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
3289         if (tb[IFLA_NUM_TX_QUEUES])
3290                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3291         else if (ops->get_num_tx_queues)
3292                 num_tx_queues = ops->get_num_tx_queues();
3293
3294         if (tb[IFLA_NUM_RX_QUEUES])
3295                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3296         else if (ops->get_num_rx_queues)
3297                 num_rx_queues = ops->get_num_rx_queues();
3298
3299         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3300                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3301                 return ERR_PTR(-EINVAL);
3302         }
3303
3304         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3305                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3306                 return ERR_PTR(-EINVAL);
3307         }
3308
3309         if (ops->alloc) {
3310                 dev = ops->alloc(tb, ifname, name_assign_type,
3311                                  num_tx_queues, num_rx_queues);
3312                 if (IS_ERR(dev))
3313                         return dev;
3314         } else {
3315                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3316                                        name_assign_type, ops->setup,
3317                                        num_tx_queues, num_rx_queues);
3318         }
3319
3320         if (!dev)
3321                 return ERR_PTR(-ENOMEM);
3322
3323         dev_net_set(dev, net);
3324         dev->rtnl_link_ops = ops;
3325         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3326
3327         if (tb[IFLA_MTU]) {
3328                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3329                 int err;
3330
3331                 err = dev_validate_mtu(dev, mtu, extack);
3332                 if (err) {
3333                         free_netdev(dev);
3334                         return ERR_PTR(err);
3335                 }
3336                 dev->mtu = mtu;
3337         }
3338         if (tb[IFLA_ADDRESS]) {
3339                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3340                                nla_len(tb[IFLA_ADDRESS]));
3341                 dev->addr_assign_type = NET_ADDR_SET;
3342         }
3343         if (tb[IFLA_BROADCAST])
3344                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3345                                 nla_len(tb[IFLA_BROADCAST]));
3346         if (tb[IFLA_TXQLEN])
3347                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3348         if (tb[IFLA_OPERSTATE])
3349                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3350         if (tb[IFLA_LINKMODE])
3351                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3352         if (tb[IFLA_GROUP])
3353                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3354         if (tb[IFLA_GSO_MAX_SIZE])
3355                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3356         if (tb[IFLA_GSO_MAX_SEGS])
3357                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3358         if (tb[IFLA_GRO_MAX_SIZE])
3359                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3360         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3361                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3362         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3363                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3364
3365         return dev;
3366 }
3367 EXPORT_SYMBOL(rtnl_create_link);
3368
3369 static int rtnl_group_changelink(const struct sk_buff *skb,
3370                 struct net *net, int group,
3371                 struct ifinfomsg *ifm,
3372                 struct netlink_ext_ack *extack,
3373                 struct nlattr **tb)
3374 {
3375         struct net_device *dev, *aux;
3376         int err;
3377
3378         for_each_netdev_safe(net, dev, aux) {
3379                 if (dev->group == group) {
3380                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3381                         if (err < 0)
3382                                 return err;
3383                 }
3384         }
3385
3386         return 0;
3387 }
3388
3389 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3390                                const struct rtnl_link_ops *ops,
3391                                const struct nlmsghdr *nlh,
3392                                struct nlattr **tb, struct nlattr **data,
3393                                struct netlink_ext_ack *extack)
3394 {
3395         unsigned char name_assign_type = NET_NAME_USER;
3396         struct net *net = sock_net(skb->sk);
3397         u32 portid = NETLINK_CB(skb).portid;
3398         struct net *dest_net, *link_net;
3399         struct net_device *dev;
3400         char ifname[IFNAMSIZ];
3401         int err;
3402
3403         if (!ops->alloc && !ops->setup)
3404                 return -EOPNOTSUPP;
3405
3406         if (tb[IFLA_IFNAME]) {
3407                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3408         } else {
3409                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3410                 name_assign_type = NET_NAME_ENUM;
3411         }
3412
3413         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3414         if (IS_ERR(dest_net))
3415                 return PTR_ERR(dest_net);
3416
3417         if (tb[IFLA_LINK_NETNSID]) {
3418                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3419
3420                 link_net = get_net_ns_by_id(dest_net, id);
3421                 if (!link_net) {
3422                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3423                         err =  -EINVAL;
3424                         goto out;
3425                 }
3426                 err = -EPERM;
3427                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3428                         goto out;
3429         } else {
3430                 link_net = NULL;
3431         }
3432
3433         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3434                                name_assign_type, ops, tb, extack);
3435         if (IS_ERR(dev)) {
3436                 err = PTR_ERR(dev);
3437                 goto out;
3438         }
3439
3440         dev->ifindex = ifm->ifi_index;
3441
3442         if (ops->newlink)
3443                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3444         else
3445                 err = register_netdevice(dev);
3446         if (err < 0) {
3447                 free_netdev(dev);
3448                 goto out;
3449         }
3450
3451         err = rtnl_configure_link(dev, ifm, portid, nlh);
3452         if (err < 0)
3453                 goto out_unregister;
3454         if (link_net) {
3455                 err = dev_change_net_namespace(dev, dest_net, ifname);
3456                 if (err < 0)
3457                         goto out_unregister;
3458         }
3459         if (tb[IFLA_MASTER]) {
3460                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3461                 if (err)
3462                         goto out_unregister;
3463         }
3464 out:
3465         if (link_net)
3466                 put_net(link_net);
3467         put_net(dest_net);
3468         return err;
3469 out_unregister:
3470         if (ops->newlink) {
3471                 LIST_HEAD(list_kill);
3472
3473                 ops->dellink(dev, &list_kill);
3474                 unregister_netdevice_many(&list_kill);
3475         } else {
3476                 unregister_netdevice(dev);
3477         }
3478         goto out;
3479 }
3480
3481 struct rtnl_newlink_tbs {
3482         struct nlattr *tb[IFLA_MAX + 1];
3483         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3484         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3485 };
3486
3487 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3488                           struct rtnl_newlink_tbs *tbs,
3489                           struct netlink_ext_ack *extack)
3490 {
3491         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3492         struct nlattr ** const tb = tbs->tb;
3493         const struct rtnl_link_ops *m_ops;
3494         struct net_device *master_dev;
3495         struct net *net = sock_net(skb->sk);
3496         const struct rtnl_link_ops *ops;
3497         struct nlattr **slave_data;
3498         char kind[MODULE_NAME_LEN];
3499         struct net_device *dev;
3500         struct ifinfomsg *ifm;
3501         struct nlattr **data;
3502         bool link_specified;
3503         int err;
3504
3505 #ifdef CONFIG_MODULES
3506 replay:
3507 #endif
3508         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3509                                      ifla_policy, extack);
3510         if (err < 0)
3511                 return err;
3512
3513         err = rtnl_ensure_unique_netns(tb, extack, false);
3514         if (err < 0)
3515                 return err;
3516
3517         ifm = nlmsg_data(nlh);
3518         if (ifm->ifi_index > 0) {
3519                 link_specified = true;
3520                 dev = __dev_get_by_index(net, ifm->ifi_index);
3521         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3522                 link_specified = true;
3523                 dev = rtnl_dev_get(net, tb);
3524         } else {
3525                 link_specified = false;
3526                 dev = NULL;
3527         }
3528
3529         master_dev = NULL;
3530         m_ops = NULL;
3531         if (dev) {
3532                 master_dev = netdev_master_upper_dev_get(dev);
3533                 if (master_dev)
3534                         m_ops = master_dev->rtnl_link_ops;
3535         }
3536
3537         err = validate_linkmsg(dev, tb, extack);
3538         if (err < 0)
3539                 return err;
3540
3541         if (tb[IFLA_LINKINFO]) {
3542                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3543                                                   tb[IFLA_LINKINFO],
3544                                                   ifla_info_policy, NULL);
3545                 if (err < 0)
3546                         return err;
3547         } else
3548                 memset(linkinfo, 0, sizeof(linkinfo));
3549
3550         if (linkinfo[IFLA_INFO_KIND]) {
3551                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3552                 ops = rtnl_link_ops_get(kind);
3553         } else {
3554                 kind[0] = '\0';
3555                 ops = NULL;
3556         }
3557
3558         data = NULL;
3559         if (ops) {
3560                 if (ops->maxtype > RTNL_MAX_TYPE)
3561                         return -EINVAL;
3562
3563                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3564                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3565                                                           linkinfo[IFLA_INFO_DATA],
3566                                                           ops->policy, extack);
3567                         if (err < 0)
3568                                 return err;
3569                         data = tbs->attr;
3570                 }
3571                 if (ops->validate) {
3572                         err = ops->validate(tb, data, extack);
3573                         if (err < 0)
3574                                 return err;
3575                 }
3576         }
3577
3578         slave_data = NULL;
3579         if (m_ops) {
3580                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3581                         return -EINVAL;
3582
3583                 if (m_ops->slave_maxtype &&
3584                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3585                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3586                                                           m_ops->slave_maxtype,
3587                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3588                                                           m_ops->slave_policy,
3589                                                           extack);
3590                         if (err < 0)
3591                                 return err;
3592                         slave_data = tbs->slave_attr;
3593                 }
3594         }
3595
3596         if (dev) {
3597                 int status = 0;
3598
3599                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3600                         return -EEXIST;
3601                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3602                         return -EOPNOTSUPP;
3603
3604                 if (linkinfo[IFLA_INFO_DATA]) {
3605                         if (!ops || ops != dev->rtnl_link_ops ||
3606                             !ops->changelink)
3607                                 return -EOPNOTSUPP;
3608
3609                         err = ops->changelink(dev, tb, data, extack);
3610                         if (err < 0)
3611                                 return err;
3612                         status |= DO_SETLINK_NOTIFY;
3613                 }
3614
3615                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3616                         if (!m_ops || !m_ops->slave_changelink)
3617                                 return -EOPNOTSUPP;
3618
3619                         err = m_ops->slave_changelink(master_dev, dev, tb,
3620                                                       slave_data, extack);
3621                         if (err < 0)
3622                                 return err;
3623                         status |= DO_SETLINK_NOTIFY;
3624                 }
3625
3626                 return do_setlink(skb, dev, ifm, extack, tb, status);
3627         }
3628
3629         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3630                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3631                  * or it's for a group
3632                 */
3633                 if (link_specified)
3634                         return -ENODEV;
3635                 if (tb[IFLA_GROUP])
3636                         return rtnl_group_changelink(skb, net,
3637                                                 nla_get_u32(tb[IFLA_GROUP]),
3638                                                 ifm, extack, tb);
3639                 return -ENODEV;
3640         }
3641
3642         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3643                 return -EOPNOTSUPP;
3644
3645         if (!ops) {
3646 #ifdef CONFIG_MODULES
3647                 if (kind[0]) {
3648                         __rtnl_unlock();
3649                         request_module("rtnl-link-%s", kind);
3650                         rtnl_lock();
3651                         ops = rtnl_link_ops_get(kind);
3652                         if (ops)
3653                                 goto replay;
3654                 }
3655 #endif
3656                 NL_SET_ERR_MSG(extack, "Unknown device type");
3657                 return -EOPNOTSUPP;
3658         }
3659
3660         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3661 }
3662
3663 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3664                         struct netlink_ext_ack *extack)
3665 {
3666         struct rtnl_newlink_tbs *tbs;
3667         int ret;
3668
3669         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3670         if (!tbs)
3671                 return -ENOMEM;
3672
3673         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3674         kfree(tbs);
3675         return ret;
3676 }
3677
3678 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3679                                   const struct nlmsghdr *nlh,
3680                                   struct nlattr **tb,
3681                                   struct netlink_ext_ack *extack)
3682 {
3683         struct ifinfomsg *ifm;
3684         int i, err;
3685
3686         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3687                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3688                 return -EINVAL;
3689         }
3690
3691         if (!netlink_strict_get_check(skb))
3692                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3693                                               ifla_policy, extack);
3694
3695         ifm = nlmsg_data(nlh);
3696         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3697             ifm->ifi_change) {
3698                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3699                 return -EINVAL;
3700         }
3701
3702         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3703                                             ifla_policy, extack);
3704         if (err)
3705                 return err;
3706
3707         for (i = 0; i <= IFLA_MAX; i++) {
3708                 if (!tb[i])
3709                         continue;
3710
3711                 switch (i) {
3712                 case IFLA_IFNAME:
3713                 case IFLA_ALT_IFNAME:
3714                 case IFLA_EXT_MASK:
3715                 case IFLA_TARGET_NETNSID:
3716                         break;
3717                 default:
3718                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3719                         return -EINVAL;
3720                 }
3721         }
3722
3723         return 0;
3724 }
3725
3726 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3727                         struct netlink_ext_ack *extack)
3728 {
3729         struct net *net = sock_net(skb->sk);
3730         struct net *tgt_net = net;
3731         struct ifinfomsg *ifm;
3732         struct nlattr *tb[IFLA_MAX+1];
3733         struct net_device *dev = NULL;
3734         struct sk_buff *nskb;
3735         int netnsid = -1;
3736         int err;
3737         u32 ext_filter_mask = 0;
3738
3739         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3740         if (err < 0)
3741                 return err;
3742
3743         err = rtnl_ensure_unique_netns(tb, extack, true);
3744         if (err < 0)
3745                 return err;
3746
3747         if (tb[IFLA_TARGET_NETNSID]) {
3748                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3749                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3750                 if (IS_ERR(tgt_net))
3751                         return PTR_ERR(tgt_net);
3752         }
3753
3754         if (tb[IFLA_EXT_MASK])
3755                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3756
3757         err = -EINVAL;
3758         ifm = nlmsg_data(nlh);
3759         if (ifm->ifi_index > 0)
3760                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3761         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3762                 dev = rtnl_dev_get(tgt_net, tb);
3763         else
3764                 goto out;
3765
3766         err = -ENODEV;
3767         if (dev == NULL)
3768                 goto out;
3769
3770         err = -ENOBUFS;
3771         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3772         if (nskb == NULL)
3773                 goto out;
3774
3775         err = rtnl_fill_ifinfo(nskb, dev, net,
3776                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3777                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3778                                0, NULL, 0, netnsid, GFP_KERNEL);
3779         if (err < 0) {
3780                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3781                 WARN_ON(err == -EMSGSIZE);
3782                 kfree_skb(nskb);
3783         } else
3784                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3785 out:
3786         if (netnsid >= 0)
3787                 put_net(tgt_net);
3788
3789         return err;
3790 }
3791
3792 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3793                            bool *changed, struct netlink_ext_ack *extack)
3794 {
3795         char *alt_ifname;
3796         size_t size;
3797         int err;
3798
3799         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3800         if (err)
3801                 return err;
3802
3803         if (cmd == RTM_NEWLINKPROP) {
3804                 size = rtnl_prop_list_size(dev);
3805                 size += nla_total_size(ALTIFNAMSIZ);
3806                 if (size >= U16_MAX) {
3807                         NL_SET_ERR_MSG(extack,
3808                                        "effective property list too long");
3809                         return -EINVAL;
3810                 }
3811         }
3812
3813         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3814         if (!alt_ifname)
3815                 return -ENOMEM;
3816
3817         if (cmd == RTM_NEWLINKPROP) {
3818                 err = netdev_name_node_alt_create(dev, alt_ifname);
3819                 if (!err)
3820                         alt_ifname = NULL;
3821         } else if (cmd == RTM_DELLINKPROP) {
3822                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3823         } else {
3824                 WARN_ON_ONCE(1);
3825                 err = -EINVAL;
3826         }
3827
3828         kfree(alt_ifname);
3829         if (!err)
3830                 *changed = true;
3831         return err;
3832 }
3833
3834 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3835                          struct netlink_ext_ack *extack)
3836 {
3837         struct net *net = sock_net(skb->sk);
3838         struct nlattr *tb[IFLA_MAX + 1];
3839         struct net_device *dev;
3840         struct ifinfomsg *ifm;
3841         bool changed = false;
3842         struct nlattr *attr;
3843         int err, rem;
3844
3845         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3846         if (err)
3847                 return err;
3848
3849         err = rtnl_ensure_unique_netns(tb, extack, true);
3850         if (err)
3851                 return err;
3852
3853         ifm = nlmsg_data(nlh);
3854         if (ifm->ifi_index > 0)
3855                 dev = __dev_get_by_index(net, ifm->ifi_index);
3856         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3857                 dev = rtnl_dev_get(net, tb);
3858         else
3859                 return -EINVAL;
3860
3861         if (!dev)
3862                 return -ENODEV;
3863
3864         if (!tb[IFLA_PROP_LIST])
3865                 return 0;
3866
3867         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3868                 switch (nla_type(attr)) {
3869                 case IFLA_ALT_IFNAME:
3870                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3871                         if (err)
3872                                 return err;
3873                         break;
3874                 }
3875         }
3876
3877         if (changed)
3878                 netdev_state_change(dev);
3879         return 0;
3880 }
3881
3882 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3883                             struct netlink_ext_ack *extack)
3884 {
3885         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3886 }
3887
3888 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3889                             struct netlink_ext_ack *extack)
3890 {
3891         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3892 }
3893
3894 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3895 {
3896         struct net *net = sock_net(skb->sk);
3897         size_t min_ifinfo_dump_size = 0;
3898         struct nlattr *tb[IFLA_MAX+1];
3899         u32 ext_filter_mask = 0;
3900         struct net_device *dev;
3901         int hdrlen;
3902
3903         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3904         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3905                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3906
3907         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3908                 if (tb[IFLA_EXT_MASK])
3909                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3910         }
3911
3912         if (!ext_filter_mask)
3913                 return NLMSG_GOODSIZE;
3914         /*
3915          * traverse the list of net devices and compute the minimum
3916          * buffer size based upon the filter mask.
3917          */
3918         rcu_read_lock();
3919         for_each_netdev_rcu(net, dev) {
3920                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3921                                            if_nlmsg_size(dev, ext_filter_mask));
3922         }
3923         rcu_read_unlock();
3924
3925         return nlmsg_total_size(min_ifinfo_dump_size);
3926 }
3927
3928 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3929 {
3930         int idx;
3931         int s_idx = cb->family;
3932         int type = cb->nlh->nlmsg_type - RTM_BASE;
3933         int ret = 0;
3934
3935         if (s_idx == 0)
3936                 s_idx = 1;
3937
3938         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3939                 struct rtnl_link __rcu **tab;
3940                 struct rtnl_link *link;
3941                 rtnl_dumpit_func dumpit;
3942
3943                 if (idx < s_idx || idx == PF_PACKET)
3944                         continue;
3945
3946                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3947                         continue;
3948
3949                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3950                 if (!tab)
3951                         continue;
3952
3953                 link = rcu_dereference_rtnl(tab[type]);
3954                 if (!link)
3955                         continue;
3956
3957                 dumpit = link->dumpit;
3958                 if (!dumpit)
3959                         continue;
3960
3961                 if (idx > s_idx) {
3962                         memset(&cb->args[0], 0, sizeof(cb->args));
3963                         cb->prev_seq = 0;
3964                         cb->seq = 0;
3965                 }
3966                 ret = dumpit(skb, cb);
3967                 if (ret)
3968                         break;
3969         }
3970         cb->family = idx;
3971
3972         return skb->len ? : ret;
3973 }
3974
3975 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3976                                        unsigned int change,
3977                                        u32 event, gfp_t flags, int *new_nsid,
3978                                        int new_ifindex, u32 portid,
3979                                        const struct nlmsghdr *nlh)
3980 {
3981         struct net *net = dev_net(dev);
3982         struct sk_buff *skb;
3983         int err = -ENOBUFS;
3984         u32 seq = 0;
3985
3986         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3987         if (skb == NULL)
3988                 goto errout;
3989
3990         if (nlmsg_report(nlh))
3991                 seq = nlmsg_seq(nlh);
3992         else
3993                 portid = 0;
3994
3995         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3996                                type, portid, seq, change, 0, 0, event,
3997                                new_nsid, new_ifindex, -1, flags);
3998         if (err < 0) {
3999                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4000                 WARN_ON(err == -EMSGSIZE);
4001                 kfree_skb(skb);
4002                 goto errout;
4003         }
4004         return skb;
4005 errout:
4006         if (err < 0)
4007                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4008         return NULL;
4009 }
4010
4011 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4012                        u32 portid, const struct nlmsghdr *nlh)
4013 {
4014         struct net *net = dev_net(dev);
4015
4016         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4017 }
4018
4019 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4020                                unsigned int change, u32 event,
4021                                gfp_t flags, int *new_nsid, int new_ifindex,
4022                                u32 portid, const struct nlmsghdr *nlh)
4023 {
4024         struct sk_buff *skb;
4025
4026         if (dev->reg_state != NETREG_REGISTERED)
4027                 return;
4028
4029         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4030                                      new_ifindex, portid, nlh);
4031         if (skb)
4032                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4033 }
4034
4035 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4036                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4037 {
4038         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4039                            NULL, 0, portid, nlh);
4040 }
4041
4042 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4043                          gfp_t flags, int *new_nsid, int new_ifindex)
4044 {
4045         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4046                            new_nsid, new_ifindex, 0, NULL);
4047 }
4048
4049 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4050                                    struct net_device *dev,
4051                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4052                                    int type, unsigned int flags,
4053                                    int nlflags, u16 ndm_state)
4054 {
4055         struct nlmsghdr *nlh;
4056         struct ndmsg *ndm;
4057
4058         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4059         if (!nlh)
4060                 return -EMSGSIZE;
4061
4062         ndm = nlmsg_data(nlh);
4063         ndm->ndm_family  = AF_BRIDGE;
4064         ndm->ndm_pad1    = 0;
4065         ndm->ndm_pad2    = 0;
4066         ndm->ndm_flags   = flags;
4067         ndm->ndm_type    = 0;
4068         ndm->ndm_ifindex = dev->ifindex;
4069         ndm->ndm_state   = ndm_state;
4070
4071         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
4072                 goto nla_put_failure;
4073         if (vid)
4074                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4075                         goto nla_put_failure;
4076
4077         nlmsg_end(skb, nlh);
4078         return 0;
4079
4080 nla_put_failure:
4081         nlmsg_cancel(skb, nlh);
4082         return -EMSGSIZE;
4083 }
4084
4085 static inline size_t rtnl_fdb_nlmsg_size(void)
4086 {
4087         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4088                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
4089                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4090                0;
4091 }
4092
4093 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4094                             u16 ndm_state)
4095 {
4096         struct net *net = dev_net(dev);
4097         struct sk_buff *skb;
4098         int err = -ENOBUFS;
4099
4100         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4101         if (!skb)
4102                 goto errout;
4103
4104         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4105                                       0, 0, type, NTF_SELF, 0, ndm_state);
4106         if (err < 0) {
4107                 kfree_skb(skb);
4108                 goto errout;
4109         }
4110
4111         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4112         return;
4113 errout:
4114         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4115 }
4116
4117 /*
4118  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4119  */
4120 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4121                      struct nlattr *tb[],
4122                      struct net_device *dev,
4123                      const unsigned char *addr, u16 vid,
4124                      u16 flags)
4125 {
4126         int err = -EINVAL;
4127
4128         /* If aging addresses are supported device will need to
4129          * implement its own handler for this.
4130          */
4131         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4132                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4133                 return err;
4134         }
4135
4136         if (tb[NDA_FLAGS_EXT]) {
4137                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4138                 return err;
4139         }
4140
4141         if (vid) {
4142                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4143                 return err;
4144         }
4145
4146         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4147                 err = dev_uc_add_excl(dev, addr);
4148         else if (is_multicast_ether_addr(addr))
4149                 err = dev_mc_add_excl(dev, addr);
4150
4151         /* Only return duplicate errors if NLM_F_EXCL is set */
4152         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4153                 err = 0;
4154
4155         return err;
4156 }
4157 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4158
4159 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4160                          struct netlink_ext_ack *extack)
4161 {
4162         u16 vid = 0;
4163
4164         if (vlan_attr) {
4165                 if (nla_len(vlan_attr) != sizeof(u16)) {
4166                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4167                         return -EINVAL;
4168                 }
4169
4170                 vid = nla_get_u16(vlan_attr);
4171
4172                 if (!vid || vid >= VLAN_VID_MASK) {
4173                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4174                         return -EINVAL;
4175                 }
4176         }
4177         *p_vid = vid;
4178         return 0;
4179 }
4180
4181 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4182                         struct netlink_ext_ack *extack)
4183 {
4184         struct net *net = sock_net(skb->sk);
4185         struct ndmsg *ndm;
4186         struct nlattr *tb[NDA_MAX+1];
4187         struct net_device *dev;
4188         u8 *addr;
4189         u16 vid;
4190         int err;
4191
4192         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4193                                      extack);
4194         if (err < 0)
4195                 return err;
4196
4197         ndm = nlmsg_data(nlh);
4198         if (ndm->ndm_ifindex == 0) {
4199                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4200                 return -EINVAL;
4201         }
4202
4203         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4204         if (dev == NULL) {
4205                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4206                 return -ENODEV;
4207         }
4208
4209         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4210                 NL_SET_ERR_MSG(extack, "invalid address");
4211                 return -EINVAL;
4212         }
4213
4214         if (dev->type != ARPHRD_ETHER) {
4215                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4216                 return -EINVAL;
4217         }
4218
4219         addr = nla_data(tb[NDA_LLADDR]);
4220
4221         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4222         if (err)
4223                 return err;
4224
4225         err = -EOPNOTSUPP;
4226
4227         /* Support fdb on master device the net/bridge default case */
4228         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4229             netif_is_bridge_port(dev)) {
4230                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4231                 const struct net_device_ops *ops = br_dev->netdev_ops;
4232
4233                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4234                                        nlh->nlmsg_flags, extack);
4235                 if (err)
4236                         goto out;
4237                 else
4238                         ndm->ndm_flags &= ~NTF_MASTER;
4239         }
4240
4241         /* Embedded bridge, macvlan, and any other device support */
4242         if ((ndm->ndm_flags & NTF_SELF)) {
4243                 if (dev->netdev_ops->ndo_fdb_add)
4244                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4245                                                            vid,
4246                                                            nlh->nlmsg_flags,
4247                                                            extack);
4248                 else
4249                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4250                                                nlh->nlmsg_flags);
4251
4252                 if (!err) {
4253                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4254                                         ndm->ndm_state);
4255                         ndm->ndm_flags &= ~NTF_SELF;
4256                 }
4257         }
4258 out:
4259         return err;
4260 }
4261
4262 /*
4263  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4264  */
4265 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4266                      struct nlattr *tb[],
4267                      struct net_device *dev,
4268                      const unsigned char *addr, u16 vid)
4269 {
4270         int err = -EINVAL;
4271
4272         /* If aging addresses are supported device will need to
4273          * implement its own handler for this.
4274          */
4275         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4276                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4277                 return err;
4278         }
4279
4280         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4281                 err = dev_uc_del(dev, addr);
4282         else if (is_multicast_ether_addr(addr))
4283                 err = dev_mc_del(dev, addr);
4284
4285         return err;
4286 }
4287 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4288
4289 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4290         [NDA_VLAN]      = { .type = NLA_U16 },
4291         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4292         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4293         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4294 };
4295
4296 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4297                         struct netlink_ext_ack *extack)
4298 {
4299         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4300         struct net *net = sock_net(skb->sk);
4301         const struct net_device_ops *ops;
4302         struct ndmsg *ndm;
4303         struct nlattr *tb[NDA_MAX+1];
4304         struct net_device *dev;
4305         __u8 *addr = NULL;
4306         int err;
4307         u16 vid;
4308
4309         if (!netlink_capable(skb, CAP_NET_ADMIN))
4310                 return -EPERM;
4311
4312         if (!del_bulk) {
4313                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4314                                              NULL, extack);
4315         } else {
4316                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4317                                   fdb_del_bulk_policy, extack);
4318         }
4319         if (err < 0)
4320                 return err;
4321
4322         ndm = nlmsg_data(nlh);
4323         if (ndm->ndm_ifindex == 0) {
4324                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4325                 return -EINVAL;
4326         }
4327
4328         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4329         if (dev == NULL) {
4330                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4331                 return -ENODEV;
4332         }
4333
4334         if (!del_bulk) {
4335                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4336                         NL_SET_ERR_MSG(extack, "invalid address");
4337                         return -EINVAL;
4338                 }
4339                 addr = nla_data(tb[NDA_LLADDR]);
4340         }
4341
4342         if (dev->type != ARPHRD_ETHER) {
4343                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4344                 return -EINVAL;
4345         }
4346
4347         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4348         if (err)
4349                 return err;
4350
4351         err = -EOPNOTSUPP;
4352
4353         /* Support fdb on master device the net/bridge default case */
4354         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4355             netif_is_bridge_port(dev)) {
4356                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4357
4358                 ops = br_dev->netdev_ops;
4359                 if (!del_bulk) {
4360                         if (ops->ndo_fdb_del)
4361                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4362                 } else {
4363                         if (ops->ndo_fdb_del_bulk)
4364                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4365                                                             extack);
4366                 }
4367
4368                 if (err)
4369                         goto out;
4370                 else
4371                         ndm->ndm_flags &= ~NTF_MASTER;
4372         }
4373
4374         /* Embedded bridge, macvlan, and any other device support */
4375         if (ndm->ndm_flags & NTF_SELF) {
4376                 ops = dev->netdev_ops;
4377                 if (!del_bulk) {
4378                         if (ops->ndo_fdb_del)
4379                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4380                         else
4381                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4382                 } else {
4383                         /* in case err was cleared by NTF_MASTER call */
4384                         err = -EOPNOTSUPP;
4385                         if (ops->ndo_fdb_del_bulk)
4386                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4387                                                             extack);
4388                 }
4389
4390                 if (!err) {
4391                         if (!del_bulk)
4392                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4393                                                 ndm->ndm_state);
4394                         ndm->ndm_flags &= ~NTF_SELF;
4395                 }
4396         }
4397 out:
4398         return err;
4399 }
4400
4401 static int nlmsg_populate_fdb(struct sk_buff *skb,
4402                               struct netlink_callback *cb,
4403                               struct net_device *dev,
4404                               int *idx,
4405                               struct netdev_hw_addr_list *list)
4406 {
4407         struct netdev_hw_addr *ha;
4408         int err;
4409         u32 portid, seq;
4410
4411         portid = NETLINK_CB(cb->skb).portid;
4412         seq = cb->nlh->nlmsg_seq;
4413
4414         list_for_each_entry(ha, &list->list, list) {
4415                 if (*idx < cb->args[2])
4416                         goto skip;
4417
4418                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4419                                               portid, seq,
4420                                               RTM_NEWNEIGH, NTF_SELF,
4421                                               NLM_F_MULTI, NUD_PERMANENT);
4422                 if (err < 0)
4423                         return err;
4424 skip:
4425                 *idx += 1;
4426         }
4427         return 0;
4428 }
4429
4430 /**
4431  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4432  * @skb: socket buffer to store message in
4433  * @cb: netlink callback
4434  * @dev: netdevice
4435  * @filter_dev: ignored
4436  * @idx: the number of FDB table entries dumped is added to *@idx
4437  *
4438  * Default netdevice operation to dump the existing unicast address list.
4439  * Returns number of addresses from list put in skb.
4440  */
4441 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4442                       struct netlink_callback *cb,
4443                       struct net_device *dev,
4444                       struct net_device *filter_dev,
4445                       int *idx)
4446 {
4447         int err;
4448
4449         if (dev->type != ARPHRD_ETHER)
4450                 return -EINVAL;
4451
4452         netif_addr_lock_bh(dev);
4453         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4454         if (err)
4455                 goto out;
4456         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4457 out:
4458         netif_addr_unlock_bh(dev);
4459         return err;
4460 }
4461 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4462
4463 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4464                                  int *br_idx, int *brport_idx,
4465                                  struct netlink_ext_ack *extack)
4466 {
4467         struct nlattr *tb[NDA_MAX + 1];
4468         struct ndmsg *ndm;
4469         int err, i;
4470
4471         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4472                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4473                 return -EINVAL;
4474         }
4475
4476         ndm = nlmsg_data(nlh);
4477         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4478             ndm->ndm_flags || ndm->ndm_type) {
4479                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4480                 return -EINVAL;
4481         }
4482
4483         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4484                                             NDA_MAX, NULL, extack);
4485         if (err < 0)
4486                 return err;
4487
4488         *brport_idx = ndm->ndm_ifindex;
4489         for (i = 0; i <= NDA_MAX; ++i) {
4490                 if (!tb[i])
4491                         continue;
4492
4493                 switch (i) {
4494                 case NDA_IFINDEX:
4495                         if (nla_len(tb[i]) != sizeof(u32)) {
4496                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4497                                 return -EINVAL;
4498                         }
4499                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4500                         break;
4501                 case NDA_MASTER:
4502                         if (nla_len(tb[i]) != sizeof(u32)) {
4503                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4504                                 return -EINVAL;
4505                         }
4506                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4507                         break;
4508                 default:
4509                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4510                         return -EINVAL;
4511                 }
4512         }
4513
4514         return 0;
4515 }
4516
4517 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4518                                  int *br_idx, int *brport_idx,
4519                                  struct netlink_ext_ack *extack)
4520 {
4521         struct nlattr *tb[IFLA_MAX+1];
4522         int err;
4523
4524         /* A hack to preserve kernel<->userspace interface.
4525          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4526          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4527          * So, check for ndmsg with an optional u32 attribute (not used here).
4528          * Fortunately these sizes don't conflict with the size of ifinfomsg
4529          * with an optional attribute.
4530          */
4531         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4532             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4533              nla_attr_size(sizeof(u32)))) {
4534                 struct ifinfomsg *ifm;
4535
4536                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4537                                              tb, IFLA_MAX, ifla_policy,
4538                                              extack);
4539                 if (err < 0) {
4540                         return -EINVAL;
4541                 } else if (err == 0) {
4542                         if (tb[IFLA_MASTER])
4543                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4544                 }
4545
4546                 ifm = nlmsg_data(nlh);
4547                 *brport_idx = ifm->ifi_index;
4548         }
4549         return 0;
4550 }
4551
4552 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4553 {
4554         struct net_device *dev;
4555         struct net_device *br_dev = NULL;
4556         const struct net_device_ops *ops = NULL;
4557         const struct net_device_ops *cops = NULL;
4558         struct net *net = sock_net(skb->sk);
4559         struct hlist_head *head;
4560         int brport_idx = 0;
4561         int br_idx = 0;
4562         int h, s_h;
4563         int idx = 0, s_idx;
4564         int err = 0;
4565         int fidx = 0;
4566
4567         if (cb->strict_check)
4568                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4569                                             cb->extack);
4570         else
4571                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4572                                             cb->extack);
4573         if (err < 0)
4574                 return err;
4575
4576         if (br_idx) {
4577                 br_dev = __dev_get_by_index(net, br_idx);
4578                 if (!br_dev)
4579                         return -ENODEV;
4580
4581                 ops = br_dev->netdev_ops;
4582         }
4583
4584         s_h = cb->args[0];
4585         s_idx = cb->args[1];
4586
4587         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4588                 idx = 0;
4589                 head = &net->dev_index_head[h];
4590                 hlist_for_each_entry(dev, head, index_hlist) {
4591
4592                         if (brport_idx && (dev->ifindex != brport_idx))
4593                                 continue;
4594
4595                         if (!br_idx) { /* user did not specify a specific bridge */
4596                                 if (netif_is_bridge_port(dev)) {
4597                                         br_dev = netdev_master_upper_dev_get(dev);
4598                                         cops = br_dev->netdev_ops;
4599                                 }
4600                         } else {
4601                                 if (dev != br_dev &&
4602                                     !netif_is_bridge_port(dev))
4603                                         continue;
4604
4605                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4606                                     !netif_is_bridge_master(dev))
4607                                         continue;
4608                                 cops = ops;
4609                         }
4610
4611                         if (idx < s_idx)
4612                                 goto cont;
4613
4614                         if (netif_is_bridge_port(dev)) {
4615                                 if (cops && cops->ndo_fdb_dump) {
4616                                         err = cops->ndo_fdb_dump(skb, cb,
4617                                                                 br_dev, dev,
4618                                                                 &fidx);
4619                                         if (err == -EMSGSIZE)
4620                                                 goto out;
4621                                 }
4622                         }
4623
4624                         if (dev->netdev_ops->ndo_fdb_dump)
4625                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4626                                                                     dev, NULL,
4627                                                                     &fidx);
4628                         else
4629                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4630                                                         &fidx);
4631                         if (err == -EMSGSIZE)
4632                                 goto out;
4633
4634                         cops = NULL;
4635
4636                         /* reset fdb offset to 0 for rest of the interfaces */
4637                         cb->args[2] = 0;
4638                         fidx = 0;
4639 cont:
4640                         idx++;
4641                 }
4642         }
4643
4644 out:
4645         cb->args[0] = h;
4646         cb->args[1] = idx;
4647         cb->args[2] = fidx;
4648
4649         return skb->len;
4650 }
4651
4652 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4653                                 struct nlattr **tb, u8 *ndm_flags,
4654                                 int *br_idx, int *brport_idx, u8 **addr,
4655                                 u16 *vid, struct netlink_ext_ack *extack)
4656 {
4657         struct ndmsg *ndm;
4658         int err, i;
4659
4660         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4661                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4662                 return -EINVAL;
4663         }
4664
4665         ndm = nlmsg_data(nlh);
4666         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4667             ndm->ndm_type) {
4668                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4669                 return -EINVAL;
4670         }
4671
4672         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4673                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4674                 return -EINVAL;
4675         }
4676
4677         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4678                                             NDA_MAX, nda_policy, extack);
4679         if (err < 0)
4680                 return err;
4681
4682         *ndm_flags = ndm->ndm_flags;
4683         *brport_idx = ndm->ndm_ifindex;
4684         for (i = 0; i <= NDA_MAX; ++i) {
4685                 if (!tb[i])
4686                         continue;
4687
4688                 switch (i) {
4689                 case NDA_MASTER:
4690                         *br_idx = nla_get_u32(tb[i]);
4691                         break;
4692                 case NDA_LLADDR:
4693                         if (nla_len(tb[i]) != ETH_ALEN) {
4694                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4695                                 return -EINVAL;
4696                         }
4697                         *addr = nla_data(tb[i]);
4698                         break;
4699                 case NDA_VLAN:
4700                         err = fdb_vid_parse(tb[i], vid, extack);
4701                         if (err)
4702                                 return err;
4703                         break;
4704                 case NDA_VNI:
4705                         break;
4706                 default:
4707                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4708                         return -EINVAL;
4709                 }
4710         }
4711
4712         return 0;
4713 }
4714
4715 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4716                         struct netlink_ext_ack *extack)
4717 {
4718         struct net_device *dev = NULL, *br_dev = NULL;
4719         const struct net_device_ops *ops = NULL;
4720         struct net *net = sock_net(in_skb->sk);
4721         struct nlattr *tb[NDA_MAX + 1];
4722         struct sk_buff *skb;
4723         int brport_idx = 0;
4724         u8 ndm_flags = 0;
4725         int br_idx = 0;
4726         u8 *addr = NULL;
4727         u16 vid = 0;
4728         int err;
4729
4730         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4731                                    &brport_idx, &addr, &vid, extack);
4732         if (err < 0)
4733                 return err;
4734
4735         if (!addr) {
4736                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4737                 return -EINVAL;
4738         }
4739
4740         if (brport_idx) {
4741                 dev = __dev_get_by_index(net, brport_idx);
4742                 if (!dev) {
4743                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4744                         return -ENODEV;
4745                 }
4746         }
4747
4748         if (br_idx) {
4749                 if (dev) {
4750                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4751                         return -EINVAL;
4752                 }
4753
4754                 br_dev = __dev_get_by_index(net, br_idx);
4755                 if (!br_dev) {
4756                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4757                         return -EINVAL;
4758                 }
4759                 ops = br_dev->netdev_ops;
4760         }
4761
4762         if (dev) {
4763                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4764                         if (!netif_is_bridge_port(dev)) {
4765                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4766                                 return -EINVAL;
4767                         }
4768                         br_dev = netdev_master_upper_dev_get(dev);
4769                         if (!br_dev) {
4770                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4771                                 return -EINVAL;
4772                         }
4773                         ops = br_dev->netdev_ops;
4774                 } else {
4775                         if (!(ndm_flags & NTF_SELF)) {
4776                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4777                                 return -EINVAL;
4778                         }
4779                         ops = dev->netdev_ops;
4780                 }
4781         }
4782
4783         if (!br_dev && !dev) {
4784                 NL_SET_ERR_MSG(extack, "No device specified");
4785                 return -ENODEV;
4786         }
4787
4788         if (!ops || !ops->ndo_fdb_get) {
4789                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4790                 return -EOPNOTSUPP;
4791         }
4792
4793         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4794         if (!skb)
4795                 return -ENOBUFS;
4796
4797         if (br_dev)
4798                 dev = br_dev;
4799         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4800                                NETLINK_CB(in_skb).portid,
4801                                nlh->nlmsg_seq, extack);
4802         if (err)
4803                 goto out;
4804
4805         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4806 out:
4807         kfree_skb(skb);
4808         return err;
4809 }
4810
4811 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4812                                unsigned int attrnum, unsigned int flag)
4813 {
4814         if (mask & flag)
4815                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4816         return 0;
4817 }
4818
4819 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4820                             struct net_device *dev, u16 mode,
4821                             u32 flags, u32 mask, int nlflags,
4822                             u32 filter_mask,
4823                             int (*vlan_fill)(struct sk_buff *skb,
4824                                              struct net_device *dev,
4825                                              u32 filter_mask))
4826 {
4827         struct nlmsghdr *nlh;
4828         struct ifinfomsg *ifm;
4829         struct nlattr *br_afspec;
4830         struct nlattr *protinfo;
4831         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4832         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4833         int err = 0;
4834
4835         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4836         if (nlh == NULL)
4837                 return -EMSGSIZE;
4838
4839         ifm = nlmsg_data(nlh);
4840         ifm->ifi_family = AF_BRIDGE;
4841         ifm->__ifi_pad = 0;
4842         ifm->ifi_type = dev->type;
4843         ifm->ifi_index = dev->ifindex;
4844         ifm->ifi_flags = dev_get_flags(dev);
4845         ifm->ifi_change = 0;
4846
4847
4848         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4849             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4850             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4851             (br_dev &&
4852              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4853             (dev->addr_len &&
4854              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4855             (dev->ifindex != dev_get_iflink(dev) &&
4856              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4857                 goto nla_put_failure;
4858
4859         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4860         if (!br_afspec)
4861                 goto nla_put_failure;
4862
4863         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4864                 nla_nest_cancel(skb, br_afspec);
4865                 goto nla_put_failure;
4866         }
4867
4868         if (mode != BRIDGE_MODE_UNDEF) {
4869                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4870                         nla_nest_cancel(skb, br_afspec);
4871                         goto nla_put_failure;
4872                 }
4873         }
4874         if (vlan_fill) {
4875                 err = vlan_fill(skb, dev, filter_mask);
4876                 if (err) {
4877                         nla_nest_cancel(skb, br_afspec);
4878                         goto nla_put_failure;
4879                 }
4880         }
4881         nla_nest_end(skb, br_afspec);
4882
4883         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4884         if (!protinfo)
4885                 goto nla_put_failure;
4886
4887         if (brport_nla_put_flag(skb, flags, mask,
4888                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4889             brport_nla_put_flag(skb, flags, mask,
4890                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4891             brport_nla_put_flag(skb, flags, mask,
4892                                 IFLA_BRPORT_FAST_LEAVE,
4893                                 BR_MULTICAST_FAST_LEAVE) ||
4894             brport_nla_put_flag(skb, flags, mask,
4895                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4896             brport_nla_put_flag(skb, flags, mask,
4897                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4898             brport_nla_put_flag(skb, flags, mask,
4899                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4900             brport_nla_put_flag(skb, flags, mask,
4901                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4902             brport_nla_put_flag(skb, flags, mask,
4903                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4904             brport_nla_put_flag(skb, flags, mask,
4905                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4906             brport_nla_put_flag(skb, flags, mask,
4907                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4908                 nla_nest_cancel(skb, protinfo);
4909                 goto nla_put_failure;
4910         }
4911
4912         nla_nest_end(skb, protinfo);
4913
4914         nlmsg_end(skb, nlh);
4915         return 0;
4916 nla_put_failure:
4917         nlmsg_cancel(skb, nlh);
4918         return err ? err : -EMSGSIZE;
4919 }
4920 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4921
4922 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4923                                     bool strict_check, u32 *filter_mask,
4924                                     struct netlink_ext_ack *extack)
4925 {
4926         struct nlattr *tb[IFLA_MAX+1];
4927         int err, i;
4928
4929         if (strict_check) {
4930                 struct ifinfomsg *ifm;
4931
4932                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4933                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4934                         return -EINVAL;
4935                 }
4936
4937                 ifm = nlmsg_data(nlh);
4938                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4939                     ifm->ifi_change || ifm->ifi_index) {
4940                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4941                         return -EINVAL;
4942                 }
4943
4944                 err = nlmsg_parse_deprecated_strict(nlh,
4945                                                     sizeof(struct ifinfomsg),
4946                                                     tb, IFLA_MAX, ifla_policy,
4947                                                     extack);
4948         } else {
4949                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4950                                              tb, IFLA_MAX, ifla_policy,
4951                                              extack);
4952         }
4953         if (err < 0)
4954                 return err;
4955
4956         /* new attributes should only be added with strict checking */
4957         for (i = 0; i <= IFLA_MAX; ++i) {
4958                 if (!tb[i])
4959                         continue;
4960
4961                 switch (i) {
4962                 case IFLA_EXT_MASK:
4963                         *filter_mask = nla_get_u32(tb[i]);
4964                         break;
4965                 default:
4966                         if (strict_check) {
4967                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4968                                 return -EINVAL;
4969                         }
4970                 }
4971         }
4972
4973         return 0;
4974 }
4975
4976 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4977 {
4978         const struct nlmsghdr *nlh = cb->nlh;
4979         struct net *net = sock_net(skb->sk);
4980         struct net_device *dev;
4981         int idx = 0;
4982         u32 portid = NETLINK_CB(cb->skb).portid;
4983         u32 seq = nlh->nlmsg_seq;
4984         u32 filter_mask = 0;
4985         int err;
4986
4987         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4988                                        cb->extack);
4989         if (err < 0 && cb->strict_check)
4990                 return err;
4991
4992         rcu_read_lock();
4993         for_each_netdev_rcu(net, dev) {
4994                 const struct net_device_ops *ops = dev->netdev_ops;
4995                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4996
4997                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4998                         if (idx >= cb->args[0]) {
4999                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5000                                                 skb, portid, seq, dev,
5001                                                 filter_mask, NLM_F_MULTI);
5002                                 if (err < 0 && err != -EOPNOTSUPP) {
5003                                         if (likely(skb->len))
5004                                                 break;
5005
5006                                         goto out_err;
5007                                 }
5008                         }
5009                         idx++;
5010                 }
5011
5012                 if (ops->ndo_bridge_getlink) {
5013                         if (idx >= cb->args[0]) {
5014                                 err = ops->ndo_bridge_getlink(skb, portid,
5015                                                               seq, dev,
5016                                                               filter_mask,
5017                                                               NLM_F_MULTI);
5018                                 if (err < 0 && err != -EOPNOTSUPP) {
5019                                         if (likely(skb->len))
5020                                                 break;
5021
5022                                         goto out_err;
5023                                 }
5024                         }
5025                         idx++;
5026                 }
5027         }
5028         err = skb->len;
5029 out_err:
5030         rcu_read_unlock();
5031         cb->args[0] = idx;
5032
5033         return err;
5034 }
5035
5036 static inline size_t bridge_nlmsg_size(void)
5037 {
5038         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5039                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5040                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5041                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5042                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5043                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5044                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5045                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5046                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5047                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5048                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5049 }
5050
5051 static int rtnl_bridge_notify(struct net_device *dev)
5052 {
5053         struct net *net = dev_net(dev);
5054         struct sk_buff *skb;
5055         int err = -EOPNOTSUPP;
5056
5057         if (!dev->netdev_ops->ndo_bridge_getlink)
5058                 return 0;
5059
5060         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5061         if (!skb) {
5062                 err = -ENOMEM;
5063                 goto errout;
5064         }
5065
5066         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5067         if (err < 0)
5068                 goto errout;
5069
5070         /* Notification info is only filled for bridge ports, not the bridge
5071          * device itself. Therefore, a zero notification length is valid and
5072          * should not result in an error.
5073          */
5074         if (!skb->len)
5075                 goto errout;
5076
5077         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5078         return 0;
5079 errout:
5080         WARN_ON(err == -EMSGSIZE);
5081         kfree_skb(skb);
5082         if (err)
5083                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5084         return err;
5085 }
5086
5087 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5088                                struct netlink_ext_ack *extack)
5089 {
5090         struct net *net = sock_net(skb->sk);
5091         struct ifinfomsg *ifm;
5092         struct net_device *dev;
5093         struct nlattr *br_spec, *attr = NULL;
5094         int rem, err = -EOPNOTSUPP;
5095         u16 flags = 0;
5096         bool have_flags = false;
5097
5098         if (nlmsg_len(nlh) < sizeof(*ifm))
5099                 return -EINVAL;
5100
5101         ifm = nlmsg_data(nlh);
5102         if (ifm->ifi_family != AF_BRIDGE)
5103                 return -EPFNOSUPPORT;
5104
5105         dev = __dev_get_by_index(net, ifm->ifi_index);
5106         if (!dev) {
5107                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5108                 return -ENODEV;
5109         }
5110
5111         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5112         if (br_spec) {
5113                 nla_for_each_nested(attr, br_spec, rem) {
5114                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5115                                 if (nla_len(attr) < sizeof(flags))
5116                                         return -EINVAL;
5117
5118                                 have_flags = true;
5119                                 flags = nla_get_u16(attr);
5120                                 break;
5121                         }
5122                 }
5123         }
5124
5125         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5126                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5127
5128                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5129                         err = -EOPNOTSUPP;
5130                         goto out;
5131                 }
5132
5133                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5134                                                              extack);
5135                 if (err)
5136                         goto out;
5137
5138                 flags &= ~BRIDGE_FLAGS_MASTER;
5139         }
5140
5141         if ((flags & BRIDGE_FLAGS_SELF)) {
5142                 if (!dev->netdev_ops->ndo_bridge_setlink)
5143                         err = -EOPNOTSUPP;
5144                 else
5145                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5146                                                                   flags,
5147                                                                   extack);
5148                 if (!err) {
5149                         flags &= ~BRIDGE_FLAGS_SELF;
5150
5151                         /* Generate event to notify upper layer of bridge
5152                          * change
5153                          */
5154                         err = rtnl_bridge_notify(dev);
5155                 }
5156         }
5157
5158         if (have_flags)
5159                 memcpy(nla_data(attr), &flags, sizeof(flags));
5160 out:
5161         return err;
5162 }
5163
5164 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5165                                struct netlink_ext_ack *extack)
5166 {
5167         struct net *net = sock_net(skb->sk);
5168         struct ifinfomsg *ifm;
5169         struct net_device *dev;
5170         struct nlattr *br_spec, *attr = NULL;
5171         int rem, err = -EOPNOTSUPP;
5172         u16 flags = 0;
5173         bool have_flags = false;
5174
5175         if (nlmsg_len(nlh) < sizeof(*ifm))
5176                 return -EINVAL;
5177
5178         ifm = nlmsg_data(nlh);
5179         if (ifm->ifi_family != AF_BRIDGE)
5180                 return -EPFNOSUPPORT;
5181
5182         dev = __dev_get_by_index(net, ifm->ifi_index);
5183         if (!dev) {
5184                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5185                 return -ENODEV;
5186         }
5187
5188         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5189         if (br_spec) {
5190                 nla_for_each_nested(attr, br_spec, rem) {
5191                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5192                                 if (nla_len(attr) < sizeof(flags))
5193                                         return -EINVAL;
5194
5195                                 have_flags = true;
5196                                 flags = nla_get_u16(attr);
5197                                 break;
5198                         }
5199                 }
5200         }
5201
5202         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5203                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5204
5205                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5206                         err = -EOPNOTSUPP;
5207                         goto out;
5208                 }
5209
5210                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5211                 if (err)
5212                         goto out;
5213
5214                 flags &= ~BRIDGE_FLAGS_MASTER;
5215         }
5216
5217         if ((flags & BRIDGE_FLAGS_SELF)) {
5218                 if (!dev->netdev_ops->ndo_bridge_dellink)
5219                         err = -EOPNOTSUPP;
5220                 else
5221                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5222                                                                   flags);
5223
5224                 if (!err) {
5225                         flags &= ~BRIDGE_FLAGS_SELF;
5226
5227                         /* Generate event to notify upper layer of bridge
5228                          * change
5229                          */
5230                         err = rtnl_bridge_notify(dev);
5231                 }
5232         }
5233
5234         if (have_flags)
5235                 memcpy(nla_data(attr), &flags, sizeof(flags));
5236 out:
5237         return err;
5238 }
5239
5240 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5241 {
5242         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5243                (!idxattr || idxattr == attrid);
5244 }
5245
5246 static bool
5247 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5248 {
5249         return dev->netdev_ops &&
5250                dev->netdev_ops->ndo_has_offload_stats &&
5251                dev->netdev_ops->ndo_get_offload_stats &&
5252                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5253 }
5254
5255 static unsigned int
5256 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5257 {
5258         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5259                sizeof(struct rtnl_link_stats64) : 0;
5260 }
5261
5262 static int
5263 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5264                              struct sk_buff *skb)
5265 {
5266         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5267         struct nlattr *attr = NULL;
5268         void *attr_data;
5269         int err;
5270
5271         if (!size)
5272                 return -ENODATA;
5273
5274         attr = nla_reserve_64bit(skb, attr_id, size,
5275                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5276         if (!attr)
5277                 return -EMSGSIZE;
5278
5279         attr_data = nla_data(attr);
5280         memset(attr_data, 0, size);
5281
5282         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5283         if (err)
5284                 return err;
5285
5286         return 0;
5287 }
5288
5289 static unsigned int
5290 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5291                                    enum netdev_offload_xstats_type type)
5292 {
5293         bool enabled = netdev_offload_xstats_enabled(dev, type);
5294
5295         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5296 }
5297
5298 struct rtnl_offload_xstats_request_used {
5299         bool request;
5300         bool used;
5301 };
5302
5303 static int
5304 rtnl_offload_xstats_get_stats(struct net_device *dev,
5305                               enum netdev_offload_xstats_type type,
5306                               struct rtnl_offload_xstats_request_used *ru,
5307                               struct rtnl_hw_stats64 *stats,
5308                               struct netlink_ext_ack *extack)
5309 {
5310         bool request;
5311         bool used;
5312         int err;
5313
5314         request = netdev_offload_xstats_enabled(dev, type);
5315         if (!request) {
5316                 used = false;
5317                 goto out;
5318         }
5319
5320         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5321         if (err)
5322                 return err;
5323
5324 out:
5325         if (ru) {
5326                 ru->request = request;
5327                 ru->used = used;
5328         }
5329         return 0;
5330 }
5331
5332 static int
5333 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5334                                        struct rtnl_offload_xstats_request_used *ru)
5335 {
5336         struct nlattr *nest;
5337
5338         nest = nla_nest_start(skb, attr_id);
5339         if (!nest)
5340                 return -EMSGSIZE;
5341
5342         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5343                 goto nla_put_failure;
5344
5345         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5346                 goto nla_put_failure;
5347
5348         nla_nest_end(skb, nest);
5349         return 0;
5350
5351 nla_put_failure:
5352         nla_nest_cancel(skb, nest);
5353         return -EMSGSIZE;
5354 }
5355
5356 static int
5357 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5358                                    struct netlink_ext_ack *extack)
5359 {
5360         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5361         struct rtnl_offload_xstats_request_used ru_l3;
5362         struct nlattr *nest;
5363         int err;
5364
5365         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5366         if (err)
5367                 return err;
5368
5369         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5370         if (!nest)
5371                 return -EMSGSIZE;
5372
5373         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5374                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5375                                                    &ru_l3))
5376                 goto nla_put_failure;
5377
5378         nla_nest_end(skb, nest);
5379         return 0;
5380
5381 nla_put_failure:
5382         nla_nest_cancel(skb, nest);
5383         return -EMSGSIZE;
5384 }
5385
5386 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5387                                     int *prividx, u32 off_filter_mask,
5388                                     struct netlink_ext_ack *extack)
5389 {
5390         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5391         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5392         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5393         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5394         bool have_data = false;
5395         int err;
5396
5397         if (*prividx <= attr_id_cpu_hit &&
5398             (off_filter_mask &
5399              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5400                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5401                 if (!err) {
5402                         have_data = true;
5403                 } else if (err != -ENODATA) {
5404                         *prividx = attr_id_cpu_hit;
5405                         return err;
5406                 }
5407         }
5408
5409         if (*prividx <= attr_id_hw_s_info &&
5410             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5411                 *prividx = attr_id_hw_s_info;
5412
5413                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5414                 if (err)
5415                         return err;
5416
5417                 have_data = true;
5418                 *prividx = 0;
5419         }
5420
5421         if (*prividx <= attr_id_l3_stats &&
5422             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5423                 unsigned int size_l3;
5424                 struct nlattr *attr;
5425
5426                 *prividx = attr_id_l3_stats;
5427
5428                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5429                 if (!size_l3)
5430                         goto skip_l3_stats;
5431                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5432                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5433                 if (!attr)
5434                         return -EMSGSIZE;
5435
5436                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5437                                                     nla_data(attr), extack);
5438                 if (err)
5439                         return err;
5440
5441                 have_data = true;
5442 skip_l3_stats:
5443                 *prividx = 0;
5444         }
5445
5446         if (!have_data)
5447                 return -ENODATA;
5448
5449         *prividx = 0;
5450         return 0;
5451 }
5452
5453 static unsigned int
5454 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5455                                            enum netdev_offload_xstats_type type)
5456 {
5457         bool enabled = netdev_offload_xstats_enabled(dev, type);
5458
5459         return nla_total_size(0) +
5460                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5461                 nla_total_size(sizeof(u8)) +
5462                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5463                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5464                 0;
5465 }
5466
5467 static unsigned int
5468 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5469 {
5470         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5471
5472         return nla_total_size(0) +
5473                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5474                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5475                 0;
5476 }
5477
5478 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5479                                         u32 off_filter_mask)
5480 {
5481         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5482         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5483         int nla_size = 0;
5484         int size;
5485
5486         if (off_filter_mask &
5487             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5488                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5489                 nla_size += nla_total_size_64bit(size);
5490         }
5491
5492         if (off_filter_mask &
5493             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5494                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5495
5496         if (off_filter_mask &
5497             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5498                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5499                 nla_size += nla_total_size_64bit(size);
5500         }
5501
5502         if (nla_size != 0)
5503                 nla_size += nla_total_size(0);
5504
5505         return nla_size;
5506 }
5507
5508 struct rtnl_stats_dump_filters {
5509         /* mask[0] filters outer attributes. Then individual nests have their
5510          * filtering mask at the index of the nested attribute.
5511          */
5512         u32 mask[IFLA_STATS_MAX + 1];
5513 };
5514
5515 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5516                                int type, u32 pid, u32 seq, u32 change,
5517                                unsigned int flags,
5518                                const struct rtnl_stats_dump_filters *filters,
5519                                int *idxattr, int *prividx,
5520                                struct netlink_ext_ack *extack)
5521 {
5522         unsigned int filter_mask = filters->mask[0];
5523         struct if_stats_msg *ifsm;
5524         struct nlmsghdr *nlh;
5525         struct nlattr *attr;
5526         int s_prividx = *prividx;
5527         int err;
5528
5529         ASSERT_RTNL();
5530
5531         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5532         if (!nlh)
5533                 return -EMSGSIZE;
5534
5535         ifsm = nlmsg_data(nlh);
5536         ifsm->family = PF_UNSPEC;
5537         ifsm->pad1 = 0;
5538         ifsm->pad2 = 0;
5539         ifsm->ifindex = dev->ifindex;
5540         ifsm->filter_mask = filter_mask;
5541
5542         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5543                 struct rtnl_link_stats64 *sp;
5544
5545                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5546                                          sizeof(struct rtnl_link_stats64),
5547                                          IFLA_STATS_UNSPEC);
5548                 if (!attr) {
5549                         err = -EMSGSIZE;
5550                         goto nla_put_failure;
5551                 }
5552
5553                 sp = nla_data(attr);
5554                 dev_get_stats(dev, sp);
5555         }
5556
5557         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5558                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5559
5560                 if (ops && ops->fill_linkxstats) {
5561                         *idxattr = IFLA_STATS_LINK_XSTATS;
5562                         attr = nla_nest_start_noflag(skb,
5563                                                      IFLA_STATS_LINK_XSTATS);
5564                         if (!attr) {
5565                                 err = -EMSGSIZE;
5566                                 goto nla_put_failure;
5567                         }
5568
5569                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5570                         nla_nest_end(skb, attr);
5571                         if (err)
5572                                 goto nla_put_failure;
5573                         *idxattr = 0;
5574                 }
5575         }
5576
5577         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5578                              *idxattr)) {
5579                 const struct rtnl_link_ops *ops = NULL;
5580                 const struct net_device *master;
5581
5582                 master = netdev_master_upper_dev_get(dev);
5583                 if (master)
5584                         ops = master->rtnl_link_ops;
5585                 if (ops && ops->fill_linkxstats) {
5586                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5587                         attr = nla_nest_start_noflag(skb,
5588                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5589                         if (!attr) {
5590                                 err = -EMSGSIZE;
5591                                 goto nla_put_failure;
5592                         }
5593
5594                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5595                         nla_nest_end(skb, attr);
5596                         if (err)
5597                                 goto nla_put_failure;
5598                         *idxattr = 0;
5599                 }
5600         }
5601
5602         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5603                              *idxattr)) {
5604                 u32 off_filter_mask;
5605
5606                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5607                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5608                 attr = nla_nest_start_noflag(skb,
5609                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5610                 if (!attr) {
5611                         err = -EMSGSIZE;
5612                         goto nla_put_failure;
5613                 }
5614
5615                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5616                                                off_filter_mask, extack);
5617                 if (err == -ENODATA)
5618                         nla_nest_cancel(skb, attr);
5619                 else
5620                         nla_nest_end(skb, attr);
5621
5622                 if (err && err != -ENODATA)
5623                         goto nla_put_failure;
5624                 *idxattr = 0;
5625         }
5626
5627         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5628                 struct rtnl_af_ops *af_ops;
5629
5630                 *idxattr = IFLA_STATS_AF_SPEC;
5631                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5632                 if (!attr) {
5633                         err = -EMSGSIZE;
5634                         goto nla_put_failure;
5635                 }
5636
5637                 rcu_read_lock();
5638                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5639                         if (af_ops->fill_stats_af) {
5640                                 struct nlattr *af;
5641
5642                                 af = nla_nest_start_noflag(skb,
5643                                                            af_ops->family);
5644                                 if (!af) {
5645                                         rcu_read_unlock();
5646                                         err = -EMSGSIZE;
5647                                         goto nla_put_failure;
5648                                 }
5649                                 err = af_ops->fill_stats_af(skb, dev);
5650
5651                                 if (err == -ENODATA) {
5652                                         nla_nest_cancel(skb, af);
5653                                 } else if (err < 0) {
5654                                         rcu_read_unlock();
5655                                         goto nla_put_failure;
5656                                 }
5657
5658                                 nla_nest_end(skb, af);
5659                         }
5660                 }
5661                 rcu_read_unlock();
5662
5663                 nla_nest_end(skb, attr);
5664
5665                 *idxattr = 0;
5666         }
5667
5668         nlmsg_end(skb, nlh);
5669
5670         return 0;
5671
5672 nla_put_failure:
5673         /* not a multi message or no progress mean a real error */
5674         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5675                 nlmsg_cancel(skb, nlh);
5676         else
5677                 nlmsg_end(skb, nlh);
5678
5679         return err;
5680 }
5681
5682 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5683                                   const struct rtnl_stats_dump_filters *filters)
5684 {
5685         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5686         unsigned int filter_mask = filters->mask[0];
5687
5688         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5689                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5690
5691         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5692                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5693                 int attr = IFLA_STATS_LINK_XSTATS;
5694
5695                 if (ops && ops->get_linkxstats_size) {
5696                         size += nla_total_size(ops->get_linkxstats_size(dev,
5697                                                                         attr));
5698                         /* for IFLA_STATS_LINK_XSTATS */
5699                         size += nla_total_size(0);
5700                 }
5701         }
5702
5703         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5704                 struct net_device *_dev = (struct net_device *)dev;
5705                 const struct rtnl_link_ops *ops = NULL;
5706                 const struct net_device *master;
5707
5708                 /* netdev_master_upper_dev_get can't take const */
5709                 master = netdev_master_upper_dev_get(_dev);
5710                 if (master)
5711                         ops = master->rtnl_link_ops;
5712                 if (ops && ops->get_linkxstats_size) {
5713                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5714
5715                         size += nla_total_size(ops->get_linkxstats_size(dev,
5716                                                                         attr));
5717                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5718                         size += nla_total_size(0);
5719                 }
5720         }
5721
5722         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5723                 u32 off_filter_mask;
5724
5725                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5726                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5727         }
5728
5729         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5730                 struct rtnl_af_ops *af_ops;
5731
5732                 /* for IFLA_STATS_AF_SPEC */
5733                 size += nla_total_size(0);
5734
5735                 rcu_read_lock();
5736                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5737                         if (af_ops->get_stats_af_size) {
5738                                 size += nla_total_size(
5739                                         af_ops->get_stats_af_size(dev));
5740
5741                                 /* for AF_* */
5742                                 size += nla_total_size(0);
5743                         }
5744                 }
5745                 rcu_read_unlock();
5746         }
5747
5748         return size;
5749 }
5750
5751 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5752
5753 static const struct nla_policy
5754 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5755         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5756                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5757 };
5758
5759 static const struct nla_policy
5760 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5761         [IFLA_STATS_GET_FILTERS] =
5762                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5763 };
5764
5765 static const struct nla_policy
5766 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5767         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5768 };
5769
5770 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5771                                         struct rtnl_stats_dump_filters *filters,
5772                                         struct netlink_ext_ack *extack)
5773 {
5774         struct nlattr *tb[IFLA_STATS_MAX + 1];
5775         int err;
5776         int at;
5777
5778         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5779                                rtnl_stats_get_policy_filters, extack);
5780         if (err < 0)
5781                 return err;
5782
5783         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5784                 if (tb[at]) {
5785                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5786                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5787                                 return -EINVAL;
5788                         }
5789                         filters->mask[at] = nla_get_u32(tb[at]);
5790                 }
5791         }
5792
5793         return 0;
5794 }
5795
5796 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5797                                 u32 filter_mask,
5798                                 struct rtnl_stats_dump_filters *filters,
5799                                 struct netlink_ext_ack *extack)
5800 {
5801         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5802         int err;
5803         int i;
5804
5805         filters->mask[0] = filter_mask;
5806         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5807                 filters->mask[i] = -1U;
5808
5809         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5810                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5811         if (err < 0)
5812                 return err;
5813
5814         if (tb[IFLA_STATS_GET_FILTERS]) {
5815                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5816                                                    filters, extack);
5817                 if (err)
5818                         return err;
5819         }
5820
5821         return 0;
5822 }
5823
5824 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5825                                 bool is_dump, struct netlink_ext_ack *extack)
5826 {
5827         struct if_stats_msg *ifsm;
5828
5829         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5830                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5831                 return -EINVAL;
5832         }
5833
5834         if (!strict_check)
5835                 return 0;
5836
5837         ifsm = nlmsg_data(nlh);
5838
5839         /* only requests using strict checks can pass data to influence
5840          * the dump. The legacy exception is filter_mask.
5841          */
5842         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5843                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5844                 return -EINVAL;
5845         }
5846         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5847                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5848                 return -EINVAL;
5849         }
5850
5851         return 0;
5852 }
5853
5854 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5855                           struct netlink_ext_ack *extack)
5856 {
5857         struct rtnl_stats_dump_filters filters;
5858         struct net *net = sock_net(skb->sk);
5859         struct net_device *dev = NULL;
5860         int idxattr = 0, prividx = 0;
5861         struct if_stats_msg *ifsm;
5862         struct sk_buff *nskb;
5863         int err;
5864
5865         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5866                                    false, extack);
5867         if (err)
5868                 return err;
5869
5870         ifsm = nlmsg_data(nlh);
5871         if (ifsm->ifindex > 0)
5872                 dev = __dev_get_by_index(net, ifsm->ifindex);
5873         else
5874                 return -EINVAL;
5875
5876         if (!dev)
5877                 return -ENODEV;
5878
5879         if (!ifsm->filter_mask) {
5880                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5881                 return -EINVAL;
5882         }
5883
5884         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5885         if (err)
5886                 return err;
5887
5888         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5889         if (!nskb)
5890                 return -ENOBUFS;
5891
5892         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5893                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5894                                   0, &filters, &idxattr, &prividx, extack);
5895         if (err < 0) {
5896                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5897                 WARN_ON(err == -EMSGSIZE);
5898                 kfree_skb(nskb);
5899         } else {
5900                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5901         }
5902
5903         return err;
5904 }
5905
5906 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5907 {
5908         struct netlink_ext_ack *extack = cb->extack;
5909         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5910         struct rtnl_stats_dump_filters filters;
5911         struct net *net = sock_net(skb->sk);
5912         unsigned int flags = NLM_F_MULTI;
5913         struct if_stats_msg *ifsm;
5914         struct hlist_head *head;
5915         struct net_device *dev;
5916         int idx = 0;
5917
5918         s_h = cb->args[0];
5919         s_idx = cb->args[1];
5920         s_idxattr = cb->args[2];
5921         s_prividx = cb->args[3];
5922
5923         cb->seq = net->dev_base_seq;
5924
5925         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5926         if (err)
5927                 return err;
5928
5929         ifsm = nlmsg_data(cb->nlh);
5930         if (!ifsm->filter_mask) {
5931                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5932                 return -EINVAL;
5933         }
5934
5935         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5936                                    extack);
5937         if (err)
5938                 return err;
5939
5940         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5941                 idx = 0;
5942                 head = &net->dev_index_head[h];
5943                 hlist_for_each_entry(dev, head, index_hlist) {
5944                         if (idx < s_idx)
5945                                 goto cont;
5946                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5947                                                   NETLINK_CB(cb->skb).portid,
5948                                                   cb->nlh->nlmsg_seq, 0,
5949                                                   flags, &filters,
5950                                                   &s_idxattr, &s_prividx,
5951                                                   extack);
5952                         /* If we ran out of room on the first message,
5953                          * we're in trouble
5954                          */
5955                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5956
5957                         if (err < 0)
5958                                 goto out;
5959                         s_prividx = 0;
5960                         s_idxattr = 0;
5961                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5962 cont:
5963                         idx++;
5964                 }
5965         }
5966 out:
5967         cb->args[3] = s_prividx;
5968         cb->args[2] = s_idxattr;
5969         cb->args[1] = idx;
5970         cb->args[0] = h;
5971
5972         return skb->len;
5973 }
5974
5975 void rtnl_offload_xstats_notify(struct net_device *dev)
5976 {
5977         struct rtnl_stats_dump_filters response_filters = {};
5978         struct net *net = dev_net(dev);
5979         int idxattr = 0, prividx = 0;
5980         struct sk_buff *skb;
5981         int err = -ENOBUFS;
5982
5983         ASSERT_RTNL();
5984
5985         response_filters.mask[0] |=
5986                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5987         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5988                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5989
5990         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
5991                         GFP_KERNEL);
5992         if (!skb)
5993                 goto errout;
5994
5995         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
5996                                   &response_filters, &idxattr, &prividx, NULL);
5997         if (err < 0) {
5998                 kfree_skb(skb);
5999                 goto errout;
6000         }
6001
6002         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6003         return;
6004
6005 errout:
6006         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6007 }
6008 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6009
6010 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6011                           struct netlink_ext_ack *extack)
6012 {
6013         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6014         struct rtnl_stats_dump_filters response_filters = {};
6015         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6016         struct net *net = sock_net(skb->sk);
6017         struct net_device *dev = NULL;
6018         struct if_stats_msg *ifsm;
6019         bool notify = false;
6020         int err;
6021
6022         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6023                                    false, extack);
6024         if (err)
6025                 return err;
6026
6027         ifsm = nlmsg_data(nlh);
6028         if (ifsm->family != AF_UNSPEC) {
6029                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6030                 return -EINVAL;
6031         }
6032
6033         if (ifsm->ifindex > 0)
6034                 dev = __dev_get_by_index(net, ifsm->ifindex);
6035         else
6036                 return -EINVAL;
6037
6038         if (!dev)
6039                 return -ENODEV;
6040
6041         if (ifsm->filter_mask) {
6042                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6043                 return -EINVAL;
6044         }
6045
6046         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6047                           ifla_stats_set_policy, extack);
6048         if (err < 0)
6049                 return err;
6050
6051         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6052                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6053
6054                 if (req)
6055                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6056                 else
6057                         err = netdev_offload_xstats_disable(dev, t_l3);
6058
6059                 if (!err)
6060                         notify = true;
6061                 else if (err != -EALREADY)
6062                         return err;
6063
6064                 response_filters.mask[0] |=
6065                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6066                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6067                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6068         }
6069
6070         if (notify)
6071                 rtnl_offload_xstats_notify(dev);
6072
6073         return 0;
6074 }
6075
6076 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6077                                    struct netlink_ext_ack *extack)
6078 {
6079         struct br_port_msg *bpm;
6080
6081         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6082                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6083                 return -EINVAL;
6084         }
6085
6086         bpm = nlmsg_data(nlh);
6087         if (bpm->ifindex) {
6088                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6089                 return -EINVAL;
6090         }
6091         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6092                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6093                 return -EINVAL;
6094         }
6095
6096         return 0;
6097 }
6098
6099 struct rtnl_mdb_dump_ctx {
6100         long idx;
6101 };
6102
6103 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6104 {
6105         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6106         struct net *net = sock_net(skb->sk);
6107         struct net_device *dev;
6108         int idx, s_idx;
6109         int err;
6110
6111         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6112
6113         if (cb->strict_check) {
6114                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6115                 if (err)
6116                         return err;
6117         }
6118
6119         s_idx = ctx->idx;
6120         idx = 0;
6121
6122         for_each_netdev(net, dev) {
6123                 if (idx < s_idx)
6124                         goto skip;
6125                 if (!dev->netdev_ops->ndo_mdb_dump)
6126                         goto skip;
6127
6128                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6129                 if (err == -EMSGSIZE)
6130                         goto out;
6131                 /* Moving on to next device, reset markers and sequence
6132                  * counters since they are all maintained per-device.
6133                  */
6134                 memset(cb->ctx, 0, sizeof(cb->ctx));
6135                 cb->prev_seq = 0;
6136                 cb->seq = 0;
6137 skip:
6138                 idx++;
6139         }
6140
6141 out:
6142         ctx->idx = idx;
6143         return skb->len;
6144 }
6145
6146 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6147                                    struct netlink_ext_ack *extack)
6148 {
6149         struct br_mdb_entry *entry = nla_data(attr);
6150
6151         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6152                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6153                 return -EINVAL;
6154         }
6155
6156         if (entry->ifindex == 0) {
6157                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6158                 return -EINVAL;
6159         }
6160
6161         if (entry->addr.proto == htons(ETH_P_IP)) {
6162                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6163                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6164                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6165                         return -EINVAL;
6166                 }
6167                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6168                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6169                         return -EINVAL;
6170                 }
6171 #if IS_ENABLED(CONFIG_IPV6)
6172         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6173                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6174                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6175                         return -EINVAL;
6176                 }
6177 #endif
6178         } else if (entry->addr.proto == 0) {
6179                 /* L2 mdb */
6180                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6181                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6182                         return -EINVAL;
6183                 }
6184         } else {
6185                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6186                 return -EINVAL;
6187         }
6188
6189         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6190                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6191                 return -EINVAL;
6192         }
6193         if (entry->vid >= VLAN_VID_MASK) {
6194                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6195                 return -EINVAL;
6196         }
6197
6198         return 0;
6199 }
6200
6201 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6202         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6203         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6204                                                   rtnl_validate_mdb_entry,
6205                                                   sizeof(struct br_mdb_entry)),
6206         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6207 };
6208
6209 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6210                         struct netlink_ext_ack *extack)
6211 {
6212         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6213         struct net *net = sock_net(skb->sk);
6214         struct br_port_msg *bpm;
6215         struct net_device *dev;
6216         int err;
6217
6218         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6219                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6220         if (err)
6221                 return err;
6222
6223         bpm = nlmsg_data(nlh);
6224         if (!bpm->ifindex) {
6225                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6226                 return -EINVAL;
6227         }
6228
6229         dev = __dev_get_by_index(net, bpm->ifindex);
6230         if (!dev) {
6231                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6232                 return -ENODEV;
6233         }
6234
6235         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6236                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6237                 return -EINVAL;
6238         }
6239
6240         if (!dev->netdev_ops->ndo_mdb_add) {
6241                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6242                 return -EOPNOTSUPP;
6243         }
6244
6245         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6246 }
6247
6248 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6249                         struct netlink_ext_ack *extack)
6250 {
6251         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6252         struct net *net = sock_net(skb->sk);
6253         struct br_port_msg *bpm;
6254         struct net_device *dev;
6255         int err;
6256
6257         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6258                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6259         if (err)
6260                 return err;
6261
6262         bpm = nlmsg_data(nlh);
6263         if (!bpm->ifindex) {
6264                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6265                 return -EINVAL;
6266         }
6267
6268         dev = __dev_get_by_index(net, bpm->ifindex);
6269         if (!dev) {
6270                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6271                 return -ENODEV;
6272         }
6273
6274         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6275                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6276                 return -EINVAL;
6277         }
6278
6279         if (!dev->netdev_ops->ndo_mdb_del) {
6280                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6281                 return -EOPNOTSUPP;
6282         }
6283
6284         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6285 }
6286
6287 /* Process one rtnetlink message. */
6288
6289 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6290                              struct netlink_ext_ack *extack)
6291 {
6292         struct net *net = sock_net(skb->sk);
6293         struct rtnl_link *link;
6294         enum rtnl_kinds kind;
6295         struct module *owner;
6296         int err = -EOPNOTSUPP;
6297         rtnl_doit_func doit;
6298         unsigned int flags;
6299         int family;
6300         int type;
6301
6302         type = nlh->nlmsg_type;
6303         if (type > RTM_MAX)
6304                 return -EOPNOTSUPP;
6305
6306         type -= RTM_BASE;
6307
6308         /* All the messages must have at least 1 byte length */
6309         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6310                 return 0;
6311
6312         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6313         kind = rtnl_msgtype_kind(type);
6314
6315         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6316                 return -EPERM;
6317
6318         rcu_read_lock();
6319         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6320                 struct sock *rtnl;
6321                 rtnl_dumpit_func dumpit;
6322                 u32 min_dump_alloc = 0;
6323
6324                 link = rtnl_get_link(family, type);
6325                 if (!link || !link->dumpit) {
6326                         family = PF_UNSPEC;
6327                         link = rtnl_get_link(family, type);
6328                         if (!link || !link->dumpit)
6329                                 goto err_unlock;
6330                 }
6331                 owner = link->owner;
6332                 dumpit = link->dumpit;
6333
6334                 if (type == RTM_GETLINK - RTM_BASE)
6335                         min_dump_alloc = rtnl_calcit(skb, nlh);
6336
6337                 err = 0;
6338                 /* need to do this before rcu_read_unlock() */
6339                 if (!try_module_get(owner))
6340                         err = -EPROTONOSUPPORT;
6341
6342                 rcu_read_unlock();
6343
6344                 rtnl = net->rtnl;
6345                 if (err == 0) {
6346                         struct netlink_dump_control c = {
6347                                 .dump           = dumpit,
6348                                 .min_dump_alloc = min_dump_alloc,
6349                                 .module         = owner,
6350                         };
6351                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6352                         /* netlink_dump_start() will keep a reference on
6353                          * module if dump is still in progress.
6354                          */
6355                         module_put(owner);
6356                 }
6357                 return err;
6358         }
6359
6360         link = rtnl_get_link(family, type);
6361         if (!link || !link->doit) {
6362                 family = PF_UNSPEC;
6363                 link = rtnl_get_link(PF_UNSPEC, type);
6364                 if (!link || !link->doit)
6365                         goto out_unlock;
6366         }
6367
6368         owner = link->owner;
6369         if (!try_module_get(owner)) {
6370                 err = -EPROTONOSUPPORT;
6371                 goto out_unlock;
6372         }
6373
6374         flags = link->flags;
6375         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6376             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6377                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6378                 module_put(owner);
6379                 goto err_unlock;
6380         }
6381
6382         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6383                 doit = link->doit;
6384                 rcu_read_unlock();
6385                 if (doit)
6386                         err = doit(skb, nlh, extack);
6387                 module_put(owner);
6388                 return err;
6389         }
6390         rcu_read_unlock();
6391
6392         rtnl_lock();
6393         link = rtnl_get_link(family, type);
6394         if (link && link->doit)
6395                 err = link->doit(skb, nlh, extack);
6396         rtnl_unlock();
6397
6398         module_put(owner);
6399
6400         return err;
6401
6402 out_unlock:
6403         rcu_read_unlock();
6404         return err;
6405
6406 err_unlock:
6407         rcu_read_unlock();
6408         return -EOPNOTSUPP;
6409 }
6410
6411 static void rtnetlink_rcv(struct sk_buff *skb)
6412 {
6413         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6414 }
6415
6416 static int rtnetlink_bind(struct net *net, int group)
6417 {
6418         switch (group) {
6419         case RTNLGRP_IPV4_MROUTE_R:
6420         case RTNLGRP_IPV6_MROUTE_R:
6421                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6422                         return -EPERM;
6423                 break;
6424         }
6425         return 0;
6426 }
6427
6428 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6429 {
6430         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6431
6432         switch (event) {
6433         case NETDEV_REBOOT:
6434         case NETDEV_CHANGEMTU:
6435         case NETDEV_CHANGEADDR:
6436         case NETDEV_CHANGENAME:
6437         case NETDEV_FEAT_CHANGE:
6438         case NETDEV_BONDING_FAILOVER:
6439         case NETDEV_POST_TYPE_CHANGE:
6440         case NETDEV_NOTIFY_PEERS:
6441         case NETDEV_CHANGEUPPER:
6442         case NETDEV_RESEND_IGMP:
6443         case NETDEV_CHANGEINFODATA:
6444         case NETDEV_CHANGELOWERSTATE:
6445         case NETDEV_CHANGE_TX_QUEUE_LEN:
6446                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6447                                    GFP_KERNEL, NULL, 0, 0, NULL);
6448                 break;
6449         default:
6450                 break;
6451         }
6452         return NOTIFY_DONE;
6453 }
6454
6455 static struct notifier_block rtnetlink_dev_notifier = {
6456         .notifier_call  = rtnetlink_event,
6457 };
6458
6459
6460 static int __net_init rtnetlink_net_init(struct net *net)
6461 {
6462         struct sock *sk;
6463         struct netlink_kernel_cfg cfg = {
6464                 .groups         = RTNLGRP_MAX,
6465                 .input          = rtnetlink_rcv,
6466                 .cb_mutex       = &rtnl_mutex,
6467                 .flags          = NL_CFG_F_NONROOT_RECV,
6468                 .bind           = rtnetlink_bind,
6469         };
6470
6471         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6472         if (!sk)
6473                 return -ENOMEM;
6474         net->rtnl = sk;
6475         return 0;
6476 }
6477
6478 static void __net_exit rtnetlink_net_exit(struct net *net)
6479 {
6480         netlink_kernel_release(net->rtnl);
6481         net->rtnl = NULL;
6482 }
6483
6484 static struct pernet_operations rtnetlink_net_ops = {
6485         .init = rtnetlink_net_init,
6486         .exit = rtnetlink_net_exit,
6487 };
6488
6489 void __init rtnetlink_init(void)
6490 {
6491         if (register_pernet_subsys(&rtnetlink_net_ops))
6492                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6493
6494         register_netdevice_notifier(&rtnetlink_dev_notifier);
6495
6496         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6497                       rtnl_dump_ifinfo, 0);
6498         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6499         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6500         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6501
6502         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6503         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6504         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6505
6506         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6507         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6508
6509         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6510         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6511                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6512         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6513
6514         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6515         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6516         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6517
6518         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6519                       0);
6520         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6521
6522         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, rtnl_mdb_dump, 0);
6523         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6524         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL, 0);
6525 }