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