         }
 507                 }
 508                 if (!unregistering)
 509                         break;
 510                 __rtnl_unlock();
 511
 512                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 513         }
 514         remove_wait_queue(&netdev_unregistering_wq, &wait);
 515 }
 516
 517 /**
 518  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 519  * @ops: struct rtnl_link_ops * to unregister
 520  */
 521 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 522 {
 523         /* Close the race with setup_net() and cleanup_net() */
 524         down_write(&pernet_ops_rwsem);
 525         rtnl_lock_unregistering_all();
 526         __rtnl_link_unregister(ops);
 527         rtnl_unlock();
 528         up_write(&pernet_ops_rwsem);
 529 }
 530 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 531
 532 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 533 {
 534         struct net_device *master_dev;
 535         const struct rtnl_link_ops *ops;
 536         size_t size = 0;
 537
 538         rcu_read_lock();
 539
 540         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 541         if (!master_dev)
 542                 goto out;
 543
 544         ops = master_dev->rtnl_link_ops;
 545         if (!ops || !ops->get_slave_size)
 546                 goto out;
 547         /* IFLA_INFO_SLAVE_DATA + nested data */
 548         size = nla_total_size(sizeof(struct nlattr)) +
 549                ops->get_slave_size(master_dev, dev);
 550
 551 out:
 552         rcu_read_unlock();
 553         return size;
 554 }
 555
 556 static size_t rtnl_link_get_size(const struct net_device *dev)
 557 {
 558         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 559         size_t size;
 560
 561         if (!ops)
 562                 return 0;
 563
 564         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 565                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 566
 567         if (ops->get_size)
 568                 /* IFLA_INFO_DATA + nested data */
 569                 size += nla_total_size(sizeof(struct nlattr)) +
 570                         ops->get_size(dev);
 571
 572         if (ops->get_xstats_size)
 573                 /* IFLA_INFO_XSTATS */
 574                 size += nla_total_size(ops->get_xstats_size(dev));
 575
 576         size += rtnl_link_get_slave_info_data_size(dev);
 577
 578         return size;
 579 }
 580
 581 static LIST_HEAD(rtnl_af_ops);
 582
 583 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 584 {
 585         const struct rtnl_af_ops *ops;
 586
 587         ASSERT_RTNL();
 588
 589         list_for_each_entry(ops, &rtnl_af_ops, list) {
 590                 if (ops->family == family)
 591                         return ops;
 592         }
 593
 594         return NULL;
 595 }
 596
 597 /**
 598  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 599  * @ops: struct rtnl_af_ops * to register
 600  *
 601  * Returns 0 on success or a negative error code.
 602  */
 603 void rtnl_af_register(struct rtnl_af_ops *ops)
 604 {
 605         rtnl_lock();
 606         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 607         rtnl_unlock();
 608 }
 609 EXPORT_SYMBOL_GPL(rtnl_af_register);
 610
 611 /**
 612  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 613  * @ops: struct rtnl_af_ops * to unregister
 614  */
 615 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 616 {
 617         rtnl_lock();
 618         list_del_rcu(&ops->list);
 619         rtnl_unlock();
 620
 621         synchronize_rcu();
 622 }
 623 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 624
 625 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 626                                     u32 ext_filter_mask)
 627 {
 628         struct rtnl_af_ops *af_ops;
 629         size_t size;
 630
 631         /* IFLA_AF_SPEC */
 632         size = nla_total_size(sizeof(struct nlattr));
 633
 634         rcu_read_lock();
 635         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 636                 if (af_ops->get_link_af_size) {
 637                         /* AF_* + nested data */
 638                         size += nla_total_size(sizeof(struct nlattr)) +
 639                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 640                 }
 641         }
 642         rcu_read_unlock();
 643
 644         return size;
 645 }
 646
 647 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 648 {
 649         struct net_device *master_dev;
 650         bool ret = false;
 651
 652         rcu_read_lock();
 653
 654         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 655         if (master_dev && master_dev->rtnl_link_ops)
 656                 ret = true;
 657         rcu_read_unlock();
 658         return ret;
 659 }
 660
 661 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 662                                      const struct net_device *dev)
 663 {
 664         struct net_device *master_dev;
 665         const struct rtnl_link_ops *ops;
 666         struct nlattr *slave_data;
 667         int err;
 668
 669         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 670         if (!master_dev)
 671                 return 0;
 672         ops = master_dev->rtnl_link_ops;
 673         if (!ops)
 674                 return 0;
 675         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 676                 return -EMSGSIZE;
 677         if (ops->fill_slave_info) {
 678                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 679                 if (!slave_data)
 680                         return -EMSGSIZE;
 681                 err = ops->fill_slave_info(skb, master_dev, dev);
 682                 if (err < 0)
 683                         goto err_cancel_slave_data;
 684                 nla_nest_end(skb, slave_data);
 685         }
 686         return 0;
 687
 688 err_cancel_slave_data:
 689         nla_nest_cancel(skb, slave_data);
 690         return err;
 691 }
 692
 693 static int rtnl_link_info_fill(struct sk_buff *skb,
 694                                const struct net_device *dev)
 695 {
 696         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 697         struct nlattr *data;
 698         int err;
 699
 700         if (!ops)
 701                 return 0;
 702         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 703                 return -EMSGSIZE;
 704         if (ops->fill_xstats) {
 705                 err = ops->fill_xstats(skb, dev);
 706                 if (err < 0)
 707                         return err;
 708         }
 709         if (ops->fill_info) {
 710                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 711                 if (data == NULL)
 712                         return -EMSGSIZE;
 713                 err = ops->fill_info(skb, dev);
 714                 if (err < 0)
 715                         goto err_cancel_data;
 716                 nla_nest_end(skb, data);
 717         }
 718         return 0;
 719
 720 err_cancel_data:
 721         nla_nest_cancel(skb, data);
 722         return err;
 723 }
 724
 725 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 726 {
 727         struct nlattr *linkinfo;
 728         int err = -EMSGSIZE;
 729
 730         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 731         if (linkinfo == NULL)
 732                 goto out;
 733
 734         err = rtnl_link_info_fill(skb, dev);
 735         if (err < 0)
 736                 goto err_cancel_link;
 737
 738         err = rtnl_link_slave_info_fill(skb, dev);
 739         if (err < 0)
 740                 goto err_cancel_link;
 741
 742         nla_nest_end(skb, linkinfo);
 743         return 0;
 744
 745 err_cancel_link:
 746         nla_nest_cancel(skb, linkinfo);
 747 out:
 748         return err;
 749 }
 750
 751 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 752 {
 753         struct sock *rtnl = net->rtnl;
 754
 755         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 756 }
 757
 758 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 759 {
 760         struct sock *rtnl = net->rtnl;
 761
 762         return nlmsg_unicast(rtnl, skb, pid);
 763 }
 764 EXPORT_SYMBOL(rtnl_unicast);
 765
 766 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 767                  const struct nlmsghdr *nlh, gfp_t flags)
 768 {
 769         struct sock *rtnl = net->rtnl;
 770
 771         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 772 }
 773 EXPORT_SYMBOL(rtnl_notify);
 774
 775 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 776 {
 777         struct sock *rtnl = net->rtnl;
 778
 779         netlink_set_err(rtnl, 0, group, error);
 780 }
 781 EXPORT_SYMBOL(rtnl_set_sk_err);
 782
 783 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 784 {
 785         struct nlattr *mx;
 786         int i, valid = 0;
 787
 788         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 789         if (metrics == dst_default_metrics.metrics)
 790                 return 0;
 791
 792         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 793         if (mx == NULL)
 794                 return -ENOBUFS;
 795
 796         for (i = 0; i < RTAX_MAX; i++) {
 797                 if (metrics[i]) {
 798                         if (i == RTAX_CC_ALGO - 1) {
 799                                 char tmp[TCP_CA_NAME_MAX], *name;
 800
 801                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 802                                 if (!name)
 803                                         continue;
 804                                 if (nla_put_string(skb, i + 1, name))
 805                                         goto nla_put_failure;
 806                         } else if (i == RTAX_FEATURES - 1) {
 807                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 808
 809                                 if (!user_features)
 810                                         continue;
 811                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 812                                 if (nla_put_u32(skb, i + 1, user_features))
 813                                         goto nla_put_failure;
 814                         } else {
 815                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 816                                         goto nla_put_failure;
 817                         }
 818                         valid++;
 819                 }
 820         }
 821
 822         if (!valid) {
 823                 nla_nest_cancel(skb, mx);
 824                 return 0;
 825         }
 826
 827         return nla_nest_end(skb, mx);
 828
 829 nla_put_failure:
 830         nla_nest_cancel(skb, mx);
 831         return -EMSGSIZE;
 832 }
 833 EXPORT_SYMBOL(rtnetlink_put_metrics);
 834
 835 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 836                        long expires, u32 error)
 837 {
 838         struct rta_cacheinfo ci = {
 839                 .rta_error = error,
 840                 .rta_id =  id,
 841         };
 842
 843         if (dst) {
 844                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 845                 ci.rta_used = dst->__use;
 846                 ci.rta_clntref = rcuref_read(&dst->__rcuref);
 847         }
 848         if (expires) {
 849                 unsigned long clock;
 850
 851                 clock = jiffies_to_clock_t(abs(expires));
 852                 clock = min_t(unsigned long, clock, INT_MAX);
 853                 ci.rta_expires = (expires > 0) ? clock : -clock;
 854         }
 855         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 856 }
 857 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 858
 859 static void set_operstate(struct net_device *dev, unsigned char transition)
 860 {
 861         unsigned char operstate = dev->operstate;
 862
 863         switch (transition) {
 864         case IF_OPER_UP:
 865                 if ((operstate == IF_OPER_DORMANT ||
 866                      operstate == IF_OPER_TESTING ||
 867                      operstate == IF_OPER_UNKNOWN) &&
 868                     !netif_dormant(dev) && !netif_testing(dev))
 869                         operstate = IF_OPER_UP;
 870                 break;
 871
 872         case IF_OPER_TESTING:
 873                 if (netif_oper_up(dev))
 874                         operstate = IF_OPER_TESTING;
 875                 break;
 876
 877         case IF_OPER_DORMANT:
 878                 if (netif_oper_up(dev))
 879                         operstate = IF_OPER_DORMANT;
 880                 break;
 881         }
 882
 883         if (dev->operstate != operstate) {
 884                 write_lock(&dev_base_lock);
 885                 dev->operstate = operstate;
 886                 write_unlock(&dev_base_lock);
 887                 netdev_state_change(dev);
 888         }
 889 }
 890
 891 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 892 {
 893         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 894                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 895 }
 896
 897 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 898                                            const struct ifinfomsg *ifm)
 899 {
 900         unsigned int flags = ifm->ifi_flags;
 901
 902         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 903         if (ifm->ifi_change)
 904                 flags = (flags & ifm->ifi_change) |
 905                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 906
 907         return flags;
 908 }
 909
 910 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 911                                  const struct rtnl_link_stats64 *b)
 912 {
 913         a->rx_packets = b->rx_packets;
 914         a->tx_packets = b->tx_packets;
 915         a->rx_bytes = b->rx_bytes;
 916         a->tx_bytes = b->tx_bytes;
 917         a->rx_errors = b->rx_errors;
 918         a->tx_errors = b->tx_errors;
 919         a->rx_dropped = b->rx_dropped;
 920         a->tx_dropped = b->tx_dropped;
 921
 922         a->multicast = b->multicast;
 923         a->collisions = b->collisions;
 924
 925         a->rx_length_errors = b->rx_length_errors;
 926         a->rx_over_errors = b->rx_over_errors;
 927         a->rx_crc_errors = b->rx_crc_errors;
 928         a->rx_frame_errors = b->rx_frame_errors;
 929         a->rx_fifo_errors = b->rx_fifo_errors;
 930         a->rx_missed_errors = b->rx_missed_errors;
 931
 932         a->tx_aborted_errors = b->tx_aborted_errors;
 933         a->tx_carrier_errors = b->tx_carrier_errors;
 934         a->tx_fifo_errors = b->tx_fifo_errors;
 935         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 936         a->tx_window_errors = b->tx_window_errors;
 937
 938         a->rx_compressed = b->rx_compressed;
 939         a->tx_compressed = b->tx_compressed;
 940
 941         a->rx_nohandler = b->rx_nohandler;
 942 }
 943
 944 /* All VF info */
 945 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 946                                    u32 ext_filter_mask)
 947 {
 948         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 949                 int num_vfs = dev_num_vf(dev->dev.parent);
 950                 size_t size = nla_total_size(0);
 951                 size += num_vfs *
 952                         (nla_total_size(0) +
 953                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 954                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 955                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 956                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 957                          nla_total_size(MAX_VLAN_LIST_LEN *
 958                                         sizeof(struct ifla_vf_vlan_info)) +
 959                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 960                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 961                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 962                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 963                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 964                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 965                          /* IFLA_VF_STATS_RX_PACKETS */
 966                          nla_total_size_64bit(sizeof(__u64)) +
 967                          /* IFLA_VF_STATS_TX_PACKETS */
 968                          nla_total_size_64bit(sizeof(__u64)) +
 969                          /* IFLA_VF_STATS_RX_BYTES */
 970                          nla_total_size_64bit(sizeof(__u64)) +
 971                          /* IFLA_VF_STATS_TX_BYTES */
 972                          nla_total_size_64bit(sizeof(__u64)) +
 973                          /* IFLA_VF_STATS_BROADCAST */
 974                          nla_total_size_64bit(sizeof(__u64)) +
 975                          /* IFLA_VF_STATS_MULTICAST */
 976                          nla_total_size_64bit(sizeof(__u64)) +
 977                          /* IFLA_VF_STATS_RX_DROPPED */
 978                          nla_total_size_64bit(sizeof(__u64)) +
 979                          /* IFLA_VF_STATS_TX_DROPPED */
 980                          nla_total_size_64bit(sizeof(__u64)) +
 981                          nla_total_size(sizeof(struct ifla_vf_trust)));
 982                 return size;
 983         } else
 984                 return 0;
 985 }
 986
 987 static size_t rtnl_port_size(const struct net_device *dev,
 988                              u32 ext_filter_mask)
 989 {
 990         size_t port_size = nla_total_size(4)            /* PORT_VF */
 991                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 992                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 993                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 994                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 995                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 996         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 997         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 998                 + port_size;
 999         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1000                 + port_size;
1001
1002         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1003             !(ext_filter_mask & RTEXT_FILTER_VF))
1004                 return 0;
1005         if (dev_num_vf(dev->dev.parent))
1006                 return port_self_size + vf_ports_size +
1007                         vf_port_size * dev_num_vf(dev->dev.parent);
1008         else
1009                 return port_self_size;
1010 }
1011
1012 static size_t rtnl_xdp_size(void)
1013 {
1014         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1015                           nla_total_size(1) +   /* XDP_ATTACHED */
1016                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1017                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1018
1019         return xdp_size;
1020 }
1021
1022 static size_t rtnl_prop_list_size(const struct net_device *dev)
1023 {
1024         struct netdev_name_node *name_node;
1025         size_t size;
1026
1027         if (list_empty(&dev->name_node->list))
1028                 return 0;
1029         size = nla_total_size(0);
1030         list_for_each_entry(name_node, &dev->name_node->list, list)
1031                 size += nla_total_size(ALTIFNAMSIZ);
1032         return size;
1033 }
1034
1035 static size_t rtnl_proto_down_size(const struct net_device *dev)
1036 {
1037         size_t size = nla_total_size(1);
1038
1039         if (dev->proto_down_reason)
1040                 size += nla_total_size(0) + nla_total_size(4);
1041
1042         return size;
1043 }
1044
1045 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1046 {
1047         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1048
1049         if (dev->devlink_port)
1050                 size += devlink_nl_port_handle_size(dev->devlink_port);
1051
1052         return size;
1053 }
1054
1055 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1056                                      u32 ext_filter_mask)
1057 {
1058         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1059                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1060                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1061                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1062                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1063                + nla_total_size(sizeof(struct rtnl_link_stats))
1064                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1065                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1066                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1067                + nla_total_size(4) /* IFLA_TXQLEN */
1068                + nla_total_size(4) /* IFLA_WEIGHT */
1069                + nla_total_size(4) /* IFLA_MTU */
1070                + nla_total_size(4) /* IFLA_LINK */
1071                + nla_total_size(4) /* IFLA_MASTER */
1072                + nla_total_size(1) /* IFLA_CARRIER */
1073                + nla_total_size(4) /* IFLA_PROMISCUITY */
1074                + nla_total_size(4) /* IFLA_ALLMULTI */
1075                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1076                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1077                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1078                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1079                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1080                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1081                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1082                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1083                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1084                + nla_total_size(1) /* IFLA_OPERSTATE */
1085                + nla_total_size(1) /* IFLA_LINKMODE */
1086                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1087                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1088                + nla_total_size(4) /* IFLA_GROUP */
1089                + nla_total_size(ext_filter_mask
1090                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1091                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1092                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1093                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1094                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1095                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1096                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1097                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1098                + rtnl_xdp_size() /* IFLA_XDP */
1099                + nla_total_size(4)  /* IFLA_EVENT */
1100                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1101                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1102                + rtnl_proto_down_size(dev)  /* proto down */
1103                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1104                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1105                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1106                + nla_total_size(4)  /* IFLA_MIN_MTU */
1107                + nla_total_size(4)  /* IFLA_MAX_MTU */
1108                + rtnl_prop_list_size(dev)
1109                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1110                + rtnl_devlink_port_size(dev)
1111                + 0;
1112 }
1113
1114 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1115 {
1116         struct nlattr *vf_ports;
1117         struct nlattr *vf_port;
1118         int vf;
1119         int err;
1120
1121         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1122         if (!vf_ports)
1123                 return -EMSGSIZE;
1124
1125         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1126                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1127                 if (!vf_port)
1128                         goto nla_put_failure;
1129                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1130                         goto nla_put_failure;
1131                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1132                 if (err == -EMSGSIZE)
1133                         goto nla_put_failure;
1134                 if (err) {
1135                         nla_nest_cancel(skb, vf_port);
1136                         continue;
1137                 }
1138                 nla_nest_end(skb, vf_port);
1139         }
1140
1141         nla_nest_end(skb, vf_ports);
1142
1143         return 0;
1144
1145 nla_put_failure:
1146         nla_nest_cancel(skb, vf_ports);
1147         return -EMSGSIZE;
1148 }
1149
1150 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1151 {
1152         struct nlattr *port_self;
1153         int err;
1154
1155         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1156         if (!port_self)
1157                 return -EMSGSIZE;
1158
1159         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1160         if (err) {
1161                 nla_nest_cancel(skb, port_self);
1162                 return (err == -EMSGSIZE) ? err : 0;
1163         }
1164
1165         nla_nest_end(skb, port_self);
1166
1167         return 0;
1168 }
1169
1170 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1171                           u32 ext_filter_mask)
1172 {
1173         int err;
1174
1175         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1176             !(ext_filter_mask & RTEXT_FILTER_VF))
1177                 return 0;
1178
1179         err = rtnl_port_self_fill(skb, dev);
1180         if (err)
1181                 return err;
1182
1183         if (dev_num_vf(dev->dev.parent)) {
1184                 err = rtnl_vf_ports_fill(skb, dev);
1185                 if (err)
1186                         return err;
1187         }
1188
1189         return 0;
1190 }
1191
1192 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1193 {
1194         int err;
1195         struct netdev_phys_item_id ppid;
1196
1197         err = dev_get_phys_port_id(dev, &ppid);
1198         if (err) {
1199                 if (err == -EOPNOTSUPP)
1200                         return 0;
1201                 return err;
1202         }
1203
1204         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1205                 return -EMSGSIZE;
1206
1207         return 0;
1208 }
1209
1210 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1211 {
1212         char name[IFNAMSIZ];
1213         int err;
1214
1215         err = dev_get_phys_port_name(dev, name, sizeof(name));
1216         if (err) {
1217                 if (err == -EOPNOTSUPP)
1218                         return 0;
1219                 return err;
1220         }
1221
1222         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1223                 return -EMSGSIZE;
1224
1225         return 0;
1226 }
1227
1228 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1229 {
1230         struct netdev_phys_item_id ppid = { };
1231         int err;
1232
1233         err = dev_get_port_parent_id(dev, &ppid, false);
1234         if (err) {
1235                 if (err == -EOPNOTSUPP)
1236                         return 0;
1237                 return err;
1238         }
1239
1240         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1241                 return -EMSGSIZE;
1242
1243         return 0;
1244 }
1245
1246 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1247                                               struct net_device *dev)
1248 {
1249         struct rtnl_link_stats64 *sp;
1250         struct nlattr *attr;
1251
1252         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1253                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1254         if (!attr)
1255                 return -EMSGSIZE;
1256
1257         sp = nla_data(attr);
1258         dev_get_stats(dev, sp);
1259
1260         attr = nla_reserve(skb, IFLA_STATS,
1261                            sizeof(struct rtnl_link_stats));
1262         if (!attr)
1263                 return -EMSGSIZE;
1264
1265         copy_rtnl_link_stats(nla_data(attr), sp);
1266
1267         return 0;
1268 }
1269
1270 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1271                                                struct net_device *dev,
1272                                                int vfs_num,
1273                                                struct nlattr *vfinfo)
1274 {
1275         struct ifla_vf_rss_query_en vf_rss_query_en;
1276         struct nlattr *vf, *vfstats, *vfvlanlist;
1277         struct ifla_vf_link_state vf_linkstate;
1278         struct ifla_vf_vlan_info vf_vlan_info;
1279         struct ifla_vf_spoofchk vf_spoofchk;
1280         struct ifla_vf_tx_rate vf_tx_rate;
1281         struct ifla_vf_stats vf_stats;
1282         struct ifla_vf_trust vf_trust;
1283         struct ifla_vf_vlan vf_vlan;
1284         struct ifla_vf_rate vf_rate;
1285         struct ifla_vf_mac vf_mac;
1286         struct ifla_vf_broadcast vf_broadcast;
1287         struct ifla_vf_info ivi;
1288         struct ifla_vf_guid node_guid;
1289         struct ifla_vf_guid port_guid;
1290
1291         memset(&ivi, 0, sizeof(ivi));
1292
1293         /* Not all SR-IOV capable drivers support the
1294          * spoofcheck and "RSS query enable" query.  Preset to
1295          * -1 so the user space tool can detect that the driver
1296          * didn't report anything.
1297          */
1298         ivi.spoofchk = -1;
1299         ivi.rss_query_en = -1;
1300         ivi.trusted = -1;
1301         /* The default value for VF link state is "auto"
1302          * IFLA_VF_LINK_STATE_AUTO which equals zero
1303          */
1304         ivi.linkstate = 0;
1305         /* VLAN Protocol by default is 802.1Q */
1306         ivi.vlan_proto = htons(ETH_P_8021Q);
1307         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1308                 return 0;
1309
1310         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1311         memset(&node_guid, 0, sizeof(node_guid));
1312         memset(&port_guid, 0, sizeof(port_guid));
1313
1314         vf_mac.vf =
1315                 vf_vlan.vf =
1316                 vf_vlan_info.vf =
1317                 vf_rate.vf =
1318                 vf_tx_rate.vf =
1319                 vf_spoofchk.vf =
1320                 vf_linkstate.vf =
1321                 vf_rss_query_en.vf =
1322                 vf_trust.vf =
1323                 node_guid.vf =
1324                 port_guid.vf = ivi.vf;
1325
1326         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1327         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1328         vf_vlan.vlan = ivi.vlan;
1329         vf_vlan.qos = ivi.qos;
1330         vf_vlan_info.vlan = ivi.vlan;
1331         vf_vlan_info.qos = ivi.qos;
1332         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1333         vf_tx_rate.rate = ivi.max_tx_rate;
1334         vf_rate.min_tx_rate = ivi.min_tx_rate;
1335         vf_rate.max_tx_rate = ivi.max_tx_rate;
1336         vf_spoofchk.setting = ivi.spoofchk;
1337         vf_linkstate.link_state = ivi.linkstate;
1338         vf_rss_query_en.setting = ivi.rss_query_en;
1339         vf_trust.setting = ivi.trusted;
1340         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1341         if (!vf)
1342                 goto nla_put_vfinfo_failure;
1343         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1344             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1345             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1346             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1347                     &vf_rate) ||
1348             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1349                     &vf_tx_rate) ||
1350             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1351                     &vf_spoofchk) ||
1352             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1353                     &vf_linkstate) ||
1354             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1355                     sizeof(vf_rss_query_en),
1356                     &vf_rss_query_en) ||
1357             nla_put(skb, IFLA_VF_TRUST,
1358                     sizeof(vf_trust), &vf_trust))
1359                 goto nla_put_vf_failure;
1360
1361         if (dev->netdev_ops->ndo_get_vf_guid &&
1362             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1363                                               &port_guid)) {
1364                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1365                             &node_guid) ||
1366                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1367                             &port_guid))
1368                         goto nla_put_vf_failure;
1369         }
1370         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1371         if (!vfvlanlist)
1372                 goto nla_put_vf_failure;
1373         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1374                     &vf_vlan_info)) {
1375                 nla_nest_cancel(skb, vfvlanlist);
1376                 goto nla_put_vf_failure;
1377         }
1378         nla_nest_end(skb, vfvlanlist);
1379         memset(&vf_stats, 0, sizeof(vf_stats));
1380         if (dev->netdev_ops->ndo_get_vf_stats)
1381                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1382                                                 &vf_stats);
1383         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1384         if (!vfstats)
1385                 goto nla_put_vf_failure;
1386         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1387                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1388             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1389                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1390             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1391                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1392             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1393                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1394             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1395                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1396             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1397                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1398             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1399                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1400             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1401                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1402                 nla_nest_cancel(skb, vfstats);
1403                 goto nla_put_vf_failure;
1404         }
1405         nla_nest_end(skb, vfstats);
1406         nla_nest_end(skb, vf);
1407         return 0;
1408
1409 nla_put_vf_failure:
1410         nla_nest_cancel(skb, vf);
1411 nla_put_vfinfo_failure:
1412         nla_nest_cancel(skb, vfinfo);
1413         return -EMSGSIZE;
1414 }
1415
1416 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1417                                            struct net_device *dev,
1418                                            u32 ext_filter_mask)
1419 {
1420         struct nlattr *vfinfo;
1421         int i, num_vfs;
1422
1423         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1424                 return 0;
1425
1426         num_vfs = dev_num_vf(dev->dev.parent);
1427         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1428                 return -EMSGSIZE;
1429
1430         if (!dev->netdev_ops->ndo_get_vf_config)
1431                 return 0;
1432
1433         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1434         if (!vfinfo)
1435                 return -EMSGSIZE;
1436
1437         for (i = 0; i < num_vfs; i++) {
1438                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1439                         return -EMSGSIZE;
1440         }
1441
1442         nla_nest_end(skb, vfinfo);
1443         return 0;
1444 }
1445
1446 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1447 {
1448         struct rtnl_link_ifmap map;
1449
1450         memset(&map, 0, sizeof(map));
1451         map.mem_start   = dev->mem_start;
1452         map.mem_end     = dev->mem_end;
1453         map.base_addr   = dev->base_addr;
1454         map.irq         = dev->irq;
1455         map.dma         = dev->dma;
1456         map.port        = dev->if_port;
1457
1458         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1459                 return -EMSGSIZE;
1460
1461         return 0;
1462 }
1463
1464 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1465 {
1466         const struct bpf_prog *generic_xdp_prog;
1467
1468         ASSERT_RTNL();
1469
1470         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1471         if (!generic_xdp_prog)
1472                 return 0;
1473         return generic_xdp_prog->aux->id;
1474 }
1475
1476 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1477 {
1478         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1479 }
1480
1481 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1482 {
1483         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1484 }
1485
1486 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1487                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1488                                u32 (*get_prog_id)(struct net_device *dev))
1489 {
1490         u32 curr_id;
1491         int err;
1492
1493         curr_id = get_prog_id(dev);
1494         if (!curr_id)
1495                 return 0;
1496
1497         *prog_id = curr_id;
1498         err = nla_put_u32(skb, attr, curr_id);
1499         if (err)
1500                 return err;
1501
1502         if (*mode != XDP_ATTACHED_NONE)
1503                 *mode = XDP_ATTACHED_MULTI;
1504         else
1505                 *mode = tgt_mode;
1506
1507         return 0;
1508 }
1509
1510 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1511 {
1512         struct nlattr *xdp;
1513         u32 prog_id;
1514         int err;
1515         u8 mode;
1516
1517         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1518         if (!xdp)
1519                 return -EMSGSIZE;
1520
1521         prog_id = 0;
1522         mode = XDP_ATTACHED_NONE;
1523         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1524                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1525         if (err)
1526                 goto err_cancel;
1527         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1528                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1529         if (err)
1530                 goto err_cancel;
1531         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1532                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1533         if (err)
1534                 goto err_cancel;
1535
1536         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1537         if (err)
1538                 goto err_cancel;
1539
1540         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1541                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1542                 if (err)
1543                         goto err_cancel;
1544         }
1545
1546         nla_nest_end(skb, xdp);
1547         return 0;
1548
1549 err_cancel:
1550         nla_nest_cancel(skb, xdp);
1551         return err;
1552 }
1553
1554 static u32 rtnl_get_event(unsigned long event)
1555 {
1556         u32 rtnl_event_type = IFLA_EVENT_NONE;
1557
1558         switch (event) {
1559         case NETDEV_REBOOT:
1560                 rtnl_event_type = IFLA_EVENT_REBOOT;
1561                 break;
1562         case NETDEV_FEAT_CHANGE:
1563                 rtnl_event_type = IFLA_EVENT_FEATURES;
1564                 break;
1565         case NETDEV_BONDING_FAILOVER:
1566                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1567                 break;
1568         case NETDEV_NOTIFY_PEERS:
1569                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1570                 break;
1571         case NETDEV_RESEND_IGMP:
1572                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1573                 break;
1574         case NETDEV_CHANGEINFODATA:
1575                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1576                 break;
1577         default:
1578                 break;
1579         }
1580
1581         return rtnl_event_type;
1582 }
1583
1584 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1585 {
1586         const struct net_device *upper_dev;
1587         int ret = 0;
1588
1589         rcu_read_lock();
1590
1591         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1592         if (upper_dev)
1593                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1594
1595         rcu_read_unlock();
1596         return ret;
1597 }
1598
1599 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1600                           bool force)
1601 {
1602         int ifindex = dev_get_iflink(dev);
1603
1604         if (force || dev->ifindex != ifindex)
1605                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1606
1607         return 0;
1608 }
1609
1610 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1611                                               struct net_device *dev)
1612 {
1613         char buf[IFALIASZ];
1614         int ret;
1615
1616         ret = dev_get_alias(dev, buf, sizeof(buf));
1617         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1618 }
1619
1620 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1621                                   const struct net_device *dev,
1622                                   struct net *src_net, gfp_t gfp)
1623 {
1624         bool put_iflink = false;
1625
1626         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1627                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1628
1629                 if (!net_eq(dev_net(dev), link_net)) {
1630                         int id = peernet2id_alloc(src_net, link_net, gfp);
1631
1632                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1633                                 return -EMSGSIZE;
1634
1635                         put_iflink = true;
1636                 }
1637         }
1638
1639         return nla_put_iflink(skb, dev, put_iflink);
1640 }
1641
1642 static int rtnl_fill_link_af(struct sk_buff *skb,
1643                              const struct net_device *dev,
1644                              u32 ext_filter_mask)
1645 {
1646         const struct rtnl_af_ops *af_ops;
1647         struct nlattr *af_spec;
1648
1649         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1650         if (!af_spec)
1651                 return -EMSGSIZE;
1652
1653         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1654                 struct nlattr *af;
1655                 int err;
1656
1657                 if (!af_ops->fill_link_af)
1658                         continue;
1659
1660                 af = nla_nest_start_noflag(skb, af_ops->family);
1661                 if (!af)
1662                         return -EMSGSIZE;
1663
1664                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1665                 /*
1666                  * Caller may return ENODATA to indicate that there
1667                  * was no data to be dumped. This is not an error, it
1668                  * means we should trim the attribute header and
1669                  * continue.
1670                  */
1671                 if (err == -ENODATA)
1672                         nla_nest_cancel(skb, af);
1673                 else if (err < 0)
1674                         return -EMSGSIZE;
1675
1676                 nla_nest_end(skb, af);
1677         }
1678
1679         nla_nest_end(skb, af_spec);
1680         return 0;
1681 }
1682
1683 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1684                                  const struct net_device *dev)
1685 {
1686         struct netdev_name_node *name_node;
1687         int count = 0;
1688
1689         list_for_each_entry(name_node, &dev->name_node->list, list) {
1690                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1691                         return -EMSGSIZE;
1692                 count++;
1693         }
1694         return count;
1695 }
1696
1697 static int rtnl_fill_prop_list(struct sk_buff *skb,
1698                                const struct net_device *dev)
1699 {
1700         struct nlattr *prop_list;
1701         int ret;
1702
1703         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1704         if (!prop_list)
1705                 return -EMSGSIZE;
1706
1707         ret = rtnl_fill_alt_ifnames(skb, dev);
1708         if (ret <= 0)
1709                 goto nest_cancel;
1710
1711         nla_nest_end(skb, prop_list);
1712         return 0;
1713
1714 nest_cancel:
1715         nla_nest_cancel(skb, prop_list);
1716         return ret;
1717 }
1718
1719 static int rtnl_fill_proto_down(struct sk_buff *skb,
1720                                 const struct net_device *dev)
1721 {
1722         struct nlattr *pr;
1723         u32 preason;
1724
1725         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1726                 goto nla_put_failure;
1727
1728         preason = dev->proto_down_reason;
1729         if (!preason)
1730                 return 0;
1731
1732         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1733         if (!pr)
1734                 return -EMSGSIZE;
1735
1736         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1737                 nla_nest_cancel(skb, pr);
1738                 goto nla_put_failure;
1739         }
1740
1741         nla_nest_end(skb, pr);
1742         return 0;
1743
1744 nla_put_failure:
1745         return -EMSGSIZE;
1746 }
1747
1748 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1749                                   const struct net_device *dev)
1750 {
1751         struct nlattr *devlink_port_nest;
1752         int ret;
1753
1754         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1755         if (!devlink_port_nest)
1756                 return -EMSGSIZE;
1757
1758         if (dev->devlink_port) {
1759                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1760                 if (ret < 0)
1761                         goto nest_cancel;
1762         }
1763
1764         nla_nest_end(skb, devlink_port_nest);
1765         return 0;
1766
1767 nest_cancel:
1768         nla_nest_cancel(skb, devlink_port_nest);
1769         return ret;
1770 }
1771
1772 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1773                             struct net_device *dev, struct net *src_net,
1774                             int type, u32 pid, u32 seq, u32 change,
1775                             unsigned int flags, u32 ext_filter_mask,
1776                             u32 event, int *new_nsid, int new_ifindex,
1777                             int tgt_netnsid, gfp_t gfp)
1778 {
1779         struct ifinfomsg *ifm;
1780         struct nlmsghdr *nlh;
1781         struct Qdisc *qdisc;
1782
1783         ASSERT_RTNL();
1784         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1785         if (nlh == NULL)
1786                 return -EMSGSIZE;
1787
1788         ifm = nlmsg_data(nlh);
1789         ifm->ifi_family = AF_UNSPEC;
1790         ifm->__ifi_pad = 0;
1791         ifm->ifi_type = dev->type;
1792         ifm->ifi_index = dev->ifindex;
1793         ifm->ifi_flags = dev_get_flags(dev);
1794         ifm->ifi_change = change;
1795
1796         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1797                 goto nla_put_failure;
1798
1799         qdisc = rtnl_dereference(dev->qdisc);
1800         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1801             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1802             nla_put_u8(skb, IFLA_OPERSTATE,
1803                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1804             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1805             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1806             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1807             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1808             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1809             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1810             nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1811             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1812             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1813             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1814             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1815             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE, dev->gso_ipv4_max_size) ||
1816             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE, dev->gro_ipv4_max_size) ||
1817             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1818             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1819 #ifdef CONFIG_RPS
1820             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1821 #endif
1822             put_master_ifindex(skb, dev) ||
1823             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1824             (qdisc &&
1825              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1826             nla_put_ifalias(skb, dev) ||
1827             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1828                         atomic_read(&dev->carrier_up_count) +
1829                         atomic_read(&dev->carrier_down_count)) ||
1830             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1831                         atomic_read(&dev->carrier_up_count)) ||
1832             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1833                         atomic_read(&dev->carrier_down_count)))
1834                 goto nla_put_failure;
1835
1836         if (rtnl_fill_proto_down(skb, dev))
1837                 goto nla_put_failure;
1838
1839         if (event != IFLA_EVENT_NONE) {
1840                 if (nla_put_u32(skb, IFLA_EVENT, event))
1841                         goto nla_put_failure;
1842         }
1843
1844         if (rtnl_fill_link_ifmap(skb, dev))
1845                 goto nla_put_failure;
1846
1847         if (dev->addr_len) {
1848                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1849                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1850                         goto nla_put_failure;
1851         }
1852
1853         if (rtnl_phys_port_id_fill(skb, dev))
1854                 goto nla_put_failure;
1855
1856         if (rtnl_phys_port_name_fill(skb, dev))
1857                 goto nla_put_failure;
1858
1859         if (rtnl_phys_switch_id_fill(skb, dev))
1860                 goto nla_put_failure;
1861
1862         if (rtnl_fill_stats(skb, dev))
1863                 goto nla_put_failure;
1864
1865         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1866                 goto nla_put_failure;
1867
1868         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1869                 goto nla_put_failure;
1870
1871         if (rtnl_xdp_fill(skb, dev))
1872                 goto nla_put_failure;
1873
1874         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1875                 if (rtnl_link_fill(skb, dev) < 0)
1876                         goto nla_put_failure;
1877         }
1878
1879         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1880                 goto nla_put_failure;
1881
1882         if (new_nsid &&
1883             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1884                 goto nla_put_failure;
1885         if (new_ifindex &&
1886             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1887                 goto nla_put_failure;
1888
1889         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1890             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1891                 goto nla_put_failure;
1892
1893         rcu_read_lock();
1894         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1895                 goto nla_put_failure_rcu;
1896         rcu_read_unlock();
1897
1898         if (rtnl_fill_prop_list(skb, dev))
1899                 goto nla_put_failure;
1900
1901         if (dev->dev.parent &&
1902             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1903                            dev_name(dev->dev.parent)))
1904                 goto nla_put_failure;
1905
1906         if (dev->dev.parent && dev->dev.parent->bus &&
1907             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1908                            dev->dev.parent->bus->name))
1909                 goto nla_put_failure;
1910
1911         if (rtnl_fill_devlink_port(skb, dev))
1912                 goto nla_put_failure;
1913
1914         nlmsg_end(skb, nlh);
1915         return 0;
1916
1917 nla_put_failure_rcu:
1918         rcu_read_unlock();
1919 nla_put_failure:
1920         nlmsg_cancel(skb, nlh);
1921         return -EMSGSIZE;
1922 }
1923
1924 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1925         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1926         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1927         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1928         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1929         [IFLA_MTU]              = { .type = NLA_U32 },
1930         [IFLA_LINK]             = { .type = NLA_U32 },
1931         [IFLA_MASTER]           = { .type = NLA_U32 },
1932         [IFLA_CARRIER]          = { .type = NLA_U8 },
1933         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1934         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1935         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1936         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1937         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1938         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1939         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1940         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1941          * allow 0-length string (needed to remove an alias).
1942          */
1943         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1944         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1945         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1946         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1947         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1948         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1949         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1950         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1951         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1952         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1953         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1954         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1955         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1956         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1957         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1958         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1959         [IFLA_XDP]              = { .type = NLA_NESTED },
1960         [IFLA_EVENT]            = { .type = NLA_U32 },
1961         [IFLA_GROUP]            = { .type = NLA_U32 },
1962         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1963         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1964         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1965         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1966         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1967         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1968         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1969                                     .len = ALTIFNAMSIZ - 1 },
1970         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1971         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1972         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1973         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1974         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
1975         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
1976         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
1977         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
1978         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1979         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1980 };
1981
1982 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1983         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1984         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1985         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1986         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1987 };
1988
1989 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1990         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1991         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1992         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1993         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1994         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1995         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1996         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1997         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1998         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1999         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2000         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2001         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2002         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2003 };
2004
2005 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2006         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2007         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2008                                     .len = PORT_PROFILE_MAX },
2009         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2010                                       .len = PORT_UUID_MAX },
2011         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2012                                     .len = PORT_UUID_MAX },
2013         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2014         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2015
2016         /* Unused, but we need to keep it here since user space could
2017          * fill it. It's also broken with regard to NLA_BINARY use in
2018          * combination with structs.
2019          */
2020         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2021                                     .len = sizeof(struct ifla_port_vsi) },
2022 };
2023
2024 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2025         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2026         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2027         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2028         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2029         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2030         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2031 };
2032
2033 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2034 {
2035         const struct rtnl_link_ops *ops = NULL;
2036         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2037
2038         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2039                 return NULL;
2040
2041         if (linfo[IFLA_INFO_KIND]) {
2042                 char kind[MODULE_NAME_LEN];
2043
2044                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2045                 ops = rtnl_link_ops_get(kind);
2046         }
2047
2048         return ops;
2049 }
2050
2051 static bool link_master_filtered(struct net_device *dev, int master_idx)
2052 {
2053         struct net_device *master;
2054
2055         if (!master_idx)
2056                 return false;
2057
2058         master = netdev_master_upper_dev_get(dev);
2059
2060         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2061          * another invalid value for ifindex to denote "no master".
2062          */
2063         if (master_idx == -1)
2064                 return !!master;
2065
2066         if (!master || master->ifindex != master_idx)
2067                 return true;
2068
2069         return false;
2070 }
2071
2072 static bool link_kind_filtered(const struct net_device *dev,
2073                                const struct rtnl_link_ops *kind_ops)
2074 {
2075         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2076                 return true;
2077
2078         return false;
2079 }
2080
2081 static bool link_dump_filtered(struct net_device *dev,
2082                                int master_idx,
2083                                const struct rtnl_link_ops *kind_ops)
2084 {
2085         if (link_master_filtered(dev, master_idx) ||
2086             link_kind_filtered(dev, kind_ops))
2087                 return true;
2088
2089         return false;
2090 }
2091
2092 /**
2093  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2094  * @sk: netlink socket
2095  * @netnsid: network namespace identifier
2096  *
2097  * Returns the network namespace identified by netnsid on success or an error
2098  * pointer on failure.
2099  */
2100 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2101 {
2102         struct net *net;
2103
2104         net = get_net_ns_by_id(sock_net(sk), netnsid);
2105         if (!net)
2106                 return ERR_PTR(-EINVAL);
2107
2108         /* For now, the caller is required to have CAP_NET_ADMIN in
2109          * the user namespace owning the target net ns.
2110          */
2111         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2112                 put_net(net);
2113                 return ERR_PTR(-EACCES);
2114         }
2115         return net;
2116 }
2117 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2118
2119 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2120                                       bool strict_check, struct nlattr **tb,
2121                                       struct netlink_ext_ack *extack)
2122 {
2123         int hdrlen;
2124
2125         if (strict_check) {
2126                 struct ifinfomsg *ifm;
2127
2128                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2129                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2130                         return -EINVAL;
2131                 }
2132
2133                 ifm = nlmsg_data(nlh);
2134                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2135                     ifm->ifi_change) {
2136                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2137                         return -EINVAL;
2138                 }
2139                 if (ifm->ifi_index) {
2140                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2141                         return -EINVAL;
2142                 }
2143
2144                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2145                                                      IFLA_MAX, ifla_policy,
2146                                                      extack);
2147         }
2148
2149         /* A hack to preserve kernel<->userspace interface.
2150          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2151          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2152          * what iproute2 < v3.9.0 used.
2153          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2154          * attribute, its netlink message is shorter than struct ifinfomsg.
2155          */
2156         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2157                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2158
2159         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2160                                       extack);
2161 }
2162
2163 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2164 {
2165         struct netlink_ext_ack *extack = cb->extack;
2166         const struct nlmsghdr *nlh = cb->nlh;
2167         struct net *net = sock_net(skb->sk);
2168         struct net *tgt_net = net;
2169         int h, s_h;
2170         int idx = 0, s_idx;
2171         struct net_device *dev;
2172         struct hlist_head *head;
2173         struct nlattr *tb[IFLA_MAX+1];
2174         u32 ext_filter_mask = 0;
2175         const struct rtnl_link_ops *kind_ops = NULL;
2176         unsigned int flags = NLM_F_MULTI;
2177         int master_idx = 0;
2178         int netnsid = -1;
2179         int err, i;
2180
2181         s_h = cb->args[0];
2182         s_idx = cb->args[1];
2183
2184         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2185         if (err < 0) {
2186                 if (cb->strict_check)
2187                         return err;
2188
2189                 goto walk_entries;
2190         }
2191
2192         for (i = 0; i <= IFLA_MAX; ++i) {
2193                 if (!tb[i])
2194                         continue;
2195
2196                 /* new attributes should only be added with strict checking */
2197                 switch (i) {
2198                 case IFLA_TARGET_NETNSID:
2199                         netnsid = nla_get_s32(tb[i]);
2200                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2201                         if (IS_ERR(tgt_net)) {
2202                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2203                                 return PTR_ERR(tgt_net);
2204                         }
2205                         break;
2206                 case IFLA_EXT_MASK:
2207                         ext_filter_mask = nla_get_u32(tb[i]);
2208                         break;
2209                 case IFLA_MASTER:
2210                         master_idx = nla_get_u32(tb[i]);
2211                         break;
2212                 case IFLA_LINKINFO:
2213                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2214                         break;
2215                 default:
2216                         if (cb->strict_check) {
2217                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2218                                 return -EINVAL;
2219                         }
2220                 }
2221         }
2222
2223         if (master_idx || kind_ops)
2224                 flags |= NLM_F_DUMP_FILTERED;
2225
2226 walk_entries:
2227         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2228                 idx = 0;
2229                 head = &tgt_net->dev_index_head[h];
2230                 hlist_for_each_entry(dev, head, index_hlist) {
2231                         if (link_dump_filtered(dev, master_idx, kind_ops))
2232                                 goto cont;
2233                         if (idx < s_idx)
2234                                 goto cont;
2235                         err = rtnl_fill_ifinfo(skb, dev, net,
2236                                                RTM_NEWLINK,
2237                                                NETLINK_CB(cb->skb).portid,
2238                                                nlh->nlmsg_seq, 0, flags,
2239                                                ext_filter_mask, 0, NULL, 0,
2240                                                netnsid, GFP_KERNEL);
2241
2242                         if (err < 0) {
2243                                 if (likely(skb->len))
2244                                         goto out;
2245
2246                                 goto out_err;
2247                         }
2248 cont:
2249                         idx++;
2250                 }
2251         }
2252 out:
2253         err = skb->len;
2254 out_err:
2255         cb->args[1] = idx;
2256         cb->args[0] = h;
2257         cb->seq = tgt_net->dev_base_seq;
2258         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2259         if (netnsid >= 0)
2260                 put_net(tgt_net);
2261
2262         return err;
2263 }
2264
2265 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2266                         struct netlink_ext_ack *exterr)
2267 {
2268         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2269                                     exterr);
2270 }
2271 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2272
2273 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2274 {
2275         struct net *net;
2276         /* Examine the link attributes and figure out which
2277          * network namespace we are talking about.
2278          */
2279         if (tb[IFLA_NET_NS_PID])
2280                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2281         else if (tb[IFLA_NET_NS_FD])
2282                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2283         else
2284                 net = get_net(src_net);
2285         return net;
2286 }
2287 EXPORT_SYMBOL(rtnl_link_get_net);
2288
2289 /* Figure out which network namespace we are talking about by
2290  * examining the link attributes in the following order:
2291  *
2292  * 1. IFLA_NET_NS_PID
2293  * 2. IFLA_NET_NS_FD
2294  * 3. IFLA_TARGET_NETNSID
2295  */
2296 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2297                                                struct nlattr *tb[])
2298 {
2299         struct net *net;
2300
2301         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2302                 return rtnl_link_get_net(src_net, tb);
2303
2304         if (!tb[IFLA_TARGET_NETNSID])
2305                 return get_net(src_net);
2306
2307         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2308         if (!net)
2309                 return ERR_PTR(-EINVAL);
2310
2311         return net;
2312 }
2313
2314 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2315                                              struct net *src_net,
2316                                              struct nlattr *tb[], int cap)
2317 {
2318         struct net *net;
2319
2320         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2321         if (IS_ERR(net))
2322                 return net;
2323
2324         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2325                 put_net(net);
2326                 return ERR_PTR(-EPERM);
2327         }
2328
2329         return net;
2330 }
2331
2332 /* Verify that rtnetlink requests do not pass additional properties
2333  * potentially referring to different network namespaces.
2334  */
2335 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2336                                     struct netlink_ext_ack *extack,
2337                                     bool netns_id_only)
2338 {
2339
2340         if (netns_id_only) {
2341                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2342                         return 0;
2343
2344                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2345                 return -EOPNOTSUPP;
2346         }
2347
2348         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2349                 goto invalid_attr;
2350
2351         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2352                 goto invalid_attr;
2353
2354         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2355                 goto invalid_attr;
2356
2357         return 0;
2358
2359 invalid_attr:
2360         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2361         return -EINVAL;
2362 }
2363
2364 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2365                              int max_tx_rate)
2366 {
2367         const struct net_device_ops *ops = dev->netdev_ops;
2368
2369         if (!ops->ndo_set_vf_rate)
2370                 return -EOPNOTSUPP;
2371         if (max_tx_rate && max_tx_rate < min_tx_rate)
2372                 return -EINVAL;
2373
2374         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2375 }
2376
2377 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2378                             struct netlink_ext_ack *extack)
2379 {
2380         if (dev) {
2381                 if (tb[IFLA_ADDRESS] &&
2382                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2383                         return -EINVAL;
2384
2385                 if (tb[IFLA_BROADCAST] &&
2386                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2387                         return -EINVAL;
2388         }
2389
2390         if (tb[IFLA_AF_SPEC]) {
2391                 struct nlattr *af;
2392                 int rem, err;
2393
2394                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2395                         const struct rtnl_af_ops *af_ops;
2396
2397                         af_ops = rtnl_af_lookup(nla_type(af));
2398                         if (!af_ops)
2399                                 return -EAFNOSUPPORT;
2400
2401                         if (!af_ops->set_link_af)
2402                                 return -EOPNOTSUPP;
2403
2404                         if (af_ops->validate_link_af) {
2405                                 err = af_ops->validate_link_af(dev, af, extack);
2406                                 if (err < 0)
2407                                         return err;
2408                         }
2409                 }
2410         }
2411
2412         return 0;
2413 }
2414
2415 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2416                                   int guid_type)
2417 {
2418         const struct net_device_ops *ops = dev->netdev_ops;
2419
2420         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2421 }
2422
2423 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2424 {
2425         if (dev->type != ARPHRD_INFINIBAND)
2426                 return -EOPNOTSUPP;
2427
2428         return handle_infiniband_guid(dev, ivt, guid_type);
2429 }
2430
2431 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2432 {
2433         const struct net_device_ops *ops = dev->netdev_ops;
2434         int err = -EINVAL;
2435
2436         if (tb[IFLA_VF_MAC]) {
2437                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2438
2439                 if (ivm->vf >= INT_MAX)
2440                         return -EINVAL;
2441                 err = -EOPNOTSUPP;
2442                 if (ops->ndo_set_vf_mac)
2443                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2444                                                   ivm->mac);
2445                 if (err < 0)
2446                         return err;
2447         }
2448
2449         if (tb[IFLA_VF_VLAN]) {
2450                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2451
2452                 if (ivv->vf >= INT_MAX)
2453                         return -EINVAL;
2454                 err = -EOPNOTSUPP;
2455                 if (ops->ndo_set_vf_vlan)
2456                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2457                                                    ivv->qos,
2458                                                    htons(ETH_P_8021Q));
2459                 if (err < 0)
2460                         return err;
2461         }
2462
2463         if (tb[IFLA_VF_VLAN_LIST]) {
2464                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2465                 struct nlattr *attr;
2466                 int rem, len = 0;
2467
2468                 err = -EOPNOTSUPP;
2469                 if (!ops->ndo_set_vf_vlan)
2470                         return err;
2471
2472                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2473                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2474                             nla_len(attr) < NLA_HDRLEN) {
2475                                 return -EINVAL;
2476                         }
2477                         if (len >= MAX_VLAN_LIST_LEN)
2478                                 return -EOPNOTSUPP;
2479                         ivvl[len] = nla_data(attr);
2480
2481                         len++;
2482                 }
2483                 if (len == 0)
2484                         return -EINVAL;
2485
2486                 if (ivvl[0]->vf >= INT_MAX)
2487                         return -EINVAL;
2488                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2489                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2490                 if (err < 0)
2491                         return err;
2492         }
2493
2494         if (tb[IFLA_VF_TX_RATE]) {
2495                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2496                 struct ifla_vf_info ivf;
2497
2498                 if (ivt->vf >= INT_MAX)
2499                         return -EINVAL;
2500                 err = -EOPNOTSUPP;
2501                 if (ops->ndo_get_vf_config)
2502                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2503                 if (err < 0)
2504                         return err;
2505
2506                 err = rtnl_set_vf_rate(dev, ivt->vf,
2507                                        ivf.min_tx_rate, ivt->rate);
2508                 if (err < 0)
2509                         return err;
2510         }
2511
2512         if (tb[IFLA_VF_RATE]) {
2513                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2514
2515                 if (ivt->vf >= INT_MAX)
2516                         return -EINVAL;
2517
2518                 err = rtnl_set_vf_rate(dev, ivt->vf,
2519                                        ivt->min_tx_rate, ivt->max_tx_rate);
2520                 if (err < 0)
2521                         return err;
2522         }
2523
2524         if (tb[IFLA_VF_SPOOFCHK]) {
2525                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2526
2527                 if (ivs->vf >= INT_MAX)
2528                         return -EINVAL;
2529                 err = -EOPNOTSUPP;
2530                 if (ops->ndo_set_vf_spoofchk)
2531                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2532                                                        ivs->setting);
2533                 if (err < 0)
2534                         return err;
2535         }
2536
2537         if (tb[IFLA_VF_LINK_STATE]) {
2538                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2539
2540                 if (ivl->vf >= INT_MAX)
2541                         return -EINVAL;
2542                 err = -EOPNOTSUPP;
2543                 if (ops->ndo_set_vf_link_state)
2544                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2545                                                          ivl->link_state);
2546                 if (err < 0)
2547                         return err;
2548         }
2549
2550         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2551                 struct ifla_vf_rss_query_en *ivrssq_en;
2552
2553                 err = -EOPNOTSUPP;
2554                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2555                 if (ivrssq_en->vf >= INT_MAX)
2556                         return -EINVAL;
2557                 if (ops->ndo_set_vf_rss_query_en)
2558                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2559                                                            ivrssq_en->setting);
2560                 if (err < 0)
2561                         return err;
2562         }
2563
2564         if (tb[IFLA_VF_TRUST]) {
2565                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2566
2567                 if (ivt->vf >= INT_MAX)
2568                         return -EINVAL;
2569                 err = -EOPNOTSUPP;
2570                 if (ops->ndo_set_vf_trust)
2571                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2572                 if (err < 0)
2573                         return err;
2574         }
2575
2576         if (tb[IFLA_VF_IB_NODE_GUID]) {
2577                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2578
2579                 if (ivt->vf >= INT_MAX)
2580                         return -EINVAL;
2581                 if (!ops->ndo_set_vf_guid)
2582                         return -EOPNOTSUPP;
2583                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2584         }
2585
2586         if (tb[IFLA_VF_IB_PORT_GUID]) {
2587                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2588
2589                 if (ivt->vf >= INT_MAX)
2590                         return -EINVAL;
2591                 if (!ops->ndo_set_vf_guid)
2592                         return -EOPNOTSUPP;
2593
2594                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2595         }
2596
2597         return err;
2598 }
2599
2600 static int do_set_master(struct net_device *dev, int ifindex,
2601                          struct netlink_ext_ack *extack)
2602 {
2603         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2604         const struct net_device_ops *ops;
2605         int err;
2606
2607         if (upper_dev) {
2608                 if (upper_dev->ifindex == ifindex)
2609                         return 0;
2610                 ops = upper_dev->netdev_ops;
2611                 if (ops->ndo_del_slave) {
2612                         err = ops->ndo_del_slave(upper_dev, dev);
2613                         if (err)
2614                                 return err;
2615                 } else {
2616                         return -EOPNOTSUPP;
2617                 }
2618         }
2619
2620         if (ifindex) {
2621                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2622                 if (!upper_dev)
2623                         return -EINVAL;
2624                 ops = upper_dev->netdev_ops;
2625                 if (ops->ndo_add_slave) {
2626                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2627                         if (err)
2628                                 return err;
2629                 } else {
2630                         return -EOPNOTSUPP;
2631                 }
2632         }
2633         return 0;
2634 }
2635
2636 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2637         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2638         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2639 };
2640
2641 static int do_set_proto_down(struct net_device *dev,
2642                              struct nlattr *nl_proto_down,
2643                              struct nlattr *nl_proto_down_reason,
2644                              struct netlink_ext_ack *extack)
2645 {
2646         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2647         unsigned long mask = 0;
2648         u32 value;
2649         bool proto_down;
2650         int err;
2651
2652         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2653                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2654                 return -EOPNOTSUPP;
2655         }
2656
2657         if (nl_proto_down_reason) {
2658                 err = nla_parse_nested_deprecated(pdreason,
2659                                                   IFLA_PROTO_DOWN_REASON_MAX,
2660                                                   nl_proto_down_reason,
2661                                                   ifla_proto_down_reason_policy,
2662                                                   NULL);
2663                 if (err < 0)
2664                         return err;
2665
2666                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2667                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2668                         return -EINVAL;
2669                 }
2670
2671                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2672
2673                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2674                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2675
2676                 dev_change_proto_down_reason(dev, mask, value);
2677         }
2678
2679         if (nl_proto_down) {
2680                 proto_down = nla_get_u8(nl_proto_down);
2681
2682                 /* Don't turn off protodown if there are active reasons */
2683                 if (!proto_down && dev->proto_down_reason) {
2684                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2685                         return -EBUSY;
2686                 }
2687                 err = dev_change_proto_down(dev,
2688                                             proto_down);
2689                 if (err)
2690                         return err;
2691         }
2692
2693         return 0;
2694 }
2695
2696 #define DO_SETLINK_MODIFIED     0x01
2697 /* notify flag means notify + modified. */
2698 #define DO_SETLINK_NOTIFY       0x03
2699 static int do_setlink(const struct sk_buff *skb,
2700                       struct net_device *dev, struct ifinfomsg *ifm,
2701                       struct netlink_ext_ack *extack,
2702                       struct nlattr **tb, int status)
2703 {
2704         const struct net_device_ops *ops = dev->netdev_ops;
2705         char ifname[IFNAMSIZ];
2706         int err;
2707
2708         err = validate_linkmsg(dev, tb, extack);
2709         if (err < 0)
2710                 return err;
2711
2712         if (tb[IFLA_IFNAME])
2713                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2714         else
2715                 ifname[0] = '\0';
2716
2717         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2718                 const char *pat = ifname[0] ? ifname : NULL;
2719                 struct net *net;
2720                 int new_ifindex;
2721
2722                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2723                                                 tb, CAP_NET_ADMIN);
2724                 if (IS_ERR(net)) {
2725                         err = PTR_ERR(net);
2726                         goto errout;
2727                 }
2728
2729                 if (tb[IFLA_NEW_IFINDEX])
2730                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2731                 else
2732                         new_ifindex = 0;
2733
2734                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2735                 put_net(net);
2736                 if (err)
2737                         goto errout;
2738                 status |= DO_SETLINK_MODIFIED;
2739         }
2740
2741         if (tb[IFLA_MAP]) {
2742                 struct rtnl_link_ifmap *u_map;
2743                 struct ifmap k_map;
2744
2745                 if (!ops->ndo_set_config) {
2746                         err = -EOPNOTSUPP;
2747                         goto errout;
2748                 }
2749
2750                 if (!netif_device_present(dev)) {
2751                         err = -ENODEV;
2752                         goto errout;
2753                 }
2754
2755                 u_map = nla_data(tb[IFLA_MAP]);
2756                 k_map.mem_start = (unsigned long) u_map->mem_start;
2757                 k_map.mem_end = (unsigned long) u_map->mem_end;
2758                 k_map.base_addr = (unsigned short) u_map->base_addr;
2759                 k_map.irq = (unsigned char) u_map->irq;
2760                 k_map.dma = (unsigned char) u_map->dma;
2761                 k_map.port = (unsigned char) u_map->port;
2762
2763                 err = ops->ndo_set_config(dev, &k_map);
2764                 if (err < 0)
2765                         goto errout;
2766
2767                 status |= DO_SETLINK_NOTIFY;
2768         }
2769
2770         if (tb[IFLA_ADDRESS]) {
2771                 struct sockaddr *sa;
2772                 int len;
2773
2774                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2775                                                   sizeof(*sa));
2776                 sa = kmalloc(len, GFP_KERNEL);
2777                 if (!sa) {
2778                         err = -ENOMEM;
2779                         goto errout;
2780                 }
2781                 sa->sa_family = dev->type;
2782                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2783                        dev->addr_len);
2784                 err = dev_set_mac_address_user(dev, sa, extack);
2785                 kfree(sa);
2786                 if (err)
2787                         goto errout;
2788                 status |= DO_SETLINK_MODIFIED;
2789         }
2790
2791         if (tb[IFLA_MTU]) {
2792                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2793                 if (err < 0)
2794                         goto errout;
2795                 status |= DO_SETLINK_MODIFIED;
2796         }
2797
2798         if (tb[IFLA_GROUP]) {
2799                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2800                 status |= DO_SETLINK_NOTIFY;
2801         }
2802
2803         /*
2804          * Interface selected by interface index but interface
2805          * name provided implies that a name change has been
2806          * requested.
2807          */
2808         if (ifm->ifi_index > 0 && ifname[0]) {
2809                 err = dev_change_name(dev, ifname);
2810                 if (err < 0)
2811                         goto errout;
2812                 status |= DO_SETLINK_MODIFIED;
2813         }
2814
2815         if (tb[IFLA_IFALIAS]) {
2816                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2817                                     nla_len(tb[IFLA_IFALIAS]));
2818                 if (err < 0)
2819                         goto errout;
2820                 status |= DO_SETLINK_NOTIFY;
2821         }
2822
2823         if (tb[IFLA_BROADCAST]) {
2824                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2825                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2826         }
2827
2828         if (tb[IFLA_MASTER]) {
2829                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2830                 if (err)
2831                         goto errout;
2832                 status |= DO_SETLINK_MODIFIED;
2833         }
2834
2835         if (ifm->ifi_flags || ifm->ifi_change) {
2836                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2837                                        extack);
2838                 if (err < 0)
2839                         goto errout;
2840         }
2841
2842         if (tb[IFLA_CARRIER]) {
2843                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2844                 if (err)
2845                         goto errout;
2846                 status |= DO_SETLINK_MODIFIED;
2847         }
2848
2849         if (tb[IFLA_TXQLEN]) {
2850                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2851
2852                 err = dev_change_tx_queue_len(dev, value);
2853                 if (err)
2854                         goto errout;
2855                 status |= DO_SETLINK_MODIFIED;
2856         }
2857
2858         if (tb[IFLA_GSO_MAX_SIZE]) {
2859                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2860
2861                 if (max_size > dev->tso_max_size) {
2862                         err = -EINVAL;
2863                         goto errout;
2864                 }
2865
2866                 if (dev->gso_max_size ^ max_size) {
2867                         netif_set_gso_max_size(dev, max_size);
2868                         status |= DO_SETLINK_MODIFIED;
2869                 }
2870         }
2871
2872         if (tb[IFLA_GSO_MAX_SEGS]) {
2873                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2874
2875                 if (max_segs > GSO_MAX_SEGS || max_segs > dev->tso_max_segs) {
2876                         err = -EINVAL;
2877                         goto errout;
2878                 }
2879
2880                 if (dev->gso_max_segs ^ max_segs) {
2881                         netif_set_gso_max_segs(dev, max_segs);
2882                         status |= DO_SETLINK_MODIFIED;
2883                 }
2884         }
2885
2886         if (tb[IFLA_GRO_MAX_SIZE]) {
2887                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2888
2889                 if (dev->gro_max_size ^ gro_max_size) {
2890                         netif_set_gro_max_size(dev, gro_max_size);
2891                         status |= DO_SETLINK_MODIFIED;
2892                 }
2893         }
2894
2895         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2896                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2897
2898                 if (max_size > dev->tso_max_size) {
2899                         err = -EINVAL;
2900                         goto errout;
2901                 }
2902
2903                 if (dev->gso_ipv4_max_size ^ max_size) {
2904                         netif_set_gso_ipv4_max_size(dev, max_size);
2905                         status |= DO_SETLINK_MODIFIED;
2906                 }
2907         }
2908
2909         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2910                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2911
2912                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2913                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2914                         status |= DO_SETLINK_MODIFIED;
2915                 }
2916         }
2917
2918         if (tb[IFLA_OPERSTATE])
2919                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2920
2921         if (tb[IFLA_LINKMODE]) {
2922                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2923
2924                 write_lock(&dev_base_lock);
2925                 if (dev->link_mode ^ value)
2926                         status |= DO_SETLINK_NOTIFY;
2927                 dev->link_mode = value;
2928                 write_unlock(&dev_base_lock);
2929         }
2930
2931         if (tb[IFLA_VFINFO_LIST]) {
2932                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2933                 struct nlattr *attr;
2934                 int rem;
2935
2936                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2937                         if (nla_type(attr) != IFLA_VF_INFO ||
2938                             nla_len(attr) < NLA_HDRLEN) {
2939                                 err = -EINVAL;
2940                                 goto errout;
2941                         }
2942                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2943                                                           attr,
2944                                                           ifla_vf_policy,
2945                                                           NULL);
2946                         if (err < 0)
2947                                 goto errout;
2948                         err = do_setvfinfo(dev, vfinfo);
2949                         if (err < 0)
2950                                 goto errout;
2951                         status |= DO_SETLINK_NOTIFY;
2952                 }
2953         }
2954         err = 0;
2955
2956         if (tb[IFLA_VF_PORTS]) {
2957                 struct nlattr *port[IFLA_PORT_MAX+1];
2958                 struct nlattr *attr;
2959                 int vf;
2960                 int rem;
2961
2962                 err = -EOPNOTSUPP;
2963                 if (!ops->ndo_set_vf_port)
2964                         goto errout;
2965
2966                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2967                         if (nla_type(attr) != IFLA_VF_PORT ||
2968                             nla_len(attr) < NLA_HDRLEN) {
2969                                 err = -EINVAL;
2970                                 goto errout;
2971                         }
2972                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2973                                                           attr,
2974                                                           ifla_port_policy,
2975                                                           NULL);
2976                         if (err < 0)
2977                                 goto errout;
2978                         if (!port[IFLA_PORT_VF]) {
2979                                 err = -EOPNOTSUPP;
2980                                 goto errout;
2981                         }
2982                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2983                         err = ops->ndo_set_vf_port(dev, vf, port);
2984                         if (err < 0)
2985                                 goto errout;
2986                         status |= DO_SETLINK_NOTIFY;
2987                 }
2988         }
2989         err = 0;
2990
2991         if (tb[IFLA_PORT_SELF]) {
2992                 struct nlattr *port[IFLA_PORT_MAX+1];
2993
2994                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2995                                                   tb[IFLA_PORT_SELF],
2996                                                   ifla_port_policy, NULL);
2997                 if (err < 0)
2998                         goto errout;
2999
3000                 err = -EOPNOTSUPP;
3001                 if (ops->ndo_set_vf_port)
3002                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3003                 if (err < 0)
3004                         goto errout;
3005                 status |= DO_SETLINK_NOTIFY;
3006         }
3007
3008         if (tb[IFLA_AF_SPEC]) {
3009                 struct nlattr *af;
3010                 int rem;
3011
3012                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3013                         const struct rtnl_af_ops *af_ops;
3014
3015                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3016
3017                         err = af_ops->set_link_af(dev, af, extack);
3018                         if (err < 0)
3019                                 goto errout;
3020
3021                         status |= DO_SETLINK_NOTIFY;
3022                 }
3023         }
3024         err = 0;
3025
3026         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3027                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3028                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3029                 if (err)
3030                         goto errout;
3031                 status |= DO_SETLINK_NOTIFY;
3032         }
3033
3034         if (tb[IFLA_XDP]) {
3035                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3036                 u32 xdp_flags = 0;
3037
3038                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3039                                                   tb[IFLA_XDP],
3040                                                   ifla_xdp_policy, NULL);
3041                 if (err < 0)
3042                         goto errout;
3043
3044                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3045                         err = -EINVAL;
3046                         goto errout;
3047                 }
3048
3049                 if (xdp[IFLA_XDP_FLAGS]) {
3050                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3051                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3052                                 err = -EINVAL;
3053                                 goto errout;
3054                         }
3055                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3056                                 err = -EINVAL;
3057                                 goto errout;
3058                         }
3059                 }
3060
3061                 if (xdp[IFLA_XDP_FD]) {
3062                         int expected_fd = -1;
3063
3064                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3065                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3066                                         err = -EINVAL;
3067                                         goto errout;
3068                                 }
3069                                 expected_fd =
3070                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3071                         }
3072
3073                         err = dev_change_xdp_fd(dev, extack,
3074                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3075                                                 expected_fd,
3076                                                 xdp_flags);
3077                         if (err)
3078                                 goto errout;
3079                         status |= DO_SETLINK_NOTIFY;
3080                 }
3081         }
3082
3083 errout:
3084         if (status & DO_SETLINK_MODIFIED) {
3085                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3086                         netdev_state_change(dev);
3087
3088                 if (err < 0)
3089                         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",
3090                                              dev->name);
3091         }
3092
3093         return err;
3094 }
3095
3096 static struct net_device *rtnl_dev_get(struct net *net,
3097                                        struct nlattr *tb[])
3098 {
3099         char ifname[ALTIFNAMSIZ];
3100
3101         if (tb[IFLA_IFNAME])
3102                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3103         else if (tb[IFLA_ALT_IFNAME])
3104                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3105         else
3106                 return NULL;
3107
3108         return __dev_get_by_name(net, ifname);
3109 }
3110
3111 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3112                         struct netlink_ext_ack *extack)
3113 {
3114         struct net *net = sock_net(skb->sk);
3115         struct ifinfomsg *ifm;
3116         struct net_device *dev;
3117         int err;
3118         struct nlattr *tb[IFLA_MAX+1];
3119
3120         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3121                                      ifla_policy, extack);
3122         if (err < 0)
3123                 goto errout;
3124
3125         err = rtnl_ensure_unique_netns(tb, extack, false);
3126         if (err < 0)
3127                 goto errout;
3128
3129         err = -EINVAL;
3130         ifm = nlmsg_data(nlh);
3131         if (ifm->ifi_index > 0)
3132                 dev = __dev_get_by_index(net, ifm->ifi_index);
3133         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3134                 dev = rtnl_dev_get(net, tb);
3135         else
3136                 goto errout;
3137
3138         if (dev == NULL) {
3139                 err = -ENODEV;
3140                 goto errout;
3141         }
3142
3143         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3144 errout:
3145         return err;
3146 }
3147
3148 static int rtnl_group_dellink(const struct net *net, int group)
3149 {
3150         struct net_device *dev, *aux;
3151         LIST_HEAD(list_kill);
3152         bool found = false;
3153
3154         if (!group)
3155                 return -EPERM;
3156
3157         for_each_netdev(net, dev) {
3158                 if (dev->group == group) {
3159                         const struct rtnl_link_ops *ops;
3160
3161                         found = true;
3162                         ops = dev->rtnl_link_ops;
3163                         if (!ops || !ops->dellink)
3164                                 return -EOPNOTSUPP;
3165                 }
3166         }
3167
3168         if (!found)
3169                 return -ENODEV;
3170
3171         for_each_netdev_safe(net, dev, aux) {
3172                 if (dev->group == group) {
3173                         const struct rtnl_link_ops *ops;
3174
3175                         ops = dev->rtnl_link_ops;
3176                         ops->dellink(dev, &list_kill);
3177                 }
3178         }
3179         unregister_netdevice_many(&list_kill);
3180
3181         return 0;
3182 }
3183
3184 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3185 {
3186         const struct rtnl_link_ops *ops;
3187         LIST_HEAD(list_kill);
3188
3189         ops = dev->rtnl_link_ops;
3190         if (!ops || !ops->dellink)
3191                 return -EOPNOTSUPP;
3192
3193         ops->dellink(dev, &list_kill);
3194         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3195
3196         return 0;
3197 }
3198 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3199
3200 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3201                         struct netlink_ext_ack *extack)
3202 {
3203         struct net *net = sock_net(skb->sk);
3204         u32 portid = NETLINK_CB(skb).portid;
3205         struct net *tgt_net = net;
3206         struct net_device *dev = NULL;
3207         struct ifinfomsg *ifm;
3208         struct nlattr *tb[IFLA_MAX+1];
3209         int err;
3210         int netnsid = -1;
3211
3212         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3213                                      ifla_policy, extack);
3214         if (err < 0)
3215                 return err;
3216
3217         err = rtnl_ensure_unique_netns(tb, extack, true);
3218         if (err < 0)
3219                 return err;
3220
3221         if (tb[IFLA_TARGET_NETNSID]) {
3222                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3223                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3224                 if (IS_ERR(tgt_net))
3225                         return PTR_ERR(tgt_net);
3226         }
3227
3228         err = -EINVAL;
3229         ifm = nlmsg_data(nlh);
3230         if (ifm->ifi_index > 0)
3231                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3232         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3233                 dev = rtnl_dev_get(net, tb);
3234         else if (tb[IFLA_GROUP])
3235                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3236         else
3237                 goto out;
3238
3239         if (!dev) {
3240                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3241                         err = -ENODEV;
3242
3243                 goto out;
3244         }
3245
3246         err = rtnl_delete_link(dev, portid, nlh);
3247
3248 out:
3249         if (netnsid >= 0)
3250                 put_net(tgt_net);
3251
3252         return err;
3253 }
3254
3255 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3256                         u32 portid, const struct nlmsghdr *nlh)
3257 {
3258         unsigned int old_flags;
3259         int err;
3260
3261         old_flags = dev->flags;
3262         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3263                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3264                                          NULL);
3265                 if (err < 0)
3266                         return err;
3267         }
3268
3269         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3270                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3271         } else {
3272                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3273                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3274         }
3275         return 0;
3276 }
3277 EXPORT_SYMBOL(rtnl_configure_link);
3278
3279 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3280                                     unsigned char name_assign_type,
3281                                     const struct rtnl_link_ops *ops,
3282                                     struct nlattr *tb[],
3283                                     struct netlink_ext_ack *extack)
3284 {
3285         struct net_device *dev;
3286         unsigned int num_tx_queues = 1;
3287         unsigned int num_rx_queues = 1;
3288         int err;
3289
3290         if (tb[IFLA_NUM_TX_QUEUES])
3291                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3292         else if (ops->get_num_tx_queues)
3293                 num_tx_queues = ops->get_num_tx_queues();
3294
3295         if (tb[IFLA_NUM_RX_QUEUES])
3296                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3297         else if (ops->get_num_rx_queues)
3298                 num_rx_queues = ops->get_num_rx_queues();
3299
3300         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3301                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3302                 return ERR_PTR(-EINVAL);
3303         }
3304
3305         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3306                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3307                 return ERR_PTR(-EINVAL);
3308         }
3309
3310         if (ops->alloc) {
3311                 dev = ops->alloc(tb, ifname, name_assign_type,
3312                                  num_tx_queues, num_rx_queues);
3313                 if (IS_ERR(dev))
3314                         return dev;
3315         } else {
3316                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3317                                        name_assign_type, ops->setup,
3318                                        num_tx_queues, num_rx_queues);
3319         }
3320
3321         if (!dev)
3322                 return ERR_PTR(-ENOMEM);
3323
3324         err = validate_linkmsg(dev, tb, extack);
3325         if (err < 0) {
3326                 free_netdev(dev);
3327                 return ERR_PTR(err);
3328         }
3329
3330         dev_net_set(dev, net);
3331         dev->rtnl_link_ops = ops;
3332         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3333
3334         if (tb[IFLA_MTU]) {
3335                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3336
3337                 err = dev_validate_mtu(dev, mtu, extack);
3338                 if (err) {
3339                         free_netdev(dev);
3340                         return ERR_PTR(err);
3341                 }
3342                 dev->mtu = mtu;
3343         }
3344         if (tb[IFLA_ADDRESS]) {
3345                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3346                                nla_len(tb[IFLA_ADDRESS]));
3347                 dev->addr_assign_type = NET_ADDR_SET;
3348         }
3349         if (tb[IFLA_BROADCAST])
3350                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3351                                 nla_len(tb[IFLA_BROADCAST]));
3352         if (tb[IFLA_TXQLEN])
3353                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3354         if (tb[IFLA_OPERSTATE])
3355                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3356         if (tb[IFLA_LINKMODE])
3357                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3358         if (tb[IFLA_GROUP])
3359                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3360         if (tb[IFLA_GSO_MAX_SIZE])
3361                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3362         if (tb[IFLA_GSO_MAX_SEGS])
3363                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3364         if (tb[IFLA_GRO_MAX_SIZE])
3365                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3366         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3367                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3368         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3369                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3370
3371         return dev;
3372 }
3373 EXPORT_SYMBOL(rtnl_create_link);
3374
3375 static int rtnl_group_changelink(const struct sk_buff *skb,
3376                 struct net *net, int group,
3377                 struct ifinfomsg *ifm,
3378                 struct netlink_ext_ack *extack,
3379                 struct nlattr **tb)
3380 {
3381         struct net_device *dev, *aux;
3382         int err;
3383
3384         for_each_netdev_safe(net, dev, aux) {
3385                 if (dev->group == group) {
3386                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3387                         if (err < 0)
3388                                 return err;
3389                 }
3390         }
3391
3392         return 0;
3393 }
3394
3395 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3396                                const struct rtnl_link_ops *ops,
3397                                const struct nlmsghdr *nlh,
3398                                struct nlattr **tb, struct nlattr **data,
3399                                struct netlink_ext_ack *extack)
3400 {
3401         unsigned char name_assign_type = NET_NAME_USER;
3402         struct net *net = sock_net(skb->sk);
3403         u32 portid = NETLINK_CB(skb).portid;
3404         struct net *dest_net, *link_net;
3405         struct net_device *dev;
3406         char ifname[IFNAMSIZ];
3407         int err;
3408
3409         if (!ops->alloc && !ops->setup)
3410                 return -EOPNOTSUPP;
3411
3412         if (tb[IFLA_IFNAME]) {
3413                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3414         } else {
3415                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3416                 name_assign_type = NET_NAME_ENUM;
3417         }
3418
3419         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3420         if (IS_ERR(dest_net))
3421                 return PTR_ERR(dest_net);
3422
3423         if (tb[IFLA_LINK_NETNSID]) {
3424                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3425
3426                 link_net = get_net_ns_by_id(dest_net, id);
3427                 if (!link_net) {
3428                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3429                         err =  -EINVAL;
3430                         goto out;
3431                 }
3432                 err = -EPERM;
3433                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3434                         goto out;
3435         } else {
3436                 link_net = NULL;
3437         }
3438
3439         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3440                                name_assign_type, ops, tb, extack);
3441         if (IS_ERR(dev)) {
3442                 err = PTR_ERR(dev);
3443                 goto out;
3444         }
3445
3446         dev->ifindex = ifm->ifi_index;
3447
3448         if (ops->newlink)
3449                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3450         else
3451                 err = register_netdevice(dev);
3452         if (err < 0) {
3453                 free_netdev(dev);
3454                 goto out;
3455         }
3456
3457         err = rtnl_configure_link(dev, ifm, portid, nlh);
3458         if (err < 0)
3459                 goto out_unregister;
3460         if (link_net) {
3461                 err = dev_change_net_namespace(dev, dest_net, ifname);
3462                 if (err < 0)
3463                         goto out_unregister;
3464         }
3465         if (tb[IFLA_MASTER]) {
3466                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3467                 if (err)
3468                         goto out_unregister;
3469         }
3470 out:
3471         if (link_net)
3472                 put_net(link_net);
3473         put_net(dest_net);
3474         return err;
3475 out_unregister:
3476         if (ops->newlink) {
3477                 LIST_HEAD(list_kill);
3478
3479                 ops->dellink(dev, &list_kill);
3480                 unregister_netdevice_many(&list_kill);
3481         } else {
3482                 unregister_netdevice(dev);
3483         }
3484         goto out;
3485 }
3486
3487 struct rtnl_newlink_tbs {
3488         struct nlattr *tb[IFLA_MAX + 1];
3489         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3490         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3491 };
3492
3493 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3494                           struct rtnl_newlink_tbs *tbs,
3495                           struct netlink_ext_ack *extack)
3496 {
3497         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3498         struct nlattr ** const tb = tbs->tb;
3499         const struct rtnl_link_ops *m_ops;
3500         struct net_device *master_dev;
3501         struct net *net = sock_net(skb->sk);
3502         const struct rtnl_link_ops *ops;
3503         struct nlattr **slave_data;
3504         char kind[MODULE_NAME_LEN];
3505         struct net_device *dev;
3506         struct ifinfomsg *ifm;
3507         struct nlattr **data;
3508         bool link_specified;
3509         int err;
3510
3511 #ifdef CONFIG_MODULES
3512 replay:
3513 #endif
3514         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3515                                      ifla_policy, extack);
3516         if (err < 0)
3517                 return err;
3518
3519         err = rtnl_ensure_unique_netns(tb, extack, false);
3520         if (err < 0)
3521                 return err;
3522
3523         ifm = nlmsg_data(nlh);
3524         if (ifm->ifi_index > 0) {
3525                 link_specified = true;
3526                 dev = __dev_get_by_index(net, ifm->ifi_index);
3527         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3528                 link_specified = true;
3529                 dev = rtnl_dev_get(net, tb);
3530         } else {
3531                 link_specified = false;
3532                 dev = NULL;
3533         }
3534
3535         master_dev = NULL;
3536         m_ops = NULL;
3537         if (dev) {
3538                 master_dev = netdev_master_upper_dev_get(dev);
3539                 if (master_dev)
3540                         m_ops = master_dev->rtnl_link_ops;
3541         }
3542
3543         err = validate_linkmsg(dev, tb, extack);
3544         if (err < 0)
3545                 return err;
3546
3547         if (tb[IFLA_LINKINFO]) {
3548                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3549                                                   tb[IFLA_LINKINFO],
3550                                                   ifla_info_policy, NULL);
3551                 if (err < 0)
3552                         return err;
3553         } else
3554                 memset(linkinfo, 0, sizeof(linkinfo));
3555
3556         if (linkinfo[IFLA_INFO_KIND]) {
3557                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3558                 ops = rtnl_link_ops_get(kind);
3559         } else {
3560                 kind[0] = '\0';
3561                 ops = NULL;
3562         }
3563
3564         data = NULL;
3565         if (ops) {
3566                 if (ops->maxtype > RTNL_MAX_TYPE)
3567                         return -EINVAL;
3568
3569                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3570                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3571                                                           linkinfo[IFLA_INFO_DATA],
3572                                                           ops->policy, extack);
3573                         if (err < 0)
3574                                 return err;
3575                         data = tbs->attr;
3576                 }
3577                 if (ops->validate) {
3578                         err = ops->validate(tb, data, extack);
3579                         if (err < 0)
3580                                 return err;
3581                 }
3582         }
3583
3584         slave_data = NULL;
3585         if (m_ops) {
3586                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3587                         return -EINVAL;
3588
3589                 if (m_ops->slave_maxtype &&
3590                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3591                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3592                                                           m_ops->slave_maxtype,
3593                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3594                                                           m_ops->slave_policy,
3595                                                           extack);
3596                         if (err < 0)
3597                                 return err;
3598                         slave_data = tbs->slave_attr;
3599                 }
3600         }
3601
3602         if (dev) {
3603                 int status = 0;
3604
3605                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3606                         return -EEXIST;
3607                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3608                         return -EOPNOTSUPP;
3609
3610                 if (linkinfo[IFLA_INFO_DATA]) {
3611                         if (!ops || ops != dev->rtnl_link_ops ||
3612                             !ops->changelink)
3613                                 return -EOPNOTSUPP;
3614
3615                         err = ops->changelink(dev, tb, data, extack);
3616                         if (err < 0)
3617                                 return err;
3618                         status |= DO_SETLINK_NOTIFY;
3619                 }
3620
3621                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3622                         if (!m_ops || !m_ops->slave_changelink)
3623                                 return -EOPNOTSUPP;
3624
3625                         err = m_ops->slave_changelink(master_dev, dev, tb,
3626                                                       slave_data, extack);
3627                         if (err < 0)
3628                                 return err;
3629                         status |= DO_SETLINK_NOTIFY;
3630                 }
3631
3632                 return do_setlink(skb, dev, ifm, extack, tb, status);
3633         }
3634
3635         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3636                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3637                  * or it's for a group
3638                 */
3639                 if (link_specified)
3640                         return -ENODEV;
3641                 if (tb[IFLA_GROUP])
3642                         return rtnl_group_changelink(skb, net,
3643                                                 nla_get_u32(tb[IFLA_GROUP]),
3644                                                 ifm, extack, tb);
3645                 return -ENODEV;
3646         }
3647
3648         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3649                 return -EOPNOTSUPP;
3650
3651         if (!ops) {
3652 #ifdef CONFIG_MODULES
3653                 if (kind[0]) {
3654                         __rtnl_unlock();
3655                         request_module("rtnl-link-%s", kind);
3656                         rtnl_lock();
3657                         ops = rtnl_link_ops_get(kind);
3658                         if (ops)
3659                                 goto replay;
3660                 }
3661 #endif
3662                 NL_SET_ERR_MSG(extack, "Unknown device type");
3663                 return -EOPNOTSUPP;
3664         }
3665
3666         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3667 }
3668
3669 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3670                         struct netlink_ext_ack *extack)
3671 {
3672         struct rtnl_newlink_tbs *tbs;
3673         int ret;
3674
3675         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3676         if (!tbs)
3677                 return -ENOMEM;
3678
3679         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3680         kfree(tbs);
3681         return ret;
3682 }
3683
3684 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3685                                   const struct nlmsghdr *nlh,
3686                                   struct nlattr **tb,
3687                                   struct netlink_ext_ack *extack)
3688 {
3689         struct ifinfomsg *ifm;
3690         int i, err;
3691
3692         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3693                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3694                 return -EINVAL;
3695         }
3696
3697         if (!netlink_strict_get_check(skb))
3698                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3699                                               ifla_policy, extack);
3700
3701         ifm = nlmsg_data(nlh);
3702         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3703             ifm->ifi_change) {
3704                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3705                 return -EINVAL;
3706         }
3707
3708         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3709                                             ifla_policy, extack);
3710         if (err)
3711                 return err;
3712
3713         for (i = 0; i <= IFLA_MAX; i++) {
3714                 if (!tb[i])
3715                         continue;
3716
3717                 switch (i) {
3718                 case IFLA_IFNAME:
3719                 case IFLA_ALT_IFNAME:
3720                 case IFLA_EXT_MASK:
3721                 case IFLA_TARGET_NETNSID:
3722                         break;
3723                 default:
3724                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3725                         return -EINVAL;
3726                 }
3727         }
3728
3729         return 0;
3730 }
3731
3732 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3733                         struct netlink_ext_ack *extack)
3734 {
3735         struct net *net = sock_net(skb->sk);
3736         struct net *tgt_net = net;
3737         struct ifinfomsg *ifm;
3738         struct nlattr *tb[IFLA_MAX+1];
3739         struct net_device *dev = NULL;
3740         struct sk_buff *nskb;
3741         int netnsid = -1;
3742         int err;
3743         u32 ext_filter_mask = 0;
3744
3745         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3746         if (err < 0)
3747                 return err;
3748
3749         err = rtnl_ensure_unique_netns(tb, extack, true);
3750         if (err < 0)
3751                 return err;
3752
3753         if (tb[IFLA_TARGET_NETNSID]) {
3754                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3755                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3756                 if (IS_ERR(tgt_net))
3757                         return PTR_ERR(tgt_net);
3758         }
3759
3760         if (tb[IFLA_EXT_MASK])
3761                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3762
3763         err = -EINVAL;
3764         ifm = nlmsg_data(nlh);
3765         if (ifm->ifi_index > 0)
3766                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3767         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3768                 dev = rtnl_dev_get(tgt_net, tb);
3769         else
3770                 goto out;
3771
3772         err = -ENODEV;
3773         if (dev == NULL)
3774                 goto out;
3775
3776         err = -ENOBUFS;
3777         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3778         if (nskb == NULL)
3779                 goto out;
3780
3781         err = rtnl_fill_ifinfo(nskb, dev, net,
3782                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3783                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3784                                0, NULL, 0, netnsid, GFP_KERNEL);
3785         if (err < 0) {
3786                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3787                 WARN_ON(err == -EMSGSIZE);
3788                 kfree_skb(nskb);
3789         } else
3790                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3791 out:
3792         if (netnsid >= 0)
3793                 put_net(tgt_net);
3794
3795         return err;
3796 }
3797
3798 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3799                            bool *changed, struct netlink_ext_ack *extack)
3800 {
3801         char *alt_ifname;
3802         size_t size;
3803         int err;
3804
3805         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3806         if (err)
3807                 return err;
3808
3809         if (cmd == RTM_NEWLINKPROP) {
3810                 size = rtnl_prop_list_size(dev);
3811                 size += nla_total_size(ALTIFNAMSIZ);
3812                 if (size >= U16_MAX) {
3813                         NL_SET_ERR_MSG(extack,
3814                                        "effective property list too long");
3815                         return -EINVAL;
3816                 }
3817         }
3818
3819         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3820         if (!alt_ifname)
3821                 return -ENOMEM;
3822
3823         if (cmd == RTM_NEWLINKPROP) {
3824                 err = netdev_name_node_alt_create(dev, alt_ifname);
3825                 if (!err)
3826                         alt_ifname = NULL;
3827         } else if (cmd == RTM_DELLINKPROP) {
3828                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3829         } else {
3830                 WARN_ON_ONCE(1);
3831                 err = -EINVAL;
3832         }
3833
3834         kfree(alt_ifname);
3835         if (!err)
3836                 *changed = true;
3837         return err;
3838 }
3839
3840 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3841                          struct netlink_ext_ack *extack)
3842 {
3843         struct net *net = sock_net(skb->sk);
3844         struct nlattr *tb[IFLA_MAX + 1];
3845         struct net_device *dev;
3846         struct ifinfomsg *ifm;
3847         bool changed = false;
3848         struct nlattr *attr;
3849         int err, rem;
3850
3851         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3852         if (err)
3853                 return err;
3854
3855         err = rtnl_ensure_unique_netns(tb, extack, true);
3856         if (err)
3857                 return err;
3858
3859         ifm = nlmsg_data(nlh);
3860         if (ifm->ifi_index > 0)
3861                 dev = __dev_get_by_index(net, ifm->ifi_index);
3862         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3863                 dev = rtnl_dev_get(net, tb);
3864         else
3865                 return -EINVAL;
3866
3867         if (!dev)
3868                 return -ENODEV;
3869
3870         if (!tb[IFLA_PROP_LIST])
3871                 return 0;
3872
3873         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3874                 switch (nla_type(attr)) {
3875                 case IFLA_ALT_IFNAME:
3876                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3877                         if (err)
3878                                 return err;
3879                         break;
3880                 }
3881         }
3882
3883         if (changed)
3884                 netdev_state_change(dev);
3885         return 0;
3886 }
3887
3888 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3889                             struct netlink_ext_ack *extack)
3890 {
3891         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3892 }
3893
3894 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3895                             struct netlink_ext_ack *extack)
3896 {
3897         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3898 }
3899
3900 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3901 {
3902         struct net *net = sock_net(skb->sk);
3903         size_t min_ifinfo_dump_size = 0;
3904         struct nlattr *tb[IFLA_MAX+1];
3905         u32 ext_filter_mask = 0;
3906         struct net_device *dev;
3907         int hdrlen;
3908
3909         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3910         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3911                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3912
3913         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3914                 if (tb[IFLA_EXT_MASK])
3915                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3916         }
3917
3918         if (!ext_filter_mask)
3919                 return NLMSG_GOODSIZE;
3920         /*
3921          * traverse the list of net devices and compute the minimum
3922          * buffer size based upon the filter mask.
3923          */
3924         rcu_read_lock();
3925         for_each_netdev_rcu(net, dev) {
3926                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3927                                            if_nlmsg_size(dev, ext_filter_mask));
3928         }
3929         rcu_read_unlock();
3930
3931         return nlmsg_total_size(min_ifinfo_dump_size);
3932 }
3933
3934 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3935 {
3936         int idx;
3937         int s_idx = cb->family;
3938         int type = cb->nlh->nlmsg_type - RTM_BASE;
3939         int ret = 0;
3940
3941         if (s_idx == 0)
3942                 s_idx = 1;
3943
3944         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3945                 struct rtnl_link __rcu **tab;
3946                 struct rtnl_link *link;
3947                 rtnl_dumpit_func dumpit;
3948
3949                 if (idx < s_idx || idx == PF_PACKET)
3950                         continue;
3951
3952                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3953                         continue;
3954
3955                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3956                 if (!tab)
3957                         continue;
3958
3959                 link = rcu_dereference_rtnl(tab[type]);
3960                 if (!link)
3961                         continue;
3962
3963                 dumpit = link->dumpit;
3964                 if (!dumpit)
3965                         continue;
3966
3967                 if (idx > s_idx) {
3968                         memset(&cb->args[0], 0, sizeof(cb->args));
3969                         cb->prev_seq = 0;
3970                         cb->seq = 0;
3971                 }
3972                 ret = dumpit(skb, cb);
3973                 if (ret)
3974                         break;
3975         }
3976         cb->family = idx;
3977
3978         return skb->len ? : ret;
3979 }
3980
3981 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3982                                        unsigned int change,
3983                                        u32 event, gfp_t flags, int *new_nsid,
3984                                        int new_ifindex, u32 portid,
3985                                        const struct nlmsghdr *nlh)
3986 {
3987         struct net *net = dev_net(dev);
3988         struct sk_buff *skb;
3989         int err = -ENOBUFS;
3990         u32 seq = 0;
3991
3992         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3993         if (skb == NULL)
3994                 goto errout;
3995
3996         if (nlmsg_report(nlh))
3997                 seq = nlmsg_seq(nlh);
3998         else
3999                 portid = 0;
4000
4001         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4002                                type, portid, seq, change, 0, 0, event,
4003                                new_nsid, new_ifindex, -1, flags);
4004         if (err < 0) {
4005                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4006                 WARN_ON(err == -EMSGSIZE);
4007                 kfree_skb(skb);
4008                 goto errout;
4009         }
4010         return skb;
4011 errout:
4012         if (err < 0)
4013                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4014         return NULL;
4015 }
4016
4017 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4018                        u32 portid, const struct nlmsghdr *nlh)
4019 {
4020         struct net *net = dev_net(dev);
4021
4022         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4023 }
4024
4025 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4026                                unsigned int change, u32 event,
4027                                gfp_t flags, int *new_nsid, int new_ifindex,
4028                                u32 portid, const struct nlmsghdr *nlh)
4029 {
4030         struct sk_buff *skb;
4031
4032         if (dev->reg_state != NETREG_REGISTERED)
4033                 return;
4034
4035         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4036                                      new_ifindex, portid, nlh);
4037         if (skb)
4038                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4039 }
4040
4041 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4042                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4043 {
4044         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4045                            NULL, 0, portid, nlh);
4046 }
4047
4048 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4049                          gfp_t flags, int *new_nsid, int new_ifindex)
4050 {
4051         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4052                            new_nsid, new_ifindex, 0, NULL);
4053 }
4054
4055 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4056                                    struct net_device *dev,
4057                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4058                                    int type, unsigned int flags,
4059                                    int nlflags, u16 ndm_state)
4060 {
4061         struct nlmsghdr *nlh;
4062         struct ndmsg *ndm;
4063
4064         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4065         if (!nlh)
4066                 return -EMSGSIZE;
4067
4068         ndm = nlmsg_data(nlh);
4069         ndm->ndm_family  = AF_BRIDGE;
4070         ndm->ndm_pad1    = 0;
4071         ndm->ndm_pad2    = 0;
4072         ndm->ndm_flags   = flags;
4073         ndm->ndm_type    = 0;
4074         ndm->ndm_ifindex = dev->ifindex;
4075         ndm->ndm_state   = ndm_state;
4076
4077         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
4078                 goto nla_put_failure;
4079         if (vid)
4080                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4081                         goto nla_put_failure;
4082
4083         nlmsg_end(skb, nlh);
4084         return 0;
4085
4086 nla_put_failure:
4087         nlmsg_cancel(skb, nlh);
4088         return -EMSGSIZE;
4089 }
4090
4091 static inline size_t rtnl_fdb_nlmsg_size(void)
4092 {
4093         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4094                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
4095                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4096                0;
4097 }
4098
4099 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4100                             u16 ndm_state)
4101 {
4102         struct net *net = dev_net(dev);
4103         struct sk_buff *skb;
4104         int err = -ENOBUFS;
4105
4106         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4107         if (!skb)
4108                 goto errout;
4109
4110         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4111                                       0, 0, type, NTF_SELF, 0, ndm_state);
4112         if (err < 0) {
4113                 kfree_skb(skb);
4114                 goto errout;
4115         }
4116
4117         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4118         return;
4119 errout:
4120         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4121 }
4122
4123 /*
4124  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4125  */
4126 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4127                      struct nlattr *tb[],
4128                      struct net_device *dev,
4129                      const unsigned char *addr, u16 vid,
4130                      u16 flags)
4131 {
4132         int err = -EINVAL;
4133
4134         /* If aging addresses are supported device will need to
4135          * implement its own handler for this.
4136          */
4137         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4138                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4139                 return err;
4140         }
4141
4142         if (tb[NDA_FLAGS_EXT]) {
4143                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4144                 return err;
4145         }
4146
4147         if (vid) {
4148                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4149                 return err;
4150         }
4151
4152         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4153                 err = dev_uc_add_excl(dev, addr);
4154         else if (is_multicast_ether_addr(addr))
4155                 err = dev_mc_add_excl(dev, addr);
4156
4157         /* Only return duplicate errors if NLM_F_EXCL is set */
4158         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4159                 err = 0;
4160
4161         return err;
4162 }
4163 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4164
4165 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4166                          struct netlink_ext_ack *extack)
4167 {
4168         u16 vid = 0;
4169
4170         if (vlan_attr) {
4171                 if (nla_len(vlan_attr) != sizeof(u16)) {
4172                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4173                         return -EINVAL;
4174                 }
4175
4176                 vid = nla_get_u16(vlan_attr);
4177
4178                 if (!vid || vid >= VLAN_VID_MASK) {
4179                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4180                         return -EINVAL;
4181                 }
4182         }
4183         *p_vid = vid;
4184         return 0;
4185 }
4186
4187 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4188                         struct netlink_ext_ack *extack)
4189 {
4190         struct net *net = sock_net(skb->sk);
4191         struct ndmsg *ndm;
4192         struct nlattr *tb[NDA_MAX+1];
4193         struct net_device *dev;
4194         u8 *addr;
4195         u16 vid;
4196         int err;
4197
4198         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4199                                      extack);
4200         if (err < 0)
4201                 return err;
4202
4203         ndm = nlmsg_data(nlh);
4204         if (ndm->ndm_ifindex == 0) {
4205                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4206                 return -EINVAL;
4207         }
4208
4209         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4210         if (dev == NULL) {
4211                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4212                 return -ENODEV;
4213         }
4214
4215         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4216                 NL_SET_ERR_MSG(extack, "invalid address");
4217                 return -EINVAL;
4218         }
4219
4220         if (dev->type != ARPHRD_ETHER) {
4221                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4222                 return -EINVAL;
4223         }
4224
4225         addr = nla_data(tb[NDA_LLADDR]);
4226
4227         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4228         if (err)
4229                 return err;
4230
4231         err = -EOPNOTSUPP;
4232
4233         /* Support fdb on master device the net/bridge default case */
4234         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4235             netif_is_bridge_port(dev)) {
4236                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4237                 const struct net_device_ops *ops = br_dev->netdev_ops;
4238
4239                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4240                                        nlh->nlmsg_flags, extack);
4241                 if (err)
4242                         goto out;
4243                 else
4244                         ndm->ndm_flags &= ~NTF_MASTER;
4245         }
4246
4247         /* Embedded bridge, macvlan, and any other device support */
4248         if ((ndm->ndm_flags & NTF_SELF)) {
4249                 if (dev->netdev_ops->ndo_fdb_add)
4250                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4251                                                            vid,
4252                                                            nlh->nlmsg_flags,
4253                                                            extack);
4254                 else
4255                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4256                                                nlh->nlmsg_flags);
4257
4258                 if (!err) {
4259                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4260                                         ndm->ndm_state);
4261                         ndm->ndm_flags &= ~NTF_SELF;
4262                 }
4263         }
4264 out:
4265         return err;
4266 }
4267
4268 /*
4269  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4270  */
4271 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4272                      struct nlattr *tb[],
4273                      struct net_device *dev,
4274                      const unsigned char *addr, u16 vid)
4275 {
4276         int err = -EINVAL;
4277
4278         /* If aging addresses are supported device will need to
4279          * implement its own handler for this.
4280          */
4281         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4282                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4283                 return err;
4284         }
4285
4286         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4287                 err = dev_uc_del(dev, addr);
4288         else if (is_multicast_ether_addr(addr))
4289                 err = dev_mc_del(dev, addr);
4290
4291         return err;
4292 }
4293 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4294
4295 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4296         [NDA_VLAN]      = { .type = NLA_U16 },
4297         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4298         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4299         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4300 };
4301
4302 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4303                         struct netlink_ext_ack *extack)
4304 {
4305         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4306         struct net *net = sock_net(skb->sk);
4307         const struct net_device_ops *ops;
4308         struct ndmsg *ndm;
4309         struct nlattr *tb[NDA_MAX+1];
4310         struct net_device *dev;
4311         __u8 *addr = NULL;
4312         int err;
4313         u16 vid;
4314
4315         if (!netlink_capable(skb, CAP_NET_ADMIN))
4316                 return -EPERM;
4317
4318         if (!del_bulk) {
4319                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4320                                              NULL, extack);
4321         } else {
4322                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4323                                   fdb_del_bulk_policy, extack);
4324         }
4325         if (err < 0)
4326                 return err;
4327
4328         ndm = nlmsg_data(nlh);
4329         if (ndm->ndm_ifindex == 0) {
4330                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4331                 return -EINVAL;
4332         }
4333
4334         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4335         if (dev == NULL) {
4336                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4337                 return -ENODEV;
4338         }
4339
4340         if (!del_bulk) {
4341                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4342                         NL_SET_ERR_MSG(extack, "invalid address");
4343                         return -EINVAL;
4344                 }
4345                 addr = nla_data(tb[NDA_LLADDR]);
4346         }
4347
4348         if (dev->type != ARPHRD_ETHER) {
4349                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4350                 return -EINVAL;
4351         }
4352
4353         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4354         if (err)
4355                 return err;
4356
4357         err = -EOPNOTSUPP;
4358
4359         /* Support fdb on master device the net/bridge default case */
4360         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4361             netif_is_bridge_port(dev)) {
4362                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4363
4364                 ops = br_dev->netdev_ops;
4365                 if (!del_bulk) {
4366                         if (ops->ndo_fdb_del)
4367                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4368                 } else {
4369                         if (ops->ndo_fdb_del_bulk)
4370                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4371                                                             extack);
4372                 }
4373
4374                 if (err)
4375                         goto out;
4376                 else
4377                         ndm->ndm_flags &= ~NTF_MASTER;
4378         }
4379
4380         /* Embedded bridge, macvlan, and any other device support */
4381         if (ndm->ndm_flags & NTF_SELF) {
4382                 ops = dev->netdev_ops;
4383                 if (!del_bulk) {
4384                         if (ops->ndo_fdb_del)
4385                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4386                         else
4387                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4388                 } else {
4389                         /* in case err was cleared by NTF_MASTER call */
4390                         err = -EOPNOTSUPP;
4391                         if (ops->ndo_fdb_del_bulk)
4392                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4393                                                             extack);
4394                 }
4395
4396                 if (!err) {
4397                         if (!del_bulk)
4398                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4399                                                 ndm->ndm_state);
4400                         ndm->ndm_flags &= ~NTF_SELF;
4401                 }
4402         }
4403 out:
4404         return err;
4405 }
4406
4407 static int nlmsg_populate_fdb(struct sk_buff *skb,
4408                               struct netlink_callback *cb,
4409                               struct net_device *dev,
4410                               int *idx,
4411                               struct netdev_hw_addr_list *list)
4412 {
4413         struct netdev_hw_addr *ha;
4414         int err;
4415         u32 portid, seq;
4416
4417         portid = NETLINK_CB(cb->skb).portid;
4418         seq = cb->nlh->nlmsg_seq;
4419
4420         list_for_each_entry(ha, &list->list, list) {
4421                 if (*idx < cb->args[2])
4422                         goto skip;
4423
4424                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4425                                               portid, seq,
4426                                               RTM_NEWNEIGH, NTF_SELF,
4427                                               NLM_F_MULTI, NUD_PERMANENT);
4428                 if (err < 0)
4429                         return err;
4430 skip:
4431                 *idx += 1;
4432         }
4433         return 0;
4434 }
4435
4436 /**
4437  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4438  * @skb: socket buffer to store message in
4439  * @cb: netlink callback
4440  * @dev: netdevice
4441  * @filter_dev: ignored
4442  * @idx: the number of FDB table entries dumped is added to *@idx
4443  *
4444  * Default netdevice operation to dump the existing unicast address list.
4445  * Returns number of addresses from list put in skb.
4446  */
4447 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4448                       struct netlink_callback *cb,
4449                       struct net_device *dev,
4450                       struct net_device *filter_dev,
4451                       int *idx)
4452 {
4453         int err;
4454
4455         if (dev->type != ARPHRD_ETHER)
4456                 return -EINVAL;
4457
4458         netif_addr_lock_bh(dev);
4459         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4460         if (err)
4461                 goto out;
4462         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4463 out:
4464         netif_addr_unlock_bh(dev);
4465         return err;
4466 }
4467 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4468
4469 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4470                                  int *br_idx, int *brport_idx,
4471                                  struct netlink_ext_ack *extack)
4472 {
4473         struct nlattr *tb[NDA_MAX + 1];
4474         struct ndmsg *ndm;
4475         int err, i;
4476
4477         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4478                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4479                 return -EINVAL;
4480         }
4481
4482         ndm = nlmsg_data(nlh);
4483         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4484             ndm->ndm_flags || ndm->ndm_type) {
4485                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4486                 return -EINVAL;
4487         }
4488
4489         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4490                                             NDA_MAX, NULL, extack);
4491         if (err < 0)
4492                 return err;
4493
4494         *brport_idx = ndm->ndm_ifindex;
4495         for (i = 0; i <= NDA_MAX; ++i) {
4496                 if (!tb[i])
4497                         continue;
4498
4499                 switch (i) {
4500                 case NDA_IFINDEX:
4501                         if (nla_len(tb[i]) != sizeof(u32)) {
4502                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4503                                 return -EINVAL;
4504                         }
4505                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4506                         break;
4507                 case NDA_MASTER:
4508                         if (nla_len(tb[i]) != sizeof(u32)) {
4509                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4510                                 return -EINVAL;
4511                         }
4512                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4513                         break;
4514                 default:
4515                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4516                         return -EINVAL;
4517                 }
4518         }
4519
4520         return 0;
4521 }
4522
4523 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4524                                  int *br_idx, int *brport_idx,
4525                                  struct netlink_ext_ack *extack)
4526 {
4527         struct nlattr *tb[IFLA_MAX+1];
4528         int err;
4529
4530         /* A hack to preserve kernel<->userspace interface.
4531          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4532          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4533          * So, check for ndmsg with an optional u32 attribute (not used here).
4534          * Fortunately these sizes don't conflict with the size of ifinfomsg
4535          * with an optional attribute.
4536          */
4537         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4538             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4539              nla_attr_size(sizeof(u32)))) {
4540                 struct ifinfomsg *ifm;
4541
4542                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4543                                              tb, IFLA_MAX, ifla_policy,
4544                                              extack);
4545                 if (err < 0) {
4546                         return -EINVAL;
4547                 } else if (err == 0) {
4548                         if (tb[IFLA_MASTER])
4549                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4550                 }
4551
4552                 ifm = nlmsg_data(nlh);
4553                 *brport_idx = ifm->ifi_index;
4554         }
4555         return 0;
4556 }
4557
4558 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4559 {
4560         struct net_device *dev;
4561         struct net_device *br_dev = NULL;
4562         const struct net_device_ops *ops = NULL;
4563         const struct net_device_ops *cops = NULL;
4564         struct net *net = sock_net(skb->sk);
4565         struct hlist_head *head;
4566         int brport_idx = 0;
4567         int br_idx = 0;
4568         int h, s_h;
4569         int idx = 0, s_idx;
4570         int err = 0;
4571         int fidx = 0;
4572
4573         if (cb->strict_check)
4574                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4575                                             cb->extack);
4576         else
4577                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4578                                             cb->extack);
4579         if (err < 0)
4580                 return err;
4581
4582         if (br_idx) {
4583                 br_dev = __dev_get_by_index(net, br_idx);
4584                 if (!br_dev)
4585                         return -ENODEV;
4586
4587                 ops = br_dev->netdev_ops;
4588         }
4589
4590         s_h = cb->args[0];
4591         s_idx = cb->args[1];
4592
4593         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4594                 idx = 0;
4595                 head = &net->dev_index_head[h];
4596                 hlist_for_each_entry(dev, head, index_hlist) {
4597
4598                         if (brport_idx && (dev->ifindex != brport_idx))
4599                                 continue;
4600
4601                         if (!br_idx) { /* user did not specify a specific bridge */
4602                                 if (netif_is_bridge_port(dev)) {
4603                                         br_dev = netdev_master_upper_dev_get(dev);
4604                                         cops = br_dev->netdev_ops;
4605                                 }
4606                         } else {
4607                                 if (dev != br_dev &&
4608                                     !netif_is_bridge_port(dev))
4609                                         continue;
4610
4611                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4612                                     !netif_is_bridge_master(dev))
4613                                         continue;
4614                                 cops = ops;
4615                         }
4616
4617                         if (idx < s_idx)
4618                                 goto cont;
4619
4620                         if (netif_is_bridge_port(dev)) {
4621                                 if (cops && cops->ndo_fdb_dump) {
4622                                         err = cops->ndo_fdb_dump(skb, cb,
4623                                                                 br_dev, dev,
4624                                                                 &fidx);
4625                                         if (err == -EMSGSIZE)
4626                                                 goto out;
4627                                 }
4628                         }
4629
4630                         if (dev->netdev_ops->ndo_fdb_dump)
4631                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4632                                                                     dev, NULL,
4633                                                                     &fidx);
4634                         else
4635                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4636                                                         &fidx);
4637                         if (err == -EMSGSIZE)
4638                                 goto out;
4639
4640                         cops = NULL;
4641
4642                         /* reset fdb offset to 0 for rest of the interfaces */
4643                         cb->args[2] = 0;
4644                         fidx = 0;
4645 cont:
4646                         idx++;
4647                 }
4648         }
4649
4650 out:
4651         cb->args[0] = h;
4652         cb->args[1] = idx;
4653         cb->args[2] = fidx;
4654
4655         return skb->len;
4656 }
4657
4658 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4659                                 struct nlattr **tb, u8 *ndm_flags,
4660                                 int *br_idx, int *brport_idx, u8 **addr,
4661                                 u16 *vid, struct netlink_ext_ack *extack)
4662 {
4663         struct ndmsg *ndm;
4664         int err, i;
4665
4666         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4667                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4668                 return -EINVAL;
4669         }
4670
4671         ndm = nlmsg_data(nlh);
4672         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4673             ndm->ndm_type) {
4674                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4675                 return -EINVAL;
4676         }
4677
4678         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4679                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4680                 return -EINVAL;
4681         }
4682
4683         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4684                                             NDA_MAX, nda_policy, extack);
4685         if (err < 0)
4686                 return err;
4687
4688         *ndm_flags = ndm->ndm_flags;
4689         *brport_idx = ndm->ndm_ifindex;
4690         for (i = 0; i <= NDA_MAX; ++i) {
4691                 if (!tb[i])
4692                         continue;
4693
4694                 switch (i) {
4695                 case NDA_MASTER:
4696                         *br_idx = nla_get_u32(tb[i]);
4697                         break;
4698                 case NDA_LLADDR:
4699                         if (nla_len(tb[i]) != ETH_ALEN) {
4700                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4701                                 return -EINVAL;
4702                         }
4703                         *addr = nla_data(tb[i]);
4704                         break;
4705                 case NDA_VLAN:
4706                         err = fdb_vid_parse(tb[i], vid, extack);
4707                         if (err)
4708                                 return err;
4709                         break;
4710                 case NDA_VNI:
4711                         break;
4712                 default:
4713                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4714                         return -EINVAL;
4715                 }
4716         }
4717
4718         return 0;
4719 }
4720
4721 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4722                         struct netlink_ext_ack *extack)
4723 {
4724         struct net_device *dev = NULL, *br_dev = NULL;
4725         const struct net_device_ops *ops = NULL;
4726         struct net *net = sock_net(in_skb->sk);
4727         struct nlattr *tb[NDA_MAX + 1];
4728         struct sk_buff *skb;
4729         int brport_idx = 0;
4730         u8 ndm_flags = 0;
4731         int br_idx = 0;
4732         u8 *addr = NULL;
4733         u16 vid = 0;
4734         int err;
4735
4736         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4737                                    &brport_idx, &addr, &vid, extack);
4738         if (err < 0)
4739                 return err;
4740
4741         if (!addr) {
4742                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4743                 return -EINVAL;
4744         }
4745
4746         if (brport_idx) {
4747                 dev = __dev_get_by_index(net, brport_idx);
4748                 if (!dev) {
4749                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4750                         return -ENODEV;
4751                 }
4752         }
4753
4754         if (br_idx) {
4755                 if (dev) {
4756                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4757                         return -EINVAL;
4758                 }
4759
4760                 br_dev = __dev_get_by_index(net, br_idx);
4761                 if (!br_dev) {
4762                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4763                         return -EINVAL;
4764                 }
4765                 ops = br_dev->netdev_ops;
4766         }
4767
4768         if (dev) {
4769                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4770                         if (!netif_is_bridge_port(dev)) {
4771                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4772                                 return -EINVAL;
4773                         }
4774                         br_dev = netdev_master_upper_dev_get(dev);
4775                         if (!br_dev) {
4776                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4777                                 return -EINVAL;
4778                         }
4779                         ops = br_dev->netdev_ops;
4780                 } else {
4781                         if (!(ndm_flags & NTF_SELF)) {
4782                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4783                                 return -EINVAL;
4784                         }
4785                         ops = dev->netdev_ops;
4786                 }
4787         }
4788
4789         if (!br_dev && !dev) {
4790                 NL_SET_ERR_MSG(extack, "No device specified");
4791                 return -ENODEV;
4792         }
4793
4794         if (!ops || !ops->ndo_fdb_get) {
4795                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4796                 return -EOPNOTSUPP;
4797         }
4798
4799         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4800         if (!skb)
4801                 return -ENOBUFS;
4802
4803         if (br_dev)
4804                 dev = br_dev;
4805         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4806                                NETLINK_CB(in_skb).portid,
4807                                nlh->nlmsg_seq, extack);
4808         if (err)
4809                 goto out;
4810
4811         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4812 out:
4813         kfree_skb(skb);
4814         return err;
4815 }
4816
4817 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4818                                unsigned int attrnum, unsigned int flag)
4819 {
4820         if (mask & flag)
4821                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4822         return 0;
4823 }
4824
4825 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4826                             struct net_device *dev, u16 mode,
4827                             u32 flags, u32 mask, int nlflags,
4828                             u32 filter_mask,
4829                             int (*vlan_fill)(struct sk_buff *skb,
4830                                              struct net_device *dev,
4831                                              u32 filter_mask))
4832 {
4833         struct nlmsghdr *nlh;
4834         struct ifinfomsg *ifm;
4835         struct nlattr *br_afspec;
4836         struct nlattr *protinfo;
4837         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4838         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4839         int err = 0;
4840
4841         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4842         if (nlh == NULL)
4843                 return -EMSGSIZE;
4844
4845         ifm = nlmsg_data(nlh);
4846         ifm->ifi_family = AF_BRIDGE;
4847         ifm->__ifi_pad = 0;
4848         ifm->ifi_type = dev->type;
4849         ifm->ifi_index = dev->ifindex;
4850         ifm->ifi_flags = dev_get_flags(dev);
4851         ifm->ifi_change = 0;
4852
4853
4854         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4855             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4856             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4857             (br_dev &&
4858              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4859             (dev->addr_len &&
4860              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4861             (dev->ifindex != dev_get_iflink(dev) &&
4862              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4863                 goto nla_put_failure;
4864
4865         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4866         if (!br_afspec)
4867                 goto nla_put_failure;
4868
4869         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4870                 nla_nest_cancel(skb, br_afspec);
4871                 goto nla_put_failure;
4872         }
4873
4874         if (mode != BRIDGE_MODE_UNDEF) {
4875                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4876                         nla_nest_cancel(skb, br_afspec);
4877                         goto nla_put_failure;
4878                 }
4879         }
4880         if (vlan_fill) {
4881                 err = vlan_fill(skb, dev, filter_mask);
4882                 if (err) {
4883                         nla_nest_cancel(skb, br_afspec);
4884                         goto nla_put_failure;
4885                 }
4886         }
4887         nla_nest_end(skb, br_afspec);
4888
4889         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4890         if (!protinfo)
4891                 goto nla_put_failure;
4892
4893         if (brport_nla_put_flag(skb, flags, mask,
4894                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4895             brport_nla_put_flag(skb, flags, mask,
4896                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4897             brport_nla_put_flag(skb, flags, mask,
4898                                 IFLA_BRPORT_FAST_LEAVE,
4899                                 BR_MULTICAST_FAST_LEAVE) ||
4900             brport_nla_put_flag(skb, flags, mask,
4901                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4902             brport_nla_put_flag(skb, flags, mask,
4903                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4904             brport_nla_put_flag(skb, flags, mask,
4905                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4906             brport_nla_put_flag(skb, flags, mask,
4907                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4908             brport_nla_put_flag(skb, flags, mask,
4909                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4910             brport_nla_put_flag(skb, flags, mask,
4911                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4912             brport_nla_put_flag(skb, flags, mask,
4913                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4914                 nla_nest_cancel(skb, protinfo);
4915                 goto nla_put_failure;
4916         }
4917
4918         nla_nest_end(skb, protinfo);
4919
4920         nlmsg_end(skb, nlh);
4921         return 0;
4922 nla_put_failure:
4923         nlmsg_cancel(skb, nlh);
4924         return err ? err : -EMSGSIZE;
4925 }
4926 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4927
4928 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4929                                     bool strict_check, u32 *filter_mask,
4930                                     struct netlink_ext_ack *extack)
4931 {
4932         struct nlattr *tb[IFLA_MAX+1];
4933         int err, i;
4934
4935         if (strict_check) {
4936                 struct ifinfomsg *ifm;
4937
4938                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4939                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4940                         return -EINVAL;
4941                 }
4942
4943                 ifm = nlmsg_data(nlh);
4944                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4945                     ifm->ifi_change || ifm->ifi_index) {
4946                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4947                         return -EINVAL;
4948                 }
4949
4950                 err = nlmsg_parse_deprecated_strict(nlh,
4951                                                     sizeof(struct ifinfomsg),
4952                                                     tb, IFLA_MAX, ifla_policy,
4953                                                     extack);
4954         } else {
4955                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4956                                              tb, IFLA_MAX, ifla_policy,
4957                                              extack);
4958         }
4959         if (err < 0)
4960                 return err;
4961
4962         /* new attributes should only be added with strict checking */
4963         for (i = 0; i <= IFLA_MAX; ++i) {
4964                 if (!tb[i])
4965                         continue;
4966
4967                 switch (i) {
4968                 case IFLA_EXT_MASK:
4969                         *filter_mask = nla_get_u32(tb[i]);
4970                         break;
4971                 default:
4972                         if (strict_check) {
4973                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4974                                 return -EINVAL;
4975                         }
4976                 }
4977         }
4978
4979         return 0;
4980 }
4981
4982 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4983 {
4984         const struct nlmsghdr *nlh = cb->nlh;
4985         struct net *net = sock_net(skb->sk);
4986         struct net_device *dev;
4987         int idx = 0;
4988         u32 portid = NETLINK_CB(cb->skb).portid;
4989         u32 seq = nlh->nlmsg_seq;
4990         u32 filter_mask = 0;
4991         int err;
4992
4993         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4994                                        cb->extack);
4995         if (err < 0 && cb->strict_check)
4996                 return err;
4997
4998         rcu_read_lock();
4999         for_each_netdev_rcu(net, dev) {
5000                 const struct net_device_ops *ops = dev->netdev_ops;
5001                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5002
5003                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5004                         if (idx >= cb->args[0]) {
5005                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5006                                                 skb, portid, seq, dev,
5007                                                 filter_mask, NLM_F_MULTI);
5008                                 if (err < 0 && err != -EOPNOTSUPP) {
5009                                         if (likely(skb->len))
5010                                                 break;
5011
5012                                         goto out_err;
5013                                 }
5014                         }
5015                         idx++;
5016                 }
5017
5018                 if (ops->ndo_bridge_getlink) {
5019                         if (idx >= cb->args[0]) {
5020                                 err = ops->ndo_bridge_getlink(skb, portid,
5021                                                               seq, dev,
5022                                                               filter_mask,
5023                                                               NLM_F_MULTI);
5024                                 if (err < 0 && err != -EOPNOTSUPP) {
5025                                         if (likely(skb->len))
5026                                                 break;
5027
5028                                         goto out_err;
5029                                 }
5030                         }
5031                         idx++;
5032                 }
5033         }
5034         err = skb->len;
5035 out_err:
5036         rcu_read_unlock();
5037         cb->args[0] = idx;
5038
5039         return err;
5040 }
5041
5042 static inline size_t bridge_nlmsg_size(void)
5043 {
5044         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5045                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5046                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5047                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5048                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5049                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5050                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5051                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5052                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5053                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5054                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5055 }
5056
5057 static int rtnl_bridge_notify(struct net_device *dev)
5058 {
5059         struct net *net = dev_net(dev);
5060         struct sk_buff *skb;
5061         int err = -EOPNOTSUPP;
5062
5063         if (!dev->netdev_ops->ndo_bridge_getlink)
5064                 return 0;
5065
5066         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5067         if (!skb) {
5068                 err = -ENOMEM;
5069                 goto errout;
5070         }
5071
5072         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5073         if (err < 0)
5074                 goto errout;
5075
5076         /* Notification info is only filled for bridge ports, not the bridge
5077          * device itself. Therefore, a zero notification length is valid and
5078          * should not result in an error.
5079          */
5080         if (!skb->len)
5081                 goto errout;
5082
5083         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5084         return 0;
5085 errout:
5086         WARN_ON(err == -EMSGSIZE);
5087         kfree_skb(skb);
5088         if (err)
5089                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5090         return err;
5091 }
5092
5093 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5094                                struct netlink_ext_ack *extack)
5095 {
5096         struct net *net = sock_net(skb->sk);
5097         struct ifinfomsg *ifm;
5098         struct net_device *dev;
5099         struct nlattr *br_spec, *attr = NULL;
5100         int rem, err = -EOPNOTSUPP;
5101         u16 flags = 0;
5102         bool have_flags = false;
5103
5104         if (nlmsg_len(nlh) < sizeof(*ifm))
5105                 return -EINVAL;
5106
5107         ifm = nlmsg_data(nlh);
5108         if (ifm->ifi_family != AF_BRIDGE)
5109                 return -EPFNOSUPPORT;
5110
5111         dev = __dev_get_by_index(net, ifm->ifi_index);
5112         if (!dev) {
5113                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5114                 return -ENODEV;
5115         }
5116
5117         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5118         if (br_spec) {
5119                 nla_for_each_nested(attr, br_spec, rem) {
5120                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5121                                 if (nla_len(attr) < sizeof(flags))
5122                                         return -EINVAL;
5123
5124                                 have_flags = true;
5125                                 flags = nla_get_u16(attr);
5126                                 break;
5127                         }
5128                 }
5129         }
5130
5131         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5132                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5133
5134                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5135                         err = -EOPNOTSUPP;
5136                         goto out;
5137                 }
5138
5139                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5140                                                              extack);
5141                 if (err)
5142                         goto out;
5143
5144                 flags &= ~BRIDGE_FLAGS_MASTER;
5145         }
5146
5147         if ((flags & BRIDGE_FLAGS_SELF)) {
5148                 if (!dev->netdev_ops->ndo_bridge_setlink)
5149                         err = -EOPNOTSUPP;
5150                 else
5151                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5152                                                                   flags,
5153                                                                   extack);
5154                 if (!err) {
5155                         flags &= ~BRIDGE_FLAGS_SELF;
5156
5157                         /* Generate event to notify upper layer of bridge
5158                          * change
5159                          */
5160                         err = rtnl_bridge_notify(dev);
5161                 }
5162         }
5163
5164         if (have_flags)
5165                 memcpy(nla_data(attr), &flags, sizeof(flags));
5166 out:
5167         return err;
5168 }
5169
5170 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5171                                struct netlink_ext_ack *extack)
5172 {
5173         struct net *net = sock_net(skb->sk);
5174         struct ifinfomsg *ifm;
5175         struct net_device *dev;
5176         struct nlattr *br_spec, *attr = NULL;
5177         int rem, err = -EOPNOTSUPP;
5178         u16 flags = 0;
5179         bool have_flags = false;
5180
5181         if (nlmsg_len(nlh) < sizeof(*ifm))
5182                 return -EINVAL;
5183
5184         ifm = nlmsg_data(nlh);
5185         if (ifm->ifi_family != AF_BRIDGE)
5186                 return -EPFNOSUPPORT;
5187
5188         dev = __dev_get_by_index(net, ifm->ifi_index);
5189         if (!dev) {
5190                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5191                 return -ENODEV;
5192         }
5193
5194         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5195         if (br_spec) {
5196                 nla_for_each_nested(attr, br_spec, rem) {
5197                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5198                                 if (nla_len(attr) < sizeof(flags))
5199                                         return -EINVAL;
5200
5201                                 have_flags = true;
5202                                 flags = nla_get_u16(attr);
5203                                 break;
5204                         }
5205                 }
5206         }
5207
5208         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5209                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5210
5211                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5212                         err = -EOPNOTSUPP;
5213                         goto out;
5214                 }
5215
5216                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5217                 if (err)
5218                         goto out;
5219
5220                 flags &= ~BRIDGE_FLAGS_MASTER;
5221         }
5222
5223         if ((flags & BRIDGE_FLAGS_SELF)) {
5224                 if (!dev->netdev_ops->ndo_bridge_dellink)
5225                         err = -EOPNOTSUPP;
5226                 else
5227                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5228                                                                   flags);
5229
5230                 if (!err) {
5231                         flags &= ~BRIDGE_FLAGS_SELF;
5232
5233                         /* Generate event to notify upper layer of bridge
5234                          * change
5235                          */
5236                         err = rtnl_bridge_notify(dev);
5237                 }
5238         }
5239
5240         if (have_flags)
5241                 memcpy(nla_data(attr), &flags, sizeof(flags));
5242 out:
5243         return err;
5244 }
5245
5246 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5247 {
5248         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5249                (!idxattr || idxattr == attrid);
5250 }
5251
5252 static bool
5253 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5254 {
5255         return dev->netdev_ops &&
5256                dev->netdev_ops->ndo_has_offload_stats &&
5257                dev->netdev_ops->ndo_get_offload_stats &&
5258                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5259 }
5260
5261 static unsigned int
5262 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5263 {
5264         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5265                sizeof(struct rtnl_link_stats64) : 0;
5266 }
5267
5268 static int
5269 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5270                              struct sk_buff *skb)
5271 {
5272         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5273         struct nlattr *attr = NULL;
5274         void *attr_data;
5275         int err;
5276
5277         if (!size)
5278                 return -ENODATA;
5279
5280         attr = nla_reserve_64bit(skb, attr_id, size,
5281                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5282         if (!attr)
5283                 return -EMSGSIZE;
5284
5285         attr_data = nla_data(attr);
5286         memset(attr_data, 0, size);
5287
5288         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5289         if (err)
5290                 return err;
5291
5292         return 0;
5293 }
5294
5295 static unsigned int
5296 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5297                                    enum netdev_offload_xstats_type type)
5298 {
5299         bool enabled = netdev_offload_xstats_enabled(dev, type);
5300
5301         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5302 }
5303
5304 struct rtnl_offload_xstats_request_used {
5305         bool request;
5306         bool used;
5307 };
5308
5309 static int
5310 rtnl_offload_xstats_get_stats(struct net_device *dev,
5311                               enum netdev_offload_xstats_type type,
5312                               struct rtnl_offload_xstats_request_used *ru,
5313                               struct rtnl_hw_stats64 *stats,
5314                               struct netlink_ext_ack *extack)
5315 {
5316         bool request;
5317         bool used;
5318         int err;
5319
5320         request = netdev_offload_xstats_enabled(dev, type);
5321         if (!request) {
5322                 used = false;
5323                 goto out;
5324         }
5325
5326         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5327         if (err)
5328                 return err;
5329
5330 out:
5331         if (ru) {
5332                 ru->request = request;
5333                 ru->used = used;
5334         }
5335         return 0;
5336 }
5337
5338 static int
5339 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5340                                        struct rtnl_offload_xstats_request_used *ru)
5341 {
5342         struct nlattr *nest;
5343
5344         nest = nla_nest_start(skb, attr_id);
5345         if (!nest)
5346                 return -EMSGSIZE;
5347
5348         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5349                 goto nla_put_failure;
5350
5351         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5352                 goto nla_put_failure;
5353
5354         nla_nest_end(skb, nest);
5355         return 0;
5356
5357 nla_put_failure:
5358         nla_nest_cancel(skb, nest);
5359         return -EMSGSIZE;
5360 }
5361
5362 static int
5363 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5364                                    struct netlink_ext_ack *extack)
5365 {
5366         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5367         struct rtnl_offload_xstats_request_used ru_l3;
5368         struct nlattr *nest;
5369         int err;
5370
5371         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5372         if (err)
5373                 return err;
5374
5375         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5376         if (!nest)
5377                 return -EMSGSIZE;
5378
5379         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5380                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5381                                                    &ru_l3))
5382                 goto nla_put_failure;
5383
5384         nla_nest_end(skb, nest);
5385         return 0;
5386
5387 nla_put_failure:
5388         nla_nest_cancel(skb, nest);
5389         return -EMSGSIZE;
5390 }
5391
5392 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5393                                     int *prividx, u32 off_filter_mask,
5394                                     struct netlink_ext_ack *extack)
5395 {
5396         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5397         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5398         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5399         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5400         bool have_data = false;
5401         int err;
5402
5403         if (*prividx <= attr_id_cpu_hit &&
5404             (off_filter_mask &
5405              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5406                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5407                 if (!err) {
5408                         have_data = true;
5409                 } else if (err != -ENODATA) {
5410                         *prividx = attr_id_cpu_hit;
5411                         return err;
5412                 }
5413         }
5414
5415         if (*prividx <= attr_id_hw_s_info &&
5416             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5417                 *prividx = attr_id_hw_s_info;
5418
5419                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5420                 if (err)
5421                         return err;
5422
5423                 have_data = true;
5424                 *prividx = 0;
5425         }
5426
5427         if (*prividx <= attr_id_l3_stats &&
5428             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5429                 unsigned int size_l3;
5430                 struct nlattr *attr;
5431
5432                 *prividx = attr_id_l3_stats;
5433
5434                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5435                 if (!size_l3)
5436                         goto skip_l3_stats;
5437                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5438                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5439                 if (!attr)
5440                         return -EMSGSIZE;
5441
5442                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5443                                                     nla_data(attr), extack);
5444                 if (err)
5445                         return err;
5446
5447                 have_data = true;
5448 skip_l3_stats:
5449                 *prividx = 0;
5450         }
5451
5452         if (!have_data)
5453                 return -ENODATA;
5454
5455         *prividx = 0;
5456         return 0;
5457 }
5458
5459 static unsigned int
5460 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5461                                            enum netdev_offload_xstats_type type)
5462 {
5463         bool enabled = netdev_offload_xstats_enabled(dev, type);
5464
5465         return nla_total_size(0) +
5466                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5467                 nla_total_size(sizeof(u8)) +
5468                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5469                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5470                 0;
5471 }
5472
5473 static unsigned int
5474 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5475 {
5476         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5477
5478         return nla_total_size(0) +
5479                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5480                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5481                 0;
5482 }
5483
5484 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5485                                         u32 off_filter_mask)
5486 {
5487         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5488         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5489         int nla_size = 0;
5490         int size;
5491
5492         if (off_filter_mask &
5493             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5494                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5495                 nla_size += nla_total_size_64bit(size);
5496         }
5497
5498         if (off_filter_mask &
5499             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5500                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5501
5502         if (off_filter_mask &
5503             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5504                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5505                 nla_size += nla_total_size_64bit(size);
5506         }
5507
5508         if (nla_size != 0)
5509                 nla_size += nla_total_size(0);
5510
5511         return nla_size;
5512 }
5513
5514 struct rtnl_stats_dump_filters {
5515         /* mask[0] filters outer attributes. Then individual nests have their
5516          * filtering mask at the index of the nested attribute.
5517          */
5518         u32 mask[IFLA_STATS_MAX + 1];
5519 };
5520
5521 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5522                                int type, u32 pid, u32 seq, u32 change,
5523                                unsigned int flags,
5524                                const struct rtnl_stats_dump_filters *filters,
5525                                int *idxattr, int *prividx,
5526                                struct netlink_ext_ack *extack)
5527 {
5528         unsigned int filter_mask = filters->mask[0];
5529         struct if_stats_msg *ifsm;
5530         struct nlmsghdr *nlh;
5531         struct nlattr *attr;
5532         int s_prividx = *prividx;
5533         int err;
5534
5535         ASSERT_RTNL();
5536
5537         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5538         if (!nlh)
5539                 return -EMSGSIZE;
5540
5541         ifsm = nlmsg_data(nlh);
5542         ifsm->family = PF_UNSPEC;
5543         ifsm->pad1 = 0;
5544         ifsm->pad2 = 0;
5545         ifsm->ifindex = dev->ifindex;
5546         ifsm->filter_mask = filter_mask;
5547
5548         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5549                 struct rtnl_link_stats64 *sp;
5550
5551                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5552                                          sizeof(struct rtnl_link_stats64),
5553                                          IFLA_STATS_UNSPEC);
5554                 if (!attr) {
5555                         err = -EMSGSIZE;
5556                         goto nla_put_failure;
5557                 }
5558
5559                 sp = nla_data(attr);
5560                 dev_get_stats(dev, sp);
5561         }
5562
5563         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5564                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5565
5566                 if (ops && ops->fill_linkxstats) {
5567                         *idxattr = IFLA_STATS_LINK_XSTATS;
5568                         attr = nla_nest_start_noflag(skb,
5569                                                      IFLA_STATS_LINK_XSTATS);
5570                         if (!attr) {
5571                                 err = -EMSGSIZE;
5572                                 goto nla_put_failure;
5573                         }
5574
5575                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5576                         nla_nest_end(skb, attr);
5577                         if (err)
5578                                 goto nla_put_failure;
5579                         *idxattr = 0;
5580                 }
5581         }
5582
5583         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5584                              *idxattr)) {
5585                 const struct rtnl_link_ops *ops = NULL;
5586                 const struct net_device *master;
5587
5588                 master = netdev_master_upper_dev_get(dev);
5589                 if (master)
5590                         ops = master->rtnl_link_ops;
5591                 if (ops && ops->fill_linkxstats) {
5592                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5593                         attr = nla_nest_start_noflag(skb,
5594                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5595                         if (!attr) {
5596                                 err = -EMSGSIZE;
5597                                 goto nla_put_failure;
5598                         }
5599
5600                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5601                         nla_nest_end(skb, attr);
5602                         if (err)
5603                                 goto nla_put_failure;
5604                         *idxattr = 0;
5605                 }
5606         }
5607
5608         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5609                              *idxattr)) {
5610                 u32 off_filter_mask;
5611
5612                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5613                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5614                 attr = nla_nest_start_noflag(skb,
5615                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5616                 if (!attr) {
5617                         err = -EMSGSIZE;
5618                         goto nla_put_failure;
5619                 }
5620
5621                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5622                                                off_filter_mask, extack);
5623                 if (err == -ENODATA)
5624                         nla_nest_cancel(skb, attr);
5625                 else
5626                         nla_nest_end(skb, attr);
5627
5628                 if (err && err != -ENODATA)
5629                         goto nla_put_failure;
5630                 *idxattr = 0;
5631         }
5632
5633         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5634                 struct rtnl_af_ops *af_ops;
5635
5636                 *idxattr = IFLA_STATS_AF_SPEC;
5637                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5638                 if (!attr) {
5639                         err = -EMSGSIZE;
5640                         goto nla_put_failure;
5641                 }
5642
5643                 rcu_read_lock();
5644                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5645                         if (af_ops->fill_stats_af) {
5646                                 struct nlattr *af;
5647
5648                                 af = nla_nest_start_noflag(skb,
5649                                                            af_ops->family);
5650                                 if (!af) {
5651                                         rcu_read_unlock();
5652                                         err = -EMSGSIZE;
5653                                         goto nla_put_failure;
5654                                 }
5655                                 err = af_ops->fill_stats_af(skb, dev);
5656
5657                                 if (err == -ENODATA) {
5658                                         nla_nest_cancel(skb, af);
5659                                 } else if (err < 0) {
5660                                         rcu_read_unlock();
5661                                         goto nla_put_failure;
5662                                 }
5663
5664                                 nla_nest_end(skb, af);
5665                         }
5666                 }
5667                 rcu_read_unlock();
5668
5669                 nla_nest_end(skb, attr);
5670
5671                 *idxattr = 0;
5672         }
5673
5674         nlmsg_end(skb, nlh);
5675
5676         return 0;
5677
5678 nla_put_failure:
5679         /* not a multi message or no progress mean a real error */
5680         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5681                 nlmsg_cancel(skb, nlh);
5682         else
5683                 nlmsg_end(skb, nlh);
5684
5685         return err;
5686 }
5687
5688 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5689                                   const struct rtnl_stats_dump_filters *filters)
5690 {
5691         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5692         unsigned int filter_mask = filters->mask[0];
5693
5694         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5695                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5696
5697         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5698                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5699                 int attr = IFLA_STATS_LINK_XSTATS;
5700
5701                 if (ops && ops->get_linkxstats_size) {
5702                         size += nla_total_size(ops->get_linkxstats_size(dev,
5703                                                                         attr));
5704                         /* for IFLA_STATS_LINK_XSTATS */
5705                         size += nla_total_size(0);
5706                 }
5707         }
5708
5709         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5710                 struct net_device *_dev = (struct net_device *)dev;
5711                 const struct rtnl_link_ops *ops = NULL;
5712                 const struct net_device *master;
5713
5714                 /* netdev_master_upper_dev_get can't take const */
5715                 master = netdev_master_upper_dev_get(_dev);
5716                 if (master)
5717                         ops = master->rtnl_link_ops;
5718                 if (ops && ops->get_linkxstats_size) {
5719                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5720
5721                         size += nla_total_size(ops->get_linkxstats_size(dev,
5722                                                                         attr));
5723                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5724                         size += nla_total_size(0);
5725                 }
5726         }
5727
5728         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5729                 u32 off_filter_mask;
5730
5731                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5732                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5733         }
5734
5735         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5736                 struct rtnl_af_ops *af_ops;
5737
5738                 /* for IFLA_STATS_AF_SPEC */
5739                 size += nla_total_size(0);
5740
5741                 rcu_read_lock();
5742                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5743                         if (af_ops->get_stats_af_size) {
5744                                 size += nla_total_size(
5745                                         af_ops->get_stats_af_size(dev));
5746
5747                                 /* for AF_* */
5748                                 size += nla_total_size(0);
5749                         }
5750                 }
5751                 rcu_read_unlock();
5752         }
5753
5754         return size;
5755 }
5756
5757 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5758
5759 static const struct nla_policy
5760 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5761         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5762                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5763 };
5764
5765 static const struct nla_policy
5766 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5767         [IFLA_STATS_GET_FILTERS] =
5768                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5769 };
5770
5771 static const struct nla_policy
5772 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5773         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5774 };
5775
5776 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5777                                         struct rtnl_stats_dump_filters *filters,
5778                                         struct netlink_ext_ack *extack)
5779 {
5780         struct nlattr *tb[IFLA_STATS_MAX + 1];
5781         int err;
5782         int at;
5783
5784         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5785                                rtnl_stats_get_policy_filters, extack);
5786         if (err < 0)
5787                 return err;
5788
5789         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5790                 if (tb[at]) {
5791                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5792                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5793                                 return -EINVAL;
5794                         }
5795                         filters->mask[at] = nla_get_u32(tb[at]);
5796                 }
5797         }
5798
5799         return 0;
5800 }
5801
5802 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5803                                 u32 filter_mask,
5804                                 struct rtnl_stats_dump_filters *filters,
5805                                 struct netlink_ext_ack *extack)
5806 {
5807         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5808         int err;
5809         int i;
5810
5811         filters->mask[0] = filter_mask;
5812         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5813                 filters->mask[i] = -1U;
5814
5815         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5816                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5817         if (err < 0)
5818                 return err;
5819
5820         if (tb[IFLA_STATS_GET_FILTERS]) {
5821                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5822                                                    filters, extack);
5823                 if (err)
5824                         return err;
5825         }
5826
5827         return 0;
5828 }
5829
5830 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5831                                 bool is_dump, struct netlink_ext_ack *extack)
5832 {
5833         struct if_stats_msg *ifsm;
5834
5835         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5836                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5837                 return -EINVAL;
5838         }
5839
5840         if (!strict_check)
5841                 return 0;
5842
5843         ifsm = nlmsg_data(nlh);
5844
5845         /* only requests using strict checks can pass data to influence
5846          * the dump. The legacy exception is filter_mask.
5847          */
5848         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5849                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5850                 return -EINVAL;
5851         }
5852         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5853                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5854                 return -EINVAL;
5855         }
5856
5857         return 0;
5858 }
5859
5860 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5861                           struct netlink_ext_ack *extack)
5862 {
5863         struct rtnl_stats_dump_filters filters;
5864         struct net *net = sock_net(skb->sk);
5865         struct net_device *dev = NULL;
5866         int idxattr = 0, prividx = 0;
5867         struct if_stats_msg *ifsm;
5868         struct sk_buff *nskb;
5869         int err;
5870
5871         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5872                                    false, extack);
5873         if (err)
5874                 return err;
5875
5876         ifsm = nlmsg_data(nlh);
5877         if (ifsm->ifindex > 0)
5878                 dev = __dev_get_by_index(net, ifsm->ifindex);
5879         else
5880                 return -EINVAL;
5881
5882         if (!dev)
5883                 return -ENODEV;
5884
5885         if (!ifsm->filter_mask) {
5886                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5887                 return -EINVAL;
5888         }
5889
5890         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5891         if (err)
5892                 return err;
5893
5894         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5895         if (!nskb)
5896                 return -ENOBUFS;
5897
5898         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5899                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5900                                   0, &filters, &idxattr, &prividx, extack);
5901         if (err < 0) {
5902                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5903                 WARN_ON(err == -EMSGSIZE);
5904                 kfree_skb(nskb);
5905         } else {
5906                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5907         }
5908
5909         return err;
5910 }
5911
5912 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5913 {
5914         struct netlink_ext_ack *extack = cb->extack;
5915         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5916         struct rtnl_stats_dump_filters filters;
5917         struct net *net = sock_net(skb->sk);
5918         unsigned int flags = NLM_F_MULTI;
5919         struct if_stats_msg *ifsm;
5920         struct hlist_head *head;
5921         struct net_device *dev;
5922         int idx = 0;
5923
5924         s_h = cb->args[0];
5925         s_idx = cb->args[1];
5926         s_idxattr = cb->args[2];
5927         s_prividx = cb->args[3];
5928
5929         cb->seq = net->dev_base_seq;
5930
5931         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5932         if (err)
5933                 return err;
5934
5935         ifsm = nlmsg_data(cb->nlh);
5936         if (!ifsm->filter_mask) {
5937                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5938                 return -EINVAL;
5939         }
5940
5941         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5942                                    extack);
5943         if (err)
5944                 return err;
5945
5946         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5947                 idx = 0;
5948                 head = &net->dev_index_head[h];
5949                 hlist_for_each_entry(dev, head, index_hlist) {
5950                         if (idx < s_idx)
5951                                 goto cont;
5952                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5953                                                   NETLINK_CB(cb->skb).portid,
5954                                                   cb->nlh->nlmsg_seq, 0,
5955                                                   flags, &filters,
5956                                                   &s_idxattr, &s_prividx,
5957                                                   extack);
5958                         /* If we ran out of room on the first message,
5959                          * we're in trouble
5960                          */
5961                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5962
5963                         if (err < 0)
5964                                 goto out;
5965                         s_prividx = 0;
5966                         s_idxattr = 0;
5967                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5968 cont:
5969                         idx++;
5970                 }
5971         }
5972 out:
5973         cb->args[3] = s_prividx;
5974         cb->args[2] = s_idxattr;
5975         cb->args[1] = idx;
5976         cb->args[0] = h;
5977
5978         return skb->len;
5979 }
5980
5981 void rtnl_offload_xstats_notify(struct net_device *dev)
5982 {
5983         struct rtnl_stats_dump_filters response_filters = {};
5984         struct net *net = dev_net(dev);
5985         int idxattr = 0, prividx = 0;
5986         struct sk_buff *skb;
5987         int err = -ENOBUFS;
5988
5989         ASSERT_RTNL();
5990
5991         response_filters.mask[0] |=
5992                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5993         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5994                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5995
5996         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
5997                         GFP_KERNEL);
5998         if (!skb)
5999                 goto errout;
6000
6001         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6002                                   &response_filters, &idxattr, &prividx, NULL);
6003         if (err < 0) {
6004                 kfree_skb(skb);
6005                 goto errout;
6006         }
6007
6008         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6009         return;
6010
6011 errout:
6012         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6013 }
6014 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6015
6016 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6017                           struct netlink_ext_ack *extack)
6018 {
6019         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6020         struct rtnl_stats_dump_filters response_filters = {};
6021         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6022         struct net *net = sock_net(skb->sk);
6023         struct net_device *dev = NULL;
6024         struct if_stats_msg *ifsm;
6025         bool notify = false;
6026         int err;
6027
6028         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6029                                    false, extack);
6030         if (err)
6031                 return err;
6032
6033         ifsm = nlmsg_data(nlh);
6034         if (ifsm->family != AF_UNSPEC) {
6035                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6036                 return -EINVAL;
6037         }
6038
6039         if (ifsm->ifindex > 0)
6040                 dev = __dev_get_by_index(net, ifsm->ifindex);
6041         else
6042                 return -EINVAL;
6043
6044         if (!dev)
6045                 return -ENODEV;
6046
6047         if (ifsm->filter_mask) {
6048                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6049                 return -EINVAL;
6050         }
6051
6052         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6053                           ifla_stats_set_policy, extack);
6054         if (err < 0)
6055                 return err;
6056
6057         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6058                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6059
6060                 if (req)
6061                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6062                 else
6063                         err = netdev_offload_xstats_disable(dev, t_l3);
6064
6065                 if (!err)
6066                         notify = true;
6067                 else if (err != -EALREADY)
6068                         return err;
6069
6070                 response_filters.mask[0] |=
6071                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6072                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6073                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6074         }
6075
6076         if (notify)
6077                 rtnl_offload_xstats_notify(dev);
6078
6079         return 0;
6080 }
6081
6082 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6083                                    struct netlink_ext_ack *extack)
6084 {
6085         struct br_port_msg *bpm;
6086
6087         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6088                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6089                 return -EINVAL;
6090         }
6091
6092         bpm = nlmsg_data(nlh);
6093         if (bpm->ifindex) {
6094                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6095                 return -EINVAL;
6096         }
6097         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6098                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6099                 return -EINVAL;
6100         }
6101
6102         return 0;
6103 }
6104
6105 struct rtnl_mdb_dump_ctx {
6106         long idx;
6107 };
6108
6109 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6110 {
6111         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6112         struct net *net = sock_net(skb->sk);
6113         struct net_device *dev;
6114         int idx, s_idx;
6115         int err;
6116
6117         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6118
6119         if (cb->strict_check) {
6120                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6121                 if (err)
6122                         return err;
6123         }
6124
6125         s_idx = ctx->idx;
6126         idx = 0;
6127
6128         for_each_netdev(net, dev) {
6129                 if (idx < s_idx)
6130                         goto skip;
6131                 if (!dev->netdev_ops->ndo_mdb_dump)
6132                         goto skip;
6133
6134                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6135                 if (err == -EMSGSIZE)
6136                         goto out;
6137                 /* Moving on to next device, reset markers and sequence
6138                  * counters since they are all maintained per-device.
6139                  */
6140                 memset(cb->ctx, 0, sizeof(cb->ctx));
6141                 cb->prev_seq = 0;
6142                 cb->seq = 0;
6143 skip:
6144                 idx++;
6145         }
6146
6147 out:
6148         ctx->idx = idx;
6149         return skb->len;
6150 }
6151
6152 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6153                                    struct netlink_ext_ack *extack)
6154 {
6155         struct br_mdb_entry *entry = nla_data(attr);
6156
6157         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6158                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6159                 return -EINVAL;
6160         }
6161
6162         if (entry->ifindex == 0) {
6163                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6164                 return -EINVAL;
6165         }
6166
6167         if (entry->addr.proto == htons(ETH_P_IP)) {
6168                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6169                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6170                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6171                         return -EINVAL;
6172                 }
6173                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6174                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6175                         return -EINVAL;
6176                 }
6177 #if IS_ENABLED(CONFIG_IPV6)
6178         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6179                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6180                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6181                         return -EINVAL;
6182                 }
6183 #endif
6184         } else if (entry->addr.proto == 0) {
6185                 /* L2 mdb */
6186                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6187                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6188                         return -EINVAL;
6189                 }
6190         } else {
6191                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6192                 return -EINVAL;
6193         }
6194
6195         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6196                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6197                 return -EINVAL;
6198         }
6199         if (entry->vid >= VLAN_VID_MASK) {
6200                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6201                 return -EINVAL;
6202         }
6203
6204         return 0;
6205 }
6206
6207 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6208         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6209         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6210                                                   rtnl_validate_mdb_entry,
6211                                                   sizeof(struct br_mdb_entry)),
6212         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6213 };
6214
6215 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6216                         struct netlink_ext_ack *extack)
6217 {
6218         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6219         struct net *net = sock_net(skb->sk);
6220         struct br_port_msg *bpm;
6221         struct net_device *dev;
6222         int err;
6223
6224         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6225                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6226         if (err)
6227                 return err;
6228
6229         bpm = nlmsg_data(nlh);
6230         if (!bpm->ifindex) {
6231                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6232                 return -EINVAL;
6233         }
6234
6235         dev = __dev_get_by_index(net, bpm->ifindex);
6236         if (!dev) {
6237                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6238                 return -ENODEV;
6239         }
6240
6241         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6242                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6243                 return -EINVAL;
6244         }
6245
6246         if (!dev->netdev_ops->ndo_mdb_add) {
6247                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6248                 return -EOPNOTSUPP;
6249         }
6250
6251         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6252 }
6253
6254 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6255                         struct netlink_ext_ack *extack)
6256 {
6257         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6258         struct net *net = sock_net(skb->sk);
6259         struct br_port_msg *bpm;
6260         struct net_device *dev;
6261         int err;
6262
6263         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6264                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6265         if (err)
6266                 return err;
6267
6268         bpm = nlmsg_data(nlh);
6269         if (!bpm->ifindex) {
6270                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6271                 return -EINVAL;
6272         }
6273
6274         dev = __dev_get_by_index(net, bpm->ifindex);
6275         if (!dev) {
6276                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6277                 return -ENODEV;
6278         }
6279
6280         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6281                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6282                 return -EINVAL;
6283         }
6284
6285         if (!dev->netdev_ops->ndo_mdb_del) {
6286                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6287                 return -EOPNOTSUPP;
6288         }
6289
6290         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6291 }
6292
6293 /* Process one rtnetlink message. */
6294
6295 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6296                              struct netlink_ext_ack *extack)
6297 {
6298         struct net *net = sock_net(skb->sk);
6299         struct rtnl_link *link;
6300         enum rtnl_kinds kind;
6301         struct module *owner;
6302         int err = -EOPNOTSUPP;
6303         rtnl_doit_func doit;
6304         unsigned int flags;
6305         int family;
6306         int type;
6307
6308         type = nlh->nlmsg_type;
6309         if (type > RTM_MAX)
6310                 return -EOPNOTSUPP;
6311
6312         type -= RTM_BASE;
6313
6314         /* All the messages must have at least 1 byte length */
6315         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6316                 return 0;
6317
6318         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6319         kind = rtnl_msgtype_kind(type);
6320
6321         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6322                 return -EPERM;
6323
6324         rcu_read_lock();
6325         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6326                 struct sock *rtnl;
6327                 rtnl_dumpit_func dumpit;
6328                 u32 min_dump_alloc = 0;
6329
6330                 link = rtnl_get_link(family, type);
6331                 if (!link || !link->dumpit) {
6332                         family = PF_UNSPEC;
6333                         link = rtnl_get_link(family, type);
6334                         if (!link || !link->dumpit)
6335                                 goto err_unlock;
6336                 }
6337                 owner = link->owner;
6338                 dumpit = link->dumpit;
6339
6340                 if (type == RTM_GETLINK - RTM_BASE)
6341                         min_dump_alloc = rtnl_calcit(skb, nlh);
6342
6343                 err = 0;
6344                 /* need to do this before rcu_read_unlock() */
6345                 if (!try_module_get(owner))
6346                         err = -EPROTONOSUPPORT;
6347
6348                 rcu_read_unlock();
6349
6350                 rtnl = net->rtnl;
6351                 if (err == 0) {
6352                         struct netlink_dump_control c = {
6353                                 .dump           = dumpit,
6354                                 .min_dump_alloc = min_dump_alloc,
6355                                 .module         = owner,
6356                         };
6357                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6358                         /* netlink_dump_start() will keep a reference on
6359                          * module if dump is still in progress.
6360                          */
6361                         module_put(owner);
6362                 }
6363                 return err;
6364         }
6365
6366         link = rtnl_get_link(family, type);
6367         if (!link || !link->doit) {
6368                 family = PF_UNSPEC;
6369                 link = rtnl_get_link(PF_UNSPEC, type);
6370                 if (!link || !link->doit)
6371                         goto out_unlock;
6372         }
6373
6374         owner = link->owner;
6375         if (!try_module_get(owner)) {
6376                 err = -EPROTONOSUPPORT;
6377                 goto out_unlock;
6378         }
6379
6380         flags = link->flags;
6381         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6382             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6383                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6384                 module_put(owner);
6385                 goto err_unlock;
6386         }
6387
6388         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6389                 doit = link->doit;
6390                 rcu_read_unlock();
6391                 if (doit)
6392                         err = doit(skb, nlh, extack);
6393                 module_put(owner);
6394                 return err;
6395         }
6396         rcu_read_unlock();
6397
6398         rtnl_lock();
6399         link = rtnl_get_link(family, type);
6400         if (link && link->doit)
6401                 err = link->doit(skb, nlh, extack);
6402         rtnl_unlock();
6403
6404         module_put(owner);
6405
6406         return err;
6407
6408 out_unlock:
6409         rcu_read_unlock();
6410         return err;
6411
6412 err_unlock:
6413         rcu_read_unlock();
6414         return -EOPNOTSUPP;
6415 }
6416
6417 static void rtnetlink_rcv(struct sk_buff *skb)
6418 {
6419         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6420 }
6421
6422 static int rtnetlink_bind(struct net *net, int group)
6423 {
6424         switch (group) {
6425         case RTNLGRP_IPV4_MROUTE_R:
6426         case RTNLGRP_IPV6_MROUTE_R:
6427                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6428                         return -EPERM;
6429                 break;
6430         }
6431         return 0;
6432 }
6433
6434 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6435 {
6436         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6437
6438         switch (event) {
6439         case NETDEV_REBOOT:
6440         case NETDEV_CHANGEMTU:
6441         case NETDEV_CHANGEADDR:
6442         case NETDEV_CHANGENAME:
6443         case NETDEV_FEAT_CHANGE:
6444         case NETDEV_BONDING_FAILOVER:
6445         case NETDEV_POST_TYPE_CHANGE:
6446         case NETDEV_NOTIFY_PEERS:
6447         case NETDEV_CHANGEUPPER:
6448         case NETDEV_RESEND_IGMP:
6449         case NETDEV_CHANGEINFODATA:
6450         case NETDEV_CHANGELOWERSTATE:
6451         case NETDEV_CHANGE_TX_QUEUE_LEN:
6452                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6453                                    GFP_KERNEL, NULL, 0, 0, NULL);
6454                 break;
6455         default:
6456                 break;
6457         }
6458         return NOTIFY_DONE;
6459 }
6460
6461 static struct notifier_block rtnetlink_dev_notifier = {
6462         .notifier_call  = rtnetlink_event,
6463 };
6464
6465
6466 static int __net_init rtnetlink_net_init(struct net *net)
6467 {
6468         struct sock *sk;
6469         struct netlink_kernel_cfg cfg = {
6470                 .groups         = RTNLGRP_MAX,
6471                 .input          = rtnetlink_rcv,
6472                 .cb_mutex       = &rtnl_mutex,
6473                 .flags          = NL_CFG_F_NONROOT_RECV,
6474                 .bind           = rtnetlink_bind,
6475         };
6476
6477         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6478         if (!sk)
6479                 return -ENOMEM;
6480         net->rtnl = sk;
6481         return 0;
6482 }
6483
6484 static void __net_exit rtnetlink_net_exit(struct net *net)
6485 {
6486         netlink_kernel_release(net->rtnl);
6487         net->rtnl = NULL;
6488 }
6489
6490 static struct pernet_operations rtnetlink_net_ops = {
6491         .init = rtnetlink_net_init,
6492         .exit = rtnetlink_net_exit,
6493 };
6494
6495 void __init rtnetlink_init(void)
6496 {
6497         if (register_pernet_subsys(&rtnetlink_net_ops))
6498                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6499
6500         register_netdevice_notifier(&rtnetlink_dev_notifier);
6501
6502         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6503                       rtnl_dump_ifinfo, 0);
6504         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6505         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6506         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6507
6508         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6509         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6510         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6511
6512         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6513         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6514
6515         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6516         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6517                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6518         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6519
6520         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6521         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6522         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6523
6524         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6525                       0);
6526         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6527
6528         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, rtnl_mdb_dump, 0);
6529         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6530         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL, 0);
6531 }