net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/bitops.h>
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/kernel.h>
  25 #include <linux/timer.h>
  26 #include <linux/string.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/fcntl.h>
  30 #include <linux/mm.h>
  31 #include <linux/slab.h>
  32 #include <linux/interrupt.h>
  33 #include <linux/capability.h>
  34 #include <linux/skbuff.h>
  35 #include <linux/init.h>
  36 #include <linux/security.h>
  37 #include <linux/mutex.h>
  38 #include <linux/if_addr.h>
  39 #include <linux/if_bridge.h>
  40 #include <linux/if_vlan.h>
  41 #include <linux/pci.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/bpf.h>
  44
  45 #include <linux/uaccess.h>
  46
  47 #include <linux/inet.h>
  48 #include <linux/netdevice.h>
  49 #include <net/switchdev.h>
  50 #include <net/ip.h>
  51 #include <net/protocol.h>
  52 #include <net/arp.h>
  53 #include <net/route.h>
  54 #include <net/udp.h>
  55 #include <net/tcp.h>
  56 #include <net/sock.h>
  57 #include <net/pkt_sched.h>
  58 #include <net/fib_rules.h>
  59 #include <net/rtnetlink.h>
  60 #include <net/net_namespace.h>
  61
  62 struct rtnl_link {
  63         rtnl_doit_func          doit;
  64         rtnl_dumpit_func        dumpit;
  65         unsigned int            flags;
  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 static struct sk_buff *defer_kfree_skb_list;
  77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  78 {
  79         if (head && tail) {
  80                 tail->next = defer_kfree_skb_list;
  81                 defer_kfree_skb_list = head;
  82         }
  83 }
  84 EXPORT_SYMBOL(rtnl_kfree_skbs);
  85
  86 void __rtnl_unlock(void)
  87 {
  88         struct sk_buff *head = defer_kfree_skb_list;
  89
  90         defer_kfree_skb_list = NULL;
  91
  92         mutex_unlock(&rtnl_mutex);
  93
  94         while (head) {
  95                 struct sk_buff *next = head->next;
  96
  97                 kfree_skb(head);
  98                 cond_resched();
  99                 head = next;
 100         }
 101 }
 102
 103 void rtnl_unlock(void)
 104 {
 105         /* This fellow will unlock it for us. */
 106         netdev_run_todo();
 107 }
 108 EXPORT_SYMBOL(rtnl_unlock);
 109
 110 int rtnl_trylock(void)
 111 {
 112         return mutex_trylock(&rtnl_mutex);
 113 }
 114 EXPORT_SYMBOL(rtnl_trylock);
 115
 116 int rtnl_is_locked(void)
 117 {
 118         return mutex_is_locked(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_is_locked);
 121
 122 #ifdef CONFIG_PROVE_LOCKING
 123 bool lockdep_rtnl_is_held(void)
 124 {
 125         return lockdep_is_held(&rtnl_mutex);
 126 }
 127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 129
 130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
 132
 133 static inline int rtm_msgindex(int msgtype)
 134 {
 135         int msgindex = msgtype - RTM_BASE;
 136
 137         /*
 138          * msgindex < 0 implies someone tried to register a netlink
 139          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 140          * the message type has not been added to linux/rtnetlink.h
 141          */
 142         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 143
 144         return msgindex;
 145 }
 146
 147 /**
 148  * __rtnl_register - Register a rtnetlink message type
 149  * @protocol: Protocol family or PF_UNSPEC
 150  * @msgtype: rtnetlink message type
 151  * @doit: Function pointer called for each request message
 152  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 153  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 154  *
 155  * Registers the specified function pointers (at least one of them has
 156  * to be non-NULL) to be called whenever a request message for the
 157  * specified protocol family and message type is received.
 158  *
 159  * The special protocol family PF_UNSPEC may be used to define fallback
 160  * function pointers for the case when no entry for the specific protocol
 161  * family exists.
 162  *
 163  * Returns 0 on success or a negative error code.
 164  */
 165 int __rtnl_register(int protocol, int msgtype,
 166                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 167                     unsigned int flags)
 168 {
 169         struct rtnl_link *tab;
 170         int msgindex;
 171
 172         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 173         msgindex = rtm_msgindex(msgtype);
 174
 175         tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
 176         if (tab == NULL) {
 177                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 178                 if (tab == NULL)
 179                         return -ENOBUFS;
 180
 181                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 182         }
 183
 184         if (doit)
 185                 tab[msgindex].doit = doit;
 186         if (dumpit)
 187                 tab[msgindex].dumpit = dumpit;
 188         tab[msgindex].flags |= flags;
 189
 190         return 0;
 191 }
 192 EXPORT_SYMBOL_GPL(__rtnl_register);
 193
 194 /**
 195  * rtnl_register - Register a rtnetlink message type
 196  *
 197  * Identical to __rtnl_register() but panics on failure. This is useful
 198  * as failure of this function is very unlikely, it can only happen due
 199  * to lack of memory when allocating the chain to store all message
 200  * handlers for a protocol. Meant for use in init functions where lack
 201  * of memory implies no sense in continuing.
 202  */
 203 void rtnl_register(int protocol, int msgtype,
 204                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 205                    unsigned int flags)
 206 {
 207         if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
 208                 panic("Unable to register rtnetlink message handler, "
 209                       "protocol = %d, message type = %d\n",
 210                       protocol, msgtype);
 211 }
 212 EXPORT_SYMBOL_GPL(rtnl_register);
 213
 214 /**
 215  * rtnl_unregister - Unregister a rtnetlink message type
 216  * @protocol: Protocol family or PF_UNSPEC
 217  * @msgtype: rtnetlink message type
 218  *
 219  * Returns 0 on success or a negative error code.
 220  */
 221 int rtnl_unregister(int protocol, int msgtype)
 222 {
 223         struct rtnl_link *handlers;
 224         int msgindex;
 225
 226         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 227         msgindex = rtm_msgindex(msgtype);
 228
 229         rtnl_lock();
 230         handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
 231         if (!handlers) {
 232                 rtnl_unlock();
 233                 return -ENOENT;
 234         }
 235
 236         handlers[msgindex].doit = NULL;
 237         handlers[msgindex].dumpit = NULL;
 238         handlers[msgindex].flags = 0;
 239         rtnl_unlock();
 240
 241         return 0;
 242 }
 243 EXPORT_SYMBOL_GPL(rtnl_unregister);
 244
 245 /**
 246  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 247  * @protocol : Protocol family or PF_UNSPEC
 248  *
 249  * Identical to calling rtnl_unregster() for all registered message types
 250  * of a certain protocol family.
 251  */
 252 void rtnl_unregister_all(int protocol)
 253 {
 254         struct rtnl_link *handlers;
 255
 256         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 257
 258         rtnl_lock();
 259         handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
 260         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 261         rtnl_unlock();
 262
 263         synchronize_net();
 264
 265         while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
 266                 schedule();
 267         kfree(handlers);
 268 }
 269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 270
 271 static LIST_HEAD(link_ops);
 272
 273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 274 {
 275         const struct rtnl_link_ops *ops;
 276
 277         list_for_each_entry(ops, &link_ops, list) {
 278                 if (!strcmp(ops->kind, kind))
 279                         return ops;
 280         }
 281         return NULL;
 282 }
 283
 284 /**
 285  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 286  * @ops: struct rtnl_link_ops * to register
 287  *
 288  * The caller must hold the rtnl_mutex. This function should be used
 289  * by drivers that create devices during module initialization. It
 290  * must be called before registering the devices.
 291  *
 292  * Returns 0 on success or a negative error code.
 293  */
 294 int __rtnl_link_register(struct rtnl_link_ops *ops)
 295 {
 296         if (rtnl_link_ops_get(ops->kind))
 297                 return -EEXIST;
 298
 299         /* The check for setup is here because if ops
 300          * does not have that filled up, it is not possible
 301          * to use the ops for creating device. So do not
 302          * fill up dellink as well. That disables rtnl_dellink.
 303          */
 304         if (ops->setup && !ops->dellink)
 305                 ops->dellink = unregister_netdevice_queue;
 306
 307         list_add_tail(&ops->list, &link_ops);
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 311
 312 /**
 313  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 314  * @ops: struct rtnl_link_ops * to register
 315  *
 316  * Returns 0 on success or a negative error code.
 317  */
 318 int rtnl_link_register(struct rtnl_link_ops *ops)
 319 {
 320         int err;
 321
 322         rtnl_lock();
 323         err = __rtnl_link_register(ops);
 324         rtnl_unlock();
 325         return err;
 326 }
 327 EXPORT_SYMBOL_GPL(rtnl_link_register);
 328
 329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 330 {
 331         struct net_device *dev;
 332         LIST_HEAD(list_kill);
 333
 334         for_each_netdev(net, dev) {
 335                 if (dev->rtnl_link_ops == ops)
 336                         ops->dellink(dev, &list_kill);
 337         }
 338         unregister_netdevice_many(&list_kill);
 339 }
 340
 341 /**
 342  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 343  * @ops: struct rtnl_link_ops * to unregister
 344  *
 345  * The caller must hold the rtnl_mutex.
 346  */
 347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 348 {
 349         struct net *net;
 350
 351         for_each_net(net) {
 352                 __rtnl_kill_links(net, ops);
 353         }
 354         list_del(&ops->list);
 355 }
 356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 357
 358 /* Return with the rtnl_lock held when there are no network
 359  * devices unregistering in any network namespace.
 360  */
 361 static void rtnl_lock_unregistering_all(void)
 362 {
 363         struct net *net;
 364         bool unregistering;
 365         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 366
 367         add_wait_queue(&netdev_unregistering_wq, &wait);
 368         for (;;) {
 369                 unregistering = false;
 370                 rtnl_lock();
 371                 for_each_net(net) {
 372                         if (net->dev_unreg_count > 0) {
 373                                 unregistering = true;
 374                                 break;
 375                         }
 376                 }
 377                 if (!unregistering)
 378                         break;
 379                 __rtnl_unlock();
 380
 381                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 382         }
 383         remove_wait_queue(&netdev_unregistering_wq, &wait);
 384 }
 385
 386 /**
 387  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 388  * @ops: struct rtnl_link_ops * to unregister
 389  */
 390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 391 {
 392         /* Close the race with cleanup_net() */
 393         mutex_lock(&net_mutex);
 394         rtnl_lock_unregistering_all();
 395         __rtnl_link_unregister(ops);
 396         rtnl_unlock();
 397         mutex_unlock(&net_mutex);
 398 }
 399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 400
 401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 402 {
 403         struct net_device *master_dev;
 404         const struct rtnl_link_ops *ops;
 405         size_t size = 0;
 406
 407         rcu_read_lock();
 408
 409         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 410         if (!master_dev)
 411                 goto out;
 412
 413         ops = master_dev->rtnl_link_ops;
 414         if (!ops || !ops->get_slave_size)
 415                 goto out;
 416         /* IFLA_INFO_SLAVE_DATA + nested data */
 417         size = nla_total_size(sizeof(struct nlattr)) +
 418                ops->get_slave_size(master_dev, dev);
 419
 420 out:
 421         rcu_read_unlock();
 422         return size;
 423 }
 424
 425 static size_t rtnl_link_get_size(const struct net_device *dev)
 426 {
 427         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 428         size_t size;
 429
 430         if (!ops)
 431                 return 0;
 432
 433         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 434                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 435
 436         if (ops->get_size)
 437                 /* IFLA_INFO_DATA + nested data */
 438                 size += nla_total_size(sizeof(struct nlattr)) +
 439                         ops->get_size(dev);
 440
 441         if (ops->get_xstats_size)
 442                 /* IFLA_INFO_XSTATS */
 443                 size += nla_total_size(ops->get_xstats_size(dev));
 444
 445         size += rtnl_link_get_slave_info_data_size(dev);
 446
 447         return size;
 448 }
 449
 450 static LIST_HEAD(rtnl_af_ops);
 451
 452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 453 {
 454         const struct rtnl_af_ops *ops;
 455
 456         list_for_each_entry(ops, &rtnl_af_ops, list) {
 457                 if (ops->family == family)
 458                         return ops;
 459         }
 460
 461         return NULL;
 462 }
 463
 464 /**
 465  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 466  * @ops: struct rtnl_af_ops * to register
 467  *
 468  * Returns 0 on success or a negative error code.
 469  */
 470 void rtnl_af_register(struct rtnl_af_ops *ops)
 471 {
 472         rtnl_lock();
 473         list_add_tail(&ops->list, &rtnl_af_ops);
 474         rtnl_unlock();
 475 }
 476 EXPORT_SYMBOL_GPL(rtnl_af_register);
 477
 478 /**
 479  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 480  * @ops: struct rtnl_af_ops * to unregister
 481  */
 482 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         rtnl_lock();
 485         list_del(&ops->list);
 486         rtnl_unlock();
 487 }
 488 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 489
 490 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 491                                     u32 ext_filter_mask)
 492 {
 493         struct rtnl_af_ops *af_ops;
 494         size_t size;
 495
 496         /* IFLA_AF_SPEC */
 497         size = nla_total_size(sizeof(struct nlattr));
 498
 499         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 500                 if (af_ops->get_link_af_size) {
 501                         /* AF_* + nested data */
 502                         size += nla_total_size(sizeof(struct nlattr)) +
 503                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 504                 }
 505         }
 506
 507         return size;
 508 }
 509
 510 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 511 {
 512         struct net_device *master_dev;
 513         bool ret = false;
 514
 515         rcu_read_lock();
 516
 517         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 518         if (master_dev && master_dev->rtnl_link_ops)
 519                 ret = true;
 520         rcu_read_unlock();
 521         return ret;
 522 }
 523
 524 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 525                                      const struct net_device *dev)
 526 {
 527         struct net_device *master_dev;
 528         const struct rtnl_link_ops *ops;
 529         struct nlattr *slave_data;
 530         int err;
 531
 532         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 533         if (!master_dev)
 534                 return 0;
 535         ops = master_dev->rtnl_link_ops;
 536         if (!ops)
 537                 return 0;
 538         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 539                 return -EMSGSIZE;
 540         if (ops->fill_slave_info) {
 541                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 542                 if (!slave_data)
 543                         return -EMSGSIZE;
 544                 err = ops->fill_slave_info(skb, master_dev, dev);
 545                 if (err < 0)
 546                         goto err_cancel_slave_data;
 547                 nla_nest_end(skb, slave_data);
 548         }
 549         return 0;
 550
 551 err_cancel_slave_data:
 552         nla_nest_cancel(skb, slave_data);
 553         return err;
 554 }
 555
 556 static int rtnl_link_info_fill(struct sk_buff *skb,
 557                                const struct net_device *dev)
 558 {
 559         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 560         struct nlattr *data;
 561         int err;
 562
 563         if (!ops)
 564                 return 0;
 565         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 566                 return -EMSGSIZE;
 567         if (ops->fill_xstats) {
 568                 err = ops->fill_xstats(skb, dev);
 569                 if (err < 0)
 570                         return err;
 571         }
 572         if (ops->fill_info) {
 573                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 574                 if (data == NULL)
 575                         return -EMSGSIZE;
 576                 err = ops->fill_info(skb, dev);
 577                 if (err < 0)
 578                         goto err_cancel_data;
 579                 nla_nest_end(skb, data);
 580         }
 581         return 0;
 582
 583 err_cancel_data:
 584         nla_nest_cancel(skb, data);
 585         return err;
 586 }
 587
 588 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 589 {
 590         struct nlattr *linkinfo;
 591         int err = -EMSGSIZE;
 592
 593         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 594         if (linkinfo == NULL)
 595                 goto out;
 596
 597         err = rtnl_link_info_fill(skb, dev);
 598         if (err < 0)
 599                 goto err_cancel_link;
 600
 601         err = rtnl_link_slave_info_fill(skb, dev);
 602         if (err < 0)
 603                 goto err_cancel_link;
 604
 605         nla_nest_end(skb, linkinfo);
 606         return 0;
 607
 608 err_cancel_link:
 609         nla_nest_cancel(skb, linkinfo);
 610 out:
 611         return err;
 612 }
 613
 614 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 615 {
 616         struct sock *rtnl = net->rtnl;
 617         int err = 0;
 618
 619         NETLINK_CB(skb).dst_group = group;
 620         if (echo)
 621                 refcount_inc(&skb->users);
 622         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 623         if (echo)
 624                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 625         return err;
 626 }
 627
 628 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 629 {
 630         struct sock *rtnl = net->rtnl;
 631
 632         return nlmsg_unicast(rtnl, skb, pid);
 633 }
 634 EXPORT_SYMBOL(rtnl_unicast);
 635
 636 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 637                  struct nlmsghdr *nlh, gfp_t flags)
 638 {
 639         struct sock *rtnl = net->rtnl;
 640         int report = 0;
 641
 642         if (nlh)
 643                 report = nlmsg_report(nlh);
 644
 645         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 646 }
 647 EXPORT_SYMBOL(rtnl_notify);
 648
 649 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 650 {
 651         struct sock *rtnl = net->rtnl;
 652
 653         netlink_set_err(rtnl, 0, group, error);
 654 }
 655 EXPORT_SYMBOL(rtnl_set_sk_err);
 656
 657 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 658 {
 659         struct nlattr *mx;
 660         int i, valid = 0;
 661
 662         mx = nla_nest_start(skb, RTA_METRICS);
 663         if (mx == NULL)
 664                 return -ENOBUFS;
 665
 666         for (i = 0; i < RTAX_MAX; i++) {
 667                 if (metrics[i]) {
 668                         if (i == RTAX_CC_ALGO - 1) {
 669                                 char tmp[TCP_CA_NAME_MAX], *name;
 670
 671                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 672                                 if (!name)
 673                                         continue;
 674                                 if (nla_put_string(skb, i + 1, name))
 675                                         goto nla_put_failure;
 676                         } else if (i == RTAX_FEATURES - 1) {
 677                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 678
 679                                 if (!user_features)
 680                                         continue;
 681                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 682                                 if (nla_put_u32(skb, i + 1, user_features))
 683                                         goto nla_put_failure;
 684                         } else {
 685                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 686                                         goto nla_put_failure;
 687                         }
 688                         valid++;
 689                 }
 690         }
 691
 692         if (!valid) {
 693                 nla_nest_cancel(skb, mx);
 694                 return 0;
 695         }
 696
 697         return nla_nest_end(skb, mx);
 698
 699 nla_put_failure:
 700         nla_nest_cancel(skb, mx);
 701         return -EMSGSIZE;
 702 }
 703 EXPORT_SYMBOL(rtnetlink_put_metrics);
 704
 705 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 706                        long expires, u32 error)
 707 {
 708         struct rta_cacheinfo ci = {
 709                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 710                 .rta_used = dst->__use,
 711                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 712                 .rta_error = error,
 713                 .rta_id =  id,
 714         };
 715
 716         if (expires) {
 717                 unsigned long clock;
 718
 719                 clock = jiffies_to_clock_t(abs(expires));
 720                 clock = min_t(unsigned long, clock, INT_MAX);
 721                 ci.rta_expires = (expires > 0) ? clock : -clock;
 722         }
 723         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 724 }
 725 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 726
 727 static void set_operstate(struct net_device *dev, unsigned char transition)
 728 {
 729         unsigned char operstate = dev->operstate;
 730
 731         switch (transition) {
 732         case IF_OPER_UP:
 733                 if ((operstate == IF_OPER_DORMANT ||
 734                      operstate == IF_OPER_UNKNOWN) &&
 735                     !netif_dormant(dev))
 736                         operstate = IF_OPER_UP;
 737                 break;
 738
 739         case IF_OPER_DORMANT:
 740                 if (operstate == IF_OPER_UP ||
 741                     operstate == IF_OPER_UNKNOWN)
 742                         operstate = IF_OPER_DORMANT;
 743                 break;
 744         }
 745
 746         if (dev->operstate != operstate) {
 747                 write_lock_bh(&dev_base_lock);
 748                 dev->operstate = operstate;
 749                 write_unlock_bh(&dev_base_lock);
 750                 netdev_state_change(dev);
 751         }
 752 }
 753
 754 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 755 {
 756         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 757                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 758 }
 759
 760 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 761                                            const struct ifinfomsg *ifm)
 762 {
 763         unsigned int flags = ifm->ifi_flags;
 764
 765         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 766         if (ifm->ifi_change)
 767                 flags = (flags & ifm->ifi_change) |
 768                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 769
 770         return flags;
 771 }
 772
 773 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 774                                  const struct rtnl_link_stats64 *b)
 775 {
 776         a->rx_packets = b->rx_packets;
 777         a->tx_packets = b->tx_packets;
 778         a->rx_bytes = b->rx_bytes;
 779         a->tx_bytes = b->tx_bytes;
 780         a->rx_errors = b->rx_errors;
 781         a->tx_errors = b->tx_errors;
 782         a->rx_dropped = b->rx_dropped;
 783         a->tx_dropped = b->tx_dropped;
 784
 785         a->multicast = b->multicast;
 786         a->collisions = b->collisions;
 787
 788         a->rx_length_errors = b->rx_length_errors;
 789         a->rx_over_errors = b->rx_over_errors;
 790         a->rx_crc_errors = b->rx_crc_errors;
 791         a->rx_frame_errors = b->rx_frame_errors;
 792         a->rx_fifo_errors = b->rx_fifo_errors;
 793         a->rx_missed_errors = b->rx_missed_errors;
 794
 795         a->tx_aborted_errors = b->tx_aborted_errors;
 796         a->tx_carrier_errors = b->tx_carrier_errors;
 797         a->tx_fifo_errors = b->tx_fifo_errors;
 798         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 799         a->tx_window_errors = b->tx_window_errors;
 800
 801         a->rx_compressed = b->rx_compressed;
 802         a->tx_compressed = b->tx_compressed;
 803
 804         a->rx_nohandler = b->rx_nohandler;
 805 }
 806
 807 /* All VF info */
 808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 809                                    u32 ext_filter_mask)
 810 {
 811         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 812                 int num_vfs = dev_num_vf(dev->dev.parent);
 813                 size_t size = nla_total_size(0);
 814                 size += num_vfs *
 815                         (nla_total_size(0) +
 816                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 817                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 818                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 819                          nla_total_size(MAX_VLAN_LIST_LEN *
 820                                         sizeof(struct ifla_vf_vlan_info)) +
 821                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 822                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 823                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 824                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 825                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 826                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 827                          /* IFLA_VF_STATS_RX_PACKETS */
 828                          nla_total_size_64bit(sizeof(__u64)) +
 829                          /* IFLA_VF_STATS_TX_PACKETS */
 830                          nla_total_size_64bit(sizeof(__u64)) +
 831                          /* IFLA_VF_STATS_RX_BYTES */
 832                          nla_total_size_64bit(sizeof(__u64)) +
 833                          /* IFLA_VF_STATS_TX_BYTES */
 834                          nla_total_size_64bit(sizeof(__u64)) +
 835                          /* IFLA_VF_STATS_BROADCAST */
 836                          nla_total_size_64bit(sizeof(__u64)) +
 837                          /* IFLA_VF_STATS_MULTICAST */
 838                          nla_total_size_64bit(sizeof(__u64)) +
 839                          nla_total_size(sizeof(struct ifla_vf_trust)));
 840                 return size;
 841         } else
 842                 return 0;
 843 }
 844
 845 static size_t rtnl_port_size(const struct net_device *dev,
 846                              u32 ext_filter_mask)
 847 {
 848         size_t port_size = nla_total_size(4)            /* PORT_VF */
 849                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 850                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 851                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 852                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 853                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 854         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 855         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 856                 + port_size;
 857         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 858                 + port_size;
 859
 860         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 861             !(ext_filter_mask & RTEXT_FILTER_VF))
 862                 return 0;
 863         if (dev_num_vf(dev->dev.parent))
 864                 return port_self_size + vf_ports_size +
 865                         vf_port_size * dev_num_vf(dev->dev.parent);
 866         else
 867                 return port_self_size;
 868 }
 869
 870 static size_t rtnl_xdp_size(void)
 871 {
 872         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 873                           nla_total_size(1) +   /* XDP_ATTACHED */
 874                           nla_total_size(4);    /* XDP_PROG_ID */
 875
 876         return xdp_size;
 877 }
 878
 879 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 880                                      u32 ext_filter_mask)
 881 {
 882         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 883                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 884                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 885                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 886                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 887                + nla_total_size(sizeof(struct rtnl_link_stats))
 888                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 889                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 890                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 891                + nla_total_size(4) /* IFLA_TXQLEN */
 892                + nla_total_size(4) /* IFLA_WEIGHT */
 893                + nla_total_size(4) /* IFLA_MTU */
 894                + nla_total_size(4) /* IFLA_LINK */
 895                + nla_total_size(4) /* IFLA_MASTER */
 896                + nla_total_size(1) /* IFLA_CARRIER */
 897                + nla_total_size(4) /* IFLA_PROMISCUITY */
 898                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 899                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 900                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 901                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 902                + nla_total_size(1) /* IFLA_OPERSTATE */
 903                + nla_total_size(1) /* IFLA_LINKMODE */
 904                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 905                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 906                + nla_total_size(4) /* IFLA_GROUP */
 907                + nla_total_size(ext_filter_mask
 908                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 909                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 910                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 911                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 912                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 913                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 914                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 915                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 916                + rtnl_xdp_size() /* IFLA_XDP */
 917                + nla_total_size(4)  /* IFLA_EVENT */
 918                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
 919                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 920
 921 }
 922
 923 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 924 {
 925         struct nlattr *vf_ports;
 926         struct nlattr *vf_port;
 927         int vf;
 928         int err;
 929
 930         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 931         if (!vf_ports)
 932                 return -EMSGSIZE;
 933
 934         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 935                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 936                 if (!vf_port)
 937                         goto nla_put_failure;
 938                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 939                         goto nla_put_failure;
 940                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 941                 if (err == -EMSGSIZE)
 942                         goto nla_put_failure;
 943                 if (err) {
 944                         nla_nest_cancel(skb, vf_port);
 945                         continue;
 946                 }
 947                 nla_nest_end(skb, vf_port);
 948         }
 949
 950         nla_nest_end(skb, vf_ports);
 951
 952         return 0;
 953
 954 nla_put_failure:
 955         nla_nest_cancel(skb, vf_ports);
 956         return -EMSGSIZE;
 957 }
 958
 959 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 960 {
 961         struct nlattr *port_self;
 962         int err;
 963
 964         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 965         if (!port_self)
 966                 return -EMSGSIZE;
 967
 968         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 969         if (err) {
 970                 nla_nest_cancel(skb, port_self);
 971                 return (err == -EMSGSIZE) ? err : 0;
 972         }
 973
 974         nla_nest_end(skb, port_self);
 975
 976         return 0;
 977 }
 978
 979 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 980                           u32 ext_filter_mask)
 981 {
 982         int err;
 983
 984         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 985             !(ext_filter_mask & RTEXT_FILTER_VF))
 986                 return 0;
 987
 988         err = rtnl_port_self_fill(skb, dev);
 989         if (err)
 990                 return err;
 991
 992         if (dev_num_vf(dev->dev.parent)) {
 993                 err = rtnl_vf_ports_fill(skb, dev);
 994                 if (err)
 995                         return err;
 996         }
 997
 998         return 0;
 999 }
1000
1001 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1002 {
1003         int err;
1004         struct netdev_phys_item_id ppid;
1005
1006         err = dev_get_phys_port_id(dev, &ppid);
1007         if (err) {
1008                 if (err == -EOPNOTSUPP)
1009                         return 0;
1010                 return err;
1011         }
1012
1013         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1014                 return -EMSGSIZE;
1015
1016         return 0;
1017 }
1018
1019 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1020 {
1021         char name[IFNAMSIZ];
1022         int err;
1023
1024         err = dev_get_phys_port_name(dev, name, sizeof(name));
1025         if (err) {
1026                 if (err == -EOPNOTSUPP)
1027                         return 0;
1028                 return err;
1029         }
1030
1031         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1032                 return -EMSGSIZE;
1033
1034         return 0;
1035 }
1036
1037 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1038 {
1039         int err;
1040         struct switchdev_attr attr = {
1041                 .orig_dev = dev,
1042                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1043                 .flags = SWITCHDEV_F_NO_RECURSE,
1044         };
1045
1046         err = switchdev_port_attr_get(dev, &attr);
1047         if (err) {
1048                 if (err == -EOPNOTSUPP)
1049                         return 0;
1050                 return err;
1051         }
1052
1053         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1054                     attr.u.ppid.id))
1055                 return -EMSGSIZE;
1056
1057         return 0;
1058 }
1059
1060 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1061                                               struct net_device *dev)
1062 {
1063         struct rtnl_link_stats64 *sp;
1064         struct nlattr *attr;
1065
1066         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1067                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1068         if (!attr)
1069                 return -EMSGSIZE;
1070
1071         sp = nla_data(attr);
1072         dev_get_stats(dev, sp);
1073
1074         attr = nla_reserve(skb, IFLA_STATS,
1075                            sizeof(struct rtnl_link_stats));
1076         if (!attr)
1077                 return -EMSGSIZE;
1078
1079         copy_rtnl_link_stats(nla_data(attr), sp);
1080
1081         return 0;
1082 }
1083
1084 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1085                                                struct net_device *dev,
1086                                                int vfs_num,
1087                                                struct nlattr *vfinfo)
1088 {
1089         struct ifla_vf_rss_query_en vf_rss_query_en;
1090         struct nlattr *vf, *vfstats, *vfvlanlist;
1091         struct ifla_vf_link_state vf_linkstate;
1092         struct ifla_vf_vlan_info vf_vlan_info;
1093         struct ifla_vf_spoofchk vf_spoofchk;
1094         struct ifla_vf_tx_rate vf_tx_rate;
1095         struct ifla_vf_stats vf_stats;
1096         struct ifla_vf_trust vf_trust;
1097         struct ifla_vf_vlan vf_vlan;
1098         struct ifla_vf_rate vf_rate;
1099         struct ifla_vf_mac vf_mac;
1100         struct ifla_vf_info ivi;
1101
1102         memset(&ivi, 0, sizeof(ivi));
1103
1104         /* Not all SR-IOV capable drivers support the
1105          * spoofcheck and "RSS query enable" query.  Preset to
1106          * -1 so the user space tool can detect that the driver
1107          * didn't report anything.
1108          */
1109         ivi.spoofchk = -1;
1110         ivi.rss_query_en = -1;
1111         ivi.trusted = -1;
1112         /* The default value for VF link state is "auto"
1113          * IFLA_VF_LINK_STATE_AUTO which equals zero
1114          */
1115         ivi.linkstate = 0;
1116         /* VLAN Protocol by default is 802.1Q */
1117         ivi.vlan_proto = htons(ETH_P_8021Q);
1118         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1119                 return 0;
1120
1121         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1122
1123         vf_mac.vf =
1124                 vf_vlan.vf =
1125                 vf_vlan_info.vf =
1126                 vf_rate.vf =
1127                 vf_tx_rate.vf =
1128                 vf_spoofchk.vf =
1129                 vf_linkstate.vf =
1130                 vf_rss_query_en.vf =
1131                 vf_trust.vf = ivi.vf;
1132
1133         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1134         vf_vlan.vlan = ivi.vlan;
1135         vf_vlan.qos = ivi.qos;
1136         vf_vlan_info.vlan = ivi.vlan;
1137         vf_vlan_info.qos = ivi.qos;
1138         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1139         vf_tx_rate.rate = ivi.max_tx_rate;
1140         vf_rate.min_tx_rate = ivi.min_tx_rate;
1141         vf_rate.max_tx_rate = ivi.max_tx_rate;
1142         vf_spoofchk.setting = ivi.spoofchk;
1143         vf_linkstate.link_state = ivi.linkstate;
1144         vf_rss_query_en.setting = ivi.rss_query_en;
1145         vf_trust.setting = ivi.trusted;
1146         vf = nla_nest_start(skb, IFLA_VF_INFO);
1147         if (!vf)
1148                 goto nla_put_vfinfo_failure;
1149         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1150             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1151             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1152                     &vf_rate) ||
1153             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1154                     &vf_tx_rate) ||
1155             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1156                     &vf_spoofchk) ||
1157             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1158                     &vf_linkstate) ||
1159             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1160                     sizeof(vf_rss_query_en),
1161                     &vf_rss_query_en) ||
1162             nla_put(skb, IFLA_VF_TRUST,
1163                     sizeof(vf_trust), &vf_trust))
1164                 goto nla_put_vf_failure;
1165         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1166         if (!vfvlanlist)
1167                 goto nla_put_vf_failure;
1168         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1169                     &vf_vlan_info)) {
1170                 nla_nest_cancel(skb, vfvlanlist);
1171                 goto nla_put_vf_failure;
1172         }
1173         nla_nest_end(skb, vfvlanlist);
1174         memset(&vf_stats, 0, sizeof(vf_stats));
1175         if (dev->netdev_ops->ndo_get_vf_stats)
1176                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1177                                                 &vf_stats);
1178         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1179         if (!vfstats)
1180                 goto nla_put_vf_failure;
1181         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1182                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1183             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1184                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1185             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1186                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1187             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1188                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1189             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1190                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1191             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1192                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1193                 nla_nest_cancel(skb, vfstats);
1194                 goto nla_put_vf_failure;
1195         }
1196         nla_nest_end(skb, vfstats);
1197         nla_nest_end(skb, vf);
1198         return 0;
1199
1200 nla_put_vf_failure:
1201         nla_nest_cancel(skb, vf);
1202 nla_put_vfinfo_failure:
1203         nla_nest_cancel(skb, vfinfo);
1204         return -EMSGSIZE;
1205 }
1206
1207 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1208                                            struct net_device *dev,
1209                                            u32 ext_filter_mask)
1210 {
1211         struct nlattr *vfinfo;
1212         int i, num_vfs;
1213
1214         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1215                 return 0;
1216
1217         num_vfs = dev_num_vf(dev->dev.parent);
1218         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1219                 return -EMSGSIZE;
1220
1221         if (!dev->netdev_ops->ndo_get_vf_config)
1222                 return 0;
1223
1224         vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1225         if (!vfinfo)
1226                 return -EMSGSIZE;
1227
1228         for (i = 0; i < num_vfs; i++) {
1229                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1230                         return -EMSGSIZE;
1231         }
1232
1233         nla_nest_end(skb, vfinfo);
1234         return 0;
1235 }
1236
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1238 {
1239         struct rtnl_link_ifmap map;
1240
1241         memset(&map, 0, sizeof(map));
1242         map.mem_start   = dev->mem_start;
1243         map.mem_end     = dev->mem_end;
1244         map.base_addr   = dev->base_addr;
1245         map.irq         = dev->irq;
1246         map.dma         = dev->dma;
1247         map.port        = dev->if_port;
1248
1249         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1250                 return -EMSGSIZE;
1251
1252         return 0;
1253 }
1254
1255 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1256 {
1257         const struct net_device_ops *ops = dev->netdev_ops;
1258         const struct bpf_prog *generic_xdp_prog;
1259
1260         ASSERT_RTNL();
1261
1262         *prog_id = 0;
1263         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1264         if (generic_xdp_prog) {
1265                 *prog_id = generic_xdp_prog->aux->id;
1266                 return XDP_ATTACHED_SKB;
1267         }
1268         if (!ops->ndo_xdp)
1269                 return XDP_ATTACHED_NONE;
1270
1271         return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1272 }
1273
1274 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1275 {
1276         struct nlattr *xdp;
1277         u32 prog_id;
1278         int err;
1279
1280         xdp = nla_nest_start(skb, IFLA_XDP);
1281         if (!xdp)
1282                 return -EMSGSIZE;
1283
1284         err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1285                          rtnl_xdp_attached_mode(dev, &prog_id));
1286         if (err)
1287                 goto err_cancel;
1288
1289         if (prog_id) {
1290                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1291                 if (err)
1292                         goto err_cancel;
1293         }
1294
1295         nla_nest_end(skb, xdp);
1296         return 0;
1297
1298 err_cancel:
1299         nla_nest_cancel(skb, xdp);
1300         return err;
1301 }
1302
1303 static u32 rtnl_get_event(unsigned long event)
1304 {
1305         u32 rtnl_event_type = IFLA_EVENT_NONE;
1306
1307         switch (event) {
1308         case NETDEV_REBOOT:
1309                 rtnl_event_type = IFLA_EVENT_REBOOT;
1310                 break;
1311         case NETDEV_FEAT_CHANGE:
1312                 rtnl_event_type = IFLA_EVENT_FEATURES;
1313                 break;
1314         case NETDEV_BONDING_FAILOVER:
1315                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1316                 break;
1317         case NETDEV_NOTIFY_PEERS:
1318                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1319                 break;
1320         case NETDEV_RESEND_IGMP:
1321                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1322                 break;
1323         case NETDEV_CHANGEINFODATA:
1324                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1325                 break;
1326         default:
1327                 break;
1328         }
1329
1330         return rtnl_event_type;
1331 }
1332
1333 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1334 {
1335         const struct net_device *upper_dev;
1336         int ret = 0;
1337
1338         rcu_read_lock();
1339
1340         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1341         if (upper_dev)
1342                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1343
1344         rcu_read_unlock();
1345         return ret;
1346 }
1347
1348 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1349 {
1350         int ifindex = dev_get_iflink(dev);
1351
1352         if (dev->ifindex == ifindex)
1353                 return 0;
1354
1355         return nla_put_u32(skb, IFLA_LINK, ifindex);
1356 }
1357
1358 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1359                                               struct net_device *dev)
1360 {
1361         char buf[IFALIASZ];
1362         int ret;
1363
1364         ret = dev_get_alias(dev, buf, sizeof(buf));
1365         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1366 }
1367
1368 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1369                                   const struct net_device *dev)
1370 {
1371         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1372                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1373
1374                 if (!net_eq(dev_net(dev), link_net)) {
1375                         int id = peernet2id_alloc(dev_net(dev), link_net);
1376
1377                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1378                                 return -EMSGSIZE;
1379                 }
1380         }
1381
1382         return 0;
1383 }
1384
1385 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1386                             int type, u32 pid, u32 seq, u32 change,
1387                             unsigned int flags, u32 ext_filter_mask,
1388                             u32 event, int *new_nsid)
1389 {
1390         struct ifinfomsg *ifm;
1391         struct nlmsghdr *nlh;
1392         struct nlattr *af_spec;
1393         struct rtnl_af_ops *af_ops;
1394
1395         ASSERT_RTNL();
1396         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1397         if (nlh == NULL)
1398                 return -EMSGSIZE;
1399
1400         ifm = nlmsg_data(nlh);
1401         ifm->ifi_family = AF_UNSPEC;
1402         ifm->__ifi_pad = 0;
1403         ifm->ifi_type = dev->type;
1404         ifm->ifi_index = dev->ifindex;
1405         ifm->ifi_flags = dev_get_flags(dev);
1406         ifm->ifi_change = change;
1407
1408         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1409             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1410             nla_put_u8(skb, IFLA_OPERSTATE,
1411                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1412             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1413             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1414             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1415             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1416             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1417             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1418             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1419 #ifdef CONFIG_RPS
1420             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1421 #endif
1422             nla_put_iflink(skb, dev) ||
1423             put_master_ifindex(skb, dev) ||
1424             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1425             (dev->qdisc &&
1426              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1427             nla_put_ifalias(skb, dev) ||
1428             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1429                         atomic_read(&dev->carrier_changes)) ||
1430             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1431                 goto nla_put_failure;
1432
1433         if (event != IFLA_EVENT_NONE) {
1434                 if (nla_put_u32(skb, IFLA_EVENT, event))
1435                         goto nla_put_failure;
1436         }
1437
1438         if (rtnl_fill_link_ifmap(skb, dev))
1439                 goto nla_put_failure;
1440
1441         if (dev->addr_len) {
1442                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1443                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1444                         goto nla_put_failure;
1445         }
1446
1447         if (rtnl_phys_port_id_fill(skb, dev))
1448                 goto nla_put_failure;
1449
1450         if (rtnl_phys_port_name_fill(skb, dev))
1451                 goto nla_put_failure;
1452
1453         if (rtnl_phys_switch_id_fill(skb, dev))
1454                 goto nla_put_failure;
1455
1456         if (rtnl_fill_stats(skb, dev))
1457                 goto nla_put_failure;
1458
1459         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1460                 goto nla_put_failure;
1461
1462         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1463                 goto nla_put_failure;
1464
1465         if (rtnl_xdp_fill(skb, dev))
1466                 goto nla_put_failure;
1467
1468         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1469                 if (rtnl_link_fill(skb, dev) < 0)
1470                         goto nla_put_failure;
1471         }
1472
1473         if (rtnl_fill_link_netnsid(skb, dev))
1474                 goto nla_put_failure;
1475
1476         if (new_nsid &&
1477             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1478                 goto nla_put_failure;
1479
1480         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1481                 goto nla_put_failure;
1482
1483         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1484                 if (af_ops->fill_link_af) {
1485                         struct nlattr *af;
1486                         int err;
1487
1488                         if (!(af = nla_nest_start(skb, af_ops->family)))
1489                                 goto nla_put_failure;
1490
1491                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1492
1493                         /*
1494                          * Caller may return ENODATA to indicate that there
1495                          * was no data to be dumped. This is not an error, it
1496                          * means we should trim the attribute header and
1497                          * continue.
1498                          */
1499                         if (err == -ENODATA)
1500                                 nla_nest_cancel(skb, af);
1501                         else if (err < 0)
1502                                 goto nla_put_failure;
1503
1504                         nla_nest_end(skb, af);
1505                 }
1506         }
1507
1508         nla_nest_end(skb, af_spec);
1509
1510         nlmsg_end(skb, nlh);
1511         return 0;
1512
1513 nla_put_failure:
1514         nlmsg_cancel(skb, nlh);
1515         return -EMSGSIZE;
1516 }
1517
1518 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1519         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1520         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1521         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1522         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1523         [IFLA_MTU]              = { .type = NLA_U32 },
1524         [IFLA_LINK]             = { .type = NLA_U32 },
1525         [IFLA_MASTER]           = { .type = NLA_U32 },
1526         [IFLA_CARRIER]          = { .type = NLA_U8 },
1527         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1528         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1529         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1530         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1531         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1532         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1533         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1534         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1535         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1536         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1537         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1538         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1539         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1540         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1541         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1542         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1543         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1544         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1545         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1546         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1547         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1548         [IFLA_XDP]              = { .type = NLA_NESTED },
1549         [IFLA_EVENT]            = { .type = NLA_U32 },
1550         [IFLA_GROUP]            = { .type = NLA_U32 },
1551 };
1552
1553 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1554         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1555         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1556         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1557         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1558 };
1559
1560 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1561         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1562         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1563         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1564         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1565         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1566         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1567         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1568         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1569         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1570         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1571         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1572         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1573 };
1574
1575 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1576         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1577         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1578                                     .len = PORT_PROFILE_MAX },
1579         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1580                                       .len = PORT_UUID_MAX },
1581         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1582                                     .len = PORT_UUID_MAX },
1583         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1584         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1585
1586         /* Unused, but we need to keep it here since user space could
1587          * fill it. It's also broken with regard to NLA_BINARY use in
1588          * combination with structs.
1589          */
1590         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1591                                     .len = sizeof(struct ifla_port_vsi) },
1592 };
1593
1594 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1595         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1596         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1597         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1598         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1599 };
1600
1601 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1602 {
1603         const struct rtnl_link_ops *ops = NULL;
1604         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1605
1606         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1607                              ifla_info_policy, NULL) < 0)
1608                 return NULL;
1609
1610         if (linfo[IFLA_INFO_KIND]) {
1611                 char kind[MODULE_NAME_LEN];
1612
1613                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1614                 ops = rtnl_link_ops_get(kind);
1615         }
1616
1617         return ops;
1618 }
1619
1620 static bool link_master_filtered(struct net_device *dev, int master_idx)
1621 {
1622         struct net_device *master;
1623
1624         if (!master_idx)
1625                 return false;
1626
1627         master = netdev_master_upper_dev_get(dev);
1628         if (!master || master->ifindex != master_idx)
1629                 return true;
1630
1631         return false;
1632 }
1633
1634 static bool link_kind_filtered(const struct net_device *dev,
1635                                const struct rtnl_link_ops *kind_ops)
1636 {
1637         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1638                 return true;
1639
1640         return false;
1641 }
1642
1643 static bool link_dump_filtered(struct net_device *dev,
1644                                int master_idx,
1645                                const struct rtnl_link_ops *kind_ops)
1646 {
1647         if (link_master_filtered(dev, master_idx) ||
1648             link_kind_filtered(dev, kind_ops))
1649                 return true;
1650
1651         return false;
1652 }
1653
1654 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1655 {
1656         struct net *net = sock_net(skb->sk);
1657         int h, s_h;
1658         int idx = 0, s_idx;
1659         struct net_device *dev;
1660         struct hlist_head *head;
1661         struct nlattr *tb[IFLA_MAX+1];
1662         u32 ext_filter_mask = 0;
1663         const struct rtnl_link_ops *kind_ops = NULL;
1664         unsigned int flags = NLM_F_MULTI;
1665         int master_idx = 0;
1666         int err;
1667         int hdrlen;
1668
1669         s_h = cb->args[0];
1670         s_idx = cb->args[1];
1671
1672         /* A hack to preserve kernel<->userspace interface.
1673          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1674          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1675          * what iproute2 < v3.9.0 used.
1676          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1677          * attribute, its netlink message is shorter than struct ifinfomsg.
1678          */
1679         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1680                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1681
1682         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1683                         ifla_policy, NULL) >= 0) {
1684                 if (tb[IFLA_EXT_MASK])
1685                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1686
1687                 if (tb[IFLA_MASTER])
1688                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1689
1690                 if (tb[IFLA_LINKINFO])
1691                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1692
1693                 if (master_idx || kind_ops)
1694                         flags |= NLM_F_DUMP_FILTERED;
1695         }
1696
1697         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1698                 idx = 0;
1699                 head = &net->dev_index_head[h];
1700                 hlist_for_each_entry(dev, head, index_hlist) {
1701                         if (link_dump_filtered(dev, master_idx, kind_ops))
1702                                 goto cont;
1703                         if (idx < s_idx)
1704                                 goto cont;
1705                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1706                                                NETLINK_CB(cb->skb).portid,
1707                                                cb->nlh->nlmsg_seq, 0,
1708                                                flags,
1709                                                ext_filter_mask, 0, NULL);
1710
1711                         if (err < 0) {
1712                                 if (likely(skb->len))
1713                                         goto out;
1714
1715                                 goto out_err;
1716                         }
1717 cont:
1718                         idx++;
1719                 }
1720         }
1721 out:
1722         err = skb->len;
1723 out_err:
1724         cb->args[1] = idx;
1725         cb->args[0] = h;
1726         cb->seq = net->dev_base_seq;
1727         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1728
1729         return err;
1730 }
1731
1732 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1733                         struct netlink_ext_ack *exterr)
1734 {
1735         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1736 }
1737 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1738
1739 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1740 {
1741         struct net *net;
1742         /* Examine the link attributes and figure out which
1743          * network namespace we are talking about.
1744          */
1745         if (tb[IFLA_NET_NS_PID])
1746                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1747         else if (tb[IFLA_NET_NS_FD])
1748                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1749         else
1750                 net = get_net(src_net);
1751         return net;
1752 }
1753 EXPORT_SYMBOL(rtnl_link_get_net);
1754
1755 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1756 {
1757         if (dev) {
1758                 if (tb[IFLA_ADDRESS] &&
1759                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1760                         return -EINVAL;
1761
1762                 if (tb[IFLA_BROADCAST] &&
1763                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1764                         return -EINVAL;
1765         }
1766
1767         if (tb[IFLA_AF_SPEC]) {
1768                 struct nlattr *af;
1769                 int rem, err;
1770
1771                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1772                         const struct rtnl_af_ops *af_ops;
1773
1774                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1775                                 return -EAFNOSUPPORT;
1776
1777                         if (!af_ops->set_link_af)
1778                                 return -EOPNOTSUPP;
1779
1780                         if (af_ops->validate_link_af) {
1781                                 err = af_ops->validate_link_af(dev, af);
1782                                 if (err < 0)
1783                                         return err;
1784                         }
1785                 }
1786         }
1787
1788         return 0;
1789 }
1790
1791 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1792                                   int guid_type)
1793 {
1794         const struct net_device_ops *ops = dev->netdev_ops;
1795
1796         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1797 }
1798
1799 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1800 {
1801         if (dev->type != ARPHRD_INFINIBAND)
1802                 return -EOPNOTSUPP;
1803
1804         return handle_infiniband_guid(dev, ivt, guid_type);
1805 }
1806
1807 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1808 {
1809         const struct net_device_ops *ops = dev->netdev_ops;
1810         int err = -EINVAL;
1811
1812         if (tb[IFLA_VF_MAC]) {
1813                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1814
1815                 err = -EOPNOTSUPP;
1816                 if (ops->ndo_set_vf_mac)
1817                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1818                                                   ivm->mac);
1819                 if (err < 0)
1820                         return err;
1821         }
1822
1823         if (tb[IFLA_VF_VLAN]) {
1824                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1825
1826                 err = -EOPNOTSUPP;
1827                 if (ops->ndo_set_vf_vlan)
1828                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1829                                                    ivv->qos,
1830                                                    htons(ETH_P_8021Q));
1831                 if (err < 0)
1832                         return err;
1833         }
1834
1835         if (tb[IFLA_VF_VLAN_LIST]) {
1836                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1837                 struct nlattr *attr;
1838                 int rem, len = 0;
1839
1840                 err = -EOPNOTSUPP;
1841                 if (!ops->ndo_set_vf_vlan)
1842                         return err;
1843
1844                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1845                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1846                             nla_len(attr) < NLA_HDRLEN) {
1847                                 return -EINVAL;
1848                         }
1849                         if (len >= MAX_VLAN_LIST_LEN)
1850                                 return -EOPNOTSUPP;
1851                         ivvl[len] = nla_data(attr);
1852
1853                         len++;
1854                 }
1855                 if (len == 0)
1856                         return -EINVAL;
1857
1858                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1859                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1860                 if (err < 0)
1861                         return err;
1862         }
1863
1864         if (tb[IFLA_VF_TX_RATE]) {
1865                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1866                 struct ifla_vf_info ivf;
1867
1868                 err = -EOPNOTSUPP;
1869                 if (ops->ndo_get_vf_config)
1870                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1871                 if (err < 0)
1872                         return err;
1873
1874                 err = -EOPNOTSUPP;
1875                 if (ops->ndo_set_vf_rate)
1876                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1877                                                    ivf.min_tx_rate,
1878                                                    ivt->rate);
1879                 if (err < 0)
1880                         return err;
1881         }
1882
1883         if (tb[IFLA_VF_RATE]) {
1884                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1885
1886                 err = -EOPNOTSUPP;
1887                 if (ops->ndo_set_vf_rate)
1888                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1889                                                    ivt->min_tx_rate,
1890                                                    ivt->max_tx_rate);
1891                 if (err < 0)
1892                         return err;
1893         }
1894
1895         if (tb[IFLA_VF_SPOOFCHK]) {
1896                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1897
1898                 err = -EOPNOTSUPP;
1899                 if (ops->ndo_set_vf_spoofchk)
1900                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1901                                                        ivs->setting);
1902                 if (err < 0)
1903                         return err;
1904         }
1905
1906         if (tb[IFLA_VF_LINK_STATE]) {
1907                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1908
1909                 err = -EOPNOTSUPP;
1910                 if (ops->ndo_set_vf_link_state)
1911                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1912                                                          ivl->link_state);
1913                 if (err < 0)
1914                         return err;
1915         }
1916
1917         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1918                 struct ifla_vf_rss_query_en *ivrssq_en;
1919
1920                 err = -EOPNOTSUPP;
1921                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1922                 if (ops->ndo_set_vf_rss_query_en)
1923                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1924                                                            ivrssq_en->setting);
1925                 if (err < 0)
1926                         return err;
1927         }
1928
1929         if (tb[IFLA_VF_TRUST]) {
1930                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1931
1932                 err = -EOPNOTSUPP;
1933                 if (ops->ndo_set_vf_trust)
1934                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1935                 if (err < 0)
1936                         return err;
1937         }
1938
1939         if (tb[IFLA_VF_IB_NODE_GUID]) {
1940                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1941
1942                 if (!ops->ndo_set_vf_guid)
1943                         return -EOPNOTSUPP;
1944
1945                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1946         }
1947
1948         if (tb[IFLA_VF_IB_PORT_GUID]) {
1949                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1950
1951                 if (!ops->ndo_set_vf_guid)
1952                         return -EOPNOTSUPP;
1953
1954                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1955         }
1956
1957         return err;
1958 }
1959
1960 static int do_set_master(struct net_device *dev, int ifindex,
1961                          struct netlink_ext_ack *extack)
1962 {
1963         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1964         const struct net_device_ops *ops;
1965         int err;
1966
1967         if (upper_dev) {
1968                 if (upper_dev->ifindex == ifindex)
1969                         return 0;
1970                 ops = upper_dev->netdev_ops;
1971                 if (ops->ndo_del_slave) {
1972                         err = ops->ndo_del_slave(upper_dev, dev);
1973                         if (err)
1974                                 return err;
1975                 } else {
1976                         return -EOPNOTSUPP;
1977                 }
1978         }
1979
1980         if (ifindex) {
1981                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1982                 if (!upper_dev)
1983                         return -EINVAL;
1984                 ops = upper_dev->netdev_ops;
1985                 if (ops->ndo_add_slave) {
1986                         err = ops->ndo_add_slave(upper_dev, dev, extack);
1987                         if (err)
1988                                 return err;
1989                 } else {
1990                         return -EOPNOTSUPP;
1991                 }
1992         }
1993         return 0;
1994 }
1995
1996 #define DO_SETLINK_MODIFIED     0x01
1997 /* notify flag means notify + modified. */
1998 #define DO_SETLINK_NOTIFY       0x03
1999 static int do_setlink(const struct sk_buff *skb,
2000                       struct net_device *dev, struct ifinfomsg *ifm,
2001                       struct netlink_ext_ack *extack,
2002                       struct nlattr **tb, char *ifname, int status)
2003 {
2004         const struct net_device_ops *ops = dev->netdev_ops;
2005         int err;
2006
2007         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
2008                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
2009                 if (IS_ERR(net)) {
2010                         err = PTR_ERR(net);
2011                         goto errout;
2012                 }
2013                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
2014                         put_net(net);
2015                         err = -EPERM;
2016                         goto errout;
2017                 }
2018                 err = dev_change_net_namespace(dev, net, ifname);
2019                 put_net(net);
2020                 if (err)
2021                         goto errout;
2022                 status |= DO_SETLINK_MODIFIED;
2023         }
2024
2025         if (tb[IFLA_MAP]) {
2026                 struct rtnl_link_ifmap *u_map;
2027                 struct ifmap k_map;
2028
2029                 if (!ops->ndo_set_config) {
2030                         err = -EOPNOTSUPP;
2031                         goto errout;
2032                 }
2033
2034                 if (!netif_device_present(dev)) {
2035                         err = -ENODEV;
2036                         goto errout;
2037                 }
2038
2039                 u_map = nla_data(tb[IFLA_MAP]);
2040                 k_map.mem_start = (unsigned long) u_map->mem_start;
2041                 k_map.mem_end = (unsigned long) u_map->mem_end;
2042                 k_map.base_addr = (unsigned short) u_map->base_addr;
2043                 k_map.irq = (unsigned char) u_map->irq;
2044                 k_map.dma = (unsigned char) u_map->dma;
2045                 k_map.port = (unsigned char) u_map->port;
2046
2047                 err = ops->ndo_set_config(dev, &k_map);
2048                 if (err < 0)
2049                         goto errout;
2050
2051                 status |= DO_SETLINK_NOTIFY;
2052         }
2053
2054         if (tb[IFLA_ADDRESS]) {
2055                 struct sockaddr *sa;
2056                 int len;
2057
2058                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2059                                                   sizeof(*sa));
2060                 sa = kmalloc(len, GFP_KERNEL);
2061                 if (!sa) {
2062                         err = -ENOMEM;
2063                         goto errout;
2064                 }
2065                 sa->sa_family = dev->type;
2066                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2067                        dev->addr_len);
2068                 err = dev_set_mac_address(dev, sa);
2069                 kfree(sa);
2070                 if (err)
2071                         goto errout;
2072                 status |= DO_SETLINK_MODIFIED;
2073         }
2074
2075         if (tb[IFLA_MTU]) {
2076                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2077                 if (err < 0)
2078                         goto errout;
2079                 status |= DO_SETLINK_MODIFIED;
2080         }
2081
2082         if (tb[IFLA_GROUP]) {
2083                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2084                 status |= DO_SETLINK_NOTIFY;
2085         }
2086
2087         /*
2088          * Interface selected by interface index but interface
2089          * name provided implies that a name change has been
2090          * requested.
2091          */
2092         if (ifm->ifi_index > 0 && ifname[0]) {
2093                 err = dev_change_name(dev, ifname);
2094                 if (err < 0)
2095                         goto errout;
2096                 status |= DO_SETLINK_MODIFIED;
2097         }
2098
2099         if (tb[IFLA_IFALIAS]) {
2100                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2101                                     nla_len(tb[IFLA_IFALIAS]));
2102                 if (err < 0)
2103                         goto errout;
2104                 status |= DO_SETLINK_NOTIFY;
2105         }
2106
2107         if (tb[IFLA_BROADCAST]) {
2108                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2109                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2110         }
2111
2112         if (ifm->ifi_flags || ifm->ifi_change) {
2113                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2114                 if (err < 0)
2115                         goto errout;
2116         }
2117
2118         if (tb[IFLA_MASTER]) {
2119                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2120                 if (err)
2121                         goto errout;
2122                 status |= DO_SETLINK_MODIFIED;
2123         }
2124
2125         if (tb[IFLA_CARRIER]) {
2126                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2127                 if (err)
2128                         goto errout;
2129                 status |= DO_SETLINK_MODIFIED;
2130         }
2131
2132         if (tb[IFLA_TXQLEN]) {
2133                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2134                 unsigned int orig_len = dev->tx_queue_len;
2135
2136                 if (dev->tx_queue_len ^ value) {
2137                         dev->tx_queue_len = value;
2138                         err = call_netdevice_notifiers(
2139                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2140                         err = notifier_to_errno(err);
2141                         if (err) {
2142                                 dev->tx_queue_len = orig_len;
2143                                 goto errout;
2144                         }
2145                         status |= DO_SETLINK_NOTIFY;
2146                 }
2147         }
2148
2149         if (tb[IFLA_OPERSTATE])
2150                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2151
2152         if (tb[IFLA_LINKMODE]) {
2153                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2154
2155                 write_lock_bh(&dev_base_lock);
2156                 if (dev->link_mode ^ value)
2157                         status |= DO_SETLINK_NOTIFY;
2158                 dev->link_mode = value;
2159                 write_unlock_bh(&dev_base_lock);
2160         }
2161
2162         if (tb[IFLA_VFINFO_LIST]) {
2163                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2164                 struct nlattr *attr;
2165                 int rem;
2166
2167                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2168                         if (nla_type(attr) != IFLA_VF_INFO ||
2169                             nla_len(attr) < NLA_HDRLEN) {
2170                                 err = -EINVAL;
2171                                 goto errout;
2172                         }
2173                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2174                                                ifla_vf_policy, NULL);
2175                         if (err < 0)
2176                                 goto errout;
2177                         err = do_setvfinfo(dev, vfinfo);
2178                         if (err < 0)
2179                                 goto errout;
2180                         status |= DO_SETLINK_NOTIFY;
2181                 }
2182         }
2183         err = 0;
2184
2185         if (tb[IFLA_VF_PORTS]) {
2186                 struct nlattr *port[IFLA_PORT_MAX+1];
2187                 struct nlattr *attr;
2188                 int vf;
2189                 int rem;
2190
2191                 err = -EOPNOTSUPP;
2192                 if (!ops->ndo_set_vf_port)
2193                         goto errout;
2194
2195                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2196                         if (nla_type(attr) != IFLA_VF_PORT ||
2197                             nla_len(attr) < NLA_HDRLEN) {
2198                                 err = -EINVAL;
2199                                 goto errout;
2200                         }
2201                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2202                                                ifla_port_policy, NULL);
2203                         if (err < 0)
2204                                 goto errout;
2205                         if (!port[IFLA_PORT_VF]) {
2206                                 err = -EOPNOTSUPP;
2207                                 goto errout;
2208                         }
2209                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2210                         err = ops->ndo_set_vf_port(dev, vf, port);
2211                         if (err < 0)
2212                                 goto errout;
2213                         status |= DO_SETLINK_NOTIFY;
2214                 }
2215         }
2216         err = 0;
2217
2218         if (tb[IFLA_PORT_SELF]) {
2219                 struct nlattr *port[IFLA_PORT_MAX+1];
2220
2221                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2222                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2223                                        NULL);
2224                 if (err < 0)
2225                         goto errout;
2226
2227                 err = -EOPNOTSUPP;
2228                 if (ops->ndo_set_vf_port)
2229                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2230                 if (err < 0)
2231                         goto errout;
2232                 status |= DO_SETLINK_NOTIFY;
2233         }
2234
2235         if (tb[IFLA_AF_SPEC]) {
2236                 struct nlattr *af;
2237                 int rem;
2238
2239                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2240                         const struct rtnl_af_ops *af_ops;
2241
2242                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2243                                 BUG();
2244
2245                         err = af_ops->set_link_af(dev, af);
2246                         if (err < 0)
2247                                 goto errout;
2248
2249                         status |= DO_SETLINK_NOTIFY;
2250                 }
2251         }
2252         err = 0;
2253
2254         if (tb[IFLA_PROTO_DOWN]) {
2255                 err = dev_change_proto_down(dev,
2256                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2257                 if (err)
2258                         goto errout;
2259                 status |= DO_SETLINK_NOTIFY;
2260         }
2261
2262         if (tb[IFLA_XDP]) {
2263                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2264                 u32 xdp_flags = 0;
2265
2266                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2267                                        ifla_xdp_policy, NULL);
2268                 if (err < 0)
2269                         goto errout;
2270
2271                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2272                         err = -EINVAL;
2273                         goto errout;
2274                 }
2275
2276                 if (xdp[IFLA_XDP_FLAGS]) {
2277                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2278                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2279                                 err = -EINVAL;
2280                                 goto errout;
2281                         }
2282                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2283                                 err = -EINVAL;
2284                                 goto errout;
2285                         }
2286                 }
2287
2288                 if (xdp[IFLA_XDP_FD]) {
2289                         err = dev_change_xdp_fd(dev, extack,
2290                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2291                                                 xdp_flags);
2292                         if (err)
2293                                 goto errout;
2294                         status |= DO_SETLINK_NOTIFY;
2295                 }
2296         }
2297
2298 errout:
2299         if (status & DO_SETLINK_MODIFIED) {
2300                 if (status & DO_SETLINK_NOTIFY)
2301                         netdev_state_change(dev);
2302
2303                 if (err < 0)
2304                         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",
2305                                              dev->name);
2306         }
2307
2308         return err;
2309 }
2310
2311 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2312                         struct netlink_ext_ack *extack)
2313 {
2314         struct net *net = sock_net(skb->sk);
2315         struct ifinfomsg *ifm;
2316         struct net_device *dev;
2317         int err;
2318         struct nlattr *tb[IFLA_MAX+1];
2319         char ifname[IFNAMSIZ];
2320
2321         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2322                           extack);
2323         if (err < 0)
2324                 goto errout;
2325
2326         if (tb[IFLA_IFNAME])
2327                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2328         else
2329                 ifname[0] = '\0';
2330
2331         err = -EINVAL;
2332         ifm = nlmsg_data(nlh);
2333         if (ifm->ifi_index > 0)
2334                 dev = __dev_get_by_index(net, ifm->ifi_index);
2335         else if (tb[IFLA_IFNAME])
2336                 dev = __dev_get_by_name(net, ifname);
2337         else
2338                 goto errout;
2339
2340         if (dev == NULL) {
2341                 err = -ENODEV;
2342                 goto errout;
2343         }
2344
2345         err = validate_linkmsg(dev, tb);
2346         if (err < 0)
2347                 goto errout;
2348
2349         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2350 errout:
2351         return err;
2352 }
2353
2354 static int rtnl_group_dellink(const struct net *net, int group)
2355 {
2356         struct net_device *dev, *aux;
2357         LIST_HEAD(list_kill);
2358         bool found = false;
2359
2360         if (!group)
2361                 return -EPERM;
2362
2363         for_each_netdev(net, dev) {
2364                 if (dev->group == group) {
2365                         const struct rtnl_link_ops *ops;
2366
2367                         found = true;
2368                         ops = dev->rtnl_link_ops;
2369                         if (!ops || !ops->dellink)
2370                                 return -EOPNOTSUPP;
2371                 }
2372         }
2373
2374         if (!found)
2375                 return -ENODEV;
2376
2377         for_each_netdev_safe(net, dev, aux) {
2378                 if (dev->group == group) {
2379                         const struct rtnl_link_ops *ops;
2380
2381                         ops = dev->rtnl_link_ops;
2382                         ops->dellink(dev, &list_kill);
2383                 }
2384         }
2385         unregister_netdevice_many(&list_kill);
2386
2387         return 0;
2388 }
2389
2390 int rtnl_delete_link(struct net_device *dev)
2391 {
2392         const struct rtnl_link_ops *ops;
2393         LIST_HEAD(list_kill);
2394
2395         ops = dev->rtnl_link_ops;
2396         if (!ops || !ops->dellink)
2397                 return -EOPNOTSUPP;
2398
2399         ops->dellink(dev, &list_kill);
2400         unregister_netdevice_many(&list_kill);
2401
2402         return 0;
2403 }
2404 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2405
2406 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2407                         struct netlink_ext_ack *extack)
2408 {
2409         struct net *net = sock_net(skb->sk);
2410         struct net_device *dev;
2411         struct ifinfomsg *ifm;
2412         char ifname[IFNAMSIZ];
2413         struct nlattr *tb[IFLA_MAX+1];
2414         int err;
2415
2416         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2417         if (err < 0)
2418                 return err;
2419
2420         if (tb[IFLA_IFNAME])
2421                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2422
2423         ifm = nlmsg_data(nlh);
2424         if (ifm->ifi_index > 0)
2425                 dev = __dev_get_by_index(net, ifm->ifi_index);
2426         else if (tb[IFLA_IFNAME])
2427                 dev = __dev_get_by_name(net, ifname);
2428         else if (tb[IFLA_GROUP])
2429                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2430         else
2431                 return -EINVAL;
2432
2433         if (!dev)
2434                 return -ENODEV;
2435
2436         return rtnl_delete_link(dev);
2437 }
2438
2439 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2440 {
2441         unsigned int old_flags;
2442         int err;
2443
2444         old_flags = dev->flags;
2445         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2446                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2447                 if (err < 0)
2448                         return err;
2449         }
2450
2451         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2452
2453         __dev_notify_flags(dev, old_flags, ~0U);
2454         return 0;
2455 }
2456 EXPORT_SYMBOL(rtnl_configure_link);
2457
2458 struct net_device *rtnl_create_link(struct net *net,
2459         const char *ifname, unsigned char name_assign_type,
2460         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2461 {
2462         struct net_device *dev;
2463         unsigned int num_tx_queues = 1;
2464         unsigned int num_rx_queues = 1;
2465
2466         if (tb[IFLA_NUM_TX_QUEUES])
2467                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2468         else if (ops->get_num_tx_queues)
2469                 num_tx_queues = ops->get_num_tx_queues();
2470
2471         if (tb[IFLA_NUM_RX_QUEUES])
2472                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2473         else if (ops->get_num_rx_queues)
2474                 num_rx_queues = ops->get_num_rx_queues();
2475
2476         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2477                                ops->setup, num_tx_queues, num_rx_queues);
2478         if (!dev)
2479                 return ERR_PTR(-ENOMEM);
2480
2481         dev_net_set(dev, net);
2482         dev->rtnl_link_ops = ops;
2483         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2484
2485         if (tb[IFLA_MTU])
2486                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2487         if (tb[IFLA_ADDRESS]) {
2488                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2489                                 nla_len(tb[IFLA_ADDRESS]));
2490                 dev->addr_assign_type = NET_ADDR_SET;
2491         }
2492         if (tb[IFLA_BROADCAST])
2493                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2494                                 nla_len(tb[IFLA_BROADCAST]));
2495         if (tb[IFLA_TXQLEN])
2496                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2497         if (tb[IFLA_OPERSTATE])
2498                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2499         if (tb[IFLA_LINKMODE])
2500                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2501         if (tb[IFLA_GROUP])
2502                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2503
2504         return dev;
2505 }
2506 EXPORT_SYMBOL(rtnl_create_link);
2507
2508 static int rtnl_group_changelink(const struct sk_buff *skb,
2509                 struct net *net, int group,
2510                 struct ifinfomsg *ifm,
2511                 struct netlink_ext_ack *extack,
2512                 struct nlattr **tb)
2513 {
2514         struct net_device *dev, *aux;
2515         int err;
2516
2517         for_each_netdev_safe(net, dev, aux) {
2518                 if (dev->group == group) {
2519                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2520                         if (err < 0)
2521                                 return err;
2522                 }
2523         }
2524
2525         return 0;
2526 }
2527
2528 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2529                         struct netlink_ext_ack *extack)
2530 {
2531         struct net *net = sock_net(skb->sk);
2532         const struct rtnl_link_ops *ops;
2533         const struct rtnl_link_ops *m_ops = NULL;
2534         struct net_device *dev;
2535         struct net_device *master_dev = NULL;
2536         struct ifinfomsg *ifm;
2537         char kind[MODULE_NAME_LEN];
2538         char ifname[IFNAMSIZ];
2539         struct nlattr *tb[IFLA_MAX+1];
2540         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2541         unsigned char name_assign_type = NET_NAME_USER;
2542         int err;
2543
2544 #ifdef CONFIG_MODULES
2545 replay:
2546 #endif
2547         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2548         if (err < 0)
2549                 return err;
2550
2551         if (tb[IFLA_IFNAME])
2552                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2553         else
2554                 ifname[0] = '\0';
2555
2556         ifm = nlmsg_data(nlh);
2557         if (ifm->ifi_index > 0)
2558                 dev = __dev_get_by_index(net, ifm->ifi_index);
2559         else {
2560                 if (ifname[0])
2561                         dev = __dev_get_by_name(net, ifname);
2562                 else
2563                         dev = NULL;
2564         }
2565
2566         if (dev) {
2567                 master_dev = netdev_master_upper_dev_get(dev);
2568                 if (master_dev)
2569                         m_ops = master_dev->rtnl_link_ops;
2570         }
2571
2572         err = validate_linkmsg(dev, tb);
2573         if (err < 0)
2574                 return err;
2575
2576         if (tb[IFLA_LINKINFO]) {
2577                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2578                                        tb[IFLA_LINKINFO], ifla_info_policy,
2579                                        NULL);
2580                 if (err < 0)
2581                         return err;
2582         } else
2583                 memset(linkinfo, 0, sizeof(linkinfo));
2584
2585         if (linkinfo[IFLA_INFO_KIND]) {
2586                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2587                 ops = rtnl_link_ops_get(kind);
2588         } else {
2589                 kind[0] = '\0';
2590                 ops = NULL;
2591         }
2592
2593         if (1) {
2594                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2595                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2596                 struct nlattr **data = NULL;
2597                 struct nlattr **slave_data = NULL;
2598                 struct net *dest_net, *link_net = NULL;
2599
2600                 if (ops) {
2601                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2602                                 err = nla_parse_nested(attr, ops->maxtype,
2603                                                        linkinfo[IFLA_INFO_DATA],
2604                                                        ops->policy, NULL);
2605                                 if (err < 0)
2606                                         return err;
2607                                 data = attr;
2608                         }
2609                         if (ops->validate) {
2610                                 err = ops->validate(tb, data, extack);
2611                                 if (err < 0)
2612                                         return err;
2613                         }
2614                 }
2615
2616                 if (m_ops) {
2617                         if (m_ops->slave_maxtype &&
2618                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2619                                 err = nla_parse_nested(slave_attr,
2620                                                        m_ops->slave_maxtype,
2621                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2622                                                        m_ops->slave_policy,
2623                                                        NULL);
2624                                 if (err < 0)
2625                                         return err;
2626                                 slave_data = slave_attr;
2627                         }
2628                 }
2629
2630                 if (dev) {
2631                         int status = 0;
2632
2633                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2634                                 return -EEXIST;
2635                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2636                                 return -EOPNOTSUPP;
2637
2638                         if (linkinfo[IFLA_INFO_DATA]) {
2639                                 if (!ops || ops != dev->rtnl_link_ops ||
2640                                     !ops->changelink)
2641                                         return -EOPNOTSUPP;
2642
2643                                 err = ops->changelink(dev, tb, data, extack);
2644                                 if (err < 0)
2645                                         return err;
2646                                 status |= DO_SETLINK_NOTIFY;
2647                         }
2648
2649                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2650                                 if (!m_ops || !m_ops->slave_changelink)
2651                                         return -EOPNOTSUPP;
2652
2653                                 err = m_ops->slave_changelink(master_dev, dev,
2654                                                               tb, slave_data,
2655                                                               extack);
2656                                 if (err < 0)
2657                                         return err;
2658                                 status |= DO_SETLINK_NOTIFY;
2659                         }
2660
2661                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2662                                           status);
2663                 }
2664
2665                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2666                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2667                                 return rtnl_group_changelink(skb, net,
2668                                                 nla_get_u32(tb[IFLA_GROUP]),
2669                                                 ifm, extack, tb);
2670                         return -ENODEV;
2671                 }
2672
2673                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2674                         return -EOPNOTSUPP;
2675
2676                 if (!ops) {
2677 #ifdef CONFIG_MODULES
2678                         if (kind[0]) {
2679                                 __rtnl_unlock();
2680                                 request_module("rtnl-link-%s", kind);
2681                                 rtnl_lock();
2682                                 ops = rtnl_link_ops_get(kind);
2683                                 if (ops)
2684                                         goto replay;
2685                         }
2686 #endif
2687                         return -EOPNOTSUPP;
2688                 }
2689
2690                 if (!ops->setup)
2691                         return -EOPNOTSUPP;
2692
2693                 if (!ifname[0]) {
2694                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2695                         name_assign_type = NET_NAME_ENUM;
2696                 }
2697
2698                 dest_net = rtnl_link_get_net(net, tb);
2699                 if (IS_ERR(dest_net))
2700                         return PTR_ERR(dest_net);
2701
2702                 err = -EPERM;
2703                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2704                         goto out;
2705
2706                 if (tb[IFLA_LINK_NETNSID]) {
2707                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2708
2709                         link_net = get_net_ns_by_id(dest_net, id);
2710                         if (!link_net) {
2711                                 err =  -EINVAL;
2712                                 goto out;
2713                         }
2714                         err = -EPERM;
2715                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2716                                 goto out;
2717                 }
2718
2719                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2720                                        name_assign_type, ops, tb);
2721                 if (IS_ERR(dev)) {
2722                         err = PTR_ERR(dev);
2723                         goto out;
2724                 }
2725
2726                 dev->ifindex = ifm->ifi_index;
2727
2728                 if (ops->newlink) {
2729                         err = ops->newlink(link_net ? : net, dev, tb, data,
2730                                            extack);
2731                         /* Drivers should call free_netdev() in ->destructor
2732                          * and unregister it on failure after registration
2733                          * so that device could be finally freed in rtnl_unlock.
2734                          */
2735                         if (err < 0) {
2736                                 /* If device is not registered at all, free it now */
2737                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2738                                         free_netdev(dev);
2739                                 goto out;
2740                         }
2741                 } else {
2742                         err = register_netdevice(dev);
2743                         if (err < 0) {
2744                                 free_netdev(dev);
2745                                 goto out;
2746                         }
2747                 }
2748                 err = rtnl_configure_link(dev, ifm);
2749                 if (err < 0)
2750                         goto out_unregister;
2751                 if (link_net) {
2752                         err = dev_change_net_namespace(dev, dest_net, ifname);
2753                         if (err < 0)
2754                                 goto out_unregister;
2755                 }
2756                 if (tb[IFLA_MASTER]) {
2757                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
2758                                             extack);
2759                         if (err)
2760                                 goto out_unregister;
2761                 }
2762 out:
2763                 if (link_net)
2764                         put_net(link_net);
2765                 put_net(dest_net);
2766                 return err;
2767 out_unregister:
2768                 if (ops->newlink) {
2769                         LIST_HEAD(list_kill);
2770
2771                         ops->dellink(dev, &list_kill);
2772                         unregister_netdevice_many(&list_kill);
2773                 } else {
2774                         unregister_netdevice(dev);
2775                 }
2776                 goto out;
2777         }
2778 }
2779
2780 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2781                         struct netlink_ext_ack *extack)
2782 {
2783         struct net *net = sock_net(skb->sk);
2784         struct ifinfomsg *ifm;
2785         char ifname[IFNAMSIZ];
2786         struct nlattr *tb[IFLA_MAX+1];
2787         struct net_device *dev = NULL;
2788         struct sk_buff *nskb;
2789         int err;
2790         u32 ext_filter_mask = 0;
2791
2792         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2793         if (err < 0)
2794                 return err;
2795
2796         if (tb[IFLA_IFNAME])
2797                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2798
2799         if (tb[IFLA_EXT_MASK])
2800                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2801
2802         ifm = nlmsg_data(nlh);
2803         if (ifm->ifi_index > 0)
2804                 dev = __dev_get_by_index(net, ifm->ifi_index);
2805         else if (tb[IFLA_IFNAME])
2806                 dev = __dev_get_by_name(net, ifname);
2807         else
2808                 return -EINVAL;
2809
2810         if (dev == NULL)
2811                 return -ENODEV;
2812
2813         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2814         if (nskb == NULL)
2815                 return -ENOBUFS;
2816
2817         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2818                                nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0, NULL);
2819         if (err < 0) {
2820                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2821                 WARN_ON(err == -EMSGSIZE);
2822                 kfree_skb(nskb);
2823         } else
2824                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2825
2826         return err;
2827 }
2828
2829 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2830 {
2831         struct net *net = sock_net(skb->sk);
2832         struct net_device *dev;
2833         struct nlattr *tb[IFLA_MAX+1];
2834         u32 ext_filter_mask = 0;
2835         u16 min_ifinfo_dump_size = 0;
2836         int hdrlen;
2837
2838         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2839         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2840                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2841
2842         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2843                 if (tb[IFLA_EXT_MASK])
2844                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2845         }
2846
2847         if (!ext_filter_mask)
2848                 return NLMSG_GOODSIZE;
2849         /*
2850          * traverse the list of net devices and compute the minimum
2851          * buffer size based upon the filter mask.
2852          */
2853         rcu_read_lock();
2854         for_each_netdev_rcu(net, dev) {
2855                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2856                                              if_nlmsg_size(dev,
2857                                                            ext_filter_mask));
2858         }
2859         rcu_read_unlock();
2860
2861         return nlmsg_total_size(min_ifinfo_dump_size);
2862 }
2863
2864 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2865 {
2866         int idx;
2867         int s_idx = cb->family;
2868
2869         if (s_idx == 0)
2870                 s_idx = 1;
2871
2872         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2873                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2874                 struct rtnl_link *handlers;
2875                 rtnl_dumpit_func dumpit;
2876
2877                 if (idx < s_idx || idx == PF_PACKET)
2878                         continue;
2879
2880                 handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2881                 if (!handlers)
2882                         continue;
2883
2884                 dumpit = READ_ONCE(handlers[type].dumpit);
2885                 if (!dumpit)
2886                         continue;
2887
2888                 if (idx > s_idx) {
2889                         memset(&cb->args[0], 0, sizeof(cb->args));
2890                         cb->prev_seq = 0;
2891                         cb->seq = 0;
2892                 }
2893                 if (dumpit(skb, cb))
2894                         break;
2895         }
2896         cb->family = idx;
2897
2898         return skb->len;
2899 }
2900
2901 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2902                                        unsigned int change,
2903                                        u32 event, gfp_t flags, int *new_nsid)
2904 {
2905         struct net *net = dev_net(dev);
2906         struct sk_buff *skb;
2907         int err = -ENOBUFS;
2908         size_t if_info_size;
2909
2910         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2911         if (skb == NULL)
2912                 goto errout;
2913
2914         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event,
2915                                new_nsid);
2916         if (err < 0) {
2917                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2918                 WARN_ON(err == -EMSGSIZE);
2919                 kfree_skb(skb);
2920                 goto errout;
2921         }
2922         return skb;
2923 errout:
2924         if (err < 0)
2925                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2926         return NULL;
2927 }
2928
2929 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2930 {
2931         struct net *net = dev_net(dev);
2932
2933         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2934 }
2935
2936 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2937                                unsigned int change, u32 event,
2938                                gfp_t flags, int *new_nsid)
2939 {
2940         struct sk_buff *skb;
2941
2942         if (dev->reg_state != NETREG_REGISTERED)
2943                 return;
2944
2945         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid);
2946         if (skb)
2947                 rtmsg_ifinfo_send(skb, dev, flags);
2948 }
2949
2950 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2951                   gfp_t flags)
2952 {
2953         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, NULL);
2954 }
2955 EXPORT_SYMBOL(rtmsg_ifinfo);
2956
2957 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
2958                          gfp_t flags, int *new_nsid)
2959 {
2960         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
2961                            new_nsid);
2962 }
2963
2964 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2965                                    struct net_device *dev,
2966                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2967                                    int type, unsigned int flags,
2968                                    int nlflags, u16 ndm_state)
2969 {
2970         struct nlmsghdr *nlh;
2971         struct ndmsg *ndm;
2972
2973         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2974         if (!nlh)
2975                 return -EMSGSIZE;
2976
2977         ndm = nlmsg_data(nlh);
2978         ndm->ndm_family  = AF_BRIDGE;
2979         ndm->ndm_pad1    = 0;
2980         ndm->ndm_pad2    = 0;
2981         ndm->ndm_flags   = flags;
2982         ndm->ndm_type    = 0;
2983         ndm->ndm_ifindex = dev->ifindex;
2984         ndm->ndm_state   = ndm_state;
2985
2986         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2987                 goto nla_put_failure;
2988         if (vid)
2989                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2990                         goto nla_put_failure;
2991
2992         nlmsg_end(skb, nlh);
2993         return 0;
2994
2995 nla_put_failure:
2996         nlmsg_cancel(skb, nlh);
2997         return -EMSGSIZE;
2998 }
2999
3000 static inline size_t rtnl_fdb_nlmsg_size(void)
3001 {
3002         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3003                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3004                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3005                0;
3006 }
3007
3008 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3009                             u16 ndm_state)
3010 {
3011         struct net *net = dev_net(dev);
3012         struct sk_buff *skb;
3013         int err = -ENOBUFS;
3014
3015         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3016         if (!skb)
3017                 goto errout;
3018
3019         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3020                                       0, 0, type, NTF_SELF, 0, ndm_state);
3021         if (err < 0) {
3022                 kfree_skb(skb);
3023                 goto errout;
3024         }
3025
3026         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3027         return;
3028 errout:
3029         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3030 }
3031
3032 /**
3033  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3034  */
3035 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3036                      struct nlattr *tb[],
3037                      struct net_device *dev,
3038                      const unsigned char *addr, u16 vid,
3039                      u16 flags)
3040 {
3041         int err = -EINVAL;
3042
3043         /* If aging addresses are supported device will need to
3044          * implement its own handler for this.
3045          */
3046         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3047                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3048                 return err;
3049         }
3050
3051         if (vid) {
3052                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3053                 return err;
3054         }
3055
3056         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3057                 err = dev_uc_add_excl(dev, addr);
3058         else if (is_multicast_ether_addr(addr))
3059                 err = dev_mc_add_excl(dev, addr);
3060
3061         /* Only return duplicate errors if NLM_F_EXCL is set */
3062         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3063                 err = 0;
3064
3065         return err;
3066 }
3067 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3068
3069 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3070 {
3071         u16 vid = 0;
3072
3073         if (vlan_attr) {
3074                 if (nla_len(vlan_attr) != sizeof(u16)) {
3075                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3076                         return -EINVAL;
3077                 }
3078
3079                 vid = nla_get_u16(vlan_attr);
3080
3081                 if (!vid || vid >= VLAN_VID_MASK) {
3082                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3083                                 vid);
3084                         return -EINVAL;
3085                 }
3086         }
3087         *p_vid = vid;
3088         return 0;
3089 }
3090
3091 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3092                         struct netlink_ext_ack *extack)
3093 {
3094         struct net *net = sock_net(skb->sk);
3095         struct ndmsg *ndm;
3096         struct nlattr *tb[NDA_MAX+1];
3097         struct net_device *dev;
3098         u8 *addr;
3099         u16 vid;
3100         int err;
3101
3102         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3103         if (err < 0)
3104                 return err;
3105
3106         ndm = nlmsg_data(nlh);
3107         if (ndm->ndm_ifindex == 0) {
3108                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3109                 return -EINVAL;
3110         }
3111
3112         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3113         if (dev == NULL) {
3114                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3115                 return -ENODEV;
3116         }
3117
3118         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3119                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3120                 return -EINVAL;
3121         }
3122
3123         addr = nla_data(tb[NDA_LLADDR]);
3124
3125         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3126         if (err)
3127                 return err;
3128
3129         err = -EOPNOTSUPP;
3130
3131         /* Support fdb on master device the net/bridge default case */
3132         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3133             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3134                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3135                 const struct net_device_ops *ops = br_dev->netdev_ops;
3136
3137                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3138                                        nlh->nlmsg_flags);
3139                 if (err)
3140                         goto out;
3141                 else
3142                         ndm->ndm_flags &= ~NTF_MASTER;
3143         }
3144
3145         /* Embedded bridge, macvlan, and any other device support */
3146         if ((ndm->ndm_flags & NTF_SELF)) {
3147                 if (dev->netdev_ops->ndo_fdb_add)
3148                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3149                                                            vid,
3150                                                            nlh->nlmsg_flags);
3151                 else
3152                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3153                                                nlh->nlmsg_flags);
3154
3155                 if (!err) {
3156                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3157                                         ndm->ndm_state);
3158                         ndm->ndm_flags &= ~NTF_SELF;
3159                 }
3160         }
3161 out:
3162         return err;
3163 }
3164
3165 /**
3166  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3167  */
3168 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3169                      struct nlattr *tb[],
3170                      struct net_device *dev,
3171                      const unsigned char *addr, u16 vid)
3172 {
3173         int err = -EINVAL;
3174
3175         /* If aging addresses are supported device will need to
3176          * implement its own handler for this.
3177          */
3178         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3179                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3180                 return err;
3181         }
3182
3183         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3184                 err = dev_uc_del(dev, addr);
3185         else if (is_multicast_ether_addr(addr))
3186                 err = dev_mc_del(dev, addr);
3187
3188         return err;
3189 }
3190 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3191
3192 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3193                         struct netlink_ext_ack *extack)
3194 {
3195         struct net *net = sock_net(skb->sk);
3196         struct ndmsg *ndm;
3197         struct nlattr *tb[NDA_MAX+1];
3198         struct net_device *dev;
3199         int err = -EINVAL;
3200         __u8 *addr;
3201         u16 vid;
3202
3203         if (!netlink_capable(skb, CAP_NET_ADMIN))
3204                 return -EPERM;
3205
3206         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3207         if (err < 0)
3208                 return err;
3209
3210         ndm = nlmsg_data(nlh);
3211         if (ndm->ndm_ifindex == 0) {
3212                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3213                 return -EINVAL;
3214         }
3215
3216         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3217         if (dev == NULL) {
3218                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3219                 return -ENODEV;
3220         }
3221
3222         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3223                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3224                 return -EINVAL;
3225         }
3226
3227         addr = nla_data(tb[NDA_LLADDR]);
3228
3229         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3230         if (err)
3231                 return err;
3232
3233         err = -EOPNOTSUPP;
3234
3235         /* Support fdb on master device the net/bridge default case */
3236         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3237             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3238                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3239                 const struct net_device_ops *ops = br_dev->netdev_ops;
3240
3241                 if (ops->ndo_fdb_del)
3242                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3243
3244                 if (err)
3245                         goto out;
3246                 else
3247                         ndm->ndm_flags &= ~NTF_MASTER;
3248         }
3249
3250         /* Embedded bridge, macvlan, and any other device support */
3251         if (ndm->ndm_flags & NTF_SELF) {
3252                 if (dev->netdev_ops->ndo_fdb_del)
3253                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3254                                                            vid);
3255                 else
3256                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3257
3258                 if (!err) {
3259                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3260                                         ndm->ndm_state);
3261                         ndm->ndm_flags &= ~NTF_SELF;
3262                 }
3263         }
3264 out:
3265         return err;
3266 }
3267
3268 static int nlmsg_populate_fdb(struct sk_buff *skb,
3269                               struct netlink_callback *cb,
3270                               struct net_device *dev,
3271                               int *idx,
3272                               struct netdev_hw_addr_list *list)
3273 {
3274         struct netdev_hw_addr *ha;
3275         int err;
3276         u32 portid, seq;
3277
3278         portid = NETLINK_CB(cb->skb).portid;
3279         seq = cb->nlh->nlmsg_seq;
3280
3281         list_for_each_entry(ha, &list->list, list) {
3282                 if (*idx < cb->args[2])
3283                         goto skip;
3284
3285                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3286                                               portid, seq,
3287                                               RTM_NEWNEIGH, NTF_SELF,
3288                                               NLM_F_MULTI, NUD_PERMANENT);
3289                 if (err < 0)
3290                         return err;
3291 skip:
3292                 *idx += 1;
3293         }
3294         return 0;
3295 }
3296
3297 /**
3298  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3299  * @nlh: netlink message header
3300  * @dev: netdevice
3301  *
3302  * Default netdevice operation to dump the existing unicast address list.
3303  * Returns number of addresses from list put in skb.
3304  */
3305 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3306                       struct netlink_callback *cb,
3307                       struct net_device *dev,
3308                       struct net_device *filter_dev,
3309                       int *idx)
3310 {
3311         int err;
3312
3313         netif_addr_lock_bh(dev);
3314         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3315         if (err)
3316                 goto out;
3317         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3318 out:
3319         netif_addr_unlock_bh(dev);
3320         return err;
3321 }
3322 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3323
3324 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3325 {
3326         struct net_device *dev;
3327         struct nlattr *tb[IFLA_MAX+1];
3328         struct net_device *br_dev = NULL;
3329         const struct net_device_ops *ops = NULL;
3330         const struct net_device_ops *cops = NULL;
3331         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3332         struct net *net = sock_net(skb->sk);
3333         struct hlist_head *head;
3334         int brport_idx = 0;
3335         int br_idx = 0;
3336         int h, s_h;
3337         int idx = 0, s_idx;
3338         int err = 0;
3339         int fidx = 0;
3340
3341         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3342                           IFLA_MAX, ifla_policy, NULL);
3343         if (err < 0) {
3344                 return -EINVAL;
3345         } else if (err == 0) {
3346                 if (tb[IFLA_MASTER])
3347                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3348         }
3349
3350         brport_idx = ifm->ifi_index;
3351
3352         if (br_idx) {
3353                 br_dev = __dev_get_by_index(net, br_idx);
3354                 if (!br_dev)
3355                         return -ENODEV;
3356
3357                 ops = br_dev->netdev_ops;
3358         }
3359
3360         s_h = cb->args[0];
3361         s_idx = cb->args[1];
3362
3363         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3364                 idx = 0;
3365                 head = &net->dev_index_head[h];
3366                 hlist_for_each_entry(dev, head, index_hlist) {
3367
3368                         if (brport_idx && (dev->ifindex != brport_idx))
3369                                 continue;
3370
3371                         if (!br_idx) { /* user did not specify a specific bridge */
3372                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3373                                         br_dev = netdev_master_upper_dev_get(dev);
3374                                         cops = br_dev->netdev_ops;
3375                                 }
3376                         } else {
3377                                 if (dev != br_dev &&
3378                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3379                                         continue;
3380
3381                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3382                                     !(dev->priv_flags & IFF_EBRIDGE))
3383                                         continue;
3384                                 cops = ops;
3385                         }
3386
3387                         if (idx < s_idx)
3388                                 goto cont;
3389
3390                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3391                                 if (cops && cops->ndo_fdb_dump) {
3392                                         err = cops->ndo_fdb_dump(skb, cb,
3393                                                                 br_dev, dev,
3394                                                                 &fidx);
3395                                         if (err == -EMSGSIZE)
3396                                                 goto out;
3397                                 }
3398                         }
3399
3400                         if (dev->netdev_ops->ndo_fdb_dump)
3401                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3402                                                                     dev, NULL,
3403                                                                     &fidx);
3404                         else
3405                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3406                                                         &fidx);
3407                         if (err == -EMSGSIZE)
3408                                 goto out;
3409
3410                         cops = NULL;
3411
3412                         /* reset fdb offset to 0 for rest of the interfaces */
3413                         cb->args[2] = 0;
3414                         fidx = 0;
3415 cont:
3416                         idx++;
3417                 }
3418         }
3419
3420 out:
3421         cb->args[0] = h;
3422         cb->args[1] = idx;
3423         cb->args[2] = fidx;
3424
3425         return skb->len;
3426 }
3427
3428 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3429                                unsigned int attrnum, unsigned int flag)
3430 {
3431         if (mask & flag)
3432                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3433         return 0;
3434 }
3435
3436 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3437                             struct net_device *dev, u16 mode,
3438                             u32 flags, u32 mask, int nlflags,
3439                             u32 filter_mask,
3440                             int (*vlan_fill)(struct sk_buff *skb,
3441                                              struct net_device *dev,
3442                                              u32 filter_mask))
3443 {
3444         struct nlmsghdr *nlh;
3445         struct ifinfomsg *ifm;
3446         struct nlattr *br_afspec;
3447         struct nlattr *protinfo;
3448         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3449         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3450         int err = 0;
3451
3452         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3453         if (nlh == NULL)
3454                 return -EMSGSIZE;
3455
3456         ifm = nlmsg_data(nlh);
3457         ifm->ifi_family = AF_BRIDGE;
3458         ifm->__ifi_pad = 0;
3459         ifm->ifi_type = dev->type;
3460         ifm->ifi_index = dev->ifindex;
3461         ifm->ifi_flags = dev_get_flags(dev);
3462         ifm->ifi_change = 0;
3463
3464
3465         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3466             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3467             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3468             (br_dev &&
3469              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3470             (dev->addr_len &&
3471              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3472             (dev->ifindex != dev_get_iflink(dev) &&
3473              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3474                 goto nla_put_failure;
3475
3476         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3477         if (!br_afspec)
3478                 goto nla_put_failure;
3479
3480         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3481                 nla_nest_cancel(skb, br_afspec);
3482                 goto nla_put_failure;
3483         }
3484
3485         if (mode != BRIDGE_MODE_UNDEF) {
3486                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3487                         nla_nest_cancel(skb, br_afspec);
3488                         goto nla_put_failure;
3489                 }
3490         }
3491         if (vlan_fill) {
3492                 err = vlan_fill(skb, dev, filter_mask);
3493                 if (err) {
3494                         nla_nest_cancel(skb, br_afspec);
3495                         goto nla_put_failure;
3496                 }
3497         }
3498         nla_nest_end(skb, br_afspec);
3499
3500         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3501         if (!protinfo)
3502                 goto nla_put_failure;
3503
3504         if (brport_nla_put_flag(skb, flags, mask,
3505                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3506             brport_nla_put_flag(skb, flags, mask,
3507                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3508             brport_nla_put_flag(skb, flags, mask,
3509                                 IFLA_BRPORT_FAST_LEAVE,
3510                                 BR_MULTICAST_FAST_LEAVE) ||
3511             brport_nla_put_flag(skb, flags, mask,
3512                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3513             brport_nla_put_flag(skb, flags, mask,
3514                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3515             brport_nla_put_flag(skb, flags, mask,
3516                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3517             brport_nla_put_flag(skb, flags, mask,
3518                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3519             brport_nla_put_flag(skb, flags, mask,
3520                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3521                 nla_nest_cancel(skb, protinfo);
3522                 goto nla_put_failure;
3523         }
3524
3525         nla_nest_end(skb, protinfo);
3526
3527         nlmsg_end(skb, nlh);
3528         return 0;
3529 nla_put_failure:
3530         nlmsg_cancel(skb, nlh);
3531         return err ? err : -EMSGSIZE;
3532 }
3533 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3534
3535 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3536 {
3537         struct net *net = sock_net(skb->sk);
3538         struct net_device *dev;
3539         int idx = 0;
3540         u32 portid = NETLINK_CB(cb->skb).portid;
3541         u32 seq = cb->nlh->nlmsg_seq;
3542         u32 filter_mask = 0;
3543         int err;
3544
3545         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3546                 struct nlattr *extfilt;
3547
3548                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3549                                           IFLA_EXT_MASK);
3550                 if (extfilt) {
3551                         if (nla_len(extfilt) < sizeof(filter_mask))
3552                                 return -EINVAL;
3553
3554                         filter_mask = nla_get_u32(extfilt);
3555                 }
3556         }
3557
3558         rcu_read_lock();
3559         for_each_netdev_rcu(net, dev) {
3560                 const struct net_device_ops *ops = dev->netdev_ops;
3561                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3562
3563                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3564                         if (idx >= cb->args[0]) {
3565                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3566                                                 skb, portid, seq, dev,
3567                                                 filter_mask, NLM_F_MULTI);
3568                                 if (err < 0 && err != -EOPNOTSUPP) {
3569                                         if (likely(skb->len))
3570                                                 break;
3571
3572                                         goto out_err;
3573                                 }
3574                         }
3575                         idx++;
3576                 }
3577
3578                 if (ops->ndo_bridge_getlink) {
3579                         if (idx >= cb->args[0]) {
3580                                 err = ops->ndo_bridge_getlink(skb, portid,
3581                                                               seq, dev,
3582                                                               filter_mask,
3583                                                               NLM_F_MULTI);
3584                                 if (err < 0 && err != -EOPNOTSUPP) {
3585                                         if (likely(skb->len))
3586                                                 break;
3587
3588                                         goto out_err;
3589                                 }
3590                         }
3591                         idx++;
3592                 }
3593         }
3594         err = skb->len;
3595 out_err:
3596         rcu_read_unlock();
3597         cb->args[0] = idx;
3598
3599         return err;
3600 }
3601
3602 static inline size_t bridge_nlmsg_size(void)
3603 {
3604         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3605                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3606                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3607                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3608                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3609                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3610                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3611                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3612                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3613                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3614                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3615 }
3616
3617 static int rtnl_bridge_notify(struct net_device *dev)
3618 {
3619         struct net *net = dev_net(dev);
3620         struct sk_buff *skb;
3621         int err = -EOPNOTSUPP;
3622
3623         if (!dev->netdev_ops->ndo_bridge_getlink)
3624                 return 0;
3625
3626         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3627         if (!skb) {
3628                 err = -ENOMEM;
3629                 goto errout;
3630         }
3631
3632         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3633         if (err < 0)
3634                 goto errout;
3635
3636         if (!skb->len)
3637                 goto errout;
3638
3639         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3640         return 0;
3641 errout:
3642         WARN_ON(err == -EMSGSIZE);
3643         kfree_skb(skb);
3644         if (err)
3645                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3646         return err;
3647 }
3648
3649 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3650                                struct netlink_ext_ack *extack)
3651 {
3652         struct net *net = sock_net(skb->sk);
3653         struct ifinfomsg *ifm;
3654         struct net_device *dev;
3655         struct nlattr *br_spec, *attr = NULL;
3656         int rem, err = -EOPNOTSUPP;
3657         u16 flags = 0;
3658         bool have_flags = false;
3659
3660         if (nlmsg_len(nlh) < sizeof(*ifm))
3661                 return -EINVAL;
3662
3663         ifm = nlmsg_data(nlh);
3664         if (ifm->ifi_family != AF_BRIDGE)
3665                 return -EPFNOSUPPORT;
3666
3667         dev = __dev_get_by_index(net, ifm->ifi_index);
3668         if (!dev) {
3669                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3670                 return -ENODEV;
3671         }
3672
3673         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3674         if (br_spec) {
3675                 nla_for_each_nested(attr, br_spec, rem) {
3676                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3677                                 if (nla_len(attr) < sizeof(flags))
3678                                         return -EINVAL;
3679
3680                                 have_flags = true;
3681                                 flags = nla_get_u16(attr);
3682                                 break;
3683                         }
3684                 }
3685         }
3686
3687         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3688                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3689
3690                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3691                         err = -EOPNOTSUPP;
3692                         goto out;
3693                 }
3694
3695                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3696                 if (err)
3697                         goto out;
3698
3699                 flags &= ~BRIDGE_FLAGS_MASTER;
3700         }
3701
3702         if ((flags & BRIDGE_FLAGS_SELF)) {
3703                 if (!dev->netdev_ops->ndo_bridge_setlink)
3704                         err = -EOPNOTSUPP;
3705                 else
3706                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3707                                                                   flags);
3708                 if (!err) {
3709                         flags &= ~BRIDGE_FLAGS_SELF;
3710
3711                         /* Generate event to notify upper layer of bridge
3712                          * change
3713                          */
3714                         err = rtnl_bridge_notify(dev);
3715                 }
3716         }
3717
3718         if (have_flags)
3719                 memcpy(nla_data(attr), &flags, sizeof(flags));
3720 out:
3721         return err;
3722 }
3723
3724 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3725                                struct netlink_ext_ack *extack)
3726 {
3727         struct net *net = sock_net(skb->sk);
3728         struct ifinfomsg *ifm;
3729         struct net_device *dev;
3730         struct nlattr *br_spec, *attr = NULL;
3731         int rem, err = -EOPNOTSUPP;
3732         u16 flags = 0;
3733         bool have_flags = false;
3734
3735         if (nlmsg_len(nlh) < sizeof(*ifm))
3736                 return -EINVAL;
3737
3738         ifm = nlmsg_data(nlh);
3739         if (ifm->ifi_family != AF_BRIDGE)
3740                 return -EPFNOSUPPORT;
3741
3742         dev = __dev_get_by_index(net, ifm->ifi_index);
3743         if (!dev) {
3744                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3745                 return -ENODEV;
3746         }
3747
3748         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3749         if (br_spec) {
3750                 nla_for_each_nested(attr, br_spec, rem) {
3751                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3752                                 if (nla_len(attr) < sizeof(flags))
3753                                         return -EINVAL;
3754
3755                                 have_flags = true;
3756                                 flags = nla_get_u16(attr);
3757                                 break;
3758                         }
3759                 }
3760         }
3761
3762         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3763                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3764
3765                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3766                         err = -EOPNOTSUPP;
3767                         goto out;
3768                 }
3769
3770                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3771                 if (err)
3772                         goto out;
3773
3774                 flags &= ~BRIDGE_FLAGS_MASTER;
3775         }
3776
3777         if ((flags & BRIDGE_FLAGS_SELF)) {
3778                 if (!dev->netdev_ops->ndo_bridge_dellink)
3779                         err = -EOPNOTSUPP;
3780                 else
3781                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3782                                                                   flags);
3783
3784                 if (!err) {
3785                         flags &= ~BRIDGE_FLAGS_SELF;
3786
3787                         /* Generate event to notify upper layer of bridge
3788                          * change
3789                          */
3790                         err = rtnl_bridge_notify(dev);
3791                 }
3792         }
3793
3794         if (have_flags)
3795                 memcpy(nla_data(attr), &flags, sizeof(flags));
3796 out:
3797         return err;
3798 }
3799
3800 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3801 {
3802         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3803                (!idxattr || idxattr == attrid);
3804 }
3805
3806 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3807 static int rtnl_get_offload_stats_attr_size(int attr_id)
3808 {
3809         switch (attr_id) {
3810         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3811                 return sizeof(struct rtnl_link_stats64);
3812         }
3813
3814         return 0;
3815 }
3816
3817 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3818                                   int *prividx)
3819 {
3820         struct nlattr *attr = NULL;
3821         int attr_id, size;
3822         void *attr_data;
3823         int err;
3824
3825         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3826               dev->netdev_ops->ndo_get_offload_stats))
3827                 return -ENODATA;
3828
3829         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3830              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3831                 if (attr_id < *prividx)
3832                         continue;
3833
3834                 size = rtnl_get_offload_stats_attr_size(attr_id);
3835                 if (!size)
3836                         continue;
3837
3838                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3839                         continue;
3840
3841                 attr = nla_reserve_64bit(skb, attr_id, size,
3842                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3843                 if (!attr)
3844                         goto nla_put_failure;
3845
3846                 attr_data = nla_data(attr);
3847                 memset(attr_data, 0, size);
3848                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3849                                                              attr_data);
3850                 if (err)
3851                         goto get_offload_stats_failure;
3852         }
3853
3854         if (!attr)
3855                 return -ENODATA;
3856
3857         *prividx = 0;
3858         return 0;
3859
3860 nla_put_failure:
3861         err = -EMSGSIZE;
3862 get_offload_stats_failure:
3863         *prividx = attr_id;
3864         return err;
3865 }
3866
3867 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3868 {
3869         int nla_size = 0;
3870         int attr_id;
3871         int size;
3872
3873         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3874               dev->netdev_ops->ndo_get_offload_stats))
3875                 return 0;
3876
3877         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3878              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3879                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3880                         continue;
3881                 size = rtnl_get_offload_stats_attr_size(attr_id);
3882                 nla_size += nla_total_size_64bit(size);
3883         }
3884
3885         if (nla_size != 0)
3886                 nla_size += nla_total_size(0);
3887
3888         return nla_size;
3889 }
3890
3891 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3892                                int type, u32 pid, u32 seq, u32 change,
3893                                unsigned int flags, unsigned int filter_mask,
3894                                int *idxattr, int *prividx)
3895 {
3896         struct if_stats_msg *ifsm;
3897         struct nlmsghdr *nlh;
3898         struct nlattr *attr;
3899         int s_prividx = *prividx;
3900         int err;
3901
3902         ASSERT_RTNL();
3903
3904         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3905         if (!nlh)
3906                 return -EMSGSIZE;
3907
3908         ifsm = nlmsg_data(nlh);
3909         ifsm->family = PF_UNSPEC;
3910         ifsm->pad1 = 0;
3911         ifsm->pad2 = 0;
3912         ifsm->ifindex = dev->ifindex;
3913         ifsm->filter_mask = filter_mask;
3914
3915         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3916                 struct rtnl_link_stats64 *sp;
3917
3918                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3919                                          sizeof(struct rtnl_link_stats64),
3920                                          IFLA_STATS_UNSPEC);
3921                 if (!attr)
3922                         goto nla_put_failure;
3923
3924                 sp = nla_data(attr);
3925                 dev_get_stats(dev, sp);
3926         }
3927
3928         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3929                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3930
3931                 if (ops && ops->fill_linkxstats) {
3932                         *idxattr = IFLA_STATS_LINK_XSTATS;
3933                         attr = nla_nest_start(skb,
3934                                               IFLA_STATS_LINK_XSTATS);
3935                         if (!attr)
3936                                 goto nla_put_failure;
3937
3938                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3939                         nla_nest_end(skb, attr);
3940                         if (err)
3941                                 goto nla_put_failure;
3942                         *idxattr = 0;
3943                 }
3944         }
3945
3946         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3947                              *idxattr)) {
3948                 const struct rtnl_link_ops *ops = NULL;
3949                 const struct net_device *master;
3950
3951                 master = netdev_master_upper_dev_get(dev);
3952                 if (master)
3953                         ops = master->rtnl_link_ops;
3954                 if (ops && ops->fill_linkxstats) {
3955                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3956                         attr = nla_nest_start(skb,
3957                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3958                         if (!attr)
3959                                 goto nla_put_failure;
3960
3961                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3962                         nla_nest_end(skb, attr);
3963                         if (err)
3964                                 goto nla_put_failure;
3965                         *idxattr = 0;
3966                 }
3967         }
3968
3969         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3970                              *idxattr)) {
3971                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3972                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3973                 if (!attr)
3974                         goto nla_put_failure;
3975
3976                 err = rtnl_get_offload_stats(skb, dev, prividx);
3977                 if (err == -ENODATA)
3978                         nla_nest_cancel(skb, attr);
3979                 else
3980                         nla_nest_end(skb, attr);
3981
3982                 if (err && err != -ENODATA)
3983                         goto nla_put_failure;
3984                 *idxattr = 0;
3985         }
3986
3987         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3988                 struct rtnl_af_ops *af_ops;
3989
3990                 *idxattr = IFLA_STATS_AF_SPEC;
3991                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3992                 if (!attr)
3993                         goto nla_put_failure;
3994
3995                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3996                         if (af_ops->fill_stats_af) {
3997                                 struct nlattr *af;
3998                                 int err;
3999
4000                                 af = nla_nest_start(skb, af_ops->family);
4001                                 if (!af)
4002                                         goto nla_put_failure;
4003
4004                                 err = af_ops->fill_stats_af(skb, dev);
4005
4006                                 if (err == -ENODATA)
4007                                         nla_nest_cancel(skb, af);
4008                                 else if (err < 0)
4009                                         goto nla_put_failure;
4010
4011                                 nla_nest_end(skb, af);
4012                         }
4013                 }
4014
4015                 nla_nest_end(skb, attr);
4016
4017                 *idxattr = 0;
4018         }
4019
4020         nlmsg_end(skb, nlh);
4021
4022         return 0;
4023
4024 nla_put_failure:
4025         /* not a multi message or no progress mean a real error */
4026         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4027                 nlmsg_cancel(skb, nlh);
4028         else
4029                 nlmsg_end(skb, nlh);
4030
4031         return -EMSGSIZE;
4032 }
4033
4034 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4035                                   u32 filter_mask)
4036 {
4037         size_t size = 0;
4038
4039         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4040                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4041
4042         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4043                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4044                 int attr = IFLA_STATS_LINK_XSTATS;
4045
4046                 if (ops && ops->get_linkxstats_size) {
4047                         size += nla_total_size(ops->get_linkxstats_size(dev,
4048                                                                         attr));
4049                         /* for IFLA_STATS_LINK_XSTATS */
4050                         size += nla_total_size(0);
4051                 }
4052         }
4053
4054         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4055                 struct net_device *_dev = (struct net_device *)dev;
4056                 const struct rtnl_link_ops *ops = NULL;
4057                 const struct net_device *master;
4058
4059                 /* netdev_master_upper_dev_get can't take const */
4060                 master = netdev_master_upper_dev_get(_dev);
4061                 if (master)
4062                         ops = master->rtnl_link_ops;
4063                 if (ops && ops->get_linkxstats_size) {
4064                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4065
4066                         size += nla_total_size(ops->get_linkxstats_size(dev,
4067                                                                         attr));
4068                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4069                         size += nla_total_size(0);
4070                 }
4071         }
4072
4073         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4074                 size += rtnl_get_offload_stats_size(dev);
4075
4076         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4077                 struct rtnl_af_ops *af_ops;
4078
4079                 /* for IFLA_STATS_AF_SPEC */
4080                 size += nla_total_size(0);
4081
4082                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4083                         if (af_ops->get_stats_af_size) {
4084                                 size += nla_total_size(
4085                                         af_ops->get_stats_af_size(dev));
4086
4087                                 /* for AF_* */
4088                                 size += nla_total_size(0);
4089                         }
4090                 }
4091         }
4092
4093         return size;
4094 }
4095
4096 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4097                           struct netlink_ext_ack *extack)
4098 {
4099         struct net *net = sock_net(skb->sk);
4100         struct net_device *dev = NULL;
4101         int idxattr = 0, prividx = 0;
4102         struct if_stats_msg *ifsm;
4103         struct sk_buff *nskb;
4104         u32 filter_mask;
4105         int err;
4106
4107         if (nlmsg_len(nlh) < sizeof(*ifsm))
4108                 return -EINVAL;
4109
4110         ifsm = nlmsg_data(nlh);
4111         if (ifsm->ifindex > 0)
4112                 dev = __dev_get_by_index(net, ifsm->ifindex);
4113         else
4114                 return -EINVAL;
4115
4116         if (!dev)
4117                 return -ENODEV;
4118
4119         filter_mask = ifsm->filter_mask;
4120         if (!filter_mask)
4121                 return -EINVAL;
4122
4123         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4124         if (!nskb)
4125                 return -ENOBUFS;
4126
4127         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4128                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4129                                   0, filter_mask, &idxattr, &prividx);
4130         if (err < 0) {
4131                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4132                 WARN_ON(err == -EMSGSIZE);
4133                 kfree_skb(nskb);
4134         } else {
4135                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4136         }
4137
4138         return err;
4139 }
4140
4141 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4142 {
4143         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4144         struct net *net = sock_net(skb->sk);
4145         unsigned int flags = NLM_F_MULTI;
4146         struct if_stats_msg *ifsm;
4147         struct hlist_head *head;
4148         struct net_device *dev;
4149         u32 filter_mask = 0;
4150         int idx = 0;
4151
4152         s_h = cb->args[0];
4153         s_idx = cb->args[1];
4154         s_idxattr = cb->args[2];
4155         s_prividx = cb->args[3];
4156
4157         cb->seq = net->dev_base_seq;
4158
4159         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4160                 return -EINVAL;
4161
4162         ifsm = nlmsg_data(cb->nlh);
4163         filter_mask = ifsm->filter_mask;
4164         if (!filter_mask)
4165                 return -EINVAL;
4166
4167         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4168                 idx = 0;
4169                 head = &net->dev_index_head[h];
4170                 hlist_for_each_entry(dev, head, index_hlist) {
4171                         if (idx < s_idx)
4172                                 goto cont;
4173                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4174                                                   NETLINK_CB(cb->skb).portid,
4175                                                   cb->nlh->nlmsg_seq, 0,
4176                                                   flags, filter_mask,
4177                                                   &s_idxattr, &s_prividx);
4178                         /* If we ran out of room on the first message,
4179                          * we're in trouble
4180                          */
4181                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4182
4183                         if (err < 0)
4184                                 goto out;
4185                         s_prividx = 0;
4186                         s_idxattr = 0;
4187                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4188 cont:
4189                         idx++;
4190                 }
4191         }
4192 out:
4193         cb->args[3] = s_prividx;
4194         cb->args[2] = s_idxattr;
4195         cb->args[1] = idx;
4196         cb->args[0] = h;
4197
4198         return skb->len;
4199 }
4200
4201 /* Process one rtnetlink message. */
4202
4203 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4204                              struct netlink_ext_ack *extack)
4205 {
4206         struct net *net = sock_net(skb->sk);
4207         struct rtnl_link *handlers;
4208         int err = -EOPNOTSUPP;
4209         rtnl_doit_func doit;
4210         unsigned int flags;
4211         int kind;
4212         int family;
4213         int type;
4214
4215         type = nlh->nlmsg_type;
4216         if (type > RTM_MAX)
4217                 return -EOPNOTSUPP;
4218
4219         type -= RTM_BASE;
4220
4221         /* All the messages must have at least 1 byte length */
4222         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4223                 return 0;
4224
4225         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4226         kind = type&3;
4227
4228         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4229                 return -EPERM;
4230
4231         if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4232                 family = PF_UNSPEC;
4233
4234         rcu_read_lock();
4235         handlers = rcu_dereference(rtnl_msg_handlers[family]);
4236         if (!handlers) {
4237                 family = PF_UNSPEC;
4238                 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4239         }
4240
4241         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4242                 struct sock *rtnl;
4243                 rtnl_dumpit_func dumpit;
4244                 u16 min_dump_alloc = 0;
4245
4246                 dumpit = READ_ONCE(handlers[type].dumpit);
4247                 if (!dumpit) {
4248                         family = PF_UNSPEC;
4249                         handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4250                         if (!handlers)
4251                                 goto err_unlock;
4252
4253                         dumpit = READ_ONCE(handlers[type].dumpit);
4254                         if (!dumpit)
4255                                 goto err_unlock;
4256                 }
4257
4258                 refcount_inc(&rtnl_msg_handlers_ref[family]);
4259
4260                 if (type == RTM_GETLINK - RTM_BASE)
4261                         min_dump_alloc = rtnl_calcit(skb, nlh);
4262
4263                 rcu_read_unlock();
4264
4265                 rtnl = net->rtnl;
4266                 {
4267                         struct netlink_dump_control c = {
4268                                 .dump           = dumpit,
4269                                 .min_dump_alloc = min_dump_alloc,
4270                         };
4271                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4272                 }
4273                 refcount_dec(&rtnl_msg_handlers_ref[family]);
4274                 return err;
4275         }
4276
4277         doit = READ_ONCE(handlers[type].doit);
4278         if (!doit) {
4279                 family = PF_UNSPEC;
4280                 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4281         }
4282
4283         flags = READ_ONCE(handlers[type].flags);
4284         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4285                 refcount_inc(&rtnl_msg_handlers_ref[family]);
4286                 doit = READ_ONCE(handlers[type].doit);
4287                 rcu_read_unlock();
4288                 if (doit)
4289                         err = doit(skb, nlh, extack);
4290                 refcount_dec(&rtnl_msg_handlers_ref[family]);
4291                 return err;
4292         }
4293
4294         rcu_read_unlock();
4295
4296         rtnl_lock();
4297         handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4298         if (handlers) {
4299                 doit = READ_ONCE(handlers[type].doit);
4300                 if (doit)
4301                         err = doit(skb, nlh, extack);
4302         }
4303         rtnl_unlock();
4304         return err;
4305
4306 err_unlock:
4307         rcu_read_unlock();
4308         return -EOPNOTSUPP;
4309 }
4310
4311 static void rtnetlink_rcv(struct sk_buff *skb)
4312 {
4313         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4314 }
4315
4316 static int rtnetlink_bind(struct net *net, int group)
4317 {
4318         switch (group) {
4319         case RTNLGRP_IPV4_MROUTE_R:
4320         case RTNLGRP_IPV6_MROUTE_R:
4321                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4322                         return -EPERM;
4323                 break;
4324         }
4325         return 0;
4326 }
4327
4328 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4329 {
4330         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4331
4332         switch (event) {
4333         case NETDEV_REBOOT:
4334         case NETDEV_CHANGEADDR:
4335         case NETDEV_CHANGENAME:
4336         case NETDEV_FEAT_CHANGE:
4337         case NETDEV_BONDING_FAILOVER:
4338         case NETDEV_NOTIFY_PEERS:
4339         case NETDEV_RESEND_IGMP:
4340         case NETDEV_CHANGEINFODATA:
4341                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4342                                    GFP_KERNEL, NULL);
4343                 break;
4344         default:
4345                 break;
4346         }
4347         return NOTIFY_DONE;
4348 }
4349
4350 static struct notifier_block rtnetlink_dev_notifier = {
4351         .notifier_call  = rtnetlink_event,
4352 };
4353
4354
4355 static int __net_init rtnetlink_net_init(struct net *net)
4356 {
4357         struct sock *sk;
4358         struct netlink_kernel_cfg cfg = {
4359                 .groups         = RTNLGRP_MAX,
4360                 .input          = rtnetlink_rcv,
4361                 .cb_mutex       = &rtnl_mutex,
4362                 .flags          = NL_CFG_F_NONROOT_RECV,
4363                 .bind           = rtnetlink_bind,
4364         };
4365
4366         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4367         if (!sk)
4368                 return -ENOMEM;
4369         net->rtnl = sk;
4370         return 0;
4371 }
4372
4373 static void __net_exit rtnetlink_net_exit(struct net *net)
4374 {
4375         netlink_kernel_release(net->rtnl);
4376         net->rtnl = NULL;
4377 }
4378
4379 static struct pernet_operations rtnetlink_net_ops = {
4380         .init = rtnetlink_net_init,
4381         .exit = rtnetlink_net_exit,
4382 };
4383
4384 void __init rtnetlink_init(void)
4385 {
4386         int i;
4387
4388         for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4389                 refcount_set(&rtnl_msg_handlers_ref[i], 1);
4390
4391         if (register_pernet_subsys(&rtnetlink_net_ops))
4392                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4393
4394         register_netdevice_notifier(&rtnetlink_dev_notifier);
4395
4396         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4397                       rtnl_dump_ifinfo, 0);
4398         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4399         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4400         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4401
4402         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4403         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4404         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4405
4406         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4407         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4408         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4409
4410         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4411         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4412         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4413
4414         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4415                       0);
4416 }