kernel/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/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 int lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256
 257         return 0;
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_unregister);
 260
 261 /**
 262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 263  * @protocol : Protocol family or PF_UNSPEC
 264  *
 265  * Identical to calling rtnl_unregster() for all registered message types
 266  * of a certain protocol family.
 267  */
 268 void rtnl_unregister_all(int protocol)
 269 {
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271
 272         kfree(rtnl_msg_handlers[protocol]);
 273         rtnl_msg_handlers[protocol] = NULL;
 274 }
 275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 276
 277 static LIST_HEAD(link_ops);
 278
 279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 280 {
 281         const struct rtnl_link_ops *ops;
 282
 283         list_for_each_entry(ops, &link_ops, list) {
 284                 if (!strcmp(ops->kind, kind))
 285                         return ops;
 286         }
 287         return NULL;
 288 }
 289
 290 /**
 291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 292  * @ops: struct rtnl_link_ops * to register
 293  *
 294  * The caller must hold the rtnl_mutex. This function should be used
 295  * by drivers that create devices during module initialization. It
 296  * must be called before registering the devices.
 297  *
 298  * Returns 0 on success or a negative error code.
 299  */
 300 int __rtnl_link_register(struct rtnl_link_ops *ops)
 301 {
 302         if (rtnl_link_ops_get(ops->kind))
 303                 return -EEXIST;
 304
 305         /* The check for setup is here because if ops
 306          * does not have that filled up, it is not possible
 307          * to use the ops for creating device. So do not
 308          * fill up dellink as well. That disables rtnl_dellink.
 309          */
 310         if (ops->setup && !ops->dellink)
 311                 ops->dellink = unregister_netdevice_queue;
 312
 313         list_add_tail(&ops->list, &link_ops);
 314         return 0;
 315 }
 316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 317
 318 /**
 319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 320  * @ops: struct rtnl_link_ops * to register
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_link_register(struct rtnl_link_ops *ops)
 325 {
 326         int err;
 327
 328         rtnl_lock();
 329         err = __rtnl_link_register(ops);
 330         rtnl_unlock();
 331         return err;
 332 }
 333 EXPORT_SYMBOL_GPL(rtnl_link_register);
 334
 335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 336 {
 337         struct net_device *dev;
 338         LIST_HEAD(list_kill);
 339
 340         for_each_netdev(net, dev) {
 341                 if (dev->rtnl_link_ops == ops)
 342                         ops->dellink(dev, &list_kill);
 343         }
 344         unregister_netdevice_many(&list_kill);
 345 }
 346
 347 /**
 348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 349  * @ops: struct rtnl_link_ops * to unregister
 350  *
 351  * The caller must hold the rtnl_mutex.
 352  */
 353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 354 {
 355         struct net *net;
 356
 357         for_each_net(net) {
 358                 __rtnl_kill_links(net, ops);
 359         }
 360         list_del(&ops->list);
 361 }
 362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 363
 364 /* Return with the rtnl_lock held when there are no network
 365  * devices unregistering in any network namespace.
 366  */
 367 static void rtnl_lock_unregistering_all(void)
 368 {
 369         struct net *net;
 370         bool unregistering;
 371         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 372
 373         add_wait_queue(&netdev_unregistering_wq, &wait);
 374         for (;;) {
 375                 unregistering = false;
 376                 rtnl_lock();
 377                 for_each_net(net) {
 378                         if (net->dev_unreg_count > 0) {
 379                                 unregistering = true;
 380                                 break;
 381                         }
 382                 }
 383                 if (!unregistering)
 384                         break;
 385                 __rtnl_unlock();
 386
 387                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 388         }
 389         remove_wait_queue(&netdev_unregistering_wq, &wait);
 390 }
 391
 392 /**
 393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 394  * @ops: struct rtnl_link_ops * to unregister
 395  */
 396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 397 {
 398         /* Close the race with cleanup_net() */
 399         mutex_lock(&net_mutex);
 400         rtnl_lock_unregistering_all();
 401         __rtnl_link_unregister(ops);
 402         rtnl_unlock();
 403         mutex_unlock(&net_mutex);
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 406
 407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 408 {
 409         struct net_device *master_dev;
 410         const struct rtnl_link_ops *ops;
 411
 412         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 413         if (!master_dev)
 414                 return 0;
 415         ops = master_dev->rtnl_link_ops;
 416         if (!ops || !ops->get_slave_size)
 417                 return 0;
 418         /* IFLA_INFO_SLAVE_DATA + nested data */
 419         return nla_total_size(sizeof(struct nlattr)) +
 420                ops->get_slave_size(master_dev, dev);
 421 }
 422
 423 static size_t rtnl_link_get_size(const struct net_device *dev)
 424 {
 425         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 426         size_t size;
 427
 428         if (!ops)
 429                 return 0;
 430
 431         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 432                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 433
 434         if (ops->get_size)
 435                 /* IFLA_INFO_DATA + nested data */
 436                 size += nla_total_size(sizeof(struct nlattr)) +
 437                         ops->get_size(dev);
 438
 439         if (ops->get_xstats_size)
 440                 /* IFLA_INFO_XSTATS */
 441                 size += nla_total_size(ops->get_xstats_size(dev));
 442
 443         size += rtnl_link_get_slave_info_data_size(dev);
 444
 445         return size;
 446 }
 447
 448 static LIST_HEAD(rtnl_af_ops);
 449
 450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 451 {
 452         const struct rtnl_af_ops *ops;
 453
 454         list_for_each_entry(ops, &rtnl_af_ops, list) {
 455                 if (ops->family == family)
 456                         return ops;
 457         }
 458
 459         return NULL;
 460 }
 461
 462 /**
 463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 464  * @ops: struct rtnl_af_ops * to register
 465  *
 466  * Returns 0 on success or a negative error code.
 467  */
 468 void rtnl_af_register(struct rtnl_af_ops *ops)
 469 {
 470         rtnl_lock();
 471         list_add_tail(&ops->list, &rtnl_af_ops);
 472         rtnl_unlock();
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_af_register);
 475
 476 /**
 477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 478  * @ops: struct rtnl_af_ops * to unregister
 479  *
 480  * The caller must hold the rtnl_mutex.
 481  */
 482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         list_del(&ops->list);
 485 }
 486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 487
 488 /**
 489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 490  * @ops: struct rtnl_af_ops * to unregister
 491  */
 492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 493 {
 494         rtnl_lock();
 495         __rtnl_af_unregister(ops);
 496         rtnl_unlock();
 497 }
 498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 499
 500 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 501 {
 502         struct rtnl_af_ops *af_ops;
 503         size_t size;
 504
 505         /* IFLA_AF_SPEC */
 506         size = nla_total_size(sizeof(struct nlattr));
 507
 508         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 509                 if (af_ops->get_link_af_size) {
 510                         /* AF_* + nested data */
 511                         size += nla_total_size(sizeof(struct nlattr)) +
 512                                 af_ops->get_link_af_size(dev);
 513                 }
 514         }
 515
 516         return size;
 517 }
 518
 519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 520 {
 521         struct net_device *master_dev;
 522
 523         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 524         if (master_dev && master_dev->rtnl_link_ops)
 525                 return true;
 526         return false;
 527 }
 528
 529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 530                                      const struct net_device *dev)
 531 {
 532         struct net_device *master_dev;
 533         const struct rtnl_link_ops *ops;
 534         struct nlattr *slave_data;
 535         int err;
 536
 537         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 538         if (!master_dev)
 539                 return 0;
 540         ops = master_dev->rtnl_link_ops;
 541         if (!ops)
 542                 return 0;
 543         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 544                 return -EMSGSIZE;
 545         if (ops->fill_slave_info) {
 546                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 547                 if (!slave_data)
 548                         return -EMSGSIZE;
 549                 err = ops->fill_slave_info(skb, master_dev, dev);
 550                 if (err < 0)
 551                         goto err_cancel_slave_data;
 552                 nla_nest_end(skb, slave_data);
 553         }
 554         return 0;
 555
 556 err_cancel_slave_data:
 557         nla_nest_cancel(skb, slave_data);
 558         return err;
 559 }
 560
 561 static int rtnl_link_info_fill(struct sk_buff *skb,
 562                                const struct net_device *dev)
 563 {
 564         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 565         struct nlattr *data;
 566         int err;
 567
 568         if (!ops)
 569                 return 0;
 570         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 571                 return -EMSGSIZE;
 572         if (ops->fill_xstats) {
 573                 err = ops->fill_xstats(skb, dev);
 574                 if (err < 0)
 575                         return err;
 576         }
 577         if (ops->fill_info) {
 578                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 579                 if (data == NULL)
 580                         return -EMSGSIZE;
 581                 err = ops->fill_info(skb, dev);
 582                 if (err < 0)
 583                         goto err_cancel_data;
 584                 nla_nest_end(skb, data);
 585         }
 586         return 0;
 587
 588 err_cancel_data:
 589         nla_nest_cancel(skb, data);
 590         return err;
 591 }
 592
 593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 594 {
 595         struct nlattr *linkinfo;
 596         int err = -EMSGSIZE;
 597
 598         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 599         if (linkinfo == NULL)
 600                 goto out;
 601
 602         err = rtnl_link_info_fill(skb, dev);
 603         if (err < 0)
 604                 goto err_cancel_link;
 605
 606         err = rtnl_link_slave_info_fill(skb, dev);
 607         if (err < 0)
 608                 goto err_cancel_link;
 609
 610         nla_nest_end(skb, linkinfo);
 611         return 0;
 612
 613 err_cancel_link:
 614         nla_nest_cancel(skb, linkinfo);
 615 out:
 616         return err;
 617 }
 618
 619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 620 {
 621         struct sock *rtnl = net->rtnl;
 622         int err = 0;
 623
 624         NETLINK_CB(skb).dst_group = group;
 625         if (echo)
 626                 atomic_inc(&skb->users);
 627         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 628         if (echo)
 629                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 630         return err;
 631 }
 632
 633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 634 {
 635         struct sock *rtnl = net->rtnl;
 636
 637         return nlmsg_unicast(rtnl, skb, pid);
 638 }
 639 EXPORT_SYMBOL(rtnl_unicast);
 640
 641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 642                  struct nlmsghdr *nlh, gfp_t flags)
 643 {
 644         struct sock *rtnl = net->rtnl;
 645         int report = 0;
 646
 647         if (nlh)
 648                 report = nlmsg_report(nlh);
 649
 650         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 651 }
 652 EXPORT_SYMBOL(rtnl_notify);
 653
 654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 655 {
 656         struct sock *rtnl = net->rtnl;
 657
 658         netlink_set_err(rtnl, 0, group, error);
 659 }
 660 EXPORT_SYMBOL(rtnl_set_sk_err);
 661
 662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 663 {
 664         struct nlattr *mx;
 665         int i, valid = 0;
 666
 667         mx = nla_nest_start(skb, RTA_METRICS);
 668         if (mx == NULL)
 669                 return -ENOBUFS;
 670
 671         for (i = 0; i < RTAX_MAX; i++) {
 672                 if (metrics[i]) {
 673                         if (i == RTAX_CC_ALGO - 1) {
 674                                 char tmp[TCP_CA_NAME_MAX], *name;
 675
 676                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 677                                 if (!name)
 678                                         continue;
 679                                 if (nla_put_string(skb, i + 1, name))
 680                                         goto nla_put_failure;
 681                         } else {
 682                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 683                                         goto nla_put_failure;
 684                         }
 685                         valid++;
 686                 }
 687         }
 688
 689         if (!valid) {
 690                 nla_nest_cancel(skb, mx);
 691                 return 0;
 692         }
 693
 694         return nla_nest_end(skb, mx);
 695
 696 nla_put_failure:
 697         nla_nest_cancel(skb, mx);
 698         return -EMSGSIZE;
 699 }
 700 EXPORT_SYMBOL(rtnetlink_put_metrics);
 701
 702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 703                        long expires, u32 error)
 704 {
 705         struct rta_cacheinfo ci = {
 706                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 707                 .rta_used = dst->__use,
 708                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 709                 .rta_error = error,
 710                 .rta_id =  id,
 711         };
 712
 713         if (expires) {
 714                 unsigned long clock;
 715
 716                 clock = jiffies_to_clock_t(abs(expires));
 717                 clock = min_t(unsigned long, clock, INT_MAX);
 718                 ci.rta_expires = (expires > 0) ? clock : -clock;
 719         }
 720         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 721 }
 722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 723
 724 static void set_operstate(struct net_device *dev, unsigned char transition)
 725 {
 726         unsigned char operstate = dev->operstate;
 727
 728         switch (transition) {
 729         case IF_OPER_UP:
 730                 if ((operstate == IF_OPER_DORMANT ||
 731                      operstate == IF_OPER_UNKNOWN) &&
 732                     !netif_dormant(dev))
 733                         operstate = IF_OPER_UP;
 734                 break;
 735
 736         case IF_OPER_DORMANT:
 737                 if (operstate == IF_OPER_UP ||
 738                     operstate == IF_OPER_UNKNOWN)
 739                         operstate = IF_OPER_DORMANT;
 740                 break;
 741         }
 742
 743         if (dev->operstate != operstate) {
 744                 write_lock_bh(&dev_base_lock);
 745                 dev->operstate = operstate;
 746                 write_unlock_bh(&dev_base_lock);
 747                 netdev_state_change(dev);
 748         }
 749 }
 750
 751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 752 {
 753         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 754                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 755 }
 756
 757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 758                                            const struct ifinfomsg *ifm)
 759 {
 760         unsigned int flags = ifm->ifi_flags;
 761
 762         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 763         if (ifm->ifi_change)
 764                 flags = (flags & ifm->ifi_change) |
 765                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 766
 767         return flags;
 768 }
 769
 770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 771                                  const struct rtnl_link_stats64 *b)
 772 {
 773         a->rx_packets = b->rx_packets;
 774         a->tx_packets = b->tx_packets;
 775         a->rx_bytes = b->rx_bytes;
 776         a->tx_bytes = b->tx_bytes;
 777         a->rx_errors = b->rx_errors;
 778         a->tx_errors = b->tx_errors;
 779         a->rx_dropped = b->rx_dropped;
 780         a->tx_dropped = b->tx_dropped;
 781
 782         a->multicast = b->multicast;
 783         a->collisions = b->collisions;
 784
 785         a->rx_length_errors = b->rx_length_errors;
 786         a->rx_over_errors = b->rx_over_errors;
 787         a->rx_crc_errors = b->rx_crc_errors;
 788         a->rx_frame_errors = b->rx_frame_errors;
 789         a->rx_fifo_errors = b->rx_fifo_errors;
 790         a->rx_missed_errors = b->rx_missed_errors;
 791
 792         a->tx_aborted_errors = b->tx_aborted_errors;
 793         a->tx_carrier_errors = b->tx_carrier_errors;
 794         a->tx_fifo_errors = b->tx_fifo_errors;
 795         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 796         a->tx_window_errors = b->tx_window_errors;
 797
 798         a->rx_compressed = b->rx_compressed;
 799         a->tx_compressed = b->tx_compressed;
 800 }
 801
 802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 803 {
 804         memcpy(v, b, sizeof(*b));
 805 }
 806
 807 /* All VF info */
 808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 809                                    u32 ext_filter_mask)
 810 {
 811         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 812             (ext_filter_mask & RTEXT_FILTER_VF)) {
 813                 int num_vfs = dev_num_vf(dev->dev.parent);
 814                 size_t size = nla_total_size(sizeof(struct nlattr));
 815                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 816                 size += num_vfs *
 817                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 818                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 819                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 820                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 821                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 822                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)));
 823                 return size;
 824         } else
 825                 return 0;
 826 }
 827
 828 static size_t rtnl_port_size(const struct net_device *dev,
 829                              u32 ext_filter_mask)
 830 {
 831         size_t port_size = nla_total_size(4)            /* PORT_VF */
 832                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 833                 + nla_total_size(sizeof(struct ifla_port_vsi))
 834                                                         /* PORT_VSI_TYPE */
 835                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 836                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 837                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 838                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 839         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 840         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 841                 + port_size;
 842         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 843                 + port_size;
 844
 845         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 846             !(ext_filter_mask & RTEXT_FILTER_VF))
 847                 return 0;
 848         if (dev_num_vf(dev->dev.parent))
 849                 return port_self_size + vf_ports_size +
 850                         vf_port_size * dev_num_vf(dev->dev.parent);
 851         else
 852                 return port_self_size;
 853 }
 854
 855 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 856                                      u32 ext_filter_mask)
 857 {
 858         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 859                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 860                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 861                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 862                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 863                + nla_total_size(sizeof(struct rtnl_link_stats))
 864                + nla_total_size(sizeof(struct rtnl_link_stats64))
 865                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 866                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 867                + nla_total_size(4) /* IFLA_TXQLEN */
 868                + nla_total_size(4) /* IFLA_WEIGHT */
 869                + nla_total_size(4) /* IFLA_MTU */
 870                + nla_total_size(4) /* IFLA_LINK */
 871                + nla_total_size(4) /* IFLA_MASTER */
 872                + nla_total_size(1) /* IFLA_CARRIER */
 873                + nla_total_size(4) /* IFLA_PROMISCUITY */
 874                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 875                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 876                + nla_total_size(1) /* IFLA_OPERSTATE */
 877                + nla_total_size(1) /* IFLA_LINKMODE */
 878                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 879                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 880                + nla_total_size(ext_filter_mask
 881                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 882                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 883                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 884                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 885                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 886                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 887                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 888 }
 889
 890 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 891 {
 892         struct nlattr *vf_ports;
 893         struct nlattr *vf_port;
 894         int vf;
 895         int err;
 896
 897         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 898         if (!vf_ports)
 899                 return -EMSGSIZE;
 900
 901         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 902                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 903                 if (!vf_port)
 904                         goto nla_put_failure;
 905                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 906                         goto nla_put_failure;
 907                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 908                 if (err == -EMSGSIZE)
 909                         goto nla_put_failure;
 910                 if (err) {
 911                         nla_nest_cancel(skb, vf_port);
 912                         continue;
 913                 }
 914                 nla_nest_end(skb, vf_port);
 915         }
 916
 917         nla_nest_end(skb, vf_ports);
 918
 919         return 0;
 920
 921 nla_put_failure:
 922         nla_nest_cancel(skb, vf_ports);
 923         return -EMSGSIZE;
 924 }
 925
 926 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 927 {
 928         struct nlattr *port_self;
 929         int err;
 930
 931         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 932         if (!port_self)
 933                 return -EMSGSIZE;
 934
 935         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 936         if (err) {
 937                 nla_nest_cancel(skb, port_self);
 938                 return (err == -EMSGSIZE) ? err : 0;
 939         }
 940
 941         nla_nest_end(skb, port_self);
 942
 943         return 0;
 944 }
 945
 946 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 947                           u32 ext_filter_mask)
 948 {
 949         int err;
 950
 951         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 952             !(ext_filter_mask & RTEXT_FILTER_VF))
 953                 return 0;
 954
 955         err = rtnl_port_self_fill(skb, dev);
 956         if (err)
 957                 return err;
 958
 959         if (dev_num_vf(dev->dev.parent)) {
 960                 err = rtnl_vf_ports_fill(skb, dev);
 961                 if (err)
 962                         return err;
 963         }
 964
 965         return 0;
 966 }
 967
 968 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 969 {
 970         int err;
 971         struct netdev_phys_item_id ppid;
 972
 973         err = dev_get_phys_port_id(dev, &ppid);
 974         if (err) {
 975                 if (err == -EOPNOTSUPP)
 976                         return 0;
 977                 return err;
 978         }
 979
 980         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 981                 return -EMSGSIZE;
 982
 983         return 0;
 984 }
 985
 986 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
 987 {
 988         char name[IFNAMSIZ];
 989         int err;
 990
 991         err = dev_get_phys_port_name(dev, name, sizeof(name));
 992         if (err) {
 993                 if (err == -EOPNOTSUPP)
 994                         return 0;
 995                 return err;
 996         }
 997
 998         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
 999                 return -EMSGSIZE;
1000
1001         return 0;
1002 }
1003
1004 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1005 {
1006         int err;
1007         struct netdev_phys_item_id psid;
1008
1009         err = netdev_switch_parent_id_get(dev, &psid);
1010         if (err) {
1011                 if (err == -EOPNOTSUPP)
1012                         return 0;
1013                 return err;
1014         }
1015
1016         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
1017                 return -EMSGSIZE;
1018
1019         return 0;
1020 }
1021
1022 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1023                             int type, u32 pid, u32 seq, u32 change,
1024                             unsigned int flags, u32 ext_filter_mask)
1025 {
1026         struct ifinfomsg *ifm;
1027         struct nlmsghdr *nlh;
1028         struct rtnl_link_stats64 temp;
1029         const struct rtnl_link_stats64 *stats;
1030         struct nlattr *attr, *af_spec;
1031         struct rtnl_af_ops *af_ops;
1032         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1033
1034         ASSERT_RTNL();
1035         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1036         if (nlh == NULL)
1037                 return -EMSGSIZE;
1038
1039         ifm = nlmsg_data(nlh);
1040         ifm->ifi_family = AF_UNSPEC;
1041         ifm->__ifi_pad = 0;
1042         ifm->ifi_type = dev->type;
1043         ifm->ifi_index = dev->ifindex;
1044         ifm->ifi_flags = dev_get_flags(dev);
1045         ifm->ifi_change = change;
1046
1047         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1048             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1049             nla_put_u8(skb, IFLA_OPERSTATE,
1050                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1051             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1052             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1053             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1054             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1055             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1056 #ifdef CONFIG_RPS
1057             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1058 #endif
1059             (dev->ifindex != dev_get_iflink(dev) &&
1060              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1061             (upper_dev &&
1062              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1063             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1064             (dev->qdisc &&
1065              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1066             (dev->ifalias &&
1067              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1068             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1069                         atomic_read(&dev->carrier_changes)))
1070                 goto nla_put_failure;
1071
1072         if (1) {
1073                 struct rtnl_link_ifmap map = {
1074                         .mem_start   = dev->mem_start,
1075                         .mem_end     = dev->mem_end,
1076                         .base_addr   = dev->base_addr,
1077                         .irq         = dev->irq,
1078                         .dma         = dev->dma,
1079                         .port        = dev->if_port,
1080                 };
1081                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1082                         goto nla_put_failure;
1083         }
1084
1085         if (dev->addr_len) {
1086                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1087                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1088                         goto nla_put_failure;
1089         }
1090
1091         if (rtnl_phys_port_id_fill(skb, dev))
1092                 goto nla_put_failure;
1093
1094         if (rtnl_phys_port_name_fill(skb, dev))
1095                 goto nla_put_failure;
1096
1097         if (rtnl_phys_switch_id_fill(skb, dev))
1098                 goto nla_put_failure;
1099
1100         attr = nla_reserve(skb, IFLA_STATS,
1101                         sizeof(struct rtnl_link_stats));
1102         if (attr == NULL)
1103                 goto nla_put_failure;
1104
1105         stats = dev_get_stats(dev, &temp);
1106         copy_rtnl_link_stats(nla_data(attr), stats);
1107
1108         attr = nla_reserve(skb, IFLA_STATS64,
1109                         sizeof(struct rtnl_link_stats64));
1110         if (attr == NULL)
1111                 goto nla_put_failure;
1112         copy_rtnl_link_stats64(nla_data(attr), stats);
1113
1114         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1115             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1116                 goto nla_put_failure;
1117
1118         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1119             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1120                 int i;
1121
1122                 struct nlattr *vfinfo, *vf;
1123                 int num_vfs = dev_num_vf(dev->dev.parent);
1124
1125                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1126                 if (!vfinfo)
1127                         goto nla_put_failure;
1128                 for (i = 0; i < num_vfs; i++) {
1129                         struct ifla_vf_info ivi;
1130                         struct ifla_vf_mac vf_mac;
1131                         struct ifla_vf_vlan vf_vlan;
1132                         struct ifla_vf_rate vf_rate;
1133                         struct ifla_vf_tx_rate vf_tx_rate;
1134                         struct ifla_vf_spoofchk vf_spoofchk;
1135                         struct ifla_vf_link_state vf_linkstate;
1136                         struct ifla_vf_rss_query_en vf_rss_query_en;
1137
1138                         /*
1139                          * Not all SR-IOV capable drivers support the
1140                          * spoofcheck and "RSS query enable" query.  Preset to
1141                          * -1 so the user space tool can detect that the driver
1142                          * didn't report anything.
1143                          */
1144                         ivi.spoofchk = -1;
1145                         ivi.rss_query_en = -1;
1146                         memset(ivi.mac, 0, sizeof(ivi.mac));
1147                         /* The default value for VF link state is "auto"
1148                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1149                          */
1150                         ivi.linkstate = 0;
1151                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1152                                 break;
1153                         vf_mac.vf =
1154                                 vf_vlan.vf =
1155                                 vf_rate.vf =
1156                                 vf_tx_rate.vf =
1157                                 vf_spoofchk.vf =
1158                                 vf_linkstate.vf =
1159                                 vf_rss_query_en.vf = ivi.vf;
1160
1161                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1162                         vf_vlan.vlan = ivi.vlan;
1163                         vf_vlan.qos = ivi.qos;
1164                         vf_tx_rate.rate = ivi.max_tx_rate;
1165                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1166                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1167                         vf_spoofchk.setting = ivi.spoofchk;
1168                         vf_linkstate.link_state = ivi.linkstate;
1169                         vf_rss_query_en.setting = ivi.rss_query_en;
1170                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1171                         if (!vf) {
1172                                 nla_nest_cancel(skb, vfinfo);
1173                                 goto nla_put_failure;
1174                         }
1175                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1176                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1177                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1178                                     &vf_rate) ||
1179                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1180                                     &vf_tx_rate) ||
1181                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1182                                     &vf_spoofchk) ||
1183                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1184                                     &vf_linkstate) ||
1185                             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1186                                     sizeof(vf_rss_query_en),
1187                                     &vf_rss_query_en))
1188                                 goto nla_put_failure;
1189                         nla_nest_end(skb, vf);
1190                 }
1191                 nla_nest_end(skb, vfinfo);
1192         }
1193
1194         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1195                 goto nla_put_failure;
1196
1197         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1198                 if (rtnl_link_fill(skb, dev) < 0)
1199                         goto nla_put_failure;
1200         }
1201
1202         if (dev->rtnl_link_ops &&
1203             dev->rtnl_link_ops->get_link_net) {
1204                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1205
1206                 if (!net_eq(dev_net(dev), link_net)) {
1207                         int id = peernet2id(dev_net(dev), link_net);
1208
1209                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1210                                 goto nla_put_failure;
1211                 }
1212         }
1213
1214         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1215                 goto nla_put_failure;
1216
1217         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1218                 if (af_ops->fill_link_af) {
1219                         struct nlattr *af;
1220                         int err;
1221
1222                         if (!(af = nla_nest_start(skb, af_ops->family)))
1223                                 goto nla_put_failure;
1224
1225                         err = af_ops->fill_link_af(skb, dev);
1226
1227                         /*
1228                          * Caller may return ENODATA to indicate that there
1229                          * was no data to be dumped. This is not an error, it
1230                          * means we should trim the attribute header and
1231                          * continue.
1232                          */
1233                         if (err == -ENODATA)
1234                                 nla_nest_cancel(skb, af);
1235                         else if (err < 0)
1236                                 goto nla_put_failure;
1237
1238                         nla_nest_end(skb, af);
1239                 }
1240         }
1241
1242         nla_nest_end(skb, af_spec);
1243
1244         nlmsg_end(skb, nlh);
1245         return 0;
1246
1247 nla_put_failure:
1248         nlmsg_cancel(skb, nlh);
1249         return -EMSGSIZE;
1250 }
1251
1252 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1253         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1254         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1255         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1256         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1257         [IFLA_MTU]              = { .type = NLA_U32 },
1258         [IFLA_LINK]             = { .type = NLA_U32 },
1259         [IFLA_MASTER]           = { .type = NLA_U32 },
1260         [IFLA_CARRIER]          = { .type = NLA_U8 },
1261         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1262         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1263         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1264         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1265         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1266         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1267         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1268         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1269         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1270         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1271         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1272         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1273         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1274         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1275         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1276         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1277         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1278         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1279         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1280         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1281 };
1282
1283 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1284         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1285         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1286         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1287         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1288 };
1289
1290 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1291         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1292         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1293         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1294         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1295         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1296         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1297         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1298 };
1299
1300 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1301         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1302         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1303                                     .len = PORT_PROFILE_MAX },
1304         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1305                                     .len = sizeof(struct ifla_port_vsi)},
1306         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1307                                       .len = PORT_UUID_MAX },
1308         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1309                                     .len = PORT_UUID_MAX },
1310         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1311         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1312 };
1313
1314 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1315 {
1316         struct net *net = sock_net(skb->sk);
1317         int h, s_h;
1318         int idx = 0, s_idx;
1319         struct net_device *dev;
1320         struct hlist_head *head;
1321         struct nlattr *tb[IFLA_MAX+1];
1322         u32 ext_filter_mask = 0;
1323         int err;
1324         int hdrlen;
1325
1326         s_h = cb->args[0];
1327         s_idx = cb->args[1];
1328
1329         cb->seq = net->dev_base_seq;
1330
1331         /* A hack to preserve kernel<->userspace interface.
1332          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1333          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1334          * what iproute2 < v3.9.0 used.
1335          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1336          * attribute, its netlink message is shorter than struct ifinfomsg.
1337          */
1338         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1339                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1340
1341         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1342
1343                 if (tb[IFLA_EXT_MASK])
1344                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1345         }
1346
1347         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1348                 idx = 0;
1349                 head = &net->dev_index_head[h];
1350                 hlist_for_each_entry(dev, head, index_hlist) {
1351                         if (idx < s_idx)
1352                                 goto cont;
1353                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1354                                                NETLINK_CB(cb->skb).portid,
1355                                                cb->nlh->nlmsg_seq, 0,
1356                                                NLM_F_MULTI,
1357                                                ext_filter_mask);
1358                         /* If we ran out of room on the first message,
1359                          * we're in trouble
1360                          */
1361                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1362
1363                         if (err < 0)
1364                                 goto out;
1365
1366                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1367 cont:
1368                         idx++;
1369                 }
1370         }
1371 out:
1372         cb->args[1] = idx;
1373         cb->args[0] = h;
1374
1375         return skb->len;
1376 }
1377
1378 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1379 {
1380         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1381 }
1382 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1383
1384 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1385 {
1386         struct net *net;
1387         /* Examine the link attributes and figure out which
1388          * network namespace we are talking about.
1389          */
1390         if (tb[IFLA_NET_NS_PID])
1391                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1392         else if (tb[IFLA_NET_NS_FD])
1393                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1394         else
1395                 net = get_net(src_net);
1396         return net;
1397 }
1398 EXPORT_SYMBOL(rtnl_link_get_net);
1399
1400 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1401 {
1402         if (dev) {
1403                 if (tb[IFLA_ADDRESS] &&
1404                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1405                         return -EINVAL;
1406
1407                 if (tb[IFLA_BROADCAST] &&
1408                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1409                         return -EINVAL;
1410         }
1411
1412         if (tb[IFLA_AF_SPEC]) {
1413                 struct nlattr *af;
1414                 int rem, err;
1415
1416                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1417                         const struct rtnl_af_ops *af_ops;
1418
1419                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1420                                 return -EAFNOSUPPORT;
1421
1422                         if (!af_ops->set_link_af)
1423                                 return -EOPNOTSUPP;
1424
1425                         if (af_ops->validate_link_af) {
1426                                 err = af_ops->validate_link_af(dev, af);
1427                                 if (err < 0)
1428                                         return err;
1429                         }
1430                 }
1431         }
1432
1433         return 0;
1434 }
1435
1436 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1437 {
1438         const struct net_device_ops *ops = dev->netdev_ops;
1439         int err = -EINVAL;
1440
1441         if (tb[IFLA_VF_MAC]) {
1442                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1443
1444                 err = -EOPNOTSUPP;
1445                 if (ops->ndo_set_vf_mac)
1446                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1447                                                   ivm->mac);
1448                 if (err < 0)
1449                         return err;
1450         }
1451
1452         if (tb[IFLA_VF_VLAN]) {
1453                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1454
1455                 err = -EOPNOTSUPP;
1456                 if (ops->ndo_set_vf_vlan)
1457                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1458                                                    ivv->qos);
1459                 if (err < 0)
1460                         return err;
1461         }
1462
1463         if (tb[IFLA_VF_TX_RATE]) {
1464                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1465                 struct ifla_vf_info ivf;
1466
1467                 err = -EOPNOTSUPP;
1468                 if (ops->ndo_get_vf_config)
1469                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1470                 if (err < 0)
1471                         return err;
1472
1473                 err = -EOPNOTSUPP;
1474                 if (ops->ndo_set_vf_rate)
1475                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1476                                                    ivf.min_tx_rate,
1477                                                    ivt->rate);
1478                 if (err < 0)
1479                         return err;
1480         }
1481
1482         if (tb[IFLA_VF_RATE]) {
1483                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1484
1485                 err = -EOPNOTSUPP;
1486                 if (ops->ndo_set_vf_rate)
1487                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1488                                                    ivt->min_tx_rate,
1489                                                    ivt->max_tx_rate);
1490                 if (err < 0)
1491                         return err;
1492         }
1493
1494         if (tb[IFLA_VF_SPOOFCHK]) {
1495                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1496
1497                 err = -EOPNOTSUPP;
1498                 if (ops->ndo_set_vf_spoofchk)
1499                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1500                                                        ivs->setting);
1501                 if (err < 0)
1502                         return err;
1503         }
1504
1505         if (tb[IFLA_VF_LINK_STATE]) {
1506                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1507
1508                 err = -EOPNOTSUPP;
1509                 if (ops->ndo_set_vf_link_state)
1510                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1511                                                          ivl->link_state);
1512                 if (err < 0)
1513                         return err;
1514         }
1515
1516         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1517                 struct ifla_vf_rss_query_en *ivrssq_en;
1518
1519                 err = -EOPNOTSUPP;
1520                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1521                 if (ops->ndo_set_vf_rss_query_en)
1522                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1523                                                            ivrssq_en->setting);
1524                 if (err < 0)
1525                         return err;
1526         }
1527
1528         return err;
1529 }
1530
1531 static int do_set_master(struct net_device *dev, int ifindex)
1532 {
1533         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1534         const struct net_device_ops *ops;
1535         int err;
1536
1537         if (upper_dev) {
1538                 if (upper_dev->ifindex == ifindex)
1539                         return 0;
1540                 ops = upper_dev->netdev_ops;
1541                 if (ops->ndo_del_slave) {
1542                         err = ops->ndo_del_slave(upper_dev, dev);
1543                         if (err)
1544                                 return err;
1545                 } else {
1546                         return -EOPNOTSUPP;
1547                 }
1548         }
1549
1550         if (ifindex) {
1551                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1552                 if (!upper_dev)
1553                         return -EINVAL;
1554                 ops = upper_dev->netdev_ops;
1555                 if (ops->ndo_add_slave) {
1556                         err = ops->ndo_add_slave(upper_dev, dev);
1557                         if (err)
1558                                 return err;
1559                 } else {
1560                         return -EOPNOTSUPP;
1561                 }
1562         }
1563         return 0;
1564 }
1565
1566 #define DO_SETLINK_MODIFIED     0x01
1567 /* notify flag means notify + modified. */
1568 #define DO_SETLINK_NOTIFY       0x03
1569 static int do_setlink(const struct sk_buff *skb,
1570                       struct net_device *dev, struct ifinfomsg *ifm,
1571                       struct nlattr **tb, char *ifname, int status)
1572 {
1573         const struct net_device_ops *ops = dev->netdev_ops;
1574         int err;
1575
1576         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1577                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1578                 if (IS_ERR(net)) {
1579                         err = PTR_ERR(net);
1580                         goto errout;
1581                 }
1582                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1583                         put_net(net);
1584                         err = -EPERM;
1585                         goto errout;
1586                 }
1587                 err = dev_change_net_namespace(dev, net, ifname);
1588                 put_net(net);
1589                 if (err)
1590                         goto errout;
1591                 status |= DO_SETLINK_MODIFIED;
1592         }
1593
1594         if (tb[IFLA_MAP]) {
1595                 struct rtnl_link_ifmap *u_map;
1596                 struct ifmap k_map;
1597
1598                 if (!ops->ndo_set_config) {
1599                         err = -EOPNOTSUPP;
1600                         goto errout;
1601                 }
1602
1603                 if (!netif_device_present(dev)) {
1604                         err = -ENODEV;
1605                         goto errout;
1606                 }
1607
1608                 u_map = nla_data(tb[IFLA_MAP]);
1609                 k_map.mem_start = (unsigned long) u_map->mem_start;
1610                 k_map.mem_end = (unsigned long) u_map->mem_end;
1611                 k_map.base_addr = (unsigned short) u_map->base_addr;
1612                 k_map.irq = (unsigned char) u_map->irq;
1613                 k_map.dma = (unsigned char) u_map->dma;
1614                 k_map.port = (unsigned char) u_map->port;
1615
1616                 err = ops->ndo_set_config(dev, &k_map);
1617                 if (err < 0)
1618                         goto errout;
1619
1620                 status |= DO_SETLINK_NOTIFY;
1621         }
1622
1623         if (tb[IFLA_ADDRESS]) {
1624                 struct sockaddr *sa;
1625                 int len;
1626
1627                 len = sizeof(sa_family_t) + dev->addr_len;
1628                 sa = kmalloc(len, GFP_KERNEL);
1629                 if (!sa) {
1630                         err = -ENOMEM;
1631                         goto errout;
1632                 }
1633                 sa->sa_family = dev->type;
1634                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1635                        dev->addr_len);
1636                 err = dev_set_mac_address(dev, sa);
1637                 kfree(sa);
1638                 if (err)
1639                         goto errout;
1640                 status |= DO_SETLINK_MODIFIED;
1641         }
1642
1643         if (tb[IFLA_MTU]) {
1644                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1645                 if (err < 0)
1646                         goto errout;
1647                 status |= DO_SETLINK_MODIFIED;
1648         }
1649
1650         if (tb[IFLA_GROUP]) {
1651                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1652                 status |= DO_SETLINK_NOTIFY;
1653         }
1654
1655         /*
1656          * Interface selected by interface index but interface
1657          * name provided implies that a name change has been
1658          * requested.
1659          */
1660         if (ifm->ifi_index > 0 && ifname[0]) {
1661                 err = dev_change_name(dev, ifname);
1662                 if (err < 0)
1663                         goto errout;
1664                 status |= DO_SETLINK_MODIFIED;
1665         }
1666
1667         if (tb[IFLA_IFALIAS]) {
1668                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1669                                     nla_len(tb[IFLA_IFALIAS]));
1670                 if (err < 0)
1671                         goto errout;
1672                 status |= DO_SETLINK_NOTIFY;
1673         }
1674
1675         if (tb[IFLA_BROADCAST]) {
1676                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1677                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1678         }
1679
1680         if (ifm->ifi_flags || ifm->ifi_change) {
1681                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1682                 if (err < 0)
1683                         goto errout;
1684         }
1685
1686         if (tb[IFLA_MASTER]) {
1687                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1688                 if (err)
1689                         goto errout;
1690                 status |= DO_SETLINK_MODIFIED;
1691         }
1692
1693         if (tb[IFLA_CARRIER]) {
1694                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1695                 if (err)
1696                         goto errout;
1697                 status |= DO_SETLINK_MODIFIED;
1698         }
1699
1700         if (tb[IFLA_TXQLEN]) {
1701                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1702
1703                 if (dev->tx_queue_len ^ value)
1704                         status |= DO_SETLINK_NOTIFY;
1705
1706                 dev->tx_queue_len = value;
1707         }
1708
1709         if (tb[IFLA_OPERSTATE])
1710                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1711
1712         if (tb[IFLA_LINKMODE]) {
1713                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1714
1715                 write_lock_bh(&dev_base_lock);
1716                 if (dev->link_mode ^ value)
1717                         status |= DO_SETLINK_NOTIFY;
1718                 dev->link_mode = value;
1719                 write_unlock_bh(&dev_base_lock);
1720         }
1721
1722         if (tb[IFLA_VFINFO_LIST]) {
1723                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1724                 struct nlattr *attr;
1725                 int rem;
1726
1727                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1728                         if (nla_type(attr) != IFLA_VF_INFO ||
1729                             nla_len(attr) < NLA_HDRLEN) {
1730                                 err = -EINVAL;
1731                                 goto errout;
1732                         }
1733                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1734                                                ifla_vf_policy);
1735                         if (err < 0)
1736                                 goto errout;
1737                         err = do_setvfinfo(dev, vfinfo);
1738                         if (err < 0)
1739                                 goto errout;
1740                         status |= DO_SETLINK_NOTIFY;
1741                 }
1742         }
1743         err = 0;
1744
1745         if (tb[IFLA_VF_PORTS]) {
1746                 struct nlattr *port[IFLA_PORT_MAX+1];
1747                 struct nlattr *attr;
1748                 int vf;
1749                 int rem;
1750
1751                 err = -EOPNOTSUPP;
1752                 if (!ops->ndo_set_vf_port)
1753                         goto errout;
1754
1755                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1756                         if (nla_type(attr) != IFLA_VF_PORT)
1757                                 continue;
1758                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1759                                 attr, ifla_port_policy);
1760                         if (err < 0)
1761                                 goto errout;
1762                         if (!port[IFLA_PORT_VF]) {
1763                                 err = -EOPNOTSUPP;
1764                                 goto errout;
1765                         }
1766                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1767                         err = ops->ndo_set_vf_port(dev, vf, port);
1768                         if (err < 0)
1769                                 goto errout;
1770                         status |= DO_SETLINK_NOTIFY;
1771                 }
1772         }
1773         err = 0;
1774
1775         if (tb[IFLA_PORT_SELF]) {
1776                 struct nlattr *port[IFLA_PORT_MAX+1];
1777
1778                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1779                         tb[IFLA_PORT_SELF], ifla_port_policy);
1780                 if (err < 0)
1781                         goto errout;
1782
1783                 err = -EOPNOTSUPP;
1784                 if (ops->ndo_set_vf_port)
1785                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1786                 if (err < 0)
1787                         goto errout;
1788                 status |= DO_SETLINK_NOTIFY;
1789         }
1790
1791         if (tb[IFLA_AF_SPEC]) {
1792                 struct nlattr *af;
1793                 int rem;
1794
1795                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1796                         const struct rtnl_af_ops *af_ops;
1797
1798                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1799                                 BUG();
1800
1801                         err = af_ops->set_link_af(dev, af);
1802                         if (err < 0)
1803                                 goto errout;
1804
1805                         status |= DO_SETLINK_NOTIFY;
1806                 }
1807         }
1808         err = 0;
1809
1810 errout:
1811         if (status & DO_SETLINK_MODIFIED) {
1812                 if (status & DO_SETLINK_NOTIFY)
1813                         netdev_state_change(dev);
1814
1815                 if (err < 0)
1816                         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",
1817                                              dev->name);
1818         }
1819
1820         return err;
1821 }
1822
1823 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1824 {
1825         struct net *net = sock_net(skb->sk);
1826         struct ifinfomsg *ifm;
1827         struct net_device *dev;
1828         int err;
1829         struct nlattr *tb[IFLA_MAX+1];
1830         char ifname[IFNAMSIZ];
1831
1832         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1833         if (err < 0)
1834                 goto errout;
1835
1836         if (tb[IFLA_IFNAME])
1837                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1838         else
1839                 ifname[0] = '\0';
1840
1841         err = -EINVAL;
1842         ifm = nlmsg_data(nlh);
1843         if (ifm->ifi_index > 0)
1844                 dev = __dev_get_by_index(net, ifm->ifi_index);
1845         else if (tb[IFLA_IFNAME])
1846                 dev = __dev_get_by_name(net, ifname);
1847         else
1848                 goto errout;
1849
1850         if (dev == NULL) {
1851                 err = -ENODEV;
1852                 goto errout;
1853         }
1854
1855         err = validate_linkmsg(dev, tb);
1856         if (err < 0)
1857                 goto errout;
1858
1859         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1860 errout:
1861         return err;
1862 }
1863
1864 static int rtnl_group_dellink(const struct net *net, int group)
1865 {
1866         struct net_device *dev, *aux;
1867         LIST_HEAD(list_kill);
1868         bool found = false;
1869
1870         if (!group)
1871                 return -EPERM;
1872
1873         for_each_netdev(net, dev) {
1874                 if (dev->group == group) {
1875                         const struct rtnl_link_ops *ops;
1876
1877                         found = true;
1878                         ops = dev->rtnl_link_ops;
1879                         if (!ops || !ops->dellink)
1880                                 return -EOPNOTSUPP;
1881                 }
1882         }
1883
1884         if (!found)
1885                 return -ENODEV;
1886
1887         for_each_netdev_safe(net, dev, aux) {
1888                 if (dev->group == group) {
1889                         const struct rtnl_link_ops *ops;
1890
1891                         ops = dev->rtnl_link_ops;
1892                         ops->dellink(dev, &list_kill);
1893                 }
1894         }
1895         unregister_netdevice_many(&list_kill);
1896
1897         return 0;
1898 }
1899
1900 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1901 {
1902         struct net *net = sock_net(skb->sk);
1903         const struct rtnl_link_ops *ops;
1904         struct net_device *dev;
1905         struct ifinfomsg *ifm;
1906         char ifname[IFNAMSIZ];
1907         struct nlattr *tb[IFLA_MAX+1];
1908         int err;
1909         LIST_HEAD(list_kill);
1910
1911         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1912         if (err < 0)
1913                 return err;
1914
1915         if (tb[IFLA_IFNAME])
1916                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1917
1918         ifm = nlmsg_data(nlh);
1919         if (ifm->ifi_index > 0)
1920                 dev = __dev_get_by_index(net, ifm->ifi_index);
1921         else if (tb[IFLA_IFNAME])
1922                 dev = __dev_get_by_name(net, ifname);
1923         else if (tb[IFLA_GROUP])
1924                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1925         else
1926                 return -EINVAL;
1927
1928         if (!dev)
1929                 return -ENODEV;
1930
1931         ops = dev->rtnl_link_ops;
1932         if (!ops || !ops->dellink)
1933                 return -EOPNOTSUPP;
1934
1935         ops->dellink(dev, &list_kill);
1936         unregister_netdevice_many(&list_kill);
1937         return 0;
1938 }
1939
1940 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1941 {
1942         unsigned int old_flags;
1943         int err;
1944
1945         old_flags = dev->flags;
1946         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1947                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1948                 if (err < 0)
1949                         return err;
1950         }
1951
1952         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1953
1954         __dev_notify_flags(dev, old_flags, ~0U);
1955         return 0;
1956 }
1957 EXPORT_SYMBOL(rtnl_configure_link);
1958
1959 struct net_device *rtnl_create_link(struct net *net,
1960         const char *ifname, unsigned char name_assign_type,
1961         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1962 {
1963         int err;
1964         struct net_device *dev;
1965         unsigned int num_tx_queues = 1;
1966         unsigned int num_rx_queues = 1;
1967
1968         if (tb[IFLA_NUM_TX_QUEUES])
1969                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1970         else if (ops->get_num_tx_queues)
1971                 num_tx_queues = ops->get_num_tx_queues();
1972
1973         if (tb[IFLA_NUM_RX_QUEUES])
1974                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1975         else if (ops->get_num_rx_queues)
1976                 num_rx_queues = ops->get_num_rx_queues();
1977
1978         err = -ENOMEM;
1979         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1980                                ops->setup, num_tx_queues, num_rx_queues);
1981         if (!dev)
1982                 goto err;
1983
1984         dev_net_set(dev, net);
1985         dev->rtnl_link_ops = ops;
1986         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1987
1988         if (tb[IFLA_MTU])
1989                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1990         if (tb[IFLA_ADDRESS]) {
1991                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1992                                 nla_len(tb[IFLA_ADDRESS]));
1993                 dev->addr_assign_type = NET_ADDR_SET;
1994         }
1995         if (tb[IFLA_BROADCAST])
1996                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1997                                 nla_len(tb[IFLA_BROADCAST]));
1998         if (tb[IFLA_TXQLEN])
1999                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2000         if (tb[IFLA_OPERSTATE])
2001                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2002         if (tb[IFLA_LINKMODE])
2003                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2004         if (tb[IFLA_GROUP])
2005                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2006
2007         return dev;
2008
2009 err:
2010         return ERR_PTR(err);
2011 }
2012 EXPORT_SYMBOL(rtnl_create_link);
2013
2014 static int rtnl_group_changelink(const struct sk_buff *skb,
2015                 struct net *net, int group,
2016                 struct ifinfomsg *ifm,
2017                 struct nlattr **tb)
2018 {
2019         struct net_device *dev, *aux;
2020         int err;
2021
2022         for_each_netdev_safe(net, dev, aux) {
2023                 if (dev->group == group) {
2024                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2025                         if (err < 0)
2026                                 return err;
2027                 }
2028         }
2029
2030         return 0;
2031 }
2032
2033 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2034 {
2035         struct net *net = sock_net(skb->sk);
2036         const struct rtnl_link_ops *ops;
2037         const struct rtnl_link_ops *m_ops = NULL;
2038         struct net_device *dev;
2039         struct net_device *master_dev = NULL;
2040         struct ifinfomsg *ifm;
2041         char kind[MODULE_NAME_LEN];
2042         char ifname[IFNAMSIZ];
2043         struct nlattr *tb[IFLA_MAX+1];
2044         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2045         unsigned char name_assign_type = NET_NAME_USER;
2046         int err;
2047
2048 #ifdef CONFIG_MODULES
2049 replay:
2050 #endif
2051         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2052         if (err < 0)
2053                 return err;
2054
2055         if (tb[IFLA_IFNAME])
2056                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2057         else
2058                 ifname[0] = '\0';
2059
2060         ifm = nlmsg_data(nlh);
2061         if (ifm->ifi_index > 0)
2062                 dev = __dev_get_by_index(net, ifm->ifi_index);
2063         else {
2064                 if (ifname[0])
2065                         dev = __dev_get_by_name(net, ifname);
2066                 else
2067                         dev = NULL;
2068         }
2069
2070         if (dev) {
2071                 master_dev = netdev_master_upper_dev_get(dev);
2072                 if (master_dev)
2073                         m_ops = master_dev->rtnl_link_ops;
2074         }
2075
2076         err = validate_linkmsg(dev, tb);
2077         if (err < 0)
2078                 return err;
2079
2080         if (tb[IFLA_LINKINFO]) {
2081                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2082                                        tb[IFLA_LINKINFO], ifla_info_policy);
2083                 if (err < 0)
2084                         return err;
2085         } else
2086                 memset(linkinfo, 0, sizeof(linkinfo));
2087
2088         if (linkinfo[IFLA_INFO_KIND]) {
2089                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2090                 ops = rtnl_link_ops_get(kind);
2091         } else {
2092                 kind[0] = '\0';
2093                 ops = NULL;
2094         }
2095
2096         if (1) {
2097                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2098                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2099                 struct nlattr **data = NULL;
2100                 struct nlattr **slave_data = NULL;
2101                 struct net *dest_net, *link_net = NULL;
2102
2103                 if (ops) {
2104                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2105                                 err = nla_parse_nested(attr, ops->maxtype,
2106                                                        linkinfo[IFLA_INFO_DATA],
2107                                                        ops->policy);
2108                                 if (err < 0)
2109                                         return err;
2110                                 data = attr;
2111                         }
2112                         if (ops->validate) {
2113                                 err = ops->validate(tb, data);
2114                                 if (err < 0)
2115                                         return err;
2116                         }
2117                 }
2118
2119                 if (m_ops) {
2120                         if (m_ops->slave_maxtype &&
2121                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2122                                 err = nla_parse_nested(slave_attr,
2123                                                        m_ops->slave_maxtype,
2124                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2125                                                        m_ops->slave_policy);
2126                                 if (err < 0)
2127                                         return err;
2128                                 slave_data = slave_attr;
2129                         }
2130                         if (m_ops->slave_validate) {
2131                                 err = m_ops->slave_validate(tb, slave_data);
2132                                 if (err < 0)
2133                                         return err;
2134                         }
2135                 }
2136
2137                 if (dev) {
2138                         int status = 0;
2139
2140                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2141                                 return -EEXIST;
2142                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2143                                 return -EOPNOTSUPP;
2144
2145                         if (linkinfo[IFLA_INFO_DATA]) {
2146                                 if (!ops || ops != dev->rtnl_link_ops ||
2147                                     !ops->changelink)
2148                                         return -EOPNOTSUPP;
2149
2150                                 err = ops->changelink(dev, tb, data);
2151                                 if (err < 0)
2152                                         return err;
2153                                 status |= DO_SETLINK_NOTIFY;
2154                         }
2155
2156                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2157                                 if (!m_ops || !m_ops->slave_changelink)
2158                                         return -EOPNOTSUPP;
2159
2160                                 err = m_ops->slave_changelink(master_dev, dev,
2161                                                               tb, slave_data);
2162                                 if (err < 0)
2163                                         return err;
2164                                 status |= DO_SETLINK_NOTIFY;
2165                         }
2166
2167                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2168                 }
2169
2170                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2171                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2172                                 return rtnl_group_changelink(skb, net,
2173                                                 nla_get_u32(tb[IFLA_GROUP]),
2174                                                 ifm, tb);
2175                         return -ENODEV;
2176                 }
2177
2178                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2179                         return -EOPNOTSUPP;
2180
2181                 if (!ops) {
2182 #ifdef CONFIG_MODULES
2183                         if (kind[0]) {
2184                                 __rtnl_unlock();
2185                                 request_module("rtnl-link-%s", kind);
2186                                 rtnl_lock();
2187                                 ops = rtnl_link_ops_get(kind);
2188                                 if (ops)
2189                                         goto replay;
2190                         }
2191 #endif
2192                         return -EOPNOTSUPP;
2193                 }
2194
2195                 if (!ops->setup)
2196                         return -EOPNOTSUPP;
2197
2198                 if (!ifname[0]) {
2199                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2200                         name_assign_type = NET_NAME_ENUM;
2201                 }
2202
2203                 dest_net = rtnl_link_get_net(net, tb);
2204                 if (IS_ERR(dest_net))
2205                         return PTR_ERR(dest_net);
2206
2207                 err = -EPERM;
2208                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2209                         goto out;
2210
2211                 if (tb[IFLA_LINK_NETNSID]) {
2212                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2213
2214                         link_net = get_net_ns_by_id(dest_net, id);
2215                         if (!link_net) {
2216                                 err =  -EINVAL;
2217                                 goto out;
2218                         }
2219                         err = -EPERM;
2220                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2221                                 goto out;
2222                 }
2223
2224                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2225                                        name_assign_type, ops, tb);
2226                 if (IS_ERR(dev)) {
2227                         err = PTR_ERR(dev);
2228                         goto out;
2229                 }
2230
2231                 dev->ifindex = ifm->ifi_index;
2232
2233                 if (ops->newlink) {
2234                         err = ops->newlink(link_net ? : net, dev, tb, data);
2235                         /* Drivers should call free_netdev() in ->destructor
2236                          * and unregister it on failure after registration
2237                          * so that device could be finally freed in rtnl_unlock.
2238                          */
2239                         if (err < 0) {
2240                                 /* If device is not registered at all, free it now */
2241                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2242                                         free_netdev(dev);
2243                                 goto out;
2244                         }
2245                 } else {
2246                         err = register_netdevice(dev);
2247                         if (err < 0) {
2248                                 free_netdev(dev);
2249                                 goto out;
2250                         }
2251                 }
2252                 err = rtnl_configure_link(dev, ifm);
2253                 if (err < 0)
2254                         goto out_unregister;
2255                 if (link_net) {
2256                         err = dev_change_net_namespace(dev, dest_net, ifname);
2257                         if (err < 0)
2258                                 goto out_unregister;
2259                 }
2260 out:
2261                 if (link_net)
2262                         put_net(link_net);
2263                 put_net(dest_net);
2264                 return err;
2265 out_unregister:
2266                 if (ops->newlink) {
2267                         LIST_HEAD(list_kill);
2268
2269                         ops->dellink(dev, &list_kill);
2270                         unregister_netdevice_many(&list_kill);
2271                 } else {
2272                         unregister_netdevice(dev);
2273                 }
2274                 goto out;
2275         }
2276 }
2277
2278 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2279 {
2280         struct net *net = sock_net(skb->sk);
2281         struct ifinfomsg *ifm;
2282         char ifname[IFNAMSIZ];
2283         struct nlattr *tb[IFLA_MAX+1];
2284         struct net_device *dev = NULL;
2285         struct sk_buff *nskb;
2286         int err;
2287         u32 ext_filter_mask = 0;
2288
2289         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2290         if (err < 0)
2291                 return err;
2292
2293         if (tb[IFLA_IFNAME])
2294                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2295
2296         if (tb[IFLA_EXT_MASK])
2297                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2298
2299         ifm = nlmsg_data(nlh);
2300         if (ifm->ifi_index > 0)
2301                 dev = __dev_get_by_index(net, ifm->ifi_index);
2302         else if (tb[IFLA_IFNAME])
2303                 dev = __dev_get_by_name(net, ifname);
2304         else
2305                 return -EINVAL;
2306
2307         if (dev == NULL)
2308                 return -ENODEV;
2309
2310         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2311         if (nskb == NULL)
2312                 return -ENOBUFS;
2313
2314         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2315                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2316         if (err < 0) {
2317                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2318                 WARN_ON(err == -EMSGSIZE);
2319                 kfree_skb(nskb);
2320         } else
2321                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2322
2323         return err;
2324 }
2325
2326 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2327 {
2328         struct net *net = sock_net(skb->sk);
2329         struct net_device *dev;
2330         struct nlattr *tb[IFLA_MAX+1];
2331         u32 ext_filter_mask = 0;
2332         u16 min_ifinfo_dump_size = 0;
2333         int hdrlen;
2334
2335         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2336         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2337                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2338
2339         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2340                 if (tb[IFLA_EXT_MASK])
2341                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2342         }
2343
2344         if (!ext_filter_mask)
2345                 return NLMSG_GOODSIZE;
2346         /*
2347          * traverse the list of net devices and compute the minimum
2348          * buffer size based upon the filter mask.
2349          */
2350         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2351                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2352                                              if_nlmsg_size(dev,
2353                                                            ext_filter_mask));
2354         }
2355
2356         return min_ifinfo_dump_size;
2357 }
2358
2359 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2360 {
2361         int idx;
2362         int s_idx = cb->family;
2363
2364         if (s_idx == 0)
2365                 s_idx = 1;
2366         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2367                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2368                 if (idx < s_idx || idx == PF_PACKET)
2369                         continue;
2370                 if (rtnl_msg_handlers[idx] == NULL ||
2371                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2372                         continue;
2373                 if (idx > s_idx) {
2374                         memset(&cb->args[0], 0, sizeof(cb->args));
2375                         cb->prev_seq = 0;
2376                         cb->seq = 0;
2377                 }
2378                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2379                         break;
2380         }
2381         cb->family = idx;
2382
2383         return skb->len;
2384 }
2385
2386 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2387                                        unsigned int change, gfp_t flags)
2388 {
2389         struct net *net = dev_net(dev);
2390         struct sk_buff *skb;
2391         int err = -ENOBUFS;
2392         size_t if_info_size;
2393
2394         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2395         if (skb == NULL)
2396                 goto errout;
2397
2398         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2399         if (err < 0) {
2400                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2401                 WARN_ON(err == -EMSGSIZE);
2402                 kfree_skb(skb);
2403                 goto errout;
2404         }
2405         return skb;
2406 errout:
2407         if (err < 0)
2408                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2409         return NULL;
2410 }
2411
2412 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2413 {
2414         struct net *net = dev_net(dev);
2415
2416         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2417 }
2418
2419 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2420                   gfp_t flags)
2421 {
2422         struct sk_buff *skb;
2423
2424         if (dev->reg_state != NETREG_REGISTERED)
2425                 return;
2426
2427         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2428         if (skb)
2429                 rtmsg_ifinfo_send(skb, dev, flags);
2430 }
2431 EXPORT_SYMBOL(rtmsg_ifinfo);
2432
2433 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2434                                    struct net_device *dev,
2435                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2436                                    int type, unsigned int flags,
2437                                    int nlflags)
2438 {
2439         struct nlmsghdr *nlh;
2440         struct ndmsg *ndm;
2441
2442         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2443         if (!nlh)
2444                 return -EMSGSIZE;
2445
2446         ndm = nlmsg_data(nlh);
2447         ndm->ndm_family  = AF_BRIDGE;
2448         ndm->ndm_pad1    = 0;
2449         ndm->ndm_pad2    = 0;
2450         ndm->ndm_flags   = flags;
2451         ndm->ndm_type    = 0;
2452         ndm->ndm_ifindex = dev->ifindex;
2453         ndm->ndm_state   = NUD_PERMANENT;
2454
2455         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2456                 goto nla_put_failure;
2457         if (vid)
2458                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2459                         goto nla_put_failure;
2460
2461         nlmsg_end(skb, nlh);
2462         return 0;
2463
2464 nla_put_failure:
2465         nlmsg_cancel(skb, nlh);
2466         return -EMSGSIZE;
2467 }
2468
2469 static inline size_t rtnl_fdb_nlmsg_size(void)
2470 {
2471         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2472 }
2473
2474 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2475 {
2476         struct net *net = dev_net(dev);
2477         struct sk_buff *skb;
2478         int err = -ENOBUFS;
2479
2480         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2481         if (!skb)
2482                 goto errout;
2483
2484         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2485                                       0, 0, type, NTF_SELF, 0);
2486         if (err < 0) {
2487                 kfree_skb(skb);
2488                 goto errout;
2489         }
2490
2491         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2492         return;
2493 errout:
2494         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2495 }
2496
2497 /**
2498  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2499  */
2500 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2501                      struct nlattr *tb[],
2502                      struct net_device *dev,
2503                      const unsigned char *addr, u16 vid,
2504                      u16 flags)
2505 {
2506         int err = -EINVAL;
2507
2508         /* If aging addresses are supported device will need to
2509          * implement its own handler for this.
2510          */
2511         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2512                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2513                 return err;
2514         }
2515
2516         if (vid) {
2517                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2518                 return err;
2519         }
2520
2521         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2522                 err = dev_uc_add_excl(dev, addr);
2523         else if (is_multicast_ether_addr(addr))
2524                 err = dev_mc_add_excl(dev, addr);
2525
2526         /* Only return duplicate errors if NLM_F_EXCL is set */
2527         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2528                 err = 0;
2529
2530         return err;
2531 }
2532 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2533
2534 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2535 {
2536         u16 vid = 0;
2537
2538         if (vlan_attr) {
2539                 if (nla_len(vlan_attr) != sizeof(u16)) {
2540                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2541                         return -EINVAL;
2542                 }
2543
2544                 vid = nla_get_u16(vlan_attr);
2545
2546                 if (!vid || vid >= VLAN_VID_MASK) {
2547                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2548                                 vid);
2549                         return -EINVAL;
2550                 }
2551         }
2552         *p_vid = vid;
2553         return 0;
2554 }
2555
2556 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2557 {
2558         struct net *net = sock_net(skb->sk);
2559         struct ndmsg *ndm;
2560         struct nlattr *tb[NDA_MAX+1];
2561         struct net_device *dev;
2562         u8 *addr;
2563         u16 vid;
2564         int err;
2565
2566         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2567         if (err < 0)
2568                 return err;
2569
2570         ndm = nlmsg_data(nlh);
2571         if (ndm->ndm_ifindex == 0) {
2572                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2573                 return -EINVAL;
2574         }
2575
2576         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2577         if (dev == NULL) {
2578                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2579                 return -ENODEV;
2580         }
2581
2582         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2583                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2584                 return -EINVAL;
2585         }
2586
2587         addr = nla_data(tb[NDA_LLADDR]);
2588
2589         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2590         if (err)
2591                 return err;
2592
2593         err = -EOPNOTSUPP;
2594
2595         /* Support fdb on master device the net/bridge default case */
2596         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2597             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2598                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2599                 const struct net_device_ops *ops = br_dev->netdev_ops;
2600
2601                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2602                                        nlh->nlmsg_flags);
2603                 if (err)
2604                         goto out;
2605                 else
2606                         ndm->ndm_flags &= ~NTF_MASTER;
2607         }
2608
2609         /* Embedded bridge, macvlan, and any other device support */
2610         if ((ndm->ndm_flags & NTF_SELF)) {
2611                 if (dev->netdev_ops->ndo_fdb_add)
2612                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2613                                                            vid,
2614                                                            nlh->nlmsg_flags);
2615                 else
2616                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2617                                                nlh->nlmsg_flags);
2618
2619                 if (!err) {
2620                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2621                         ndm->ndm_flags &= ~NTF_SELF;
2622                 }
2623         }
2624 out:
2625         return err;
2626 }
2627
2628 /**
2629  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2630  */
2631 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2632                      struct nlattr *tb[],
2633                      struct net_device *dev,
2634                      const unsigned char *addr, u16 vid)
2635 {
2636         int err = -EINVAL;
2637
2638         /* If aging addresses are supported device will need to
2639          * implement its own handler for this.
2640          */
2641         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2642                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2643                 return err;
2644         }
2645
2646         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2647                 err = dev_uc_del(dev, addr);
2648         else if (is_multicast_ether_addr(addr))
2649                 err = dev_mc_del(dev, addr);
2650
2651         return err;
2652 }
2653 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2654
2655 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2656 {
2657         struct net *net = sock_net(skb->sk);
2658         struct ndmsg *ndm;
2659         struct nlattr *tb[NDA_MAX+1];
2660         struct net_device *dev;
2661         int err = -EINVAL;
2662         __u8 *addr;
2663         u16 vid;
2664
2665         if (!netlink_capable(skb, CAP_NET_ADMIN))
2666                 return -EPERM;
2667
2668         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2669         if (err < 0)
2670                 return err;
2671
2672         ndm = nlmsg_data(nlh);
2673         if (ndm->ndm_ifindex == 0) {
2674                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2675                 return -EINVAL;
2676         }
2677
2678         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2679         if (dev == NULL) {
2680                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2681                 return -ENODEV;
2682         }
2683
2684         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2685                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2686                 return -EINVAL;
2687         }
2688
2689         addr = nla_data(tb[NDA_LLADDR]);
2690
2691         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2692         if (err)
2693                 return err;
2694
2695         err = -EOPNOTSUPP;
2696
2697         /* Support fdb on master device the net/bridge default case */
2698         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2699             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2700                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2701                 const struct net_device_ops *ops = br_dev->netdev_ops;
2702
2703                 if (ops->ndo_fdb_del)
2704                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2705
2706                 if (err)
2707                         goto out;
2708                 else
2709                         ndm->ndm_flags &= ~NTF_MASTER;
2710         }
2711
2712         /* Embedded bridge, macvlan, and any other device support */
2713         if (ndm->ndm_flags & NTF_SELF) {
2714                 if (dev->netdev_ops->ndo_fdb_del)
2715                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2716                                                            vid);
2717                 else
2718                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2719
2720                 if (!err) {
2721                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2722                         ndm->ndm_flags &= ~NTF_SELF;
2723                 }
2724         }
2725 out:
2726         return err;
2727 }
2728
2729 static int nlmsg_populate_fdb(struct sk_buff *skb,
2730                               struct netlink_callback *cb,
2731                               struct net_device *dev,
2732                               int *idx,
2733                               struct netdev_hw_addr_list *list)
2734 {
2735         struct netdev_hw_addr *ha;
2736         int err;
2737         u32 portid, seq;
2738
2739         portid = NETLINK_CB(cb->skb).portid;
2740         seq = cb->nlh->nlmsg_seq;
2741
2742         list_for_each_entry(ha, &list->list, list) {
2743                 if (*idx < cb->args[0])
2744                         goto skip;
2745
2746                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2747                                               portid, seq,
2748                                               RTM_NEWNEIGH, NTF_SELF,
2749                                               NLM_F_MULTI);
2750                 if (err < 0)
2751                         return err;
2752 skip:
2753                 *idx += 1;
2754         }
2755         return 0;
2756 }
2757
2758 /**
2759  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2760  * @nlh: netlink message header
2761  * @dev: netdevice
2762  *
2763  * Default netdevice operation to dump the existing unicast address list.
2764  * Returns number of addresses from list put in skb.
2765  */
2766 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2767                       struct netlink_callback *cb,
2768                       struct net_device *dev,
2769                       struct net_device *filter_dev,
2770                       int idx)
2771 {
2772         int err;
2773
2774         netif_addr_lock_bh(dev);
2775         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2776         if (err)
2777                 goto out;
2778         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2779 out:
2780         netif_addr_unlock_bh(dev);
2781         return idx;
2782 }
2783 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2784
2785 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2786 {
2787         struct net_device *dev;
2788         struct nlattr *tb[IFLA_MAX+1];
2789         struct net_device *br_dev = NULL;
2790         const struct net_device_ops *ops = NULL;
2791         const struct net_device_ops *cops = NULL;
2792         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2793         struct net *net = sock_net(skb->sk);
2794         int brport_idx = 0;
2795         int br_idx = 0;
2796         int idx = 0;
2797
2798         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2799                         ifla_policy) == 0) {
2800                 if (tb[IFLA_MASTER])
2801                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2802         }
2803
2804         brport_idx = ifm->ifi_index;
2805
2806         if (br_idx) {
2807                 br_dev = __dev_get_by_index(net, br_idx);
2808                 if (!br_dev)
2809                         return -ENODEV;
2810
2811                 ops = br_dev->netdev_ops;
2812         }
2813
2814         for_each_netdev(net, dev) {
2815                 if (brport_idx && (dev->ifindex != brport_idx))
2816                         continue;
2817
2818                 if (!br_idx) { /* user did not specify a specific bridge */
2819                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2820                                 br_dev = netdev_master_upper_dev_get(dev);
2821                                 cops = br_dev->netdev_ops;
2822                         }
2823
2824                 } else {
2825                         if (dev != br_dev &&
2826                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2827                                 continue;
2828
2829                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2830                             !(dev->priv_flags & IFF_EBRIDGE))
2831                                 continue;
2832
2833                         cops = ops;
2834                 }
2835
2836                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2837                         if (cops && cops->ndo_fdb_dump)
2838                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2839                                                          idx);
2840                 }
2841
2842                 if (dev->netdev_ops->ndo_fdb_dump)
2843                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2844                                                             idx);
2845                 else
2846                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2847
2848                 cops = NULL;
2849         }
2850
2851         cb->args[0] = idx;
2852         return skb->len;
2853 }
2854
2855 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2856                                unsigned int attrnum, unsigned int flag)
2857 {
2858         if (mask & flag)
2859                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2860         return 0;
2861 }
2862
2863 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2864                             struct net_device *dev, u16 mode,
2865                             u32 flags, u32 mask, int nlflags)
2866 {
2867         struct nlmsghdr *nlh;
2868         struct ifinfomsg *ifm;
2869         struct nlattr *br_afspec;
2870         struct nlattr *protinfo;
2871         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2872         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2873
2874         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2875         if (nlh == NULL)
2876                 return -EMSGSIZE;
2877
2878         ifm = nlmsg_data(nlh);
2879         ifm->ifi_family = AF_BRIDGE;
2880         ifm->__ifi_pad = 0;
2881         ifm->ifi_type = dev->type;
2882         ifm->ifi_index = dev->ifindex;
2883         ifm->ifi_flags = dev_get_flags(dev);
2884         ifm->ifi_change = 0;
2885
2886
2887         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2888             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2889             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2890             (br_dev &&
2891              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2892             (dev->addr_len &&
2893              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2894             (dev->ifindex != dev_get_iflink(dev) &&
2895              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2896                 goto nla_put_failure;
2897
2898         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2899         if (!br_afspec)
2900                 goto nla_put_failure;
2901
2902         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2903                 nla_nest_cancel(skb, br_afspec);
2904                 goto nla_put_failure;
2905         }
2906
2907         if (mode != BRIDGE_MODE_UNDEF) {
2908                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2909                         nla_nest_cancel(skb, br_afspec);
2910                         goto nla_put_failure;
2911                 }
2912         }
2913         nla_nest_end(skb, br_afspec);
2914
2915         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2916         if (!protinfo)
2917                 goto nla_put_failure;
2918
2919         if (brport_nla_put_flag(skb, flags, mask,
2920                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2921             brport_nla_put_flag(skb, flags, mask,
2922                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2923             brport_nla_put_flag(skb, flags, mask,
2924                                 IFLA_BRPORT_FAST_LEAVE,
2925                                 BR_MULTICAST_FAST_LEAVE) ||
2926             brport_nla_put_flag(skb, flags, mask,
2927                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2928             brport_nla_put_flag(skb, flags, mask,
2929                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2930             brport_nla_put_flag(skb, flags, mask,
2931                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2932             brport_nla_put_flag(skb, flags, mask,
2933                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2934             brport_nla_put_flag(skb, flags, mask,
2935                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2936                 nla_nest_cancel(skb, protinfo);
2937                 goto nla_put_failure;
2938         }
2939
2940         nla_nest_end(skb, protinfo);
2941
2942         nlmsg_end(skb, nlh);
2943         return 0;
2944 nla_put_failure:
2945         nlmsg_cancel(skb, nlh);
2946         return -EMSGSIZE;
2947 }
2948 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2949
2950 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2951 {
2952         struct net *net = sock_net(skb->sk);
2953         struct net_device *dev;
2954         int idx = 0;
2955         u32 portid = NETLINK_CB(cb->skb).portid;
2956         u32 seq = cb->nlh->nlmsg_seq;
2957         u32 filter_mask = 0;
2958
2959         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2960                 struct nlattr *extfilt;
2961
2962                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2963                                           IFLA_EXT_MASK);
2964                 if (extfilt) {
2965                         if (nla_len(extfilt) < sizeof(filter_mask))
2966                                 return -EINVAL;
2967
2968                         filter_mask = nla_get_u32(extfilt);
2969                 }
2970         }
2971
2972         rcu_read_lock();
2973         for_each_netdev_rcu(net, dev) {
2974                 const struct net_device_ops *ops = dev->netdev_ops;
2975                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2976
2977                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2978                         if (idx >= cb->args[0] &&
2979                             br_dev->netdev_ops->ndo_bridge_getlink(
2980                                     skb, portid, seq, dev, filter_mask,
2981                                     NLM_F_MULTI) < 0)
2982                                 break;
2983                         idx++;
2984                 }
2985
2986                 if (ops->ndo_bridge_getlink) {
2987                         if (idx >= cb->args[0] &&
2988                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2989                                                     filter_mask,
2990                                                     NLM_F_MULTI) < 0)
2991                                 break;
2992                         idx++;
2993                 }
2994         }
2995         rcu_read_unlock();
2996         cb->args[0] = idx;
2997
2998         return skb->len;
2999 }
3000
3001 static inline size_t bridge_nlmsg_size(void)
3002 {
3003         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3004                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3005                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3006                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3007                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3008                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3009                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3010                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3011                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3012                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3013                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3014 }
3015
3016 static int rtnl_bridge_notify(struct net_device *dev)
3017 {
3018         struct net *net = dev_net(dev);
3019         struct sk_buff *skb;
3020         int err = -EOPNOTSUPP;
3021
3022         if (!dev->netdev_ops->ndo_bridge_getlink)
3023                 return 0;
3024
3025         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3026         if (!skb) {
3027                 err = -ENOMEM;
3028                 goto errout;
3029         }
3030
3031         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3032         if (err < 0)
3033                 goto errout;
3034
3035         if (!skb->len)
3036                 goto errout;
3037
3038         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3039         return 0;
3040 errout:
3041         WARN_ON(err == -EMSGSIZE);
3042         kfree_skb(skb);
3043         if (err)
3044                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3045         return err;
3046 }
3047
3048 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3049 {
3050         struct net *net = sock_net(skb->sk);
3051         struct ifinfomsg *ifm;
3052         struct net_device *dev;
3053         struct nlattr *br_spec, *attr = NULL;
3054         int rem, err = -EOPNOTSUPP;
3055         u16 flags = 0;
3056         bool have_flags = false;
3057
3058         if (nlmsg_len(nlh) < sizeof(*ifm))
3059                 return -EINVAL;
3060
3061         ifm = nlmsg_data(nlh);
3062         if (ifm->ifi_family != AF_BRIDGE)
3063                 return -EPFNOSUPPORT;
3064
3065         dev = __dev_get_by_index(net, ifm->ifi_index);
3066         if (!dev) {
3067                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3068                 return -ENODEV;
3069         }
3070
3071         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3072         if (br_spec) {
3073                 nla_for_each_nested(attr, br_spec, rem) {
3074                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3075                                 if (nla_len(attr) < sizeof(flags))
3076                                         return -EINVAL;
3077
3078                                 have_flags = true;
3079                                 flags = nla_get_u16(attr);
3080                                 break;
3081                         }
3082                 }
3083         }
3084
3085         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3086                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3087
3088                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3089                         err = -EOPNOTSUPP;
3090                         goto out;
3091                 }
3092
3093                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3094                 if (err)
3095                         goto out;
3096
3097                 flags &= ~BRIDGE_FLAGS_MASTER;
3098         }
3099
3100         if ((flags & BRIDGE_FLAGS_SELF)) {
3101                 if (!dev->netdev_ops->ndo_bridge_setlink)
3102                         err = -EOPNOTSUPP;
3103                 else
3104                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3105                                                                   flags);
3106                 if (!err) {
3107                         flags &= ~BRIDGE_FLAGS_SELF;
3108
3109                         /* Generate event to notify upper layer of bridge
3110                          * change
3111                          */
3112                         err = rtnl_bridge_notify(dev);
3113                 }
3114         }
3115
3116         if (have_flags)
3117                 memcpy(nla_data(attr), &flags, sizeof(flags));
3118 out:
3119         return err;
3120 }
3121
3122 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3123 {
3124         struct net *net = sock_net(skb->sk);
3125         struct ifinfomsg *ifm;
3126         struct net_device *dev;
3127         struct nlattr *br_spec, *attr = NULL;
3128         int rem, err = -EOPNOTSUPP;
3129         u16 flags = 0;
3130         bool have_flags = false;
3131
3132         if (nlmsg_len(nlh) < sizeof(*ifm))
3133                 return -EINVAL;
3134
3135         ifm = nlmsg_data(nlh);
3136         if (ifm->ifi_family != AF_BRIDGE)
3137                 return -EPFNOSUPPORT;
3138
3139         dev = __dev_get_by_index(net, ifm->ifi_index);
3140         if (!dev) {
3141                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3142                 return -ENODEV;
3143         }
3144
3145         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3146         if (br_spec) {
3147                 nla_for_each_nested(attr, br_spec, rem) {
3148                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3149                                 if (nla_len(attr) < sizeof(flags))
3150                                         return -EINVAL;
3151
3152                                 have_flags = true;
3153                                 flags = nla_get_u16(attr);
3154                                 break;
3155                         }
3156                 }
3157         }
3158
3159         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3160                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3161
3162                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3163                         err = -EOPNOTSUPP;
3164                         goto out;
3165                 }
3166
3167                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3168                 if (err)
3169                         goto out;
3170
3171                 flags &= ~BRIDGE_FLAGS_MASTER;
3172         }
3173
3174         if ((flags & BRIDGE_FLAGS_SELF)) {
3175                 if (!dev->netdev_ops->ndo_bridge_dellink)
3176                         err = -EOPNOTSUPP;
3177                 else
3178                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3179                                                                   flags);
3180
3181                 if (!err) {
3182                         flags &= ~BRIDGE_FLAGS_SELF;
3183
3184                         /* Generate event to notify upper layer of bridge
3185                          * change
3186                          */
3187                         err = rtnl_bridge_notify(dev);
3188                 }
3189         }
3190
3191         if (have_flags)
3192                 memcpy(nla_data(attr), &flags, sizeof(flags));
3193 out:
3194         return err;
3195 }
3196
3197 /* Process one rtnetlink message. */
3198
3199 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3200 {
3201         struct net *net = sock_net(skb->sk);
3202         rtnl_doit_func doit;
3203         int sz_idx, kind;
3204         int family;
3205         int type;
3206         int err;
3207
3208         type = nlh->nlmsg_type;
3209         if (type > RTM_MAX)
3210                 return -EOPNOTSUPP;
3211
3212         type -= RTM_BASE;
3213
3214         /* All the messages must have at least 1 byte length */
3215         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3216                 return 0;
3217
3218         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3219         sz_idx = type>>2;
3220         kind = type&3;
3221
3222         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3223                 return -EPERM;
3224
3225         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3226                 struct sock *rtnl;
3227                 rtnl_dumpit_func dumpit;
3228                 rtnl_calcit_func calcit;
3229                 u16 min_dump_alloc = 0;
3230
3231                 dumpit = rtnl_get_dumpit(family, type);
3232                 if (dumpit == NULL)
3233                         return -EOPNOTSUPP;
3234                 calcit = rtnl_get_calcit(family, type);
3235                 if (calcit)
3236                         min_dump_alloc = calcit(skb, nlh);
3237
3238                 __rtnl_unlock();
3239                 rtnl = net->rtnl;
3240                 {
3241                         struct netlink_dump_control c = {
3242                                 .dump           = dumpit,
3243                                 .min_dump_alloc = min_dump_alloc,
3244                         };
3245                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3246                 }
3247                 rtnl_lock();
3248                 return err;
3249         }
3250
3251         doit = rtnl_get_doit(family, type);
3252         if (doit == NULL)
3253                 return -EOPNOTSUPP;
3254
3255         return doit(skb, nlh);
3256 }
3257
3258 static void rtnetlink_rcv(struct sk_buff *skb)
3259 {
3260         rtnl_lock();
3261         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3262         rtnl_unlock();
3263 }
3264
3265 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3266 {
3267         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3268
3269         switch (event) {
3270         case NETDEV_UP:
3271         case NETDEV_DOWN:
3272         case NETDEV_PRE_UP:
3273         case NETDEV_POST_INIT:
3274         case NETDEV_REGISTER:
3275         case NETDEV_CHANGE:
3276         case NETDEV_PRE_TYPE_CHANGE:
3277         case NETDEV_GOING_DOWN:
3278         case NETDEV_UNREGISTER:
3279         case NETDEV_UNREGISTER_FINAL:
3280         case NETDEV_RELEASE:
3281         case NETDEV_JOIN:
3282         case NETDEV_BONDING_INFO:
3283                 break;
3284         default:
3285                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3286                 break;
3287         }
3288         return NOTIFY_DONE;
3289 }
3290
3291 static struct notifier_block rtnetlink_dev_notifier = {
3292         .notifier_call  = rtnetlink_event,
3293 };
3294
3295
3296 static int __net_init rtnetlink_net_init(struct net *net)
3297 {
3298         struct sock *sk;
3299         struct netlink_kernel_cfg cfg = {
3300                 .groups         = RTNLGRP_MAX,
3301                 .input          = rtnetlink_rcv,
3302                 .cb_mutex       = &rtnl_mutex,
3303                 .flags          = NL_CFG_F_NONROOT_RECV,
3304         };
3305
3306         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3307         if (!sk)
3308                 return -ENOMEM;
3309         net->rtnl = sk;
3310         return 0;
3311 }
3312
3313 static void __net_exit rtnetlink_net_exit(struct net *net)
3314 {
3315         netlink_kernel_release(net->rtnl);
3316         net->rtnl = NULL;
3317 }
3318
3319 static struct pernet_operations rtnetlink_net_ops = {
3320         .init = rtnetlink_net_init,
3321         .exit = rtnetlink_net_exit,
3322 };
3323
3324 void __init rtnetlink_init(void)
3325 {
3326         if (register_pernet_subsys(&rtnetlink_net_ops))
3327                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3328
3329         register_netdevice_notifier(&rtnetlink_dev_notifier);
3330
3331         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3332                       rtnl_dump_ifinfo, rtnl_calcit);
3333         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3334         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3335         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3336
3337         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3338         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3339
3340         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3341         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3342         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3343
3344         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3345         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3346         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3347 }
3348