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