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     44
  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(sizeof(struct ifla_vf_trust)));
 965                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 966                         size += num_vfs *
 967                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 968                                  /* IFLA_VF_STATS_RX_PACKETS */
 969                                  nla_total_size_64bit(sizeof(__u64)) +
 970                                  /* IFLA_VF_STATS_TX_PACKETS */
 971                                  nla_total_size_64bit(sizeof(__u64)) +
 972                                  /* IFLA_VF_STATS_RX_BYTES */
 973                                  nla_total_size_64bit(sizeof(__u64)) +
 974                                  /* IFLA_VF_STATS_TX_BYTES */
 975                                  nla_total_size_64bit(sizeof(__u64)) +
 976                                  /* IFLA_VF_STATS_BROADCAST */
 977                                  nla_total_size_64bit(sizeof(__u64)) +
 978                                  /* IFLA_VF_STATS_MULTICAST */
 979                                  nla_total_size_64bit(sizeof(__u64)) +
 980                                  /* IFLA_VF_STATS_RX_DROPPED */
 981                                  nla_total_size_64bit(sizeof(__u64)) +
 982                                  /* IFLA_VF_STATS_TX_DROPPED */
 983                                  nla_total_size_64bit(sizeof(__u64)));
 984                 }
 985                 return size;
 986         } else
 987                 return 0;
 988 }
 989
 990 static size_t rtnl_port_size(const struct net_device *dev,
 991                              u32 ext_filter_mask)
 992 {
 993         size_t port_size = nla_total_size(4)            /* PORT_VF */
 994                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 995                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 996                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 997                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 998                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 999         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
1000         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
1001                 + port_size;
1002         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1003                 + port_size;
1004
1005         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1006             !(ext_filter_mask & RTEXT_FILTER_VF))
1007                 return 0;
1008         if (dev_num_vf(dev->dev.parent))
1009                 return port_self_size + vf_ports_size +
1010                         vf_port_size * dev_num_vf(dev->dev.parent);
1011         else
1012                 return port_self_size;
1013 }
1014
1015 static size_t rtnl_xdp_size(void)
1016 {
1017         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1018                           nla_total_size(1) +   /* XDP_ATTACHED */
1019                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1020                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1021
1022         return xdp_size;
1023 }
1024
1025 static size_t rtnl_prop_list_size(const struct net_device *dev)
1026 {
1027         struct netdev_name_node *name_node;
1028         size_t size;
1029
1030         if (list_empty(&dev->name_node->list))
1031                 return 0;
1032         size = nla_total_size(0);
1033         list_for_each_entry(name_node, &dev->name_node->list, list)
1034                 size += nla_total_size(ALTIFNAMSIZ);
1035         return size;
1036 }
1037
1038 static size_t rtnl_proto_down_size(const struct net_device *dev)
1039 {
1040         size_t size = nla_total_size(1);
1041
1042         if (dev->proto_down_reason)
1043                 size += nla_total_size(0) + nla_total_size(4);
1044
1045         return size;
1046 }
1047
1048 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1049 {
1050         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1051
1052         if (dev->devlink_port)
1053                 size += devlink_nl_port_handle_size(dev->devlink_port);
1054
1055         return size;
1056 }
1057
1058 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1059                                      u32 ext_filter_mask)
1060 {
1061         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1062                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1063                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1064                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1065                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1066                + nla_total_size(sizeof(struct rtnl_link_stats))
1067                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1068                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1069                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1070                + nla_total_size(4) /* IFLA_TXQLEN */
1071                + nla_total_size(4) /* IFLA_WEIGHT */
1072                + nla_total_size(4) /* IFLA_MTU */
1073                + nla_total_size(4) /* IFLA_LINK */
1074                + nla_total_size(4) /* IFLA_MASTER */
1075                + nla_total_size(1) /* IFLA_CARRIER */
1076                + nla_total_size(4) /* IFLA_PROMISCUITY */
1077                + nla_total_size(4) /* IFLA_ALLMULTI */
1078                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1079                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1080                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1081                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1082                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1083                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1084                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1085                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1086                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1087                + nla_total_size(1) /* IFLA_OPERSTATE */
1088                + nla_total_size(1) /* IFLA_LINKMODE */
1089                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1090                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1091                + nla_total_size(4) /* IFLA_GROUP */
1092                + nla_total_size(ext_filter_mask
1093                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1094                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1095                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1096                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1097                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1098                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1099                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1100                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1101                + rtnl_xdp_size() /* IFLA_XDP */
1102                + nla_total_size(4)  /* IFLA_EVENT */
1103                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1104                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1105                + rtnl_proto_down_size(dev)  /* proto down */
1106                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1107                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1108                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1109                + nla_total_size(4)  /* IFLA_MIN_MTU */
1110                + nla_total_size(4)  /* IFLA_MAX_MTU */
1111                + rtnl_prop_list_size(dev)
1112                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1113                + rtnl_devlink_port_size(dev)
1114                + 0;
1115 }
1116
1117 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1118 {
1119         struct nlattr *vf_ports;
1120         struct nlattr *vf_port;
1121         int vf;
1122         int err;
1123
1124         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1125         if (!vf_ports)
1126                 return -EMSGSIZE;
1127
1128         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1129                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1130                 if (!vf_port)
1131                         goto nla_put_failure;
1132                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1133                         goto nla_put_failure;
1134                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1135                 if (err == -EMSGSIZE)
1136                         goto nla_put_failure;
1137                 if (err) {
1138                         nla_nest_cancel(skb, vf_port);
1139                         continue;
1140                 }
1141                 nla_nest_end(skb, vf_port);
1142         }
1143
1144         nla_nest_end(skb, vf_ports);
1145
1146         return 0;
1147
1148 nla_put_failure:
1149         nla_nest_cancel(skb, vf_ports);
1150         return -EMSGSIZE;
1151 }
1152
1153 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1154 {
1155         struct nlattr *port_self;
1156         int err;
1157
1158         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1159         if (!port_self)
1160                 return -EMSGSIZE;
1161
1162         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1163         if (err) {
1164                 nla_nest_cancel(skb, port_self);
1165                 return (err == -EMSGSIZE) ? err : 0;
1166         }
1167
1168         nla_nest_end(skb, port_self);
1169
1170         return 0;
1171 }
1172
1173 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1174                           u32 ext_filter_mask)
1175 {
1176         int err;
1177
1178         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1179             !(ext_filter_mask & RTEXT_FILTER_VF))
1180                 return 0;
1181
1182         err = rtnl_port_self_fill(skb, dev);
1183         if (err)
1184                 return err;
1185
1186         if (dev_num_vf(dev->dev.parent)) {
1187                 err = rtnl_vf_ports_fill(skb, dev);
1188                 if (err)
1189                         return err;
1190         }
1191
1192         return 0;
1193 }
1194
1195 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1196 {
1197         int err;
1198         struct netdev_phys_item_id ppid;
1199
1200         err = dev_get_phys_port_id(dev, &ppid);
1201         if (err) {
1202                 if (err == -EOPNOTSUPP)
1203                         return 0;
1204                 return err;
1205         }
1206
1207         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1208                 return -EMSGSIZE;
1209
1210         return 0;
1211 }
1212
1213 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1214 {
1215         char name[IFNAMSIZ];
1216         int err;
1217
1218         err = dev_get_phys_port_name(dev, name, sizeof(name));
1219         if (err) {
1220                 if (err == -EOPNOTSUPP)
1221                         return 0;
1222                 return err;
1223         }
1224
1225         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1226                 return -EMSGSIZE;
1227
1228         return 0;
1229 }
1230
1231 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1232 {
1233         struct netdev_phys_item_id ppid = { };
1234         int err;
1235
1236         err = dev_get_port_parent_id(dev, &ppid, false);
1237         if (err) {
1238                 if (err == -EOPNOTSUPP)
1239                         return 0;
1240                 return err;
1241         }
1242
1243         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1244                 return -EMSGSIZE;
1245
1246         return 0;
1247 }
1248
1249 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1250                                               struct net_device *dev)
1251 {
1252         struct rtnl_link_stats64 *sp;
1253         struct nlattr *attr;
1254
1255         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1256                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1257         if (!attr)
1258                 return -EMSGSIZE;
1259
1260         sp = nla_data(attr);
1261         dev_get_stats(dev, sp);
1262
1263         attr = nla_reserve(skb, IFLA_STATS,
1264                            sizeof(struct rtnl_link_stats));
1265         if (!attr)
1266                 return -EMSGSIZE;
1267
1268         copy_rtnl_link_stats(nla_data(attr), sp);
1269
1270         return 0;
1271 }
1272
1273 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1274                                                struct net_device *dev,
1275                                                int vfs_num,
1276                                                u32 ext_filter_mask)
1277 {
1278         struct ifla_vf_rss_query_en vf_rss_query_en;
1279         struct nlattr *vf, *vfstats, *vfvlanlist;
1280         struct ifla_vf_link_state vf_linkstate;
1281         struct ifla_vf_vlan_info vf_vlan_info;
1282         struct ifla_vf_spoofchk vf_spoofchk;
1283         struct ifla_vf_tx_rate vf_tx_rate;
1284         struct ifla_vf_stats vf_stats;
1285         struct ifla_vf_trust vf_trust;
1286         struct ifla_vf_vlan vf_vlan;
1287         struct ifla_vf_rate vf_rate;
1288         struct ifla_vf_mac vf_mac;
1289         struct ifla_vf_broadcast vf_broadcast;
1290         struct ifla_vf_info ivi;
1291         struct ifla_vf_guid node_guid;
1292         struct ifla_vf_guid port_guid;
1293
1294         memset(&ivi, 0, sizeof(ivi));
1295
1296         /* Not all SR-IOV capable drivers support the
1297          * spoofcheck and "RSS query enable" query.  Preset to
1298          * -1 so the user space tool can detect that the driver
1299          * didn't report anything.
1300          */
1301         ivi.spoofchk = -1;
1302         ivi.rss_query_en = -1;
1303         ivi.trusted = -1;
1304         /* The default value for VF link state is "auto"
1305          * IFLA_VF_LINK_STATE_AUTO which equals zero
1306          */
1307         ivi.linkstate = 0;
1308         /* VLAN Protocol by default is 802.1Q */
1309         ivi.vlan_proto = htons(ETH_P_8021Q);
1310         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1311                 return 0;
1312
1313         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1314         memset(&node_guid, 0, sizeof(node_guid));
1315         memset(&port_guid, 0, sizeof(port_guid));
1316
1317         vf_mac.vf =
1318                 vf_vlan.vf =
1319                 vf_vlan_info.vf =
1320                 vf_rate.vf =
1321                 vf_tx_rate.vf =
1322                 vf_spoofchk.vf =
1323                 vf_linkstate.vf =
1324                 vf_rss_query_en.vf =
1325                 vf_trust.vf =
1326                 node_guid.vf =
1327                 port_guid.vf = ivi.vf;
1328
1329         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1330         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1331         vf_vlan.vlan = ivi.vlan;
1332         vf_vlan.qos = ivi.qos;
1333         vf_vlan_info.vlan = ivi.vlan;
1334         vf_vlan_info.qos = ivi.qos;
1335         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1336         vf_tx_rate.rate = ivi.max_tx_rate;
1337         vf_rate.min_tx_rate = ivi.min_tx_rate;
1338         vf_rate.max_tx_rate = ivi.max_tx_rate;
1339         vf_spoofchk.setting = ivi.spoofchk;
1340         vf_linkstate.link_state = ivi.linkstate;
1341         vf_rss_query_en.setting = ivi.rss_query_en;
1342         vf_trust.setting = ivi.trusted;
1343         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1344         if (!vf)
1345                 return -EMSGSIZE;
1346         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1347             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1348             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1349             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1350                     &vf_rate) ||
1351             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1352                     &vf_tx_rate) ||
1353             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1354                     &vf_spoofchk) ||
1355             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1356                     &vf_linkstate) ||
1357             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1358                     sizeof(vf_rss_query_en),
1359                     &vf_rss_query_en) ||
1360             nla_put(skb, IFLA_VF_TRUST,
1361                     sizeof(vf_trust), &vf_trust))
1362                 goto nla_put_vf_failure;
1363
1364         if (dev->netdev_ops->ndo_get_vf_guid &&
1365             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1366                                               &port_guid)) {
1367                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1368                             &node_guid) ||
1369                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1370                             &port_guid))
1371                         goto nla_put_vf_failure;
1372         }
1373         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1374         if (!vfvlanlist)
1375                 goto nla_put_vf_failure;
1376         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1377                     &vf_vlan_info)) {
1378                 nla_nest_cancel(skb, vfvlanlist);
1379                 goto nla_put_vf_failure;
1380         }
1381         nla_nest_end(skb, vfvlanlist);
1382         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1383                 memset(&vf_stats, 0, sizeof(vf_stats));
1384                 if (dev->netdev_ops->ndo_get_vf_stats)
1385                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1386                                                           &vf_stats);
1387                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1388                 if (!vfstats)
1389                         goto nla_put_vf_failure;
1390                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1391                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1392                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1393                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1394                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1395                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1396                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1397                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1398                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1399                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1400                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1401                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1402                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1403                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1404                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1405                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1406                         nla_nest_cancel(skb, vfstats);
1407                         goto nla_put_vf_failure;
1408                 }
1409                 nla_nest_end(skb, vfstats);
1410         }
1411         nla_nest_end(skb, vf);
1412         return 0;
1413
1414 nla_put_vf_failure:
1415         nla_nest_cancel(skb, vf);
1416         return -EMSGSIZE;
1417 }
1418
1419 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1420                                            struct net_device *dev,
1421                                            u32 ext_filter_mask)
1422 {
1423         struct nlattr *vfinfo;
1424         int i, num_vfs;
1425
1426         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1427                 return 0;
1428
1429         num_vfs = dev_num_vf(dev->dev.parent);
1430         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1431                 return -EMSGSIZE;
1432
1433         if (!dev->netdev_ops->ndo_get_vf_config)
1434                 return 0;
1435
1436         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1437         if (!vfinfo)
1438                 return -EMSGSIZE;
1439
1440         for (i = 0; i < num_vfs; i++) {
1441                 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1442                         nla_nest_cancel(skb, vfinfo);
1443                         return -EMSGSIZE;
1444                 }
1445         }
1446
1447         nla_nest_end(skb, vfinfo);
1448         return 0;
1449 }
1450
1451 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1452 {
1453         struct rtnl_link_ifmap map;
1454
1455         memset(&map, 0, sizeof(map));
1456         map.mem_start   = dev->mem_start;
1457         map.mem_end     = dev->mem_end;
1458         map.base_addr   = dev->base_addr;
1459         map.irq         = dev->irq;
1460         map.dma         = dev->dma;
1461         map.port        = dev->if_port;
1462
1463         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1464                 return -EMSGSIZE;
1465
1466         return 0;
1467 }
1468
1469 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1470 {
1471         const struct bpf_prog *generic_xdp_prog;
1472
1473         ASSERT_RTNL();
1474
1475         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1476         if (!generic_xdp_prog)
1477                 return 0;
1478         return generic_xdp_prog->aux->id;
1479 }
1480
1481 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1482 {
1483         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1484 }
1485
1486 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1487 {
1488         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1489 }
1490
1491 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1492                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1493                                u32 (*get_prog_id)(struct net_device *dev))
1494 {
1495         u32 curr_id;
1496         int err;
1497
1498         curr_id = get_prog_id(dev);
1499         if (!curr_id)
1500                 return 0;
1501
1502         *prog_id = curr_id;
1503         err = nla_put_u32(skb, attr, curr_id);
1504         if (err)
1505                 return err;
1506
1507         if (*mode != XDP_ATTACHED_NONE)
1508                 *mode = XDP_ATTACHED_MULTI;
1509         else
1510                 *mode = tgt_mode;
1511
1512         return 0;
1513 }
1514
1515 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1516 {
1517         struct nlattr *xdp;
1518         u32 prog_id;
1519         int err;
1520         u8 mode;
1521
1522         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1523         if (!xdp)
1524                 return -EMSGSIZE;
1525
1526         prog_id = 0;
1527         mode = XDP_ATTACHED_NONE;
1528         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1529                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1530         if (err)
1531                 goto err_cancel;
1532         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1533                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1534         if (err)
1535                 goto err_cancel;
1536         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1537                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1538         if (err)
1539                 goto err_cancel;
1540
1541         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1542         if (err)
1543                 goto err_cancel;
1544
1545         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1546                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1547                 if (err)
1548                         goto err_cancel;
1549         }
1550
1551         nla_nest_end(skb, xdp);
1552         return 0;
1553
1554 err_cancel:
1555         nla_nest_cancel(skb, xdp);
1556         return err;
1557 }
1558
1559 static u32 rtnl_get_event(unsigned long event)
1560 {
1561         u32 rtnl_event_type = IFLA_EVENT_NONE;
1562
1563         switch (event) {
1564         case NETDEV_REBOOT:
1565                 rtnl_event_type = IFLA_EVENT_REBOOT;
1566                 break;
1567         case NETDEV_FEAT_CHANGE:
1568                 rtnl_event_type = IFLA_EVENT_FEATURES;
1569                 break;
1570         case NETDEV_BONDING_FAILOVER:
1571                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1572                 break;
1573         case NETDEV_NOTIFY_PEERS:
1574                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1575                 break;
1576         case NETDEV_RESEND_IGMP:
1577                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1578                 break;
1579         case NETDEV_CHANGEINFODATA:
1580                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1581                 break;
1582         default:
1583                 break;
1584         }
1585
1586         return rtnl_event_type;
1587 }
1588
1589 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1590 {
1591         const struct net_device *upper_dev;
1592         int ret = 0;
1593
1594         rcu_read_lock();
1595
1596         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1597         if (upper_dev)
1598                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1599
1600         rcu_read_unlock();
1601         return ret;
1602 }
1603
1604 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1605                           bool force)
1606 {
1607         int ifindex = dev_get_iflink(dev);
1608
1609         if (force || dev->ifindex != ifindex)
1610                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1611
1612         return 0;
1613 }
1614
1615 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1616                                               struct net_device *dev)
1617 {
1618         char buf[IFALIASZ];
1619         int ret;
1620
1621         ret = dev_get_alias(dev, buf, sizeof(buf));
1622         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1623 }
1624
1625 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1626                                   const struct net_device *dev,
1627                                   struct net *src_net, gfp_t gfp)
1628 {
1629         bool put_iflink = false;
1630
1631         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1632                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1633
1634                 if (!net_eq(dev_net(dev), link_net)) {
1635                         int id = peernet2id_alloc(src_net, link_net, gfp);
1636
1637                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1638                                 return -EMSGSIZE;
1639
1640                         put_iflink = true;
1641                 }
1642         }
1643
1644         return nla_put_iflink(skb, dev, put_iflink);
1645 }
1646
1647 static int rtnl_fill_link_af(struct sk_buff *skb,
1648                              const struct net_device *dev,
1649                              u32 ext_filter_mask)
1650 {
1651         const struct rtnl_af_ops *af_ops;
1652         struct nlattr *af_spec;
1653
1654         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1655         if (!af_spec)
1656                 return -EMSGSIZE;
1657
1658         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1659                 struct nlattr *af;
1660                 int err;
1661
1662                 if (!af_ops->fill_link_af)
1663                         continue;
1664
1665                 af = nla_nest_start_noflag(skb, af_ops->family);
1666                 if (!af)
1667                         return -EMSGSIZE;
1668
1669                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1670                 /*
1671                  * Caller may return ENODATA to indicate that there
1672                  * was no data to be dumped. This is not an error, it
1673                  * means we should trim the attribute header and
1674                  * continue.
1675                  */
1676                 if (err == -ENODATA)
1677                         nla_nest_cancel(skb, af);
1678                 else if (err < 0)
1679                         return -EMSGSIZE;
1680
1681                 nla_nest_end(skb, af);
1682         }
1683
1684         nla_nest_end(skb, af_spec);
1685         return 0;
1686 }
1687
1688 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1689                                  const struct net_device *dev)
1690 {
1691         struct netdev_name_node *name_node;
1692         int count = 0;
1693
1694         list_for_each_entry(name_node, &dev->name_node->list, list) {
1695                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1696                         return -EMSGSIZE;
1697                 count++;
1698         }
1699         return count;
1700 }
1701
1702 static int rtnl_fill_prop_list(struct sk_buff *skb,
1703                                const struct net_device *dev)
1704 {
1705         struct nlattr *prop_list;
1706         int ret;
1707
1708         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1709         if (!prop_list)
1710                 return -EMSGSIZE;
1711
1712         ret = rtnl_fill_alt_ifnames(skb, dev);
1713         if (ret <= 0)
1714                 goto nest_cancel;
1715
1716         nla_nest_end(skb, prop_list);
1717         return 0;
1718
1719 nest_cancel:
1720         nla_nest_cancel(skb, prop_list);
1721         return ret;
1722 }
1723
1724 static int rtnl_fill_proto_down(struct sk_buff *skb,
1725                                 const struct net_device *dev)
1726 {
1727         struct nlattr *pr;
1728         u32 preason;
1729
1730         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1731                 goto nla_put_failure;
1732
1733         preason = dev->proto_down_reason;
1734         if (!preason)
1735                 return 0;
1736
1737         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1738         if (!pr)
1739                 return -EMSGSIZE;
1740
1741         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1742                 nla_nest_cancel(skb, pr);
1743                 goto nla_put_failure;
1744         }
1745
1746         nla_nest_end(skb, pr);
1747         return 0;
1748
1749 nla_put_failure:
1750         return -EMSGSIZE;
1751 }
1752
1753 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1754                                   const struct net_device *dev)
1755 {
1756         struct nlattr *devlink_port_nest;
1757         int ret;
1758
1759         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1760         if (!devlink_port_nest)
1761                 return -EMSGSIZE;
1762
1763         if (dev->devlink_port) {
1764                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1765                 if (ret < 0)
1766                         goto nest_cancel;
1767         }
1768
1769         nla_nest_end(skb, devlink_port_nest);
1770         return 0;
1771
1772 nest_cancel:
1773         nla_nest_cancel(skb, devlink_port_nest);
1774         return ret;
1775 }
1776
1777 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1778                             struct net_device *dev, struct net *src_net,
1779                             int type, u32 pid, u32 seq, u32 change,
1780                             unsigned int flags, u32 ext_filter_mask,
1781                             u32 event, int *new_nsid, int new_ifindex,
1782                             int tgt_netnsid, gfp_t gfp)
1783 {
1784         struct ifinfomsg *ifm;
1785         struct nlmsghdr *nlh;
1786         struct Qdisc *qdisc;
1787
1788         ASSERT_RTNL();
1789         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1790         if (nlh == NULL)
1791                 return -EMSGSIZE;
1792
1793         ifm = nlmsg_data(nlh);
1794         ifm->ifi_family = AF_UNSPEC;
1795         ifm->__ifi_pad = 0;
1796         ifm->ifi_type = dev->type;
1797         ifm->ifi_index = dev->ifindex;
1798         ifm->ifi_flags = dev_get_flags(dev);
1799         ifm->ifi_change = change;
1800
1801         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1802                 goto nla_put_failure;
1803
1804         qdisc = rtnl_dereference(dev->qdisc);
1805         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1806             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1807             nla_put_u8(skb, IFLA_OPERSTATE,
1808                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1809             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1810             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1811             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1812             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1813             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1814             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1815             nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1816             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1817             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1818             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1819             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1820             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE, dev->gso_ipv4_max_size) ||
1821             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE, dev->gro_ipv4_max_size) ||
1822             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1823             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1824 #ifdef CONFIG_RPS
1825             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1826 #endif
1827             put_master_ifindex(skb, dev) ||
1828             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1829             (qdisc &&
1830              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1831             nla_put_ifalias(skb, dev) ||
1832             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1833                         atomic_read(&dev->carrier_up_count) +
1834                         atomic_read(&dev->carrier_down_count)) ||
1835             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1836                         atomic_read(&dev->carrier_up_count)) ||
1837             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1838                         atomic_read(&dev->carrier_down_count)))
1839                 goto nla_put_failure;
1840
1841         if (rtnl_fill_proto_down(skb, dev))
1842                 goto nla_put_failure;
1843
1844         if (event != IFLA_EVENT_NONE) {
1845                 if (nla_put_u32(skb, IFLA_EVENT, event))
1846                         goto nla_put_failure;
1847         }
1848
1849         if (rtnl_fill_link_ifmap(skb, dev))
1850                 goto nla_put_failure;
1851
1852         if (dev->addr_len) {
1853                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1854                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1855                         goto nla_put_failure;
1856         }
1857
1858         if (rtnl_phys_port_id_fill(skb, dev))
1859                 goto nla_put_failure;
1860
1861         if (rtnl_phys_port_name_fill(skb, dev))
1862                 goto nla_put_failure;
1863
1864         if (rtnl_phys_switch_id_fill(skb, dev))
1865                 goto nla_put_failure;
1866
1867         if (rtnl_fill_stats(skb, dev))
1868                 goto nla_put_failure;
1869
1870         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1871                 goto nla_put_failure;
1872
1873         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1874                 goto nla_put_failure;
1875
1876         if (rtnl_xdp_fill(skb, dev))
1877                 goto nla_put_failure;
1878
1879         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1880                 if (rtnl_link_fill(skb, dev) < 0)
1881                         goto nla_put_failure;
1882         }
1883
1884         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1885                 goto nla_put_failure;
1886
1887         if (new_nsid &&
1888             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1889                 goto nla_put_failure;
1890         if (new_ifindex &&
1891             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1892                 goto nla_put_failure;
1893
1894         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1895             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1896                 goto nla_put_failure;
1897
1898         rcu_read_lock();
1899         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1900                 goto nla_put_failure_rcu;
1901         rcu_read_unlock();
1902
1903         if (rtnl_fill_prop_list(skb, dev))
1904                 goto nla_put_failure;
1905
1906         if (dev->dev.parent &&
1907             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1908                            dev_name(dev->dev.parent)))
1909                 goto nla_put_failure;
1910
1911         if (dev->dev.parent && dev->dev.parent->bus &&
1912             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1913                            dev->dev.parent->bus->name))
1914                 goto nla_put_failure;
1915
1916         if (rtnl_fill_devlink_port(skb, dev))
1917                 goto nla_put_failure;
1918
1919         nlmsg_end(skb, nlh);
1920         return 0;
1921
1922 nla_put_failure_rcu:
1923         rcu_read_unlock();
1924 nla_put_failure:
1925         nlmsg_cancel(skb, nlh);
1926         return -EMSGSIZE;
1927 }
1928
1929 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1930         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1931         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1932         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1933         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1934         [IFLA_MTU]              = { .type = NLA_U32 },
1935         [IFLA_LINK]             = { .type = NLA_U32 },
1936         [IFLA_MASTER]           = { .type = NLA_U32 },
1937         [IFLA_CARRIER]          = { .type = NLA_U8 },
1938         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1939         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1940         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1941         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1942         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1943         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1944         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1945         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1946          * allow 0-length string (needed to remove an alias).
1947          */
1948         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1949         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1950         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1951         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1952         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1953         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1954         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1955         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1956         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1957         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1958         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1959         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1960         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1961         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1962         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1963         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1964         [IFLA_XDP]              = { .type = NLA_NESTED },
1965         [IFLA_EVENT]            = { .type = NLA_U32 },
1966         [IFLA_GROUP]            = { .type = NLA_U32 },
1967         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1968         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1969         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1970         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1971         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1972         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1973         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1974                                     .len = ALTIFNAMSIZ - 1 },
1975         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1976         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1977         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1978         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1979         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
1980         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
1981         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
1982         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
1983         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1984         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1985 };
1986
1987 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1988         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1989         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1990         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1991         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1992 };
1993
1994 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1995         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1996         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1997         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1998         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1999         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
2000         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
2001         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
2002         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
2003         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
2004         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2005         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2006         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2007         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2008 };
2009
2010 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2011         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2012         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2013                                     .len = PORT_PROFILE_MAX },
2014         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2015                                       .len = PORT_UUID_MAX },
2016         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2017                                     .len = PORT_UUID_MAX },
2018         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2019         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2020
2021         /* Unused, but we need to keep it here since user space could
2022          * fill it. It's also broken with regard to NLA_BINARY use in
2023          * combination with structs.
2024          */
2025         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2026                                     .len = sizeof(struct ifla_port_vsi) },
2027 };
2028
2029 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2030         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2031         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2032         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2033         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2034         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2035         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2036 };
2037
2038 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2039 {
2040         const struct rtnl_link_ops *ops = NULL;
2041         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2042
2043         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2044                 return NULL;
2045
2046         if (linfo[IFLA_INFO_KIND]) {
2047                 char kind[MODULE_NAME_LEN];
2048
2049                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2050                 ops = rtnl_link_ops_get(kind);
2051         }
2052
2053         return ops;
2054 }
2055
2056 static bool link_master_filtered(struct net_device *dev, int master_idx)
2057 {
2058         struct net_device *master;
2059
2060         if (!master_idx)
2061                 return false;
2062
2063         master = netdev_master_upper_dev_get(dev);
2064
2065         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2066          * another invalid value for ifindex to denote "no master".
2067          */
2068         if (master_idx == -1)
2069                 return !!master;
2070
2071         if (!master || master->ifindex != master_idx)
2072                 return true;
2073
2074         return false;
2075 }
2076
2077 static bool link_kind_filtered(const struct net_device *dev,
2078                                const struct rtnl_link_ops *kind_ops)
2079 {
2080         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2081                 return true;
2082
2083         return false;
2084 }
2085
2086 static bool link_dump_filtered(struct net_device *dev,
2087                                int master_idx,
2088                                const struct rtnl_link_ops *kind_ops)
2089 {
2090         if (link_master_filtered(dev, master_idx) ||
2091             link_kind_filtered(dev, kind_ops))
2092                 return true;
2093
2094         return false;
2095 }
2096
2097 /**
2098  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2099  * @sk: netlink socket
2100  * @netnsid: network namespace identifier
2101  *
2102  * Returns the network namespace identified by netnsid on success or an error
2103  * pointer on failure.
2104  */
2105 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2106 {
2107         struct net *net;
2108
2109         net = get_net_ns_by_id(sock_net(sk), netnsid);
2110         if (!net)
2111                 return ERR_PTR(-EINVAL);
2112
2113         /* For now, the caller is required to have CAP_NET_ADMIN in
2114          * the user namespace owning the target net ns.
2115          */
2116         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2117                 put_net(net);
2118                 return ERR_PTR(-EACCES);
2119         }
2120         return net;
2121 }
2122 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2123
2124 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2125                                       bool strict_check, struct nlattr **tb,
2126                                       struct netlink_ext_ack *extack)
2127 {
2128         int hdrlen;
2129
2130         if (strict_check) {
2131                 struct ifinfomsg *ifm;
2132
2133                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2134                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2135                         return -EINVAL;
2136                 }
2137
2138                 ifm = nlmsg_data(nlh);
2139                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2140                     ifm->ifi_change) {
2141                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2142                         return -EINVAL;
2143                 }
2144                 if (ifm->ifi_index) {
2145                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2146                         return -EINVAL;
2147                 }
2148
2149                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2150                                                      IFLA_MAX, ifla_policy,
2151                                                      extack);
2152         }
2153
2154         /* A hack to preserve kernel<->userspace interface.
2155          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2156          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2157          * what iproute2 < v3.9.0 used.
2158          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2159          * attribute, its netlink message is shorter than struct ifinfomsg.
2160          */
2161         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2162                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2163
2164         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2165                                       extack);
2166 }
2167
2168 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2169 {
2170         struct netlink_ext_ack *extack = cb->extack;
2171         const struct nlmsghdr *nlh = cb->nlh;
2172         struct net *net = sock_net(skb->sk);
2173         struct net *tgt_net = net;
2174         int h, s_h;
2175         int idx = 0, s_idx;
2176         struct net_device *dev;
2177         struct hlist_head *head;
2178         struct nlattr *tb[IFLA_MAX+1];
2179         u32 ext_filter_mask = 0;
2180         const struct rtnl_link_ops *kind_ops = NULL;
2181         unsigned int flags = NLM_F_MULTI;
2182         int master_idx = 0;
2183         int netnsid = -1;
2184         int err, i;
2185
2186         s_h = cb->args[0];
2187         s_idx = cb->args[1];
2188
2189         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2190         if (err < 0) {
2191                 if (cb->strict_check)
2192                         return err;
2193
2194                 goto walk_entries;
2195         }
2196
2197         for (i = 0; i <= IFLA_MAX; ++i) {
2198                 if (!tb[i])
2199                         continue;
2200
2201                 /* new attributes should only be added with strict checking */
2202                 switch (i) {
2203                 case IFLA_TARGET_NETNSID:
2204                         netnsid = nla_get_s32(tb[i]);
2205                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2206                         if (IS_ERR(tgt_net)) {
2207                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2208                                 return PTR_ERR(tgt_net);
2209                         }
2210                         break;
2211                 case IFLA_EXT_MASK:
2212                         ext_filter_mask = nla_get_u32(tb[i]);
2213                         break;
2214                 case IFLA_MASTER:
2215                         master_idx = nla_get_u32(tb[i]);
2216                         break;
2217                 case IFLA_LINKINFO:
2218                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2219                         break;
2220                 default:
2221                         if (cb->strict_check) {
2222                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2223                                 return -EINVAL;
2224                         }
2225                 }
2226         }
2227
2228         if (master_idx || kind_ops)
2229                 flags |= NLM_F_DUMP_FILTERED;
2230
2231 walk_entries:
2232         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2233                 idx = 0;
2234                 head = &tgt_net->dev_index_head[h];
2235                 hlist_for_each_entry(dev, head, index_hlist) {
2236                         if (link_dump_filtered(dev, master_idx, kind_ops))
2237                                 goto cont;
2238                         if (idx < s_idx)
2239                                 goto cont;
2240                         err = rtnl_fill_ifinfo(skb, dev, net,
2241                                                RTM_NEWLINK,
2242                                                NETLINK_CB(cb->skb).portid,
2243                                                nlh->nlmsg_seq, 0, flags,
2244                                                ext_filter_mask, 0, NULL, 0,
2245                                                netnsid, GFP_KERNEL);
2246
2247                         if (err < 0) {
2248                                 if (likely(skb->len))
2249                                         goto out;
2250
2251                                 goto out_err;
2252                         }
2253 cont:
2254                         idx++;
2255                 }
2256         }
2257 out:
2258         err = skb->len;
2259 out_err:
2260         cb->args[1] = idx;
2261         cb->args[0] = h;
2262         cb->seq = tgt_net->dev_base_seq;
2263         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2264         if (netnsid >= 0)
2265                 put_net(tgt_net);
2266
2267         return err;
2268 }
2269
2270 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2271                              struct netlink_ext_ack *exterr)
2272 {
2273         const struct ifinfomsg *ifmp;
2274         const struct nlattr *attrs;
2275         size_t len;
2276
2277         ifmp = nla_data(nla_peer);
2278         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2279         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2280
2281         if (ifmp->ifi_index < 0) {
2282                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2283                                     "ifindex can't be negative");
2284                 return -EINVAL;
2285         }
2286
2287         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2288                                     exterr);
2289 }
2290 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2291
2292 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2293 {
2294         struct net *net;
2295         /* Examine the link attributes and figure out which
2296          * network namespace we are talking about.
2297          */
2298         if (tb[IFLA_NET_NS_PID])
2299                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2300         else if (tb[IFLA_NET_NS_FD])
2301                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2302         else
2303                 net = get_net(src_net);
2304         return net;
2305 }
2306 EXPORT_SYMBOL(rtnl_link_get_net);
2307
2308 /* Figure out which network namespace we are talking about by
2309  * examining the link attributes in the following order:
2310  *
2311  * 1. IFLA_NET_NS_PID
2312  * 2. IFLA_NET_NS_FD
2313  * 3. IFLA_TARGET_NETNSID
2314  */
2315 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2316                                                struct nlattr *tb[])
2317 {
2318         struct net *net;
2319
2320         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2321                 return rtnl_link_get_net(src_net, tb);
2322
2323         if (!tb[IFLA_TARGET_NETNSID])
2324                 return get_net(src_net);
2325
2326         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2327         if (!net)
2328                 return ERR_PTR(-EINVAL);
2329
2330         return net;
2331 }
2332
2333 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2334                                              struct net *src_net,
2335                                              struct nlattr *tb[], int cap)
2336 {
2337         struct net *net;
2338
2339         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2340         if (IS_ERR(net))
2341                 return net;
2342
2343         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2344                 put_net(net);
2345                 return ERR_PTR(-EPERM);
2346         }
2347
2348         return net;
2349 }
2350
2351 /* Verify that rtnetlink requests do not pass additional properties
2352  * potentially referring to different network namespaces.
2353  */
2354 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2355                                     struct netlink_ext_ack *extack,
2356                                     bool netns_id_only)
2357 {
2358
2359         if (netns_id_only) {
2360                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2361                         return 0;
2362
2363                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2364                 return -EOPNOTSUPP;
2365         }
2366
2367         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2368                 goto invalid_attr;
2369
2370         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2371                 goto invalid_attr;
2372
2373         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2374                 goto invalid_attr;
2375
2376         return 0;
2377
2378 invalid_attr:
2379         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2380         return -EINVAL;
2381 }
2382
2383 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2384                              int max_tx_rate)
2385 {
2386         const struct net_device_ops *ops = dev->netdev_ops;
2387
2388         if (!ops->ndo_set_vf_rate)
2389                 return -EOPNOTSUPP;
2390         if (max_tx_rate && max_tx_rate < min_tx_rate)
2391                 return -EINVAL;
2392
2393         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2394 }
2395
2396 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2397                             struct netlink_ext_ack *extack)
2398 {
2399         if (tb[IFLA_ADDRESS] &&
2400             nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2401                 return -EINVAL;
2402
2403         if (tb[IFLA_BROADCAST] &&
2404             nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2405                 return -EINVAL;
2406
2407         if (tb[IFLA_GSO_MAX_SIZE] &&
2408             nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2409                 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2410                 return -EINVAL;
2411         }
2412
2413         if (tb[IFLA_GSO_MAX_SEGS] &&
2414             (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2415              nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2416                 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2417                 return -EINVAL;
2418         }
2419
2420         if (tb[IFLA_GRO_MAX_SIZE] &&
2421             nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2422                 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2423                 return -EINVAL;
2424         }
2425
2426         if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2427             nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2428                 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2429                 return -EINVAL;
2430         }
2431
2432         if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2433             nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2434                 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2435                 return -EINVAL;
2436         }
2437
2438         if (tb[IFLA_AF_SPEC]) {
2439                 struct nlattr *af;
2440                 int rem, err;
2441
2442                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2443                         const struct rtnl_af_ops *af_ops;
2444
2445                         af_ops = rtnl_af_lookup(nla_type(af));
2446                         if (!af_ops)
2447                                 return -EAFNOSUPPORT;
2448
2449                         if (!af_ops->set_link_af)
2450                                 return -EOPNOTSUPP;
2451
2452                         if (af_ops->validate_link_af) {
2453                                 err = af_ops->validate_link_af(dev, af, extack);
2454                                 if (err < 0)
2455                                         return err;
2456                         }
2457                 }
2458         }
2459
2460         return 0;
2461 }
2462
2463 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2464                                   int guid_type)
2465 {
2466         const struct net_device_ops *ops = dev->netdev_ops;
2467
2468         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2469 }
2470
2471 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2472 {
2473         if (dev->type != ARPHRD_INFINIBAND)
2474                 return -EOPNOTSUPP;
2475
2476         return handle_infiniband_guid(dev, ivt, guid_type);
2477 }
2478
2479 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2480 {
2481         const struct net_device_ops *ops = dev->netdev_ops;
2482         int err = -EINVAL;
2483
2484         if (tb[IFLA_VF_MAC]) {
2485                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2486
2487                 if (ivm->vf >= INT_MAX)
2488                         return -EINVAL;
2489                 err = -EOPNOTSUPP;
2490                 if (ops->ndo_set_vf_mac)
2491                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2492                                                   ivm->mac);
2493                 if (err < 0)
2494                         return err;
2495         }
2496
2497         if (tb[IFLA_VF_VLAN]) {
2498                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2499
2500                 if (ivv->vf >= INT_MAX)
2501                         return -EINVAL;
2502                 err = -EOPNOTSUPP;
2503                 if (ops->ndo_set_vf_vlan)
2504                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2505                                                    ivv->qos,
2506                                                    htons(ETH_P_8021Q));
2507                 if (err < 0)
2508                         return err;
2509         }
2510
2511         if (tb[IFLA_VF_VLAN_LIST]) {
2512                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2513                 struct nlattr *attr;
2514                 int rem, len = 0;
2515
2516                 err = -EOPNOTSUPP;
2517                 if (!ops->ndo_set_vf_vlan)
2518                         return err;
2519
2520                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2521                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2522                             nla_len(attr) < NLA_HDRLEN) {
2523                                 return -EINVAL;
2524                         }
2525                         if (len >= MAX_VLAN_LIST_LEN)
2526                                 return -EOPNOTSUPP;
2527                         ivvl[len] = nla_data(attr);
2528
2529                         len++;
2530                 }
2531                 if (len == 0)
2532                         return -EINVAL;
2533
2534                 if (ivvl[0]->vf >= INT_MAX)
2535                         return -EINVAL;
2536                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2537                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2538                 if (err < 0)
2539                         return err;
2540         }
2541
2542         if (tb[IFLA_VF_TX_RATE]) {
2543                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2544                 struct ifla_vf_info ivf;
2545
2546                 if (ivt->vf >= INT_MAX)
2547                         return -EINVAL;
2548                 err = -EOPNOTSUPP;
2549                 if (ops->ndo_get_vf_config)
2550                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2551                 if (err < 0)
2552                         return err;
2553
2554                 err = rtnl_set_vf_rate(dev, ivt->vf,
2555                                        ivf.min_tx_rate, ivt->rate);
2556                 if (err < 0)
2557                         return err;
2558         }
2559
2560         if (tb[IFLA_VF_RATE]) {
2561                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2562
2563                 if (ivt->vf >= INT_MAX)
2564                         return -EINVAL;
2565
2566                 err = rtnl_set_vf_rate(dev, ivt->vf,
2567                                        ivt->min_tx_rate, ivt->max_tx_rate);
2568                 if (err < 0)
2569                         return err;
2570         }
2571
2572         if (tb[IFLA_VF_SPOOFCHK]) {
2573                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2574
2575                 if (ivs->vf >= INT_MAX)
2576                         return -EINVAL;
2577                 err = -EOPNOTSUPP;
2578                 if (ops->ndo_set_vf_spoofchk)
2579                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2580                                                        ivs->setting);
2581                 if (err < 0)
2582                         return err;
2583         }
2584
2585         if (tb[IFLA_VF_LINK_STATE]) {
2586                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2587
2588                 if (ivl->vf >= INT_MAX)
2589                         return -EINVAL;
2590                 err = -EOPNOTSUPP;
2591                 if (ops->ndo_set_vf_link_state)
2592                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2593                                                          ivl->link_state);
2594                 if (err < 0)
2595                         return err;
2596         }
2597
2598         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2599                 struct ifla_vf_rss_query_en *ivrssq_en;
2600
2601                 err = -EOPNOTSUPP;
2602                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2603                 if (ivrssq_en->vf >= INT_MAX)
2604                         return -EINVAL;
2605                 if (ops->ndo_set_vf_rss_query_en)
2606                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2607                                                            ivrssq_en->setting);
2608                 if (err < 0)
2609                         return err;
2610         }
2611
2612         if (tb[IFLA_VF_TRUST]) {
2613                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2614
2615                 if (ivt->vf >= INT_MAX)
2616                         return -EINVAL;
2617                 err = -EOPNOTSUPP;
2618                 if (ops->ndo_set_vf_trust)
2619                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2620                 if (err < 0)
2621                         return err;
2622         }
2623
2624         if (tb[IFLA_VF_IB_NODE_GUID]) {
2625                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2626
2627                 if (ivt->vf >= INT_MAX)
2628                         return -EINVAL;
2629                 if (!ops->ndo_set_vf_guid)
2630                         return -EOPNOTSUPP;
2631                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2632         }
2633
2634         if (tb[IFLA_VF_IB_PORT_GUID]) {
2635                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2636
2637                 if (ivt->vf >= INT_MAX)
2638                         return -EINVAL;
2639                 if (!ops->ndo_set_vf_guid)
2640                         return -EOPNOTSUPP;
2641
2642                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2643         }
2644
2645         return err;
2646 }
2647
2648 static int do_set_master(struct net_device *dev, int ifindex,
2649                          struct netlink_ext_ack *extack)
2650 {
2651         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2652         const struct net_device_ops *ops;
2653         int err;
2654
2655         if (upper_dev) {
2656                 if (upper_dev->ifindex == ifindex)
2657                         return 0;
2658                 ops = upper_dev->netdev_ops;
2659                 if (ops->ndo_del_slave) {
2660                         err = ops->ndo_del_slave(upper_dev, dev);
2661                         if (err)
2662                                 return err;
2663                 } else {
2664                         return -EOPNOTSUPP;
2665                 }
2666         }
2667
2668         if (ifindex) {
2669                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2670                 if (!upper_dev)
2671                         return -EINVAL;
2672                 ops = upper_dev->netdev_ops;
2673                 if (ops->ndo_add_slave) {
2674                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2675                         if (err)
2676                                 return err;
2677                 } else {
2678                         return -EOPNOTSUPP;
2679                 }
2680         }
2681         return 0;
2682 }
2683
2684 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2685         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2686         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2687 };
2688
2689 static int do_set_proto_down(struct net_device *dev,
2690                              struct nlattr *nl_proto_down,
2691                              struct nlattr *nl_proto_down_reason,
2692                              struct netlink_ext_ack *extack)
2693 {
2694         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2695         unsigned long mask = 0;
2696         u32 value;
2697         bool proto_down;
2698         int err;
2699
2700         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2701                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2702                 return -EOPNOTSUPP;
2703         }
2704
2705         if (nl_proto_down_reason) {
2706                 err = nla_parse_nested_deprecated(pdreason,
2707                                                   IFLA_PROTO_DOWN_REASON_MAX,
2708                                                   nl_proto_down_reason,
2709                                                   ifla_proto_down_reason_policy,
2710                                                   NULL);
2711                 if (err < 0)
2712                         return err;
2713
2714                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2715                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2716                         return -EINVAL;
2717                 }
2718
2719                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2720
2721                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2722                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2723
2724                 dev_change_proto_down_reason(dev, mask, value);
2725         }
2726
2727         if (nl_proto_down) {
2728                 proto_down = nla_get_u8(nl_proto_down);
2729
2730                 /* Don't turn off protodown if there are active reasons */
2731                 if (!proto_down && dev->proto_down_reason) {
2732                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2733                         return -EBUSY;
2734                 }
2735                 err = dev_change_proto_down(dev,
2736                                             proto_down);
2737                 if (err)
2738                         return err;
2739         }
2740
2741         return 0;
2742 }
2743
2744 #define DO_SETLINK_MODIFIED     0x01
2745 /* notify flag means notify + modified. */
2746 #define DO_SETLINK_NOTIFY       0x03
2747 static int do_setlink(const struct sk_buff *skb,
2748                       struct net_device *dev, struct ifinfomsg *ifm,
2749                       struct netlink_ext_ack *extack,
2750                       struct nlattr **tb, int status)
2751 {
2752         const struct net_device_ops *ops = dev->netdev_ops;
2753         char ifname[IFNAMSIZ];
2754         int err;
2755
2756         if (tb[IFLA_IFNAME])
2757                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2758         else
2759                 ifname[0] = '\0';
2760
2761         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2762                 const char *pat = ifname[0] ? ifname : NULL;
2763                 struct net *net;
2764                 int new_ifindex;
2765
2766                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2767                                                 tb, CAP_NET_ADMIN);
2768                 if (IS_ERR(net)) {
2769                         err = PTR_ERR(net);
2770                         goto errout;
2771                 }
2772
2773                 if (tb[IFLA_NEW_IFINDEX])
2774                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2775                 else
2776                         new_ifindex = 0;
2777
2778                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2779                 put_net(net);
2780                 if (err)
2781                         goto errout;
2782                 status |= DO_SETLINK_MODIFIED;
2783         }
2784
2785         if (tb[IFLA_MAP]) {
2786                 struct rtnl_link_ifmap *u_map;
2787                 struct ifmap k_map;
2788
2789                 if (!ops->ndo_set_config) {
2790                         err = -EOPNOTSUPP;
2791                         goto errout;
2792                 }
2793
2794                 if (!netif_device_present(dev)) {
2795                         err = -ENODEV;
2796                         goto errout;
2797                 }
2798
2799                 u_map = nla_data(tb[IFLA_MAP]);
2800                 k_map.mem_start = (unsigned long) u_map->mem_start;
2801                 k_map.mem_end = (unsigned long) u_map->mem_end;
2802                 k_map.base_addr = (unsigned short) u_map->base_addr;
2803                 k_map.irq = (unsigned char) u_map->irq;
2804                 k_map.dma = (unsigned char) u_map->dma;
2805                 k_map.port = (unsigned char) u_map->port;
2806
2807                 err = ops->ndo_set_config(dev, &k_map);
2808                 if (err < 0)
2809                         goto errout;
2810
2811                 status |= DO_SETLINK_NOTIFY;
2812         }
2813
2814         if (tb[IFLA_ADDRESS]) {
2815                 struct sockaddr *sa;
2816                 int len;
2817
2818                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2819                                                   sizeof(*sa));
2820                 sa = kmalloc(len, GFP_KERNEL);
2821                 if (!sa) {
2822                         err = -ENOMEM;
2823                         goto errout;
2824                 }
2825                 sa->sa_family = dev->type;
2826                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2827                        dev->addr_len);
2828                 err = dev_set_mac_address_user(dev, sa, extack);
2829                 kfree(sa);
2830                 if (err)
2831                         goto errout;
2832                 status |= DO_SETLINK_MODIFIED;
2833         }
2834
2835         if (tb[IFLA_MTU]) {
2836                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2837                 if (err < 0)
2838                         goto errout;
2839                 status |= DO_SETLINK_MODIFIED;
2840         }
2841
2842         if (tb[IFLA_GROUP]) {
2843                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2844                 status |= DO_SETLINK_NOTIFY;
2845         }
2846
2847         /*
2848          * Interface selected by interface index but interface
2849          * name provided implies that a name change has been
2850          * requested.
2851          */
2852         if (ifm->ifi_index > 0 && ifname[0]) {
2853                 err = dev_change_name(dev, ifname);
2854                 if (err < 0)
2855                         goto errout;
2856                 status |= DO_SETLINK_MODIFIED;
2857         }
2858
2859         if (tb[IFLA_IFALIAS]) {
2860                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2861                                     nla_len(tb[IFLA_IFALIAS]));
2862                 if (err < 0)
2863                         goto errout;
2864                 status |= DO_SETLINK_NOTIFY;
2865         }
2866
2867         if (tb[IFLA_BROADCAST]) {
2868                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2869                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2870         }
2871
2872         if (tb[IFLA_MASTER]) {
2873                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2874                 if (err)
2875                         goto errout;
2876                 status |= DO_SETLINK_MODIFIED;
2877         }
2878
2879         if (ifm->ifi_flags || ifm->ifi_change) {
2880                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2881                                        extack);
2882                 if (err < 0)
2883                         goto errout;
2884         }
2885
2886         if (tb[IFLA_CARRIER]) {
2887                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2888                 if (err)
2889                         goto errout;
2890                 status |= DO_SETLINK_MODIFIED;
2891         }
2892
2893         if (tb[IFLA_TXQLEN]) {
2894                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2895
2896                 err = dev_change_tx_queue_len(dev, value);
2897                 if (err)
2898                         goto errout;
2899                 status |= DO_SETLINK_MODIFIED;
2900         }
2901
2902         if (tb[IFLA_GSO_MAX_SIZE]) {
2903                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2904
2905                 if (dev->gso_max_size ^ max_size) {
2906                         netif_set_gso_max_size(dev, max_size);
2907                         status |= DO_SETLINK_MODIFIED;
2908                 }
2909         }
2910
2911         if (tb[IFLA_GSO_MAX_SEGS]) {
2912                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2913
2914                 if (dev->gso_max_segs ^ max_segs) {
2915                         netif_set_gso_max_segs(dev, max_segs);
2916                         status |= DO_SETLINK_MODIFIED;
2917                 }
2918         }
2919
2920         if (tb[IFLA_GRO_MAX_SIZE]) {
2921                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2922
2923                 if (dev->gro_max_size ^ gro_max_size) {
2924                         netif_set_gro_max_size(dev, gro_max_size);
2925                         status |= DO_SETLINK_MODIFIED;
2926                 }
2927         }
2928
2929         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2930                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2931
2932                 if (dev->gso_ipv4_max_size ^ max_size) {
2933                         netif_set_gso_ipv4_max_size(dev, max_size);
2934                         status |= DO_SETLINK_MODIFIED;
2935                 }
2936         }
2937
2938         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2939                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2940
2941                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2942                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2943                         status |= DO_SETLINK_MODIFIED;
2944                 }
2945         }
2946
2947         if (tb[IFLA_OPERSTATE])
2948                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2949
2950         if (tb[IFLA_LINKMODE]) {
2951                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2952
2953                 write_lock(&dev_base_lock);
2954                 if (dev->link_mode ^ value)
2955                         status |= DO_SETLINK_NOTIFY;
2956                 dev->link_mode = value;
2957                 write_unlock(&dev_base_lock);
2958         }
2959
2960         if (tb[IFLA_VFINFO_LIST]) {
2961                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2962                 struct nlattr *attr;
2963                 int rem;
2964
2965                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2966                         if (nla_type(attr) != IFLA_VF_INFO ||
2967                             nla_len(attr) < NLA_HDRLEN) {
2968                                 err = -EINVAL;
2969                                 goto errout;
2970                         }
2971                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2972                                                           attr,
2973                                                           ifla_vf_policy,
2974                                                           NULL);
2975                         if (err < 0)
2976                                 goto errout;
2977                         err = do_setvfinfo(dev, vfinfo);
2978                         if (err < 0)
2979                                 goto errout;
2980                         status |= DO_SETLINK_NOTIFY;
2981                 }
2982         }
2983         err = 0;
2984
2985         if (tb[IFLA_VF_PORTS]) {
2986                 struct nlattr *port[IFLA_PORT_MAX+1];
2987                 struct nlattr *attr;
2988                 int vf;
2989                 int rem;
2990
2991                 err = -EOPNOTSUPP;
2992                 if (!ops->ndo_set_vf_port)
2993                         goto errout;
2994
2995                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2996                         if (nla_type(attr) != IFLA_VF_PORT ||
2997                             nla_len(attr) < NLA_HDRLEN) {
2998                                 err = -EINVAL;
2999                                 goto errout;
3000                         }
3001                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3002                                                           attr,
3003                                                           ifla_port_policy,
3004                                                           NULL);
3005                         if (err < 0)
3006                                 goto errout;
3007                         if (!port[IFLA_PORT_VF]) {
3008                                 err = -EOPNOTSUPP;
3009                                 goto errout;
3010                         }
3011                         vf = nla_get_u32(port[IFLA_PORT_VF]);
3012                         err = ops->ndo_set_vf_port(dev, vf, port);
3013                         if (err < 0)
3014                                 goto errout;
3015                         status |= DO_SETLINK_NOTIFY;
3016                 }
3017         }
3018         err = 0;
3019
3020         if (tb[IFLA_PORT_SELF]) {
3021                 struct nlattr *port[IFLA_PORT_MAX+1];
3022
3023                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3024                                                   tb[IFLA_PORT_SELF],
3025                                                   ifla_port_policy, NULL);
3026                 if (err < 0)
3027                         goto errout;
3028
3029                 err = -EOPNOTSUPP;
3030                 if (ops->ndo_set_vf_port)
3031                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3032                 if (err < 0)
3033                         goto errout;
3034                 status |= DO_SETLINK_NOTIFY;
3035         }
3036
3037         if (tb[IFLA_AF_SPEC]) {
3038                 struct nlattr *af;
3039                 int rem;
3040
3041                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3042                         const struct rtnl_af_ops *af_ops;
3043
3044                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3045
3046                         err = af_ops->set_link_af(dev, af, extack);
3047                         if (err < 0)
3048                                 goto errout;
3049
3050                         status |= DO_SETLINK_NOTIFY;
3051                 }
3052         }
3053         err = 0;
3054
3055         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3056                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3057                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3058                 if (err)
3059                         goto errout;
3060                 status |= DO_SETLINK_NOTIFY;
3061         }
3062
3063         if (tb[IFLA_XDP]) {
3064                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3065                 u32 xdp_flags = 0;
3066
3067                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3068                                                   tb[IFLA_XDP],
3069                                                   ifla_xdp_policy, NULL);
3070                 if (err < 0)
3071                         goto errout;
3072
3073                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3074                         err = -EINVAL;
3075                         goto errout;
3076                 }
3077
3078                 if (xdp[IFLA_XDP_FLAGS]) {
3079                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3080                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3081                                 err = -EINVAL;
3082                                 goto errout;
3083                         }
3084                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3085                                 err = -EINVAL;
3086                                 goto errout;
3087                         }
3088                 }
3089
3090                 if (xdp[IFLA_XDP_FD]) {
3091                         int expected_fd = -1;
3092
3093                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3094                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3095                                         err = -EINVAL;
3096                                         goto errout;
3097                                 }
3098                                 expected_fd =
3099                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3100                         }
3101
3102                         err = dev_change_xdp_fd(dev, extack,
3103                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3104                                                 expected_fd,
3105                                                 xdp_flags);
3106                         if (err)
3107                                 goto errout;
3108                         status |= DO_SETLINK_NOTIFY;
3109                 }
3110         }
3111
3112 errout:
3113         if (status & DO_SETLINK_MODIFIED) {
3114                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3115                         netdev_state_change(dev);
3116
3117                 if (err < 0)
3118                         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",
3119                                              dev->name);
3120         }
3121
3122         return err;
3123 }
3124
3125 static struct net_device *rtnl_dev_get(struct net *net,
3126                                        struct nlattr *tb[])
3127 {
3128         char ifname[ALTIFNAMSIZ];
3129
3130         if (tb[IFLA_IFNAME])
3131                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3132         else if (tb[IFLA_ALT_IFNAME])
3133                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3134         else
3135                 return NULL;
3136
3137         return __dev_get_by_name(net, ifname);
3138 }
3139
3140 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3141                         struct netlink_ext_ack *extack)
3142 {
3143         struct net *net = sock_net(skb->sk);
3144         struct ifinfomsg *ifm;
3145         struct net_device *dev;
3146         int err;
3147         struct nlattr *tb[IFLA_MAX+1];
3148
3149         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3150                                      ifla_policy, extack);
3151         if (err < 0)
3152                 goto errout;
3153
3154         err = rtnl_ensure_unique_netns(tb, extack, false);
3155         if (err < 0)
3156                 goto errout;
3157
3158         err = -EINVAL;
3159         ifm = nlmsg_data(nlh);
3160         if (ifm->ifi_index > 0)
3161                 dev = __dev_get_by_index(net, ifm->ifi_index);
3162         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3163                 dev = rtnl_dev_get(net, tb);
3164         else
3165                 goto errout;
3166
3167         if (dev == NULL) {
3168                 err = -ENODEV;
3169                 goto errout;
3170         }
3171
3172         err = validate_linkmsg(dev, tb, extack);
3173         if (err < 0)
3174                 goto errout;
3175
3176         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3177 errout:
3178         return err;
3179 }
3180
3181 static int rtnl_group_dellink(const struct net *net, int group)
3182 {
3183         struct net_device *dev, *aux;
3184         LIST_HEAD(list_kill);
3185         bool found = false;
3186
3187         if (!group)
3188                 return -EPERM;
3189
3190         for_each_netdev(net, dev) {
3191                 if (dev->group == group) {
3192                         const struct rtnl_link_ops *ops;
3193
3194                         found = true;
3195                         ops = dev->rtnl_link_ops;
3196                         if (!ops || !ops->dellink)
3197                                 return -EOPNOTSUPP;
3198                 }
3199         }
3200
3201         if (!found)
3202                 return -ENODEV;
3203
3204         for_each_netdev_safe(net, dev, aux) {
3205                 if (dev->group == group) {
3206                         const struct rtnl_link_ops *ops;
3207
3208                         ops = dev->rtnl_link_ops;
3209                         ops->dellink(dev, &list_kill);
3210                 }
3211         }
3212         unregister_netdevice_many(&list_kill);
3213
3214         return 0;
3215 }
3216
3217 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3218 {
3219         const struct rtnl_link_ops *ops;
3220         LIST_HEAD(list_kill);
3221
3222         ops = dev->rtnl_link_ops;
3223         if (!ops || !ops->dellink)
3224                 return -EOPNOTSUPP;
3225
3226         ops->dellink(dev, &list_kill);
3227         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3228
3229         return 0;
3230 }
3231 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3232
3233 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3234                         struct netlink_ext_ack *extack)
3235 {
3236         struct net *net = sock_net(skb->sk);
3237         u32 portid = NETLINK_CB(skb).portid;
3238         struct net *tgt_net = net;
3239         struct net_device *dev = NULL;
3240         struct ifinfomsg *ifm;
3241         struct nlattr *tb[IFLA_MAX+1];
3242         int err;
3243         int netnsid = -1;
3244
3245         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3246                                      ifla_policy, extack);
3247         if (err < 0)
3248                 return err;
3249
3250         err = rtnl_ensure_unique_netns(tb, extack, true);
3251         if (err < 0)
3252                 return err;
3253
3254         if (tb[IFLA_TARGET_NETNSID]) {
3255                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3256                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3257                 if (IS_ERR(tgt_net))
3258                         return PTR_ERR(tgt_net);
3259         }
3260
3261         err = -EINVAL;
3262         ifm = nlmsg_data(nlh);
3263         if (ifm->ifi_index > 0)
3264                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3265         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3266                 dev = rtnl_dev_get(net, tb);
3267         else if (tb[IFLA_GROUP])
3268                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3269         else
3270                 goto out;
3271
3272         if (!dev) {
3273                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3274                         err = -ENODEV;
3275
3276                 goto out;
3277         }
3278
3279         err = rtnl_delete_link(dev, portid, nlh);
3280
3281 out:
3282         if (netnsid >= 0)
3283                 put_net(tgt_net);
3284
3285         return err;
3286 }
3287
3288 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3289                         u32 portid, const struct nlmsghdr *nlh)
3290 {
3291         unsigned int old_flags;
3292         int err;
3293
3294         old_flags = dev->flags;
3295         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3296                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3297                                          NULL);
3298                 if (err < 0)
3299                         return err;
3300         }
3301
3302         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3303                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3304         } else {
3305                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3306                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3307         }
3308         return 0;
3309 }
3310 EXPORT_SYMBOL(rtnl_configure_link);
3311
3312 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3313                                     unsigned char name_assign_type,
3314                                     const struct rtnl_link_ops *ops,
3315                                     struct nlattr *tb[],
3316                                     struct netlink_ext_ack *extack)
3317 {
3318         struct net_device *dev;
3319         unsigned int num_tx_queues = 1;
3320         unsigned int num_rx_queues = 1;
3321         int err;
3322
3323         if (tb[IFLA_NUM_TX_QUEUES])
3324                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3325         else if (ops->get_num_tx_queues)
3326                 num_tx_queues = ops->get_num_tx_queues();
3327
3328         if (tb[IFLA_NUM_RX_QUEUES])
3329                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3330         else if (ops->get_num_rx_queues)
3331                 num_rx_queues = ops->get_num_rx_queues();
3332
3333         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3334                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3335                 return ERR_PTR(-EINVAL);
3336         }
3337
3338         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3339                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3340                 return ERR_PTR(-EINVAL);
3341         }
3342
3343         if (ops->alloc) {
3344                 dev = ops->alloc(tb, ifname, name_assign_type,
3345                                  num_tx_queues, num_rx_queues);
3346                 if (IS_ERR(dev))
3347                         return dev;
3348         } else {
3349                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3350                                        name_assign_type, ops->setup,
3351                                        num_tx_queues, num_rx_queues);
3352         }
3353
3354         if (!dev)
3355                 return ERR_PTR(-ENOMEM);
3356
3357         err = validate_linkmsg(dev, tb, extack);
3358         if (err < 0) {
3359                 free_netdev(dev);
3360                 return ERR_PTR(err);
3361         }
3362
3363         dev_net_set(dev, net);
3364         dev->rtnl_link_ops = ops;
3365         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3366
3367         if (tb[IFLA_MTU]) {
3368                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3369
3370                 err = dev_validate_mtu(dev, mtu, extack);
3371                 if (err) {
3372                         free_netdev(dev);
3373                         return ERR_PTR(err);
3374                 }
3375                 dev->mtu = mtu;
3376         }
3377         if (tb[IFLA_ADDRESS]) {
3378                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3379                                nla_len(tb[IFLA_ADDRESS]));
3380                 dev->addr_assign_type = NET_ADDR_SET;
3381         }
3382         if (tb[IFLA_BROADCAST])
3383                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3384                                 nla_len(tb[IFLA_BROADCAST]));
3385         if (tb[IFLA_TXQLEN])
3386                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3387         if (tb[IFLA_OPERSTATE])
3388                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3389         if (tb[IFLA_LINKMODE])
3390                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3391         if (tb[IFLA_GROUP])
3392                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3393         if (tb[IFLA_GSO_MAX_SIZE])
3394                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3395         if (tb[IFLA_GSO_MAX_SEGS])
3396                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3397         if (tb[IFLA_GRO_MAX_SIZE])
3398                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3399         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3400                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3401         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3402                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3403
3404         return dev;
3405 }
3406 EXPORT_SYMBOL(rtnl_create_link);
3407
3408 static int rtnl_group_changelink(const struct sk_buff *skb,
3409                 struct net *net, int group,
3410                 struct ifinfomsg *ifm,
3411                 struct netlink_ext_ack *extack,
3412                 struct nlattr **tb)
3413 {
3414         struct net_device *dev, *aux;
3415         int err;
3416
3417         for_each_netdev_safe(net, dev, aux) {
3418                 if (dev->group == group) {
3419                         err = validate_linkmsg(dev, tb, extack);
3420                         if (err < 0)
3421                                 return err;
3422                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3423                         if (err < 0)
3424                                 return err;
3425                 }
3426         }
3427
3428         return 0;
3429 }
3430
3431 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3432                                const struct rtnl_link_ops *ops,
3433                                const struct nlmsghdr *nlh,
3434                                struct nlattr **tb, struct nlattr **data,
3435                                struct netlink_ext_ack *extack)
3436 {
3437         unsigned char name_assign_type = NET_NAME_USER;
3438         struct net *net = sock_net(skb->sk);
3439         u32 portid = NETLINK_CB(skb).portid;
3440         struct net *dest_net, *link_net;
3441         struct net_device *dev;
3442         char ifname[IFNAMSIZ];
3443         int err;
3444
3445         if (!ops->alloc && !ops->setup)
3446                 return -EOPNOTSUPP;
3447
3448         if (tb[IFLA_IFNAME]) {
3449                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3450         } else {
3451                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3452                 name_assign_type = NET_NAME_ENUM;
3453         }
3454
3455         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3456         if (IS_ERR(dest_net))
3457                 return PTR_ERR(dest_net);
3458
3459         if (tb[IFLA_LINK_NETNSID]) {
3460                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3461
3462                 link_net = get_net_ns_by_id(dest_net, id);
3463                 if (!link_net) {
3464                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3465                         err =  -EINVAL;
3466                         goto out;
3467                 }
3468                 err = -EPERM;
3469                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3470                         goto out;
3471         } else {
3472                 link_net = NULL;
3473         }
3474
3475         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3476                                name_assign_type, ops, tb, extack);
3477         if (IS_ERR(dev)) {
3478                 err = PTR_ERR(dev);
3479                 goto out;
3480         }
3481
3482         dev->ifindex = ifm->ifi_index;
3483
3484         if (ops->newlink)
3485                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3486         else
3487                 err = register_netdevice(dev);
3488         if (err < 0) {
3489                 free_netdev(dev);
3490                 goto out;
3491         }
3492
3493         err = rtnl_configure_link(dev, ifm, portid, nlh);
3494         if (err < 0)
3495                 goto out_unregister;
3496         if (link_net) {
3497                 err = dev_change_net_namespace(dev, dest_net, ifname);
3498                 if (err < 0)
3499                         goto out_unregister;
3500         }
3501         if (tb[IFLA_MASTER]) {
3502                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3503                 if (err)
3504                         goto out_unregister;
3505         }
3506 out:
3507         if (link_net)
3508                 put_net(link_net);
3509         put_net(dest_net);
3510         return err;
3511 out_unregister:
3512         if (ops->newlink) {
3513                 LIST_HEAD(list_kill);
3514
3515                 ops->dellink(dev, &list_kill);
3516                 unregister_netdevice_many(&list_kill);
3517         } else {
3518                 unregister_netdevice(dev);
3519         }
3520         goto out;
3521 }
3522
3523 struct rtnl_newlink_tbs {
3524         struct nlattr *tb[IFLA_MAX + 1];
3525         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3526         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3527 };
3528
3529 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3530                           struct rtnl_newlink_tbs *tbs,
3531                           struct netlink_ext_ack *extack)
3532 {
3533         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3534         struct nlattr ** const tb = tbs->tb;
3535         const struct rtnl_link_ops *m_ops;
3536         struct net_device *master_dev;
3537         struct net *net = sock_net(skb->sk);
3538         const struct rtnl_link_ops *ops;
3539         struct nlattr **slave_data;
3540         char kind[MODULE_NAME_LEN];
3541         struct net_device *dev;
3542         struct ifinfomsg *ifm;
3543         struct nlattr **data;
3544         bool link_specified;
3545         int err;
3546
3547 #ifdef CONFIG_MODULES
3548 replay:
3549 #endif
3550         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3551                                      ifla_policy, extack);
3552         if (err < 0)
3553                 return err;
3554
3555         err = rtnl_ensure_unique_netns(tb, extack, false);
3556         if (err < 0)
3557                 return err;
3558
3559         ifm = nlmsg_data(nlh);
3560         if (ifm->ifi_index > 0) {
3561                 link_specified = true;
3562                 dev = __dev_get_by_index(net, ifm->ifi_index);
3563         } else if (ifm->ifi_index < 0) {
3564                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3565                 return -EINVAL;
3566         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3567                 link_specified = true;
3568                 dev = rtnl_dev_get(net, tb);
3569         } else {
3570                 link_specified = false;
3571                 dev = NULL;
3572         }
3573
3574         master_dev = NULL;
3575         m_ops = NULL;
3576         if (dev) {
3577                 master_dev = netdev_master_upper_dev_get(dev);
3578                 if (master_dev)
3579                         m_ops = master_dev->rtnl_link_ops;
3580         }
3581
3582         if (tb[IFLA_LINKINFO]) {
3583                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3584                                                   tb[IFLA_LINKINFO],
3585                                                   ifla_info_policy, NULL);
3586                 if (err < 0)
3587                         return err;
3588         } else
3589                 memset(linkinfo, 0, sizeof(linkinfo));
3590
3591         if (linkinfo[IFLA_INFO_KIND]) {
3592                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3593                 ops = rtnl_link_ops_get(kind);
3594         } else {
3595                 kind[0] = '\0';
3596                 ops = NULL;
3597         }
3598
3599         data = NULL;
3600         if (ops) {
3601                 if (ops->maxtype > RTNL_MAX_TYPE)
3602                         return -EINVAL;
3603
3604                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3605                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3606                                                           linkinfo[IFLA_INFO_DATA],
3607                                                           ops->policy, extack);
3608                         if (err < 0)
3609                                 return err;
3610                         data = tbs->attr;
3611                 }
3612                 if (ops->validate) {
3613                         err = ops->validate(tb, data, extack);
3614                         if (err < 0)
3615                                 return err;
3616                 }
3617         }
3618
3619         slave_data = NULL;
3620         if (m_ops) {
3621                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3622                         return -EINVAL;
3623
3624                 if (m_ops->slave_maxtype &&
3625                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3626                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3627                                                           m_ops->slave_maxtype,
3628                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3629                                                           m_ops->slave_policy,
3630                                                           extack);
3631                         if (err < 0)
3632                                 return err;
3633                         slave_data = tbs->slave_attr;
3634                 }
3635         }
3636
3637         if (dev) {
3638                 int status = 0;
3639
3640                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3641                         return -EEXIST;
3642                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3643                         return -EOPNOTSUPP;
3644
3645                 err = validate_linkmsg(dev, tb, extack);
3646                 if (err < 0)
3647                         return err;
3648
3649                 if (linkinfo[IFLA_INFO_DATA]) {
3650                         if (!ops || ops != dev->rtnl_link_ops ||
3651                             !ops->changelink)
3652                                 return -EOPNOTSUPP;
3653
3654                         err = ops->changelink(dev, tb, data, extack);
3655                         if (err < 0)
3656                                 return err;
3657                         status |= DO_SETLINK_NOTIFY;
3658                 }
3659
3660                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3661                         if (!m_ops || !m_ops->slave_changelink)
3662                                 return -EOPNOTSUPP;
3663
3664                         err = m_ops->slave_changelink(master_dev, dev, tb,
3665                                                       slave_data, extack);
3666                         if (err < 0)
3667                                 return err;
3668                         status |= DO_SETLINK_NOTIFY;
3669                 }
3670
3671                 return do_setlink(skb, dev, ifm, extack, tb, status);
3672         }
3673
3674         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3675                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3676                  * or it's for a group
3677                 */
3678                 if (link_specified)
3679                         return -ENODEV;
3680                 if (tb[IFLA_GROUP])
3681                         return rtnl_group_changelink(skb, net,
3682                                                 nla_get_u32(tb[IFLA_GROUP]),
3683                                                 ifm, extack, tb);
3684                 return -ENODEV;
3685         }
3686
3687         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3688                 return -EOPNOTSUPP;
3689
3690         if (!ops) {
3691 #ifdef CONFIG_MODULES
3692                 if (kind[0]) {
3693                         __rtnl_unlock();
3694                         request_module("rtnl-link-%s", kind);
3695                         rtnl_lock();
3696                         ops = rtnl_link_ops_get(kind);
3697                         if (ops)
3698                                 goto replay;
3699                 }
3700 #endif
3701                 NL_SET_ERR_MSG(extack, "Unknown device type");
3702                 return -EOPNOTSUPP;
3703         }
3704
3705         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3706 }
3707
3708 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3709                         struct netlink_ext_ack *extack)
3710 {
3711         struct rtnl_newlink_tbs *tbs;
3712         int ret;
3713
3714         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3715         if (!tbs)
3716                 return -ENOMEM;
3717
3718         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3719         kfree(tbs);
3720         return ret;
3721 }
3722
3723 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3724                                   const struct nlmsghdr *nlh,
3725                                   struct nlattr **tb,
3726                                   struct netlink_ext_ack *extack)
3727 {
3728         struct ifinfomsg *ifm;
3729         int i, err;
3730
3731         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3732                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3733                 return -EINVAL;
3734         }
3735
3736         if (!netlink_strict_get_check(skb))
3737                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3738                                               ifla_policy, extack);
3739
3740         ifm = nlmsg_data(nlh);
3741         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3742             ifm->ifi_change) {
3743                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3744                 return -EINVAL;
3745         }
3746
3747         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3748                                             ifla_policy, extack);
3749         if (err)
3750                 return err;
3751
3752         for (i = 0; i <= IFLA_MAX; i++) {
3753                 if (!tb[i])
3754                         continue;
3755
3756                 switch (i) {
3757                 case IFLA_IFNAME:
3758                 case IFLA_ALT_IFNAME:
3759                 case IFLA_EXT_MASK:
3760                 case IFLA_TARGET_NETNSID:
3761                         break;
3762                 default:
3763                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3764                         return -EINVAL;
3765                 }
3766         }
3767
3768         return 0;
3769 }
3770
3771 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3772                         struct netlink_ext_ack *extack)
3773 {
3774         struct net *net = sock_net(skb->sk);
3775         struct net *tgt_net = net;
3776         struct ifinfomsg *ifm;
3777         struct nlattr *tb[IFLA_MAX+1];
3778         struct net_device *dev = NULL;
3779         struct sk_buff *nskb;
3780         int netnsid = -1;
3781         int err;
3782         u32 ext_filter_mask = 0;
3783
3784         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3785         if (err < 0)
3786                 return err;
3787
3788         err = rtnl_ensure_unique_netns(tb, extack, true);
3789         if (err < 0)
3790                 return err;
3791
3792         if (tb[IFLA_TARGET_NETNSID]) {
3793                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3794                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3795                 if (IS_ERR(tgt_net))
3796                         return PTR_ERR(tgt_net);
3797         }
3798
3799         if (tb[IFLA_EXT_MASK])
3800                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3801
3802         err = -EINVAL;
3803         ifm = nlmsg_data(nlh);
3804         if (ifm->ifi_index > 0)
3805                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3806         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3807                 dev = rtnl_dev_get(tgt_net, tb);
3808         else
3809                 goto out;
3810
3811         err = -ENODEV;
3812         if (dev == NULL)
3813                 goto out;
3814
3815         err = -ENOBUFS;
3816         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3817         if (nskb == NULL)
3818                 goto out;
3819
3820         err = rtnl_fill_ifinfo(nskb, dev, net,
3821                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3822                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3823                                0, NULL, 0, netnsid, GFP_KERNEL);
3824         if (err < 0) {
3825                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3826                 WARN_ON(err == -EMSGSIZE);
3827                 kfree_skb(nskb);
3828         } else
3829                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3830 out:
3831         if (netnsid >= 0)
3832                 put_net(tgt_net);
3833
3834         return err;
3835 }
3836
3837 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3838                            bool *changed, struct netlink_ext_ack *extack)
3839 {
3840         char *alt_ifname;
3841         size_t size;
3842         int err;
3843
3844         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3845         if (err)
3846                 return err;
3847
3848         if (cmd == RTM_NEWLINKPROP) {
3849                 size = rtnl_prop_list_size(dev);
3850                 size += nla_total_size(ALTIFNAMSIZ);
3851                 if (size >= U16_MAX) {
3852                         NL_SET_ERR_MSG(extack,
3853                                        "effective property list too long");
3854                         return -EINVAL;
3855                 }
3856         }
3857
3858         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3859         if (!alt_ifname)
3860                 return -ENOMEM;
3861
3862         if (cmd == RTM_NEWLINKPROP) {
3863                 err = netdev_name_node_alt_create(dev, alt_ifname);
3864                 if (!err)
3865                         alt_ifname = NULL;
3866         } else if (cmd == RTM_DELLINKPROP) {
3867                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3868         } else {
3869                 WARN_ON_ONCE(1);
3870                 err = -EINVAL;
3871         }
3872
3873         kfree(alt_ifname);
3874         if (!err)
3875                 *changed = true;
3876         return err;
3877 }
3878
3879 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3880                          struct netlink_ext_ack *extack)
3881 {
3882         struct net *net = sock_net(skb->sk);
3883         struct nlattr *tb[IFLA_MAX + 1];
3884         struct net_device *dev;
3885         struct ifinfomsg *ifm;
3886         bool changed = false;
3887         struct nlattr *attr;
3888         int err, rem;
3889
3890         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3891         if (err)
3892                 return err;
3893
3894         err = rtnl_ensure_unique_netns(tb, extack, true);
3895         if (err)
3896                 return err;
3897
3898         ifm = nlmsg_data(nlh);
3899         if (ifm->ifi_index > 0)
3900                 dev = __dev_get_by_index(net, ifm->ifi_index);
3901         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3902                 dev = rtnl_dev_get(net, tb);
3903         else
3904                 return -EINVAL;
3905
3906         if (!dev)
3907                 return -ENODEV;
3908
3909         if (!tb[IFLA_PROP_LIST])
3910                 return 0;
3911
3912         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3913                 switch (nla_type(attr)) {
3914                 case IFLA_ALT_IFNAME:
3915                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3916                         if (err)
3917                                 return err;
3918                         break;
3919                 }
3920         }
3921
3922         if (changed)
3923                 netdev_state_change(dev);
3924         return 0;
3925 }
3926
3927 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3928                             struct netlink_ext_ack *extack)
3929 {
3930         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3931 }
3932
3933 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3934                             struct netlink_ext_ack *extack)
3935 {
3936         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3937 }
3938
3939 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3940 {
3941         struct net *net = sock_net(skb->sk);
3942         size_t min_ifinfo_dump_size = 0;
3943         struct nlattr *tb[IFLA_MAX+1];
3944         u32 ext_filter_mask = 0;
3945         struct net_device *dev;
3946         int hdrlen;
3947
3948         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3949         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3950                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3951
3952         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3953                 if (tb[IFLA_EXT_MASK])
3954                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3955         }
3956
3957         if (!ext_filter_mask)
3958                 return NLMSG_GOODSIZE;
3959         /*
3960          * traverse the list of net devices and compute the minimum
3961          * buffer size based upon the filter mask.
3962          */
3963         rcu_read_lock();
3964         for_each_netdev_rcu(net, dev) {
3965                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3966                                            if_nlmsg_size(dev, ext_filter_mask));
3967         }
3968         rcu_read_unlock();
3969
3970         return nlmsg_total_size(min_ifinfo_dump_size);
3971 }
3972
3973 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3974 {
3975         int idx;
3976         int s_idx = cb->family;
3977         int type = cb->nlh->nlmsg_type - RTM_BASE;
3978         int ret = 0;
3979
3980         if (s_idx == 0)
3981                 s_idx = 1;
3982
3983         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3984                 struct rtnl_link __rcu **tab;
3985                 struct rtnl_link *link;
3986                 rtnl_dumpit_func dumpit;
3987
3988                 if (idx < s_idx || idx == PF_PACKET)
3989                         continue;
3990
3991                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3992                         continue;
3993
3994                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3995                 if (!tab)
3996                         continue;
3997
3998                 link = rcu_dereference_rtnl(tab[type]);
3999                 if (!link)
4000                         continue;
4001
4002                 dumpit = link->dumpit;
4003                 if (!dumpit)
4004                         continue;
4005
4006                 if (idx > s_idx) {
4007                         memset(&cb->args[0], 0, sizeof(cb->args));
4008                         cb->prev_seq = 0;
4009                         cb->seq = 0;
4010                 }
4011                 ret = dumpit(skb, cb);
4012                 if (ret)
4013                         break;
4014         }
4015         cb->family = idx;
4016
4017         return skb->len ? : ret;
4018 }
4019
4020 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4021                                        unsigned int change,
4022                                        u32 event, gfp_t flags, int *new_nsid,
4023                                        int new_ifindex, u32 portid,
4024                                        const struct nlmsghdr *nlh)
4025 {
4026         struct net *net = dev_net(dev);
4027         struct sk_buff *skb;
4028         int err = -ENOBUFS;
4029         u32 seq = 0;
4030
4031         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4032         if (skb == NULL)
4033                 goto errout;
4034
4035         if (nlmsg_report(nlh))
4036                 seq = nlmsg_seq(nlh);
4037         else
4038                 portid = 0;
4039
4040         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4041                                type, portid, seq, change, 0, 0, event,
4042                                new_nsid, new_ifindex, -1, flags);
4043         if (err < 0) {
4044                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4045                 WARN_ON(err == -EMSGSIZE);
4046                 kfree_skb(skb);
4047                 goto errout;
4048         }
4049         return skb;
4050 errout:
4051         if (err < 0)
4052                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4053         return NULL;
4054 }
4055
4056 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4057                        u32 portid, const struct nlmsghdr *nlh)
4058 {
4059         struct net *net = dev_net(dev);
4060
4061         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4062 }
4063
4064 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4065                                unsigned int change, u32 event,
4066                                gfp_t flags, int *new_nsid, int new_ifindex,
4067                                u32 portid, const struct nlmsghdr *nlh)
4068 {
4069         struct sk_buff *skb;
4070
4071         if (dev->reg_state != NETREG_REGISTERED)
4072                 return;
4073
4074         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4075                                      new_ifindex, portid, nlh);
4076         if (skb)
4077                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4078 }
4079
4080 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4081                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4082 {
4083         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4084                            NULL, 0, portid, nlh);
4085 }
4086
4087 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4088                          gfp_t flags, int *new_nsid, int new_ifindex)
4089 {
4090         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4091                            new_nsid, new_ifindex, 0, NULL);
4092 }
4093
4094 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4095                                    struct net_device *dev,
4096                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4097                                    int type, unsigned int flags,
4098                                    int nlflags, u16 ndm_state)
4099 {
4100         struct nlmsghdr *nlh;
4101         struct ndmsg *ndm;
4102
4103         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4104         if (!nlh)
4105                 return -EMSGSIZE;
4106
4107         ndm = nlmsg_data(nlh);
4108         ndm->ndm_family  = AF_BRIDGE;
4109         ndm->ndm_pad1    = 0;
4110         ndm->ndm_pad2    = 0;
4111         ndm->ndm_flags   = flags;
4112         ndm->ndm_type    = 0;
4113         ndm->ndm_ifindex = dev->ifindex;
4114         ndm->ndm_state   = ndm_state;
4115
4116         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4117                 goto nla_put_failure;
4118         if (vid)
4119                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4120                         goto nla_put_failure;
4121
4122         nlmsg_end(skb, nlh);
4123         return 0;
4124
4125 nla_put_failure:
4126         nlmsg_cancel(skb, nlh);
4127         return -EMSGSIZE;
4128 }
4129
4130 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4131 {
4132         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4133                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
4134                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4135                0;
4136 }
4137
4138 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4139                             u16 ndm_state)
4140 {
4141         struct net *net = dev_net(dev);
4142         struct sk_buff *skb;
4143         int err = -ENOBUFS;
4144
4145         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4146         if (!skb)
4147                 goto errout;
4148
4149         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4150                                       0, 0, type, NTF_SELF, 0, ndm_state);
4151         if (err < 0) {
4152                 kfree_skb(skb);
4153                 goto errout;
4154         }
4155
4156         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4157         return;
4158 errout:
4159         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4160 }
4161
4162 /*
4163  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4164  */
4165 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4166                      struct nlattr *tb[],
4167                      struct net_device *dev,
4168                      const unsigned char *addr, u16 vid,
4169                      u16 flags)
4170 {
4171         int err = -EINVAL;
4172
4173         /* If aging addresses are supported device will need to
4174          * implement its own handler for this.
4175          */
4176         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4177                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4178                 return err;
4179         }
4180
4181         if (tb[NDA_FLAGS_EXT]) {
4182                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4183                 return err;
4184         }
4185
4186         if (vid) {
4187                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4188                 return err;
4189         }
4190
4191         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4192                 err = dev_uc_add_excl(dev, addr);
4193         else if (is_multicast_ether_addr(addr))
4194                 err = dev_mc_add_excl(dev, addr);
4195
4196         /* Only return duplicate errors if NLM_F_EXCL is set */
4197         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4198                 err = 0;
4199
4200         return err;
4201 }
4202 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4203
4204 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4205                          struct netlink_ext_ack *extack)
4206 {
4207         u16 vid = 0;
4208
4209         if (vlan_attr) {
4210                 if (nla_len(vlan_attr) != sizeof(u16)) {
4211                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4212                         return -EINVAL;
4213                 }
4214
4215                 vid = nla_get_u16(vlan_attr);
4216
4217                 if (!vid || vid >= VLAN_VID_MASK) {
4218                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4219                         return -EINVAL;
4220                 }
4221         }
4222         *p_vid = vid;
4223         return 0;
4224 }
4225
4226 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4227                         struct netlink_ext_ack *extack)
4228 {
4229         struct net *net = sock_net(skb->sk);
4230         struct ndmsg *ndm;
4231         struct nlattr *tb[NDA_MAX+1];
4232         struct net_device *dev;
4233         u8 *addr;
4234         u16 vid;
4235         int err;
4236
4237         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4238                                      extack);
4239         if (err < 0)
4240                 return err;
4241
4242         ndm = nlmsg_data(nlh);
4243         if (ndm->ndm_ifindex == 0) {
4244                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4245                 return -EINVAL;
4246         }
4247
4248         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4249         if (dev == NULL) {
4250                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4251                 return -ENODEV;
4252         }
4253
4254         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4255                 NL_SET_ERR_MSG(extack, "invalid address");
4256                 return -EINVAL;
4257         }
4258
4259         if (dev->type != ARPHRD_ETHER) {
4260                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4261                 return -EINVAL;
4262         }
4263
4264         addr = nla_data(tb[NDA_LLADDR]);
4265
4266         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4267         if (err)
4268                 return err;
4269
4270         err = -EOPNOTSUPP;
4271
4272         /* Support fdb on master device the net/bridge default case */
4273         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4274             netif_is_bridge_port(dev)) {
4275                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4276                 const struct net_device_ops *ops = br_dev->netdev_ops;
4277
4278                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4279                                        nlh->nlmsg_flags, extack);
4280                 if (err)
4281                         goto out;
4282                 else
4283                         ndm->ndm_flags &= ~NTF_MASTER;
4284         }
4285
4286         /* Embedded bridge, macvlan, and any other device support */
4287         if ((ndm->ndm_flags & NTF_SELF)) {
4288                 if (dev->netdev_ops->ndo_fdb_add)
4289                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4290                                                            vid,
4291                                                            nlh->nlmsg_flags,
4292                                                            extack);
4293                 else
4294                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4295                                                nlh->nlmsg_flags);
4296
4297                 if (!err) {
4298                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4299                                         ndm->ndm_state);
4300                         ndm->ndm_flags &= ~NTF_SELF;
4301                 }
4302         }
4303 out:
4304         return err;
4305 }
4306
4307 /*
4308  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4309  */
4310 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4311                      struct nlattr *tb[],
4312                      struct net_device *dev,
4313                      const unsigned char *addr, u16 vid)
4314 {
4315         int err = -EINVAL;
4316
4317         /* If aging addresses are supported device will need to
4318          * implement its own handler for this.
4319          */
4320         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4321                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4322                 return err;
4323         }
4324
4325         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4326                 err = dev_uc_del(dev, addr);
4327         else if (is_multicast_ether_addr(addr))
4328                 err = dev_mc_del(dev, addr);
4329
4330         return err;
4331 }
4332 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4333
4334 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4335         [NDA_VLAN]      = { .type = NLA_U16 },
4336         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4337         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4338         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4339 };
4340
4341 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4342                         struct netlink_ext_ack *extack)
4343 {
4344         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4345         struct net *net = sock_net(skb->sk);
4346         const struct net_device_ops *ops;
4347         struct ndmsg *ndm;
4348         struct nlattr *tb[NDA_MAX+1];
4349         struct net_device *dev;
4350         __u8 *addr = NULL;
4351         int err;
4352         u16 vid;
4353
4354         if (!netlink_capable(skb, CAP_NET_ADMIN))
4355                 return -EPERM;
4356
4357         if (!del_bulk) {
4358                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4359                                              NULL, extack);
4360         } else {
4361                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4362                                   fdb_del_bulk_policy, extack);
4363         }
4364         if (err < 0)
4365                 return err;
4366
4367         ndm = nlmsg_data(nlh);
4368         if (ndm->ndm_ifindex == 0) {
4369                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4370                 return -EINVAL;
4371         }
4372
4373         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4374         if (dev == NULL) {
4375                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4376                 return -ENODEV;
4377         }
4378
4379         if (!del_bulk) {
4380                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4381                         NL_SET_ERR_MSG(extack, "invalid address");
4382                         return -EINVAL;
4383                 }
4384                 addr = nla_data(tb[NDA_LLADDR]);
4385         }
4386
4387         if (dev->type != ARPHRD_ETHER) {
4388                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4389                 return -EINVAL;
4390         }
4391
4392         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4393         if (err)
4394                 return err;
4395
4396         err = -EOPNOTSUPP;
4397
4398         /* Support fdb on master device the net/bridge default case */
4399         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4400             netif_is_bridge_port(dev)) {
4401                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4402
4403                 ops = br_dev->netdev_ops;
4404                 if (!del_bulk) {
4405                         if (ops->ndo_fdb_del)
4406                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4407                 } else {
4408                         if (ops->ndo_fdb_del_bulk)
4409                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4410                                                             extack);
4411                 }
4412
4413                 if (err)
4414                         goto out;
4415                 else
4416                         ndm->ndm_flags &= ~NTF_MASTER;
4417         }
4418
4419         /* Embedded bridge, macvlan, and any other device support */
4420         if (ndm->ndm_flags & NTF_SELF) {
4421                 ops = dev->netdev_ops;
4422                 if (!del_bulk) {
4423                         if (ops->ndo_fdb_del)
4424                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4425                         else
4426                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4427                 } else {
4428                         /* in case err was cleared by NTF_MASTER call */
4429                         err = -EOPNOTSUPP;
4430                         if (ops->ndo_fdb_del_bulk)
4431                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4432                                                             extack);
4433                 }
4434
4435                 if (!err) {
4436                         if (!del_bulk)
4437                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4438                                                 ndm->ndm_state);
4439                         ndm->ndm_flags &= ~NTF_SELF;
4440                 }
4441         }
4442 out:
4443         return err;
4444 }
4445
4446 static int nlmsg_populate_fdb(struct sk_buff *skb,
4447                               struct netlink_callback *cb,
4448                               struct net_device *dev,
4449                               int *idx,
4450                               struct netdev_hw_addr_list *list)
4451 {
4452         struct netdev_hw_addr *ha;
4453         int err;
4454         u32 portid, seq;
4455
4456         portid = NETLINK_CB(cb->skb).portid;
4457         seq = cb->nlh->nlmsg_seq;
4458
4459         list_for_each_entry(ha, &list->list, list) {
4460                 if (*idx < cb->args[2])
4461                         goto skip;
4462
4463                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4464                                               portid, seq,
4465                                               RTM_NEWNEIGH, NTF_SELF,
4466                                               NLM_F_MULTI, NUD_PERMANENT);
4467                 if (err < 0)
4468                         return err;
4469 skip:
4470                 *idx += 1;
4471         }
4472         return 0;
4473 }
4474
4475 /**
4476  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4477  * @skb: socket buffer to store message in
4478  * @cb: netlink callback
4479  * @dev: netdevice
4480  * @filter_dev: ignored
4481  * @idx: the number of FDB table entries dumped is added to *@idx
4482  *
4483  * Default netdevice operation to dump the existing unicast address list.
4484  * Returns number of addresses from list put in skb.
4485  */
4486 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4487                       struct netlink_callback *cb,
4488                       struct net_device *dev,
4489                       struct net_device *filter_dev,
4490                       int *idx)
4491 {
4492         int err;
4493
4494         if (dev->type != ARPHRD_ETHER)
4495                 return -EINVAL;
4496
4497         netif_addr_lock_bh(dev);
4498         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4499         if (err)
4500                 goto out;
4501         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4502 out:
4503         netif_addr_unlock_bh(dev);
4504         return err;
4505 }
4506 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4507
4508 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4509                                  int *br_idx, int *brport_idx,
4510                                  struct netlink_ext_ack *extack)
4511 {
4512         struct nlattr *tb[NDA_MAX + 1];
4513         struct ndmsg *ndm;
4514         int err, i;
4515
4516         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4517                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4518                 return -EINVAL;
4519         }
4520
4521         ndm = nlmsg_data(nlh);
4522         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4523             ndm->ndm_flags || ndm->ndm_type) {
4524                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4525                 return -EINVAL;
4526         }
4527
4528         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4529                                             NDA_MAX, NULL, extack);
4530         if (err < 0)
4531                 return err;
4532
4533         *brport_idx = ndm->ndm_ifindex;
4534         for (i = 0; i <= NDA_MAX; ++i) {
4535                 if (!tb[i])
4536                         continue;
4537
4538                 switch (i) {
4539                 case NDA_IFINDEX:
4540                         if (nla_len(tb[i]) != sizeof(u32)) {
4541                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4542                                 return -EINVAL;
4543                         }
4544                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4545                         break;
4546                 case NDA_MASTER:
4547                         if (nla_len(tb[i]) != sizeof(u32)) {
4548                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4549                                 return -EINVAL;
4550                         }
4551                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4552                         break;
4553                 default:
4554                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4555                         return -EINVAL;
4556                 }
4557         }
4558
4559         return 0;
4560 }
4561
4562 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4563                                  int *br_idx, int *brport_idx,
4564                                  struct netlink_ext_ack *extack)
4565 {
4566         struct nlattr *tb[IFLA_MAX+1];
4567         int err;
4568
4569         /* A hack to preserve kernel<->userspace interface.
4570          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4571          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4572          * So, check for ndmsg with an optional u32 attribute (not used here).
4573          * Fortunately these sizes don't conflict with the size of ifinfomsg
4574          * with an optional attribute.
4575          */
4576         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4577             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4578              nla_attr_size(sizeof(u32)))) {
4579                 struct ifinfomsg *ifm;
4580
4581                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4582                                              tb, IFLA_MAX, ifla_policy,
4583                                              extack);
4584                 if (err < 0) {
4585                         return -EINVAL;
4586                 } else if (err == 0) {
4587                         if (tb[IFLA_MASTER])
4588                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4589                 }
4590
4591                 ifm = nlmsg_data(nlh);
4592                 *brport_idx = ifm->ifi_index;
4593         }
4594         return 0;
4595 }
4596
4597 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4598 {
4599         struct net_device *dev;
4600         struct net_device *br_dev = NULL;
4601         const struct net_device_ops *ops = NULL;
4602         const struct net_device_ops *cops = NULL;
4603         struct net *net = sock_net(skb->sk);
4604         struct hlist_head *head;
4605         int brport_idx = 0;
4606         int br_idx = 0;
4607         int h, s_h;
4608         int idx = 0, s_idx;
4609         int err = 0;
4610         int fidx = 0;
4611
4612         if (cb->strict_check)
4613                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4614                                             cb->extack);
4615         else
4616                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4617                                             cb->extack);
4618         if (err < 0)
4619                 return err;
4620
4621         if (br_idx) {
4622                 br_dev = __dev_get_by_index(net, br_idx);
4623                 if (!br_dev)
4624                         return -ENODEV;
4625
4626                 ops = br_dev->netdev_ops;
4627         }
4628
4629         s_h = cb->args[0];
4630         s_idx = cb->args[1];
4631
4632         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4633                 idx = 0;
4634                 head = &net->dev_index_head[h];
4635                 hlist_for_each_entry(dev, head, index_hlist) {
4636
4637                         if (brport_idx && (dev->ifindex != brport_idx))
4638                                 continue;
4639
4640                         if (!br_idx) { /* user did not specify a specific bridge */
4641                                 if (netif_is_bridge_port(dev)) {
4642                                         br_dev = netdev_master_upper_dev_get(dev);
4643                                         cops = br_dev->netdev_ops;
4644                                 }
4645                         } else {
4646                                 if (dev != br_dev &&
4647                                     !netif_is_bridge_port(dev))
4648                                         continue;
4649
4650                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4651                                     !netif_is_bridge_master(dev))
4652                                         continue;
4653                                 cops = ops;
4654                         }
4655
4656                         if (idx < s_idx)
4657                                 goto cont;
4658
4659                         if (netif_is_bridge_port(dev)) {
4660                                 if (cops && cops->ndo_fdb_dump) {
4661                                         err = cops->ndo_fdb_dump(skb, cb,
4662                                                                 br_dev, dev,
4663                                                                 &fidx);
4664                                         if (err == -EMSGSIZE)
4665                                                 goto out;
4666                                 }
4667                         }
4668
4669                         if (dev->netdev_ops->ndo_fdb_dump)
4670                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4671                                                                     dev, NULL,
4672                                                                     &fidx);
4673                         else
4674                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4675                                                         &fidx);
4676                         if (err == -EMSGSIZE)
4677                                 goto out;
4678
4679                         cops = NULL;
4680
4681                         /* reset fdb offset to 0 for rest of the interfaces */
4682                         cb->args[2] = 0;
4683                         fidx = 0;
4684 cont:
4685                         idx++;
4686                 }
4687         }
4688
4689 out:
4690         cb->args[0] = h;
4691         cb->args[1] = idx;
4692         cb->args[2] = fidx;
4693
4694         return skb->len;
4695 }
4696
4697 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4698                                 struct nlattr **tb, u8 *ndm_flags,
4699                                 int *br_idx, int *brport_idx, u8 **addr,
4700                                 u16 *vid, struct netlink_ext_ack *extack)
4701 {
4702         struct ndmsg *ndm;
4703         int err, i;
4704
4705         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4706                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4707                 return -EINVAL;
4708         }
4709
4710         ndm = nlmsg_data(nlh);
4711         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4712             ndm->ndm_type) {
4713                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4714                 return -EINVAL;
4715         }
4716
4717         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4718                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4719                 return -EINVAL;
4720         }
4721
4722         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4723                                             NDA_MAX, nda_policy, extack);
4724         if (err < 0)
4725                 return err;
4726
4727         *ndm_flags = ndm->ndm_flags;
4728         *brport_idx = ndm->ndm_ifindex;
4729         for (i = 0; i <= NDA_MAX; ++i) {
4730                 if (!tb[i])
4731                         continue;
4732
4733                 switch (i) {
4734                 case NDA_MASTER:
4735                         *br_idx = nla_get_u32(tb[i]);
4736                         break;
4737                 case NDA_LLADDR:
4738                         if (nla_len(tb[i]) != ETH_ALEN) {
4739                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4740                                 return -EINVAL;
4741                         }
4742                         *addr = nla_data(tb[i]);
4743                         break;
4744                 case NDA_VLAN:
4745                         err = fdb_vid_parse(tb[i], vid, extack);
4746                         if (err)
4747                                 return err;
4748                         break;
4749                 case NDA_VNI:
4750                         break;
4751                 default:
4752                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4753                         return -EINVAL;
4754                 }
4755         }
4756
4757         return 0;
4758 }
4759
4760 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4761                         struct netlink_ext_ack *extack)
4762 {
4763         struct net_device *dev = NULL, *br_dev = NULL;
4764         const struct net_device_ops *ops = NULL;
4765         struct net *net = sock_net(in_skb->sk);
4766         struct nlattr *tb[NDA_MAX + 1];
4767         struct sk_buff *skb;
4768         int brport_idx = 0;
4769         u8 ndm_flags = 0;
4770         int br_idx = 0;
4771         u8 *addr = NULL;
4772         u16 vid = 0;
4773         int err;
4774
4775         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4776                                    &brport_idx, &addr, &vid, extack);
4777         if (err < 0)
4778                 return err;
4779
4780         if (!addr) {
4781                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4782                 return -EINVAL;
4783         }
4784
4785         if (brport_idx) {
4786                 dev = __dev_get_by_index(net, brport_idx);
4787                 if (!dev) {
4788                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4789                         return -ENODEV;
4790                 }
4791         }
4792
4793         if (br_idx) {
4794                 if (dev) {
4795                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4796                         return -EINVAL;
4797                 }
4798
4799                 br_dev = __dev_get_by_index(net, br_idx);
4800                 if (!br_dev) {
4801                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4802                         return -EINVAL;
4803                 }
4804                 ops = br_dev->netdev_ops;
4805         }
4806
4807         if (dev) {
4808                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4809                         if (!netif_is_bridge_port(dev)) {
4810                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4811                                 return -EINVAL;
4812                         }
4813                         br_dev = netdev_master_upper_dev_get(dev);
4814                         if (!br_dev) {
4815                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4816                                 return -EINVAL;
4817                         }
4818                         ops = br_dev->netdev_ops;
4819                 } else {
4820                         if (!(ndm_flags & NTF_SELF)) {
4821                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4822                                 return -EINVAL;
4823                         }
4824                         ops = dev->netdev_ops;
4825                 }
4826         }
4827
4828         if (!br_dev && !dev) {
4829                 NL_SET_ERR_MSG(extack, "No device specified");
4830                 return -ENODEV;
4831         }
4832
4833         if (!ops || !ops->ndo_fdb_get) {
4834                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4835                 return -EOPNOTSUPP;
4836         }
4837
4838         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4839         if (!skb)
4840                 return -ENOBUFS;
4841
4842         if (br_dev)
4843                 dev = br_dev;
4844         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4845                                NETLINK_CB(in_skb).portid,
4846                                nlh->nlmsg_seq, extack);
4847         if (err)
4848                 goto out;
4849
4850         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4851 out:
4852         kfree_skb(skb);
4853         return err;
4854 }
4855
4856 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4857                                unsigned int attrnum, unsigned int flag)
4858 {
4859         if (mask & flag)
4860                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4861         return 0;
4862 }
4863
4864 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4865                             struct net_device *dev, u16 mode,
4866                             u32 flags, u32 mask, int nlflags,
4867                             u32 filter_mask,
4868                             int (*vlan_fill)(struct sk_buff *skb,
4869                                              struct net_device *dev,
4870                                              u32 filter_mask))
4871 {
4872         struct nlmsghdr *nlh;
4873         struct ifinfomsg *ifm;
4874         struct nlattr *br_afspec;
4875         struct nlattr *protinfo;
4876         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4877         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4878         int err = 0;
4879
4880         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4881         if (nlh == NULL)
4882                 return -EMSGSIZE;
4883
4884         ifm = nlmsg_data(nlh);
4885         ifm->ifi_family = AF_BRIDGE;
4886         ifm->__ifi_pad = 0;
4887         ifm->ifi_type = dev->type;
4888         ifm->ifi_index = dev->ifindex;
4889         ifm->ifi_flags = dev_get_flags(dev);
4890         ifm->ifi_change = 0;
4891
4892
4893         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4894             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4895             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4896             (br_dev &&
4897              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4898             (dev->addr_len &&
4899              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4900             (dev->ifindex != dev_get_iflink(dev) &&
4901              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4902                 goto nla_put_failure;
4903
4904         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4905         if (!br_afspec)
4906                 goto nla_put_failure;
4907
4908         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4909                 nla_nest_cancel(skb, br_afspec);
4910                 goto nla_put_failure;
4911         }
4912
4913         if (mode != BRIDGE_MODE_UNDEF) {
4914                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4915                         nla_nest_cancel(skb, br_afspec);
4916                         goto nla_put_failure;
4917                 }
4918         }
4919         if (vlan_fill) {
4920                 err = vlan_fill(skb, dev, filter_mask);
4921                 if (err) {
4922                         nla_nest_cancel(skb, br_afspec);
4923                         goto nla_put_failure;
4924                 }
4925         }
4926         nla_nest_end(skb, br_afspec);
4927
4928         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4929         if (!protinfo)
4930                 goto nla_put_failure;
4931
4932         if (brport_nla_put_flag(skb, flags, mask,
4933                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4934             brport_nla_put_flag(skb, flags, mask,
4935                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4936             brport_nla_put_flag(skb, flags, mask,
4937                                 IFLA_BRPORT_FAST_LEAVE,
4938                                 BR_MULTICAST_FAST_LEAVE) ||
4939             brport_nla_put_flag(skb, flags, mask,
4940                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4941             brport_nla_put_flag(skb, flags, mask,
4942                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4943             brport_nla_put_flag(skb, flags, mask,
4944                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4945             brport_nla_put_flag(skb, flags, mask,
4946                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4947             brport_nla_put_flag(skb, flags, mask,
4948                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4949             brport_nla_put_flag(skb, flags, mask,
4950                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4951             brport_nla_put_flag(skb, flags, mask,
4952                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4953                 nla_nest_cancel(skb, protinfo);
4954                 goto nla_put_failure;
4955         }
4956
4957         nla_nest_end(skb, protinfo);
4958
4959         nlmsg_end(skb, nlh);
4960         return 0;
4961 nla_put_failure:
4962         nlmsg_cancel(skb, nlh);
4963         return err ? err : -EMSGSIZE;
4964 }
4965 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4966
4967 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4968                                     bool strict_check, u32 *filter_mask,
4969                                     struct netlink_ext_ack *extack)
4970 {
4971         struct nlattr *tb[IFLA_MAX+1];
4972         int err, i;
4973
4974         if (strict_check) {
4975                 struct ifinfomsg *ifm;
4976
4977                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4978                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4979                         return -EINVAL;
4980                 }
4981
4982                 ifm = nlmsg_data(nlh);
4983                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4984                     ifm->ifi_change || ifm->ifi_index) {
4985                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4986                         return -EINVAL;
4987                 }
4988
4989                 err = nlmsg_parse_deprecated_strict(nlh,
4990                                                     sizeof(struct ifinfomsg),
4991                                                     tb, IFLA_MAX, ifla_policy,
4992                                                     extack);
4993         } else {
4994                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4995                                              tb, IFLA_MAX, ifla_policy,
4996                                              extack);
4997         }
4998         if (err < 0)
4999                 return err;
5000
5001         /* new attributes should only be added with strict checking */
5002         for (i = 0; i <= IFLA_MAX; ++i) {
5003                 if (!tb[i])
5004                         continue;
5005
5006                 switch (i) {
5007                 case IFLA_EXT_MASK:
5008                         *filter_mask = nla_get_u32(tb[i]);
5009                         break;
5010                 default:
5011                         if (strict_check) {
5012                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5013                                 return -EINVAL;
5014                         }
5015                 }
5016         }
5017
5018         return 0;
5019 }
5020
5021 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5022 {
5023         const struct nlmsghdr *nlh = cb->nlh;
5024         struct net *net = sock_net(skb->sk);
5025         struct net_device *dev;
5026         int idx = 0;
5027         u32 portid = NETLINK_CB(cb->skb).portid;
5028         u32 seq = nlh->nlmsg_seq;
5029         u32 filter_mask = 0;
5030         int err;
5031
5032         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5033                                        cb->extack);
5034         if (err < 0 && cb->strict_check)
5035                 return err;
5036
5037         rcu_read_lock();
5038         for_each_netdev_rcu(net, dev) {
5039                 const struct net_device_ops *ops = dev->netdev_ops;
5040                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5041
5042                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5043                         if (idx >= cb->args[0]) {
5044                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5045                                                 skb, portid, seq, dev,
5046                                                 filter_mask, NLM_F_MULTI);
5047                                 if (err < 0 && err != -EOPNOTSUPP) {
5048                                         if (likely(skb->len))
5049                                                 break;
5050
5051                                         goto out_err;
5052                                 }
5053                         }
5054                         idx++;
5055                 }
5056
5057                 if (ops->ndo_bridge_getlink) {
5058                         if (idx >= cb->args[0]) {
5059                                 err = ops->ndo_bridge_getlink(skb, portid,
5060                                                               seq, dev,
5061                                                               filter_mask,
5062                                                               NLM_F_MULTI);
5063                                 if (err < 0 && err != -EOPNOTSUPP) {
5064                                         if (likely(skb->len))
5065                                                 break;
5066
5067                                         goto out_err;
5068                                 }
5069                         }
5070                         idx++;
5071                 }
5072         }
5073         err = skb->len;
5074 out_err:
5075         rcu_read_unlock();
5076         cb->args[0] = idx;
5077
5078         return err;
5079 }
5080
5081 static inline size_t bridge_nlmsg_size(void)
5082 {
5083         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5084                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5085                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5086                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5087                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5088                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5089                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5090                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5091                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5092                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5093                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5094 }
5095
5096 static int rtnl_bridge_notify(struct net_device *dev)
5097 {
5098         struct net *net = dev_net(dev);
5099         struct sk_buff *skb;
5100         int err = -EOPNOTSUPP;
5101
5102         if (!dev->netdev_ops->ndo_bridge_getlink)
5103                 return 0;
5104
5105         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5106         if (!skb) {
5107                 err = -ENOMEM;
5108                 goto errout;
5109         }
5110
5111         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5112         if (err < 0)
5113                 goto errout;
5114
5115         /* Notification info is only filled for bridge ports, not the bridge
5116          * device itself. Therefore, a zero notification length is valid and
5117          * should not result in an error.
5118          */
5119         if (!skb->len)
5120                 goto errout;
5121
5122         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5123         return 0;
5124 errout:
5125         WARN_ON(err == -EMSGSIZE);
5126         kfree_skb(skb);
5127         if (err)
5128                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5129         return err;
5130 }
5131
5132 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5133                                struct netlink_ext_ack *extack)
5134 {
5135         struct net *net = sock_net(skb->sk);
5136         struct ifinfomsg *ifm;
5137         struct net_device *dev;
5138         struct nlattr *br_spec, *attr = NULL;
5139         int rem, err = -EOPNOTSUPP;
5140         u16 flags = 0;
5141         bool have_flags = false;
5142
5143         if (nlmsg_len(nlh) < sizeof(*ifm))
5144                 return -EINVAL;
5145
5146         ifm = nlmsg_data(nlh);
5147         if (ifm->ifi_family != AF_BRIDGE)
5148                 return -EPFNOSUPPORT;
5149
5150         dev = __dev_get_by_index(net, ifm->ifi_index);
5151         if (!dev) {
5152                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5153                 return -ENODEV;
5154         }
5155
5156         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5157         if (br_spec) {
5158                 nla_for_each_nested(attr, br_spec, rem) {
5159                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
5160                                 if (nla_len(attr) < sizeof(flags))
5161                                         return -EINVAL;
5162
5163                                 have_flags = true;
5164                                 flags = nla_get_u16(attr);
5165                         }
5166
5167                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5168                                 if (nla_len(attr) < sizeof(u16))
5169                                         return -EINVAL;
5170                         }
5171                 }
5172         }
5173
5174         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5175                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5176
5177                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5178                         err = -EOPNOTSUPP;
5179                         goto out;
5180                 }
5181
5182                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5183                                                              extack);
5184                 if (err)
5185                         goto out;
5186
5187                 flags &= ~BRIDGE_FLAGS_MASTER;
5188         }
5189
5190         if ((flags & BRIDGE_FLAGS_SELF)) {
5191                 if (!dev->netdev_ops->ndo_bridge_setlink)
5192                         err = -EOPNOTSUPP;
5193                 else
5194                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5195                                                                   flags,
5196                                                                   extack);
5197                 if (!err) {
5198                         flags &= ~BRIDGE_FLAGS_SELF;
5199
5200                         /* Generate event to notify upper layer of bridge
5201                          * change
5202                          */
5203                         err = rtnl_bridge_notify(dev);
5204                 }
5205         }
5206
5207         if (have_flags)
5208                 memcpy(nla_data(attr), &flags, sizeof(flags));
5209 out:
5210         return err;
5211 }
5212
5213 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5214                                struct netlink_ext_ack *extack)
5215 {
5216         struct net *net = sock_net(skb->sk);
5217         struct ifinfomsg *ifm;
5218         struct net_device *dev;
5219         struct nlattr *br_spec, *attr = NULL;
5220         int rem, err = -EOPNOTSUPP;
5221         u16 flags = 0;
5222         bool have_flags = false;
5223
5224         if (nlmsg_len(nlh) < sizeof(*ifm))
5225                 return -EINVAL;
5226
5227         ifm = nlmsg_data(nlh);
5228         if (ifm->ifi_family != AF_BRIDGE)
5229                 return -EPFNOSUPPORT;
5230
5231         dev = __dev_get_by_index(net, ifm->ifi_index);
5232         if (!dev) {
5233                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5234                 return -ENODEV;
5235         }
5236
5237         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5238         if (br_spec) {
5239                 nla_for_each_nested(attr, br_spec, rem) {
5240                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5241                                 if (nla_len(attr) < sizeof(flags))
5242                                         return -EINVAL;
5243
5244                                 have_flags = true;
5245                                 flags = nla_get_u16(attr);
5246                                 break;
5247                         }
5248                 }
5249         }
5250
5251         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5252                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5253
5254                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5255                         err = -EOPNOTSUPP;
5256                         goto out;
5257                 }
5258
5259                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5260                 if (err)
5261                         goto out;
5262
5263                 flags &= ~BRIDGE_FLAGS_MASTER;
5264         }
5265
5266         if ((flags & BRIDGE_FLAGS_SELF)) {
5267                 if (!dev->netdev_ops->ndo_bridge_dellink)
5268                         err = -EOPNOTSUPP;
5269                 else
5270                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5271                                                                   flags);
5272
5273                 if (!err) {
5274                         flags &= ~BRIDGE_FLAGS_SELF;
5275
5276                         /* Generate event to notify upper layer of bridge
5277                          * change
5278                          */
5279                         err = rtnl_bridge_notify(dev);
5280                 }
5281         }
5282
5283         if (have_flags)
5284                 memcpy(nla_data(attr), &flags, sizeof(flags));
5285 out:
5286         return err;
5287 }
5288
5289 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5290 {
5291         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5292                (!idxattr || idxattr == attrid);
5293 }
5294
5295 static bool
5296 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5297 {
5298         return dev->netdev_ops &&
5299                dev->netdev_ops->ndo_has_offload_stats &&
5300                dev->netdev_ops->ndo_get_offload_stats &&
5301                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5302 }
5303
5304 static unsigned int
5305 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5306 {
5307         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5308                sizeof(struct rtnl_link_stats64) : 0;
5309 }
5310
5311 static int
5312 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5313                              struct sk_buff *skb)
5314 {
5315         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5316         struct nlattr *attr = NULL;
5317         void *attr_data;
5318         int err;
5319
5320         if (!size)
5321                 return -ENODATA;
5322
5323         attr = nla_reserve_64bit(skb, attr_id, size,
5324                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5325         if (!attr)
5326                 return -EMSGSIZE;
5327
5328         attr_data = nla_data(attr);
5329         memset(attr_data, 0, size);
5330
5331         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5332         if (err)
5333                 return err;
5334
5335         return 0;
5336 }
5337
5338 static unsigned int
5339 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5340                                    enum netdev_offload_xstats_type type)
5341 {
5342         bool enabled = netdev_offload_xstats_enabled(dev, type);
5343
5344         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5345 }
5346
5347 struct rtnl_offload_xstats_request_used {
5348         bool request;
5349         bool used;
5350 };
5351
5352 static int
5353 rtnl_offload_xstats_get_stats(struct net_device *dev,
5354                               enum netdev_offload_xstats_type type,
5355                               struct rtnl_offload_xstats_request_used *ru,
5356                               struct rtnl_hw_stats64 *stats,
5357                               struct netlink_ext_ack *extack)
5358 {
5359         bool request;
5360         bool used;
5361         int err;
5362
5363         request = netdev_offload_xstats_enabled(dev, type);
5364         if (!request) {
5365                 used = false;
5366                 goto out;
5367         }
5368
5369         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5370         if (err)
5371                 return err;
5372
5373 out:
5374         if (ru) {
5375                 ru->request = request;
5376                 ru->used = used;
5377         }
5378         return 0;
5379 }
5380
5381 static int
5382 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5383                                        struct rtnl_offload_xstats_request_used *ru)
5384 {
5385         struct nlattr *nest;
5386
5387         nest = nla_nest_start(skb, attr_id);
5388         if (!nest)
5389                 return -EMSGSIZE;
5390
5391         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5392                 goto nla_put_failure;
5393
5394         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5395                 goto nla_put_failure;
5396
5397         nla_nest_end(skb, nest);
5398         return 0;
5399
5400 nla_put_failure:
5401         nla_nest_cancel(skb, nest);
5402         return -EMSGSIZE;
5403 }
5404
5405 static int
5406 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5407                                    struct netlink_ext_ack *extack)
5408 {
5409         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5410         struct rtnl_offload_xstats_request_used ru_l3;
5411         struct nlattr *nest;
5412         int err;
5413
5414         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5415         if (err)
5416                 return err;
5417
5418         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5419         if (!nest)
5420                 return -EMSGSIZE;
5421
5422         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5423                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5424                                                    &ru_l3))
5425                 goto nla_put_failure;
5426
5427         nla_nest_end(skb, nest);
5428         return 0;
5429
5430 nla_put_failure:
5431         nla_nest_cancel(skb, nest);
5432         return -EMSGSIZE;
5433 }
5434
5435 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5436                                     int *prividx, u32 off_filter_mask,
5437                                     struct netlink_ext_ack *extack)
5438 {
5439         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5440         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5441         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5442         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5443         bool have_data = false;
5444         int err;
5445
5446         if (*prividx <= attr_id_cpu_hit &&
5447             (off_filter_mask &
5448              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5449                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5450                 if (!err) {
5451                         have_data = true;
5452                 } else if (err != -ENODATA) {
5453                         *prividx = attr_id_cpu_hit;
5454                         return err;
5455                 }
5456         }
5457
5458         if (*prividx <= attr_id_hw_s_info &&
5459             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5460                 *prividx = attr_id_hw_s_info;
5461
5462                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5463                 if (err)
5464                         return err;
5465
5466                 have_data = true;
5467                 *prividx = 0;
5468         }
5469
5470         if (*prividx <= attr_id_l3_stats &&
5471             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5472                 unsigned int size_l3;
5473                 struct nlattr *attr;
5474
5475                 *prividx = attr_id_l3_stats;
5476
5477                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5478                 if (!size_l3)
5479                         goto skip_l3_stats;
5480                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5481                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5482                 if (!attr)
5483                         return -EMSGSIZE;
5484
5485                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5486                                                     nla_data(attr), extack);
5487                 if (err)
5488                         return err;
5489
5490                 have_data = true;
5491 skip_l3_stats:
5492                 *prividx = 0;
5493         }
5494
5495         if (!have_data)
5496                 return -ENODATA;
5497
5498         *prividx = 0;
5499         return 0;
5500 }
5501
5502 static unsigned int
5503 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5504                                            enum netdev_offload_xstats_type type)
5505 {
5506         return nla_total_size(0) +
5507                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5508                 nla_total_size(sizeof(u8)) +
5509                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5510                 nla_total_size(sizeof(u8)) +
5511                 0;
5512 }
5513
5514 static unsigned int
5515 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5516 {
5517         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5518
5519         return nla_total_size(0) +
5520                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5521                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5522                 0;
5523 }
5524
5525 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5526                                         u32 off_filter_mask)
5527 {
5528         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5529         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5530         int nla_size = 0;
5531         int size;
5532
5533         if (off_filter_mask &
5534             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5535                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5536                 nla_size += nla_total_size_64bit(size);
5537         }
5538
5539         if (off_filter_mask &
5540             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5541                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5542
5543         if (off_filter_mask &
5544             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5545                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5546                 nla_size += nla_total_size_64bit(size);
5547         }
5548
5549         if (nla_size != 0)
5550                 nla_size += nla_total_size(0);
5551
5552         return nla_size;
5553 }
5554
5555 struct rtnl_stats_dump_filters {
5556         /* mask[0] filters outer attributes. Then individual nests have their
5557          * filtering mask at the index of the nested attribute.
5558          */
5559         u32 mask[IFLA_STATS_MAX + 1];
5560 };
5561
5562 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5563                                int type, u32 pid, u32 seq, u32 change,
5564                                unsigned int flags,
5565                                const struct rtnl_stats_dump_filters *filters,
5566                                int *idxattr, int *prividx,
5567                                struct netlink_ext_ack *extack)
5568 {
5569         unsigned int filter_mask = filters->mask[0];
5570         struct if_stats_msg *ifsm;
5571         struct nlmsghdr *nlh;
5572         struct nlattr *attr;
5573         int s_prividx = *prividx;
5574         int err;
5575
5576         ASSERT_RTNL();
5577
5578         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5579         if (!nlh)
5580                 return -EMSGSIZE;
5581
5582         ifsm = nlmsg_data(nlh);
5583         ifsm->family = PF_UNSPEC;
5584         ifsm->pad1 = 0;
5585         ifsm->pad2 = 0;
5586         ifsm->ifindex = dev->ifindex;
5587         ifsm->filter_mask = filter_mask;
5588
5589         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5590                 struct rtnl_link_stats64 *sp;
5591
5592                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5593                                          sizeof(struct rtnl_link_stats64),
5594                                          IFLA_STATS_UNSPEC);
5595                 if (!attr) {
5596                         err = -EMSGSIZE;
5597                         goto nla_put_failure;
5598                 }
5599
5600                 sp = nla_data(attr);
5601                 dev_get_stats(dev, sp);
5602         }
5603
5604         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5605                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5606
5607                 if (ops && ops->fill_linkxstats) {
5608                         *idxattr = IFLA_STATS_LINK_XSTATS;
5609                         attr = nla_nest_start_noflag(skb,
5610                                                      IFLA_STATS_LINK_XSTATS);
5611                         if (!attr) {
5612                                 err = -EMSGSIZE;
5613                                 goto nla_put_failure;
5614                         }
5615
5616                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5617                         nla_nest_end(skb, attr);
5618                         if (err)
5619                                 goto nla_put_failure;
5620                         *idxattr = 0;
5621                 }
5622         }
5623
5624         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5625                              *idxattr)) {
5626                 const struct rtnl_link_ops *ops = NULL;
5627                 const struct net_device *master;
5628
5629                 master = netdev_master_upper_dev_get(dev);
5630                 if (master)
5631                         ops = master->rtnl_link_ops;
5632                 if (ops && ops->fill_linkxstats) {
5633                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5634                         attr = nla_nest_start_noflag(skb,
5635                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5636                         if (!attr) {
5637                                 err = -EMSGSIZE;
5638                                 goto nla_put_failure;
5639                         }
5640
5641                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5642                         nla_nest_end(skb, attr);
5643                         if (err)
5644                                 goto nla_put_failure;
5645                         *idxattr = 0;
5646                 }
5647         }
5648
5649         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5650                              *idxattr)) {
5651                 u32 off_filter_mask;
5652
5653                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5654                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5655                 attr = nla_nest_start_noflag(skb,
5656                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5657                 if (!attr) {
5658                         err = -EMSGSIZE;
5659                         goto nla_put_failure;
5660                 }
5661
5662                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5663                                                off_filter_mask, extack);
5664                 if (err == -ENODATA)
5665                         nla_nest_cancel(skb, attr);
5666                 else
5667                         nla_nest_end(skb, attr);
5668
5669                 if (err && err != -ENODATA)
5670                         goto nla_put_failure;
5671                 *idxattr = 0;
5672         }
5673
5674         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5675                 struct rtnl_af_ops *af_ops;
5676
5677                 *idxattr = IFLA_STATS_AF_SPEC;
5678                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5679                 if (!attr) {
5680                         err = -EMSGSIZE;
5681                         goto nla_put_failure;
5682                 }
5683
5684                 rcu_read_lock();
5685                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5686                         if (af_ops->fill_stats_af) {
5687                                 struct nlattr *af;
5688
5689                                 af = nla_nest_start_noflag(skb,
5690                                                            af_ops->family);
5691                                 if (!af) {
5692                                         rcu_read_unlock();
5693                                         err = -EMSGSIZE;
5694                                         goto nla_put_failure;
5695                                 }
5696                                 err = af_ops->fill_stats_af(skb, dev);
5697
5698                                 if (err == -ENODATA) {
5699                                         nla_nest_cancel(skb, af);
5700                                 } else if (err < 0) {
5701                                         rcu_read_unlock();
5702                                         goto nla_put_failure;
5703                                 }
5704
5705                                 nla_nest_end(skb, af);
5706                         }
5707                 }
5708                 rcu_read_unlock();
5709
5710                 nla_nest_end(skb, attr);
5711
5712                 *idxattr = 0;
5713         }
5714
5715         nlmsg_end(skb, nlh);
5716
5717         return 0;
5718
5719 nla_put_failure:
5720         /* not a multi message or no progress mean a real error */
5721         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5722                 nlmsg_cancel(skb, nlh);
5723         else
5724                 nlmsg_end(skb, nlh);
5725
5726         return err;
5727 }
5728
5729 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5730                                   const struct rtnl_stats_dump_filters *filters)
5731 {
5732         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5733         unsigned int filter_mask = filters->mask[0];
5734
5735         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5736                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5737
5738         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5739                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5740                 int attr = IFLA_STATS_LINK_XSTATS;
5741
5742                 if (ops && ops->get_linkxstats_size) {
5743                         size += nla_total_size(ops->get_linkxstats_size(dev,
5744                                                                         attr));
5745                         /* for IFLA_STATS_LINK_XSTATS */
5746                         size += nla_total_size(0);
5747                 }
5748         }
5749
5750         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5751                 struct net_device *_dev = (struct net_device *)dev;
5752                 const struct rtnl_link_ops *ops = NULL;
5753                 const struct net_device *master;
5754
5755                 /* netdev_master_upper_dev_get can't take const */
5756                 master = netdev_master_upper_dev_get(_dev);
5757                 if (master)
5758                         ops = master->rtnl_link_ops;
5759                 if (ops && ops->get_linkxstats_size) {
5760                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5761
5762                         size += nla_total_size(ops->get_linkxstats_size(dev,
5763                                                                         attr));
5764                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5765                         size += nla_total_size(0);
5766                 }
5767         }
5768
5769         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5770                 u32 off_filter_mask;
5771
5772                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5773                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5774         }
5775
5776         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5777                 struct rtnl_af_ops *af_ops;
5778
5779                 /* for IFLA_STATS_AF_SPEC */
5780                 size += nla_total_size(0);
5781
5782                 rcu_read_lock();
5783                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5784                         if (af_ops->get_stats_af_size) {
5785                                 size += nla_total_size(
5786                                         af_ops->get_stats_af_size(dev));
5787
5788                                 /* for AF_* */
5789                                 size += nla_total_size(0);
5790                         }
5791                 }
5792                 rcu_read_unlock();
5793         }
5794
5795         return size;
5796 }
5797
5798 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5799
5800 static const struct nla_policy
5801 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5802         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5803                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5804 };
5805
5806 static const struct nla_policy
5807 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5808         [IFLA_STATS_GET_FILTERS] =
5809                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5810 };
5811
5812 static const struct nla_policy
5813 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5814         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5815 };
5816
5817 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5818                                         struct rtnl_stats_dump_filters *filters,
5819                                         struct netlink_ext_ack *extack)
5820 {
5821         struct nlattr *tb[IFLA_STATS_MAX + 1];
5822         int err;
5823         int at;
5824
5825         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5826                                rtnl_stats_get_policy_filters, extack);
5827         if (err < 0)
5828                 return err;
5829
5830         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5831                 if (tb[at]) {
5832                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5833                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5834                                 return -EINVAL;
5835                         }
5836                         filters->mask[at] = nla_get_u32(tb[at]);
5837                 }
5838         }
5839
5840         return 0;
5841 }
5842
5843 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5844                                 u32 filter_mask,
5845                                 struct rtnl_stats_dump_filters *filters,
5846                                 struct netlink_ext_ack *extack)
5847 {
5848         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5849         int err;
5850         int i;
5851
5852         filters->mask[0] = filter_mask;
5853         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5854                 filters->mask[i] = -1U;
5855
5856         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5857                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5858         if (err < 0)
5859                 return err;
5860
5861         if (tb[IFLA_STATS_GET_FILTERS]) {
5862                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5863                                                    filters, extack);
5864                 if (err)
5865                         return err;
5866         }
5867
5868         return 0;
5869 }
5870
5871 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5872                                 bool is_dump, struct netlink_ext_ack *extack)
5873 {
5874         struct if_stats_msg *ifsm;
5875
5876         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5877                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5878                 return -EINVAL;
5879         }
5880
5881         if (!strict_check)
5882                 return 0;
5883
5884         ifsm = nlmsg_data(nlh);
5885
5886         /* only requests using strict checks can pass data to influence
5887          * the dump. The legacy exception is filter_mask.
5888          */
5889         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5890                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5891                 return -EINVAL;
5892         }
5893         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5894                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5895                 return -EINVAL;
5896         }
5897
5898         return 0;
5899 }
5900
5901 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5902                           struct netlink_ext_ack *extack)
5903 {
5904         struct rtnl_stats_dump_filters filters;
5905         struct net *net = sock_net(skb->sk);
5906         struct net_device *dev = NULL;
5907         int idxattr = 0, prividx = 0;
5908         struct if_stats_msg *ifsm;
5909         struct sk_buff *nskb;
5910         int err;
5911
5912         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5913                                    false, extack);
5914         if (err)
5915                 return err;
5916
5917         ifsm = nlmsg_data(nlh);
5918         if (ifsm->ifindex > 0)
5919                 dev = __dev_get_by_index(net, ifsm->ifindex);
5920         else
5921                 return -EINVAL;
5922
5923         if (!dev)
5924                 return -ENODEV;
5925
5926         if (!ifsm->filter_mask) {
5927                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5928                 return -EINVAL;
5929         }
5930
5931         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5932         if (err)
5933                 return err;
5934
5935         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5936         if (!nskb)
5937                 return -ENOBUFS;
5938
5939         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5940                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5941                                   0, &filters, &idxattr, &prividx, extack);
5942         if (err < 0) {
5943                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5944                 WARN_ON(err == -EMSGSIZE);
5945                 kfree_skb(nskb);
5946         } else {
5947                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5948         }
5949
5950         return err;
5951 }
5952
5953 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5954 {
5955         struct netlink_ext_ack *extack = cb->extack;
5956         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5957         struct rtnl_stats_dump_filters filters;
5958         struct net *net = sock_net(skb->sk);
5959         unsigned int flags = NLM_F_MULTI;
5960         struct if_stats_msg *ifsm;
5961         struct hlist_head *head;
5962         struct net_device *dev;
5963         int idx = 0;
5964
5965         s_h = cb->args[0];
5966         s_idx = cb->args[1];
5967         s_idxattr = cb->args[2];
5968         s_prividx = cb->args[3];
5969
5970         cb->seq = net->dev_base_seq;
5971
5972         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5973         if (err)
5974                 return err;
5975
5976         ifsm = nlmsg_data(cb->nlh);
5977         if (!ifsm->filter_mask) {
5978                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5979                 return -EINVAL;
5980         }
5981
5982         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5983                                    extack);
5984         if (err)
5985                 return err;
5986
5987         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5988                 idx = 0;
5989                 head = &net->dev_index_head[h];
5990                 hlist_for_each_entry(dev, head, index_hlist) {
5991                         if (idx < s_idx)
5992                                 goto cont;
5993                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5994                                                   NETLINK_CB(cb->skb).portid,
5995                                                   cb->nlh->nlmsg_seq, 0,
5996                                                   flags, &filters,
5997                                                   &s_idxattr, &s_prividx,
5998                                                   extack);
5999                         /* If we ran out of room on the first message,
6000                          * we're in trouble
6001                          */
6002                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6003
6004                         if (err < 0)
6005                                 goto out;
6006                         s_prividx = 0;
6007                         s_idxattr = 0;
6008                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6009 cont:
6010                         idx++;
6011                 }
6012         }
6013 out:
6014         cb->args[3] = s_prividx;
6015         cb->args[2] = s_idxattr;
6016         cb->args[1] = idx;
6017         cb->args[0] = h;
6018
6019         return skb->len;
6020 }
6021
6022 void rtnl_offload_xstats_notify(struct net_device *dev)
6023 {
6024         struct rtnl_stats_dump_filters response_filters = {};
6025         struct net *net = dev_net(dev);
6026         int idxattr = 0, prividx = 0;
6027         struct sk_buff *skb;
6028         int err = -ENOBUFS;
6029
6030         ASSERT_RTNL();
6031
6032         response_filters.mask[0] |=
6033                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6034         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6035                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6036
6037         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6038                         GFP_KERNEL);
6039         if (!skb)
6040                 goto errout;
6041
6042         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6043                                   &response_filters, &idxattr, &prividx, NULL);
6044         if (err < 0) {
6045                 kfree_skb(skb);
6046                 goto errout;
6047         }
6048
6049         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6050         return;
6051
6052 errout:
6053         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6054 }
6055 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6056
6057 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6058                           struct netlink_ext_ack *extack)
6059 {
6060         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6061         struct rtnl_stats_dump_filters response_filters = {};
6062         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6063         struct net *net = sock_net(skb->sk);
6064         struct net_device *dev = NULL;
6065         struct if_stats_msg *ifsm;
6066         bool notify = false;
6067         int err;
6068
6069         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6070                                    false, extack);
6071         if (err)
6072                 return err;
6073
6074         ifsm = nlmsg_data(nlh);
6075         if (ifsm->family != AF_UNSPEC) {
6076                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6077                 return -EINVAL;
6078         }
6079
6080         if (ifsm->ifindex > 0)
6081                 dev = __dev_get_by_index(net, ifsm->ifindex);
6082         else
6083                 return -EINVAL;
6084
6085         if (!dev)
6086                 return -ENODEV;
6087
6088         if (ifsm->filter_mask) {
6089                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6090                 return -EINVAL;
6091         }
6092
6093         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6094                           ifla_stats_set_policy, extack);
6095         if (err < 0)
6096                 return err;
6097
6098         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6099                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6100
6101                 if (req)
6102                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6103                 else
6104                         err = netdev_offload_xstats_disable(dev, t_l3);
6105
6106                 if (!err)
6107                         notify = true;
6108                 else if (err != -EALREADY)
6109                         return err;
6110
6111                 response_filters.mask[0] |=
6112                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6113                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6114                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6115         }
6116
6117         if (notify)
6118                 rtnl_offload_xstats_notify(dev);
6119
6120         return 0;
6121 }
6122
6123 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6124                                    struct netlink_ext_ack *extack)
6125 {
6126         struct br_port_msg *bpm;
6127
6128         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6129                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6130                 return -EINVAL;
6131         }
6132
6133         bpm = nlmsg_data(nlh);
6134         if (bpm->ifindex) {
6135                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6136                 return -EINVAL;
6137         }
6138         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6139                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6140                 return -EINVAL;
6141         }
6142
6143         return 0;
6144 }
6145
6146 struct rtnl_mdb_dump_ctx {
6147         long idx;
6148 };
6149
6150 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6151 {
6152         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6153         struct net *net = sock_net(skb->sk);
6154         struct net_device *dev;
6155         int idx, s_idx;
6156         int err;
6157
6158         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6159
6160         if (cb->strict_check) {
6161                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6162                 if (err)
6163                         return err;
6164         }
6165
6166         s_idx = ctx->idx;
6167         idx = 0;
6168
6169         for_each_netdev(net, dev) {
6170                 if (idx < s_idx)
6171                         goto skip;
6172                 if (!dev->netdev_ops->ndo_mdb_dump)
6173                         goto skip;
6174
6175                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6176                 if (err == -EMSGSIZE)
6177                         goto out;
6178                 /* Moving on to next device, reset markers and sequence
6179                  * counters since they are all maintained per-device.
6180                  */
6181                 memset(cb->ctx, 0, sizeof(cb->ctx));
6182                 cb->prev_seq = 0;
6183                 cb->seq = 0;
6184 skip:
6185                 idx++;
6186         }
6187
6188 out:
6189         ctx->idx = idx;
6190         return skb->len;
6191 }
6192
6193 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6194                                    struct netlink_ext_ack *extack)
6195 {
6196         struct br_mdb_entry *entry = nla_data(attr);
6197
6198         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6199                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6200                 return -EINVAL;
6201         }
6202
6203         if (entry->ifindex == 0) {
6204                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6205                 return -EINVAL;
6206         }
6207
6208         if (entry->addr.proto == htons(ETH_P_IP)) {
6209                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6210                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6211                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6212                         return -EINVAL;
6213                 }
6214                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6215                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6216                         return -EINVAL;
6217                 }
6218 #if IS_ENABLED(CONFIG_IPV6)
6219         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6220                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6221                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6222                         return -EINVAL;
6223                 }
6224 #endif
6225         } else if (entry->addr.proto == 0) {
6226                 /* L2 mdb */
6227                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6228                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6229                         return -EINVAL;
6230                 }
6231         } else {
6232                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6233                 return -EINVAL;
6234         }
6235
6236         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6237                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6238                 return -EINVAL;
6239         }
6240         if (entry->vid >= VLAN_VID_MASK) {
6241                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6242                 return -EINVAL;
6243         }
6244
6245         return 0;
6246 }
6247
6248 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6249         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6250         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6251                                                   rtnl_validate_mdb_entry,
6252                                                   sizeof(struct br_mdb_entry)),
6253         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6254 };
6255
6256 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6257                         struct netlink_ext_ack *extack)
6258 {
6259         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6260         struct net *net = sock_net(skb->sk);
6261         struct br_port_msg *bpm;
6262         struct net_device *dev;
6263         int err;
6264
6265         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6266                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6267         if (err)
6268                 return err;
6269
6270         bpm = nlmsg_data(nlh);
6271         if (!bpm->ifindex) {
6272                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6273                 return -EINVAL;
6274         }
6275
6276         dev = __dev_get_by_index(net, bpm->ifindex);
6277         if (!dev) {
6278                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6279                 return -ENODEV;
6280         }
6281
6282         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6283                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6284                 return -EINVAL;
6285         }
6286
6287         if (!dev->netdev_ops->ndo_mdb_add) {
6288                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6289                 return -EOPNOTSUPP;
6290         }
6291
6292         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6293 }
6294
6295 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6296                         struct netlink_ext_ack *extack)
6297 {
6298         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6299         struct net *net = sock_net(skb->sk);
6300         struct br_port_msg *bpm;
6301         struct net_device *dev;
6302         int err;
6303
6304         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6305                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6306         if (err)
6307                 return err;
6308
6309         bpm = nlmsg_data(nlh);
6310         if (!bpm->ifindex) {
6311                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6312                 return -EINVAL;
6313         }
6314
6315         dev = __dev_get_by_index(net, bpm->ifindex);
6316         if (!dev) {
6317                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6318                 return -ENODEV;
6319         }
6320
6321         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6322                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6323                 return -EINVAL;
6324         }
6325
6326         if (!dev->netdev_ops->ndo_mdb_del) {
6327                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6328                 return -EOPNOTSUPP;
6329         }
6330
6331         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6332 }
6333
6334 /* Process one rtnetlink message. */
6335
6336 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6337                              struct netlink_ext_ack *extack)
6338 {
6339         struct net *net = sock_net(skb->sk);
6340         struct rtnl_link *link;
6341         enum rtnl_kinds kind;
6342         struct module *owner;
6343         int err = -EOPNOTSUPP;
6344         rtnl_doit_func doit;
6345         unsigned int flags;
6346         int family;
6347         int type;
6348
6349         type = nlh->nlmsg_type;
6350         if (type > RTM_MAX)
6351                 return -EOPNOTSUPP;
6352
6353         type -= RTM_BASE;
6354
6355         /* All the messages must have at least 1 byte length */
6356         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6357                 return 0;
6358
6359         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6360         kind = rtnl_msgtype_kind(type);
6361
6362         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6363                 return -EPERM;
6364
6365         rcu_read_lock();
6366         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6367                 struct sock *rtnl;
6368                 rtnl_dumpit_func dumpit;
6369                 u32 min_dump_alloc = 0;
6370
6371                 link = rtnl_get_link(family, type);
6372                 if (!link || !link->dumpit) {
6373                         family = PF_UNSPEC;
6374                         link = rtnl_get_link(family, type);
6375                         if (!link || !link->dumpit)
6376                                 goto err_unlock;
6377                 }
6378                 owner = link->owner;
6379                 dumpit = link->dumpit;
6380
6381                 if (type == RTM_GETLINK - RTM_BASE)
6382                         min_dump_alloc = rtnl_calcit(skb, nlh);
6383
6384                 err = 0;
6385                 /* need to do this before rcu_read_unlock() */
6386                 if (!try_module_get(owner))
6387                         err = -EPROTONOSUPPORT;
6388
6389                 rcu_read_unlock();
6390
6391                 rtnl = net->rtnl;
6392                 if (err == 0) {
6393                         struct netlink_dump_control c = {
6394                                 .dump           = dumpit,
6395                                 .min_dump_alloc = min_dump_alloc,
6396                                 .module         = owner,
6397                         };
6398                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6399                         /* netlink_dump_start() will keep a reference on
6400                          * module if dump is still in progress.
6401                          */
6402                         module_put(owner);
6403                 }
6404                 return err;
6405         }
6406
6407         link = rtnl_get_link(family, type);
6408         if (!link || !link->doit) {
6409                 family = PF_UNSPEC;
6410                 link = rtnl_get_link(PF_UNSPEC, type);
6411                 if (!link || !link->doit)
6412                         goto out_unlock;
6413         }
6414
6415         owner = link->owner;
6416         if (!try_module_get(owner)) {
6417                 err = -EPROTONOSUPPORT;
6418                 goto out_unlock;
6419         }
6420
6421         flags = link->flags;
6422         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6423             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6424                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6425                 module_put(owner);
6426                 goto err_unlock;
6427         }
6428
6429         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6430                 doit = link->doit;
6431                 rcu_read_unlock();
6432                 if (doit)
6433                         err = doit(skb, nlh, extack);
6434                 module_put(owner);
6435                 return err;
6436         }
6437         rcu_read_unlock();
6438
6439         rtnl_lock();
6440         link = rtnl_get_link(family, type);
6441         if (link && link->doit)
6442                 err = link->doit(skb, nlh, extack);
6443         rtnl_unlock();
6444
6445         module_put(owner);
6446
6447         return err;
6448
6449 out_unlock:
6450         rcu_read_unlock();
6451         return err;
6452
6453 err_unlock:
6454         rcu_read_unlock();
6455         return -EOPNOTSUPP;
6456 }
6457
6458 static void rtnetlink_rcv(struct sk_buff *skb)
6459 {
6460         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6461 }
6462
6463 static int rtnetlink_bind(struct net *net, int group)
6464 {
6465         switch (group) {
6466         case RTNLGRP_IPV4_MROUTE_R:
6467         case RTNLGRP_IPV6_MROUTE_R:
6468                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6469                         return -EPERM;
6470                 break;
6471         }
6472         return 0;
6473 }
6474
6475 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6476 {
6477         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6478
6479         switch (event) {
6480         case NETDEV_REBOOT:
6481         case NETDEV_CHANGEMTU:
6482         case NETDEV_CHANGEADDR:
6483         case NETDEV_CHANGENAME:
6484         case NETDEV_FEAT_CHANGE:
6485         case NETDEV_BONDING_FAILOVER:
6486         case NETDEV_POST_TYPE_CHANGE:
6487         case NETDEV_NOTIFY_PEERS:
6488         case NETDEV_CHANGEUPPER:
6489         case NETDEV_RESEND_IGMP:
6490         case NETDEV_CHANGEINFODATA:
6491         case NETDEV_CHANGELOWERSTATE:
6492         case NETDEV_CHANGE_TX_QUEUE_LEN:
6493                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6494                                    GFP_KERNEL, NULL, 0, 0, NULL);
6495                 break;
6496         default:
6497                 break;
6498         }
6499         return NOTIFY_DONE;
6500 }
6501
6502 static struct notifier_block rtnetlink_dev_notifier = {
6503         .notifier_call  = rtnetlink_event,
6504 };
6505
6506
6507 static int __net_init rtnetlink_net_init(struct net *net)
6508 {
6509         struct sock *sk;
6510         struct netlink_kernel_cfg cfg = {
6511                 .groups         = RTNLGRP_MAX,
6512                 .input          = rtnetlink_rcv,
6513                 .cb_mutex       = &rtnl_mutex,
6514                 .flags          = NL_CFG_F_NONROOT_RECV,
6515                 .bind           = rtnetlink_bind,
6516         };
6517
6518         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6519         if (!sk)
6520                 return -ENOMEM;
6521         net->rtnl = sk;
6522         return 0;
6523 }
6524
6525 static void __net_exit rtnetlink_net_exit(struct net *net)
6526 {
6527         netlink_kernel_release(net->rtnl);
6528         net->rtnl = NULL;
6529 }
6530
6531 static struct pernet_operations rtnetlink_net_ops = {
6532         .init = rtnetlink_net_init,
6533         .exit = rtnetlink_net_exit,
6534 };
6535
6536 void __init rtnetlink_init(void)
6537 {
6538         if (register_pernet_subsys(&rtnetlink_net_ops))
6539                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6540
6541         register_netdevice_notifier(&rtnetlink_dev_notifier);
6542
6543         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6544                       rtnl_dump_ifinfo, 0);
6545         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6546         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6547         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6548
6549         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6550         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6551         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6552
6553         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6554         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6555
6556         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6557         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6558                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6559         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6560
6561         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6562         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6563         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6564
6565         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6566                       0);
6567         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6568
6569         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, rtnl_mdb_dump, 0);
6570         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6571         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL, 0);
6572 }