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                 return size;
 823         } else
 824                 return 0;
 825 }
 826
 827 static size_t rtnl_port_size(const struct net_device *dev,
 828                              u32 ext_filter_mask)
 829 {
 830         size_t port_size = nla_total_size(4)            /* PORT_VF */
 831                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 832                 + nla_total_size(sizeof(struct ifla_port_vsi))
 833                                                         /* PORT_VSI_TYPE */
 834                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 835                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 836                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 837                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 838         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 839         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 840                 + port_size;
 841         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 842                 + port_size;
 843
 844         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 845             !(ext_filter_mask & RTEXT_FILTER_VF))
 846                 return 0;
 847         if (dev_num_vf(dev->dev.parent))
 848                 return port_self_size + vf_ports_size +
 849                         vf_port_size * dev_num_vf(dev->dev.parent);
 850         else
 851                 return port_self_size;
 852 }
 853
 854 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 855                                      u32 ext_filter_mask)
 856 {
 857         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 858                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 859                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 860                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 861                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 862                + nla_total_size(sizeof(struct rtnl_link_stats))
 863                + nla_total_size(sizeof(struct rtnl_link_stats64))
 864                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 865                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 866                + nla_total_size(4) /* IFLA_TXQLEN */
 867                + nla_total_size(4) /* IFLA_WEIGHT */
 868                + nla_total_size(4) /* IFLA_MTU */
 869                + nla_total_size(4) /* IFLA_LINK */
 870                + nla_total_size(4) /* IFLA_MASTER */
 871                + nla_total_size(1) /* IFLA_CARRIER */
 872                + nla_total_size(4) /* IFLA_PROMISCUITY */
 873                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 874                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 875                + nla_total_size(1) /* IFLA_OPERSTATE */
 876                + nla_total_size(1) /* IFLA_LINKMODE */
 877                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 878                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 879                + nla_total_size(ext_filter_mask
 880                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 881                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 882                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 883                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 884                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 885                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 886                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 887 }
 888
 889 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 890 {
 891         struct nlattr *vf_ports;
 892         struct nlattr *vf_port;
 893         int vf;
 894         int err;
 895
 896         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 897         if (!vf_ports)
 898                 return -EMSGSIZE;
 899
 900         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 901                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 902                 if (!vf_port)
 903                         goto nla_put_failure;
 904                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 905                         goto nla_put_failure;
 906                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 907                 if (err == -EMSGSIZE)
 908                         goto nla_put_failure;
 909                 if (err) {
 910                         nla_nest_cancel(skb, vf_port);
 911                         continue;
 912                 }
 913                 nla_nest_end(skb, vf_port);
 914         }
 915
 916         nla_nest_end(skb, vf_ports);
 917
 918         return 0;
 919
 920 nla_put_failure:
 921         nla_nest_cancel(skb, vf_ports);
 922         return -EMSGSIZE;
 923 }
 924
 925 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 926 {
 927         struct nlattr *port_self;
 928         int err;
 929
 930         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 931         if (!port_self)
 932                 return -EMSGSIZE;
 933
 934         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 935         if (err) {
 936                 nla_nest_cancel(skb, port_self);
 937                 return (err == -EMSGSIZE) ? err : 0;
 938         }
 939
 940         nla_nest_end(skb, port_self);
 941
 942         return 0;
 943 }
 944
 945 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 946                           u32 ext_filter_mask)
 947 {
 948         int err;
 949
 950         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 951             !(ext_filter_mask & RTEXT_FILTER_VF))
 952                 return 0;
 953
 954         err = rtnl_port_self_fill(skb, dev);
 955         if (err)
 956                 return err;
 957
 958         if (dev_num_vf(dev->dev.parent)) {
 959                 err = rtnl_vf_ports_fill(skb, dev);
 960                 if (err)
 961                         return err;
 962         }
 963
 964         return 0;
 965 }
 966
 967 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 968 {
 969         int err;
 970         struct netdev_phys_item_id ppid;
 971
 972         err = dev_get_phys_port_id(dev, &ppid);
 973         if (err) {
 974                 if (err == -EOPNOTSUPP)
 975                         return 0;
 976                 return err;
 977         }
 978
 979         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 980                 return -EMSGSIZE;
 981
 982         return 0;
 983 }
 984
 985 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
 986 {
 987         int err;
 988         struct netdev_phys_item_id psid;
 989
 990         err = netdev_switch_parent_id_get(dev, &psid);
 991         if (err) {
 992                 if (err == -EOPNOTSUPP)
 993                         return 0;
 994                 return err;
 995         }
 996
 997         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
 998                 return -EMSGSIZE;
 999
1000         return 0;
1001 }
1002
1003 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1004                             int type, u32 pid, u32 seq, u32 change,
1005                             unsigned int flags, u32 ext_filter_mask)
1006 {
1007         struct ifinfomsg *ifm;
1008         struct nlmsghdr *nlh;
1009         struct rtnl_link_stats64 temp;
1010         const struct rtnl_link_stats64 *stats;
1011         struct nlattr *attr, *af_spec;
1012         struct rtnl_af_ops *af_ops;
1013         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1014
1015         ASSERT_RTNL();
1016         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1017         if (nlh == NULL)
1018                 return -EMSGSIZE;
1019
1020         ifm = nlmsg_data(nlh);
1021         ifm->ifi_family = AF_UNSPEC;
1022         ifm->__ifi_pad = 0;
1023         ifm->ifi_type = dev->type;
1024         ifm->ifi_index = dev->ifindex;
1025         ifm->ifi_flags = dev_get_flags(dev);
1026         ifm->ifi_change = change;
1027
1028         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1029             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1030             nla_put_u8(skb, IFLA_OPERSTATE,
1031                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1032             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1033             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1034             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1035             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1036             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1037 #ifdef CONFIG_RPS
1038             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1039 #endif
1040             (dev->ifindex != dev->iflink &&
1041              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1042             (upper_dev &&
1043              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1044             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1045             (dev->qdisc &&
1046              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1047             (dev->ifalias &&
1048              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1049             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1050                         atomic_read(&dev->carrier_changes)))
1051                 goto nla_put_failure;
1052
1053         if (1) {
1054                 struct rtnl_link_ifmap map = {
1055                         .mem_start   = dev->mem_start,
1056                         .mem_end     = dev->mem_end,
1057                         .base_addr   = dev->base_addr,
1058                         .irq         = dev->irq,
1059                         .dma         = dev->dma,
1060                         .port        = dev->if_port,
1061                 };
1062                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1063                         goto nla_put_failure;
1064         }
1065
1066         if (dev->addr_len) {
1067                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1068                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1069                         goto nla_put_failure;
1070         }
1071
1072         if (rtnl_phys_port_id_fill(skb, dev))
1073                 goto nla_put_failure;
1074
1075         if (rtnl_phys_switch_id_fill(skb, dev))
1076                 goto nla_put_failure;
1077
1078         attr = nla_reserve(skb, IFLA_STATS,
1079                         sizeof(struct rtnl_link_stats));
1080         if (attr == NULL)
1081                 goto nla_put_failure;
1082
1083         stats = dev_get_stats(dev, &temp);
1084         copy_rtnl_link_stats(nla_data(attr), stats);
1085
1086         attr = nla_reserve(skb, IFLA_STATS64,
1087                         sizeof(struct rtnl_link_stats64));
1088         if (attr == NULL)
1089                 goto nla_put_failure;
1090         copy_rtnl_link_stats64(nla_data(attr), stats);
1091
1092         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1093             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1094                 goto nla_put_failure;
1095
1096         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1097             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1098                 int i;
1099
1100                 struct nlattr *vfinfo, *vf;
1101                 int num_vfs = dev_num_vf(dev->dev.parent);
1102
1103                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1104                 if (!vfinfo)
1105                         goto nla_put_failure;
1106                 for (i = 0; i < num_vfs; i++) {
1107                         struct ifla_vf_info ivi;
1108                         struct ifla_vf_mac vf_mac;
1109                         struct ifla_vf_vlan vf_vlan;
1110                         struct ifla_vf_rate vf_rate;
1111                         struct ifla_vf_tx_rate vf_tx_rate;
1112                         struct ifla_vf_spoofchk vf_spoofchk;
1113                         struct ifla_vf_link_state vf_linkstate;
1114
1115                         /*
1116                          * Not all SR-IOV capable drivers support the
1117                          * spoofcheck query.  Preset to -1 so the user
1118                          * space tool can detect that the driver didn't
1119                          * report anything.
1120                          */
1121                         ivi.spoofchk = -1;
1122                         memset(ivi.mac, 0, sizeof(ivi.mac));
1123                         /* The default value for VF link state is "auto"
1124                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1125                          */
1126                         ivi.linkstate = 0;
1127                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1128                                 break;
1129                         vf_mac.vf =
1130                                 vf_vlan.vf =
1131                                 vf_rate.vf =
1132                                 vf_tx_rate.vf =
1133                                 vf_spoofchk.vf =
1134                                 vf_linkstate.vf = ivi.vf;
1135
1136                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1137                         vf_vlan.vlan = ivi.vlan;
1138                         vf_vlan.qos = ivi.qos;
1139                         vf_tx_rate.rate = ivi.max_tx_rate;
1140                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1141                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1142                         vf_spoofchk.setting = ivi.spoofchk;
1143                         vf_linkstate.link_state = ivi.linkstate;
1144                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1145                         if (!vf) {
1146                                 nla_nest_cancel(skb, vfinfo);
1147                                 goto nla_put_failure;
1148                         }
1149                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1150                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1151                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1152                                     &vf_rate) ||
1153                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1154                                     &vf_tx_rate) ||
1155                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1156                                     &vf_spoofchk) ||
1157                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1158                                     &vf_linkstate))
1159                                 goto nla_put_failure;
1160                         nla_nest_end(skb, vf);
1161                 }
1162                 nla_nest_end(skb, vfinfo);
1163         }
1164
1165         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1166                 goto nla_put_failure;
1167
1168         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1169                 if (rtnl_link_fill(skb, dev) < 0)
1170                         goto nla_put_failure;
1171         }
1172
1173         if (dev->rtnl_link_ops &&
1174             dev->rtnl_link_ops->get_link_net) {
1175                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1176
1177                 if (!net_eq(dev_net(dev), link_net)) {
1178                         int id = peernet2id(dev_net(dev), link_net);
1179
1180                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1181                                 goto nla_put_failure;
1182                 }
1183         }
1184
1185         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1186                 goto nla_put_failure;
1187
1188         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1189                 if (af_ops->fill_link_af) {
1190                         struct nlattr *af;
1191                         int err;
1192
1193                         if (!(af = nla_nest_start(skb, af_ops->family)))
1194                                 goto nla_put_failure;
1195
1196                         err = af_ops->fill_link_af(skb, dev);
1197
1198                         /*
1199                          * Caller may return ENODATA to indicate that there
1200                          * was no data to be dumped. This is not an error, it
1201                          * means we should trim the attribute header and
1202                          * continue.
1203                          */
1204                         if (err == -ENODATA)
1205                                 nla_nest_cancel(skb, af);
1206                         else if (err < 0)
1207                                 goto nla_put_failure;
1208
1209                         nla_nest_end(skb, af);
1210                 }
1211         }
1212
1213         nla_nest_end(skb, af_spec);
1214
1215         nlmsg_end(skb, nlh);
1216         return 0;
1217
1218 nla_put_failure:
1219         nlmsg_cancel(skb, nlh);
1220         return -EMSGSIZE;
1221 }
1222
1223 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1224         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1225         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1226         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1227         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1228         [IFLA_MTU]              = { .type = NLA_U32 },
1229         [IFLA_LINK]             = { .type = NLA_U32 },
1230         [IFLA_MASTER]           = { .type = NLA_U32 },
1231         [IFLA_CARRIER]          = { .type = NLA_U8 },
1232         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1233         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1234         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1235         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1236         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1237         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1238         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1239         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1240         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1241         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1242         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1243         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1244         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1245         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1246         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1247         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1248         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1249         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1250         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1251         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1252 };
1253
1254 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1255         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1256         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1257         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1258         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1259 };
1260
1261 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1262         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1263 };
1264
1265 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1266         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1267         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1268         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1269         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1270         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1271         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1272 };
1273
1274 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1275         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1276         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1277                                     .len = PORT_PROFILE_MAX },
1278         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1279                                     .len = sizeof(struct ifla_port_vsi)},
1280         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1281                                       .len = PORT_UUID_MAX },
1282         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1283                                     .len = PORT_UUID_MAX },
1284         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1285         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1286 };
1287
1288 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1289 {
1290         struct net *net = sock_net(skb->sk);
1291         int h, s_h;
1292         int idx = 0, s_idx;
1293         struct net_device *dev;
1294         struct hlist_head *head;
1295         struct nlattr *tb[IFLA_MAX+1];
1296         u32 ext_filter_mask = 0;
1297         int err;
1298         int hdrlen;
1299
1300         s_h = cb->args[0];
1301         s_idx = cb->args[1];
1302
1303         rcu_read_lock();
1304         cb->seq = net->dev_base_seq;
1305
1306         /* A hack to preserve kernel<->userspace interface.
1307          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1308          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1309          * what iproute2 < v3.9.0 used.
1310          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1311          * attribute, its netlink message is shorter than struct ifinfomsg.
1312          */
1313         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1314                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1315
1316         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1317
1318                 if (tb[IFLA_EXT_MASK])
1319                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1320         }
1321
1322         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1323                 idx = 0;
1324                 head = &net->dev_index_head[h];
1325                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1326                         if (idx < s_idx)
1327                                 goto cont;
1328                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1329                                                NETLINK_CB(cb->skb).portid,
1330                                                cb->nlh->nlmsg_seq, 0,
1331                                                NLM_F_MULTI,
1332                                                ext_filter_mask);
1333                         /* If we ran out of room on the first message,
1334                          * we're in trouble
1335                          */
1336                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1337
1338                         if (err < 0)
1339                                 goto out;
1340
1341                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1342 cont:
1343                         idx++;
1344                 }
1345         }
1346 out:
1347         rcu_read_unlock();
1348         cb->args[1] = idx;
1349         cb->args[0] = h;
1350
1351         return skb->len;
1352 }
1353
1354 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1355 {
1356         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1357 }
1358 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1359
1360 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1361 {
1362         struct net *net;
1363         /* Examine the link attributes and figure out which
1364          * network namespace we are talking about.
1365          */
1366         if (tb[IFLA_NET_NS_PID])
1367                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1368         else if (tb[IFLA_NET_NS_FD])
1369                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1370         else
1371                 net = get_net(src_net);
1372         return net;
1373 }
1374 EXPORT_SYMBOL(rtnl_link_get_net);
1375
1376 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1377 {
1378         if (dev) {
1379                 if (tb[IFLA_ADDRESS] &&
1380                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1381                         return -EINVAL;
1382
1383                 if (tb[IFLA_BROADCAST] &&
1384                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1385                         return -EINVAL;
1386         }
1387
1388         if (tb[IFLA_AF_SPEC]) {
1389                 struct nlattr *af;
1390                 int rem, err;
1391
1392                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1393                         const struct rtnl_af_ops *af_ops;
1394
1395                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1396                                 return -EAFNOSUPPORT;
1397
1398                         if (!af_ops->set_link_af)
1399                                 return -EOPNOTSUPP;
1400
1401                         if (af_ops->validate_link_af) {
1402                                 err = af_ops->validate_link_af(dev, af);
1403                                 if (err < 0)
1404                                         return err;
1405                         }
1406                 }
1407         }
1408
1409         return 0;
1410 }
1411
1412 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1413 {
1414         int rem, err = -EINVAL;
1415         struct nlattr *vf;
1416         const struct net_device_ops *ops = dev->netdev_ops;
1417
1418         nla_for_each_nested(vf, attr, rem) {
1419                 switch (nla_type(vf)) {
1420                 case IFLA_VF_MAC: {
1421                         struct ifla_vf_mac *ivm;
1422                         ivm = nla_data(vf);
1423                         err = -EOPNOTSUPP;
1424                         if (ops->ndo_set_vf_mac)
1425                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1426                                                           ivm->mac);
1427                         break;
1428                 }
1429                 case IFLA_VF_VLAN: {
1430                         struct ifla_vf_vlan *ivv;
1431                         ivv = nla_data(vf);
1432                         err = -EOPNOTSUPP;
1433                         if (ops->ndo_set_vf_vlan)
1434                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1435                                                            ivv->vlan,
1436                                                            ivv->qos);
1437                         break;
1438                 }
1439                 case IFLA_VF_TX_RATE: {
1440                         struct ifla_vf_tx_rate *ivt;
1441                         struct ifla_vf_info ivf;
1442                         ivt = nla_data(vf);
1443                         err = -EOPNOTSUPP;
1444                         if (ops->ndo_get_vf_config)
1445                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1446                                                              &ivf);
1447                         if (err)
1448                                 break;
1449                         err = -EOPNOTSUPP;
1450                         if (ops->ndo_set_vf_rate)
1451                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1452                                                            ivf.min_tx_rate,
1453                                                            ivt->rate);
1454                         break;
1455                 }
1456                 case IFLA_VF_RATE: {
1457                         struct ifla_vf_rate *ivt;
1458                         ivt = nla_data(vf);
1459                         err = -EOPNOTSUPP;
1460                         if (ops->ndo_set_vf_rate)
1461                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1462                                                            ivt->min_tx_rate,
1463                                                            ivt->max_tx_rate);
1464                         break;
1465                 }
1466                 case IFLA_VF_SPOOFCHK: {
1467                         struct ifla_vf_spoofchk *ivs;
1468                         ivs = nla_data(vf);
1469                         err = -EOPNOTSUPP;
1470                         if (ops->ndo_set_vf_spoofchk)
1471                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1472                                                                ivs->setting);
1473                         break;
1474                 }
1475                 case IFLA_VF_LINK_STATE: {
1476                         struct ifla_vf_link_state *ivl;
1477                         ivl = nla_data(vf);
1478                         err = -EOPNOTSUPP;
1479                         if (ops->ndo_set_vf_link_state)
1480                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1481                                                                  ivl->link_state);
1482                         break;
1483                 }
1484                 default:
1485                         err = -EINVAL;
1486                         break;
1487                 }
1488                 if (err)
1489                         break;
1490         }
1491         return err;
1492 }
1493
1494 static int do_set_master(struct net_device *dev, int ifindex)
1495 {
1496         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1497         const struct net_device_ops *ops;
1498         int err;
1499
1500         if (upper_dev) {
1501                 if (upper_dev->ifindex == ifindex)
1502                         return 0;
1503                 ops = upper_dev->netdev_ops;
1504                 if (ops->ndo_del_slave) {
1505                         err = ops->ndo_del_slave(upper_dev, dev);
1506                         if (err)
1507                                 return err;
1508                 } else {
1509                         return -EOPNOTSUPP;
1510                 }
1511         }
1512
1513         if (ifindex) {
1514                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1515                 if (!upper_dev)
1516                         return -EINVAL;
1517                 ops = upper_dev->netdev_ops;
1518                 if (ops->ndo_add_slave) {
1519                         err = ops->ndo_add_slave(upper_dev, dev);
1520                         if (err)
1521                                 return err;
1522                 } else {
1523                         return -EOPNOTSUPP;
1524                 }
1525         }
1526         return 0;
1527 }
1528
1529 #define DO_SETLINK_MODIFIED     0x01
1530 /* notify flag means notify + modified. */
1531 #define DO_SETLINK_NOTIFY       0x03
1532 static int do_setlink(const struct sk_buff *skb,
1533                       struct net_device *dev, struct ifinfomsg *ifm,
1534                       struct nlattr **tb, char *ifname, int status)
1535 {
1536         const struct net_device_ops *ops = dev->netdev_ops;
1537         int err;
1538
1539         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1540                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1541                 if (IS_ERR(net)) {
1542                         err = PTR_ERR(net);
1543                         goto errout;
1544                 }
1545                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1546                         put_net(net);
1547                         err = -EPERM;
1548                         goto errout;
1549                 }
1550                 err = dev_change_net_namespace(dev, net, ifname);
1551                 put_net(net);
1552                 if (err)
1553                         goto errout;
1554                 status |= DO_SETLINK_MODIFIED;
1555         }
1556
1557         if (tb[IFLA_MAP]) {
1558                 struct rtnl_link_ifmap *u_map;
1559                 struct ifmap k_map;
1560
1561                 if (!ops->ndo_set_config) {
1562                         err = -EOPNOTSUPP;
1563                         goto errout;
1564                 }
1565
1566                 if (!netif_device_present(dev)) {
1567                         err = -ENODEV;
1568                         goto errout;
1569                 }
1570
1571                 u_map = nla_data(tb[IFLA_MAP]);
1572                 k_map.mem_start = (unsigned long) u_map->mem_start;
1573                 k_map.mem_end = (unsigned long) u_map->mem_end;
1574                 k_map.base_addr = (unsigned short) u_map->base_addr;
1575                 k_map.irq = (unsigned char) u_map->irq;
1576                 k_map.dma = (unsigned char) u_map->dma;
1577                 k_map.port = (unsigned char) u_map->port;
1578
1579                 err = ops->ndo_set_config(dev, &k_map);
1580                 if (err < 0)
1581                         goto errout;
1582
1583                 status |= DO_SETLINK_NOTIFY;
1584         }
1585
1586         if (tb[IFLA_ADDRESS]) {
1587                 struct sockaddr *sa;
1588                 int len;
1589
1590                 len = sizeof(sa_family_t) + dev->addr_len;
1591                 sa = kmalloc(len, GFP_KERNEL);
1592                 if (!sa) {
1593                         err = -ENOMEM;
1594                         goto errout;
1595                 }
1596                 sa->sa_family = dev->type;
1597                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1598                        dev->addr_len);
1599                 err = dev_set_mac_address(dev, sa);
1600                 kfree(sa);
1601                 if (err)
1602                         goto errout;
1603                 status |= DO_SETLINK_MODIFIED;
1604         }
1605
1606         if (tb[IFLA_MTU]) {
1607                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1608                 if (err < 0)
1609                         goto errout;
1610                 status |= DO_SETLINK_MODIFIED;
1611         }
1612
1613         if (tb[IFLA_GROUP]) {
1614                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1615                 status |= DO_SETLINK_NOTIFY;
1616         }
1617
1618         /*
1619          * Interface selected by interface index but interface
1620          * name provided implies that a name change has been
1621          * requested.
1622          */
1623         if (ifm->ifi_index > 0 && ifname[0]) {
1624                 err = dev_change_name(dev, ifname);
1625                 if (err < 0)
1626                         goto errout;
1627                 status |= DO_SETLINK_MODIFIED;
1628         }
1629
1630         if (tb[IFLA_IFALIAS]) {
1631                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1632                                     nla_len(tb[IFLA_IFALIAS]));
1633                 if (err < 0)
1634                         goto errout;
1635                 status |= DO_SETLINK_NOTIFY;
1636         }
1637
1638         if (tb[IFLA_BROADCAST]) {
1639                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1640                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1641         }
1642
1643         if (ifm->ifi_flags || ifm->ifi_change) {
1644                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1645                 if (err < 0)
1646                         goto errout;
1647         }
1648
1649         if (tb[IFLA_MASTER]) {
1650                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1651                 if (err)
1652                         goto errout;
1653                 status |= DO_SETLINK_MODIFIED;
1654         }
1655
1656         if (tb[IFLA_CARRIER]) {
1657                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1658                 if (err)
1659                         goto errout;
1660                 status |= DO_SETLINK_MODIFIED;
1661         }
1662
1663         if (tb[IFLA_TXQLEN]) {
1664                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1665
1666                 if (dev->tx_queue_len ^ value)
1667                         status |= DO_SETLINK_NOTIFY;
1668
1669                 dev->tx_queue_len = value;
1670         }
1671
1672         if (tb[IFLA_OPERSTATE])
1673                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1674
1675         if (tb[IFLA_LINKMODE]) {
1676                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1677
1678                 write_lock_bh(&dev_base_lock);
1679                 if (dev->link_mode ^ value)
1680                         status |= DO_SETLINK_NOTIFY;
1681                 dev->link_mode = value;
1682                 write_unlock_bh(&dev_base_lock);
1683         }
1684
1685         if (tb[IFLA_VFINFO_LIST]) {
1686                 struct nlattr *attr;
1687                 int rem;
1688                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1689                         if (nla_type(attr) != IFLA_VF_INFO) {
1690                                 err = -EINVAL;
1691                                 goto errout;
1692                         }
1693                         err = do_setvfinfo(dev, attr);
1694                         if (err < 0)
1695                                 goto errout;
1696                         status |= DO_SETLINK_NOTIFY;
1697                 }
1698         }
1699         err = 0;
1700
1701         if (tb[IFLA_VF_PORTS]) {
1702                 struct nlattr *port[IFLA_PORT_MAX+1];
1703                 struct nlattr *attr;
1704                 int vf;
1705                 int rem;
1706
1707                 err = -EOPNOTSUPP;
1708                 if (!ops->ndo_set_vf_port)
1709                         goto errout;
1710
1711                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1712                         if (nla_type(attr) != IFLA_VF_PORT)
1713                                 continue;
1714                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1715                                 attr, ifla_port_policy);
1716                         if (err < 0)
1717                                 goto errout;
1718                         if (!port[IFLA_PORT_VF]) {
1719                                 err = -EOPNOTSUPP;
1720                                 goto errout;
1721                         }
1722                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1723                         err = ops->ndo_set_vf_port(dev, vf, port);
1724                         if (err < 0)
1725                                 goto errout;
1726                         status |= DO_SETLINK_NOTIFY;
1727                 }
1728         }
1729         err = 0;
1730
1731         if (tb[IFLA_PORT_SELF]) {
1732                 struct nlattr *port[IFLA_PORT_MAX+1];
1733
1734                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1735                         tb[IFLA_PORT_SELF], ifla_port_policy);
1736                 if (err < 0)
1737                         goto errout;
1738
1739                 err = -EOPNOTSUPP;
1740                 if (ops->ndo_set_vf_port)
1741                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1742                 if (err < 0)
1743                         goto errout;
1744                 status |= DO_SETLINK_NOTIFY;
1745         }
1746
1747         if (tb[IFLA_AF_SPEC]) {
1748                 struct nlattr *af;
1749                 int rem;
1750
1751                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1752                         const struct rtnl_af_ops *af_ops;
1753
1754                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1755                                 BUG();
1756
1757                         err = af_ops->set_link_af(dev, af);
1758                         if (err < 0)
1759                                 goto errout;
1760
1761                         status |= DO_SETLINK_NOTIFY;
1762                 }
1763         }
1764         err = 0;
1765
1766 errout:
1767         if (status & DO_SETLINK_MODIFIED) {
1768                 if (status & DO_SETLINK_NOTIFY)
1769                         netdev_state_change(dev);
1770
1771                 if (err < 0)
1772                         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",
1773                                              dev->name);
1774         }
1775
1776         return err;
1777 }
1778
1779 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1780 {
1781         struct net *net = sock_net(skb->sk);
1782         struct ifinfomsg *ifm;
1783         struct net_device *dev;
1784         int err;
1785         struct nlattr *tb[IFLA_MAX+1];
1786         char ifname[IFNAMSIZ];
1787
1788         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1789         if (err < 0)
1790                 goto errout;
1791
1792         if (tb[IFLA_IFNAME])
1793                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1794         else
1795                 ifname[0] = '\0';
1796
1797         err = -EINVAL;
1798         ifm = nlmsg_data(nlh);
1799         if (ifm->ifi_index > 0)
1800                 dev = __dev_get_by_index(net, ifm->ifi_index);
1801         else if (tb[IFLA_IFNAME])
1802                 dev = __dev_get_by_name(net, ifname);
1803         else
1804                 goto errout;
1805
1806         if (dev == NULL) {
1807                 err = -ENODEV;
1808                 goto errout;
1809         }
1810
1811         err = validate_linkmsg(dev, tb);
1812         if (err < 0)
1813                 goto errout;
1814
1815         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1816 errout:
1817         return err;
1818 }
1819
1820 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1821 {
1822         struct net *net = sock_net(skb->sk);
1823         const struct rtnl_link_ops *ops;
1824         struct net_device *dev;
1825         struct ifinfomsg *ifm;
1826         char ifname[IFNAMSIZ];
1827         struct nlattr *tb[IFLA_MAX+1];
1828         int err;
1829         LIST_HEAD(list_kill);
1830
1831         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1832         if (err < 0)
1833                 return err;
1834
1835         if (tb[IFLA_IFNAME])
1836                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1837
1838         ifm = nlmsg_data(nlh);
1839         if (ifm->ifi_index > 0)
1840                 dev = __dev_get_by_index(net, ifm->ifi_index);
1841         else if (tb[IFLA_IFNAME])
1842                 dev = __dev_get_by_name(net, ifname);
1843         else
1844                 return -EINVAL;
1845
1846         if (!dev)
1847                 return -ENODEV;
1848
1849         ops = dev->rtnl_link_ops;
1850         if (!ops || !ops->dellink)
1851                 return -EOPNOTSUPP;
1852
1853         ops->dellink(dev, &list_kill);
1854         unregister_netdevice_many(&list_kill);
1855         return 0;
1856 }
1857
1858 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1859 {
1860         unsigned int old_flags;
1861         int err;
1862
1863         old_flags = dev->flags;
1864         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1865                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1866                 if (err < 0)
1867                         return err;
1868         }
1869
1870         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1871
1872         __dev_notify_flags(dev, old_flags, ~0U);
1873         return 0;
1874 }
1875 EXPORT_SYMBOL(rtnl_configure_link);
1876
1877 struct net_device *rtnl_create_link(struct net *net,
1878         char *ifname, unsigned char name_assign_type,
1879         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1880 {
1881         int err;
1882         struct net_device *dev;
1883         unsigned int num_tx_queues = 1;
1884         unsigned int num_rx_queues = 1;
1885
1886         if (tb[IFLA_NUM_TX_QUEUES])
1887                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1888         else if (ops->get_num_tx_queues)
1889                 num_tx_queues = ops->get_num_tx_queues();
1890
1891         if (tb[IFLA_NUM_RX_QUEUES])
1892                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1893         else if (ops->get_num_rx_queues)
1894                 num_rx_queues = ops->get_num_rx_queues();
1895
1896         err = -ENOMEM;
1897         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1898                                ops->setup, num_tx_queues, num_rx_queues);
1899         if (!dev)
1900                 goto err;
1901
1902         dev_net_set(dev, net);
1903         dev->rtnl_link_ops = ops;
1904         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1905
1906         if (tb[IFLA_MTU])
1907                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1908         if (tb[IFLA_ADDRESS]) {
1909                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1910                                 nla_len(tb[IFLA_ADDRESS]));
1911                 dev->addr_assign_type = NET_ADDR_SET;
1912         }
1913         if (tb[IFLA_BROADCAST])
1914                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1915                                 nla_len(tb[IFLA_BROADCAST]));
1916         if (tb[IFLA_TXQLEN])
1917                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1918         if (tb[IFLA_OPERSTATE])
1919                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1920         if (tb[IFLA_LINKMODE])
1921                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1922         if (tb[IFLA_GROUP])
1923                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1924
1925         return dev;
1926
1927 err:
1928         return ERR_PTR(err);
1929 }
1930 EXPORT_SYMBOL(rtnl_create_link);
1931
1932 static int rtnl_group_changelink(const struct sk_buff *skb,
1933                 struct net *net, int group,
1934                 struct ifinfomsg *ifm,
1935                 struct nlattr **tb)
1936 {
1937         struct net_device *dev;
1938         int err;
1939
1940         for_each_netdev(net, dev) {
1941                 if (dev->group == group) {
1942                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1943                         if (err < 0)
1944                                 return err;
1945                 }
1946         }
1947
1948         return 0;
1949 }
1950
1951 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1952 {
1953         struct net *net = sock_net(skb->sk);
1954         const struct rtnl_link_ops *ops;
1955         const struct rtnl_link_ops *m_ops = NULL;
1956         struct net_device *dev;
1957         struct net_device *master_dev = NULL;
1958         struct ifinfomsg *ifm;
1959         char kind[MODULE_NAME_LEN];
1960         char ifname[IFNAMSIZ];
1961         struct nlattr *tb[IFLA_MAX+1];
1962         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1963         unsigned char name_assign_type = NET_NAME_USER;
1964         int err;
1965
1966 #ifdef CONFIG_MODULES
1967 replay:
1968 #endif
1969         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1970         if (err < 0)
1971                 return err;
1972
1973         if (tb[IFLA_IFNAME])
1974                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1975         else
1976                 ifname[0] = '\0';
1977
1978         ifm = nlmsg_data(nlh);
1979         if (ifm->ifi_index > 0)
1980                 dev = __dev_get_by_index(net, ifm->ifi_index);
1981         else {
1982                 if (ifname[0])
1983                         dev = __dev_get_by_name(net, ifname);
1984                 else
1985                         dev = NULL;
1986         }
1987
1988         if (dev) {
1989                 master_dev = netdev_master_upper_dev_get(dev);
1990                 if (master_dev)
1991                         m_ops = master_dev->rtnl_link_ops;
1992         }
1993
1994         err = validate_linkmsg(dev, tb);
1995         if (err < 0)
1996                 return err;
1997
1998         if (tb[IFLA_LINKINFO]) {
1999                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2000                                        tb[IFLA_LINKINFO], ifla_info_policy);
2001                 if (err < 0)
2002                         return err;
2003         } else
2004                 memset(linkinfo, 0, sizeof(linkinfo));
2005
2006         if (linkinfo[IFLA_INFO_KIND]) {
2007                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2008                 ops = rtnl_link_ops_get(kind);
2009         } else {
2010                 kind[0] = '\0';
2011                 ops = NULL;
2012         }
2013
2014         if (1) {
2015                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
2016                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
2017                 struct nlattr **data = NULL;
2018                 struct nlattr **slave_data = NULL;
2019                 struct net *dest_net, *link_net = NULL;
2020
2021                 if (ops) {
2022                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2023                                 err = nla_parse_nested(attr, ops->maxtype,
2024                                                        linkinfo[IFLA_INFO_DATA],
2025                                                        ops->policy);
2026                                 if (err < 0)
2027                                         return err;
2028                                 data = attr;
2029                         }
2030                         if (ops->validate) {
2031                                 err = ops->validate(tb, data);
2032                                 if (err < 0)
2033                                         return err;
2034                         }
2035                 }
2036
2037                 if (m_ops) {
2038                         if (m_ops->slave_maxtype &&
2039                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2040                                 err = nla_parse_nested(slave_attr,
2041                                                        m_ops->slave_maxtype,
2042                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2043                                                        m_ops->slave_policy);
2044                                 if (err < 0)
2045                                         return err;
2046                                 slave_data = slave_attr;
2047                         }
2048                         if (m_ops->slave_validate) {
2049                                 err = m_ops->slave_validate(tb, slave_data);
2050                                 if (err < 0)
2051                                         return err;
2052                         }
2053                 }
2054
2055                 if (dev) {
2056                         int status = 0;
2057
2058                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2059                                 return -EEXIST;
2060                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2061                                 return -EOPNOTSUPP;
2062
2063                         if (linkinfo[IFLA_INFO_DATA]) {
2064                                 if (!ops || ops != dev->rtnl_link_ops ||
2065                                     !ops->changelink)
2066                                         return -EOPNOTSUPP;
2067
2068                                 err = ops->changelink(dev, tb, data);
2069                                 if (err < 0)
2070                                         return err;
2071                                 status |= DO_SETLINK_NOTIFY;
2072                         }
2073
2074                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2075                                 if (!m_ops || !m_ops->slave_changelink)
2076                                         return -EOPNOTSUPP;
2077
2078                                 err = m_ops->slave_changelink(master_dev, dev,
2079                                                               tb, slave_data);
2080                                 if (err < 0)
2081                                         return err;
2082                                 status |= DO_SETLINK_NOTIFY;
2083                         }
2084
2085                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2086                 }
2087
2088                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2089                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2090                                 return rtnl_group_changelink(skb, net,
2091                                                 nla_get_u32(tb[IFLA_GROUP]),
2092                                                 ifm, tb);
2093                         return -ENODEV;
2094                 }
2095
2096                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2097                         return -EOPNOTSUPP;
2098
2099                 if (!ops) {
2100 #ifdef CONFIG_MODULES
2101                         if (kind[0]) {
2102                                 __rtnl_unlock();
2103                                 request_module("rtnl-link-%s", kind);
2104                                 rtnl_lock();
2105                                 ops = rtnl_link_ops_get(kind);
2106                                 if (ops)
2107                                         goto replay;
2108                         }
2109 #endif
2110                         return -EOPNOTSUPP;
2111                 }
2112
2113                 if (!ops->setup)
2114                         return -EOPNOTSUPP;
2115
2116                 if (!ifname[0]) {
2117                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2118                         name_assign_type = NET_NAME_ENUM;
2119                 }
2120
2121                 dest_net = rtnl_link_get_net(net, tb);
2122                 if (IS_ERR(dest_net))
2123                         return PTR_ERR(dest_net);
2124
2125                 if (tb[IFLA_LINK_NETNSID]) {
2126                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2127
2128                         link_net = get_net_ns_by_id(dest_net, id);
2129                         if (!link_net) {
2130                                 err =  -EINVAL;
2131                                 goto out;
2132                         }
2133                 }
2134
2135                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2136                                        name_assign_type, ops, tb);
2137                 if (IS_ERR(dev)) {
2138                         err = PTR_ERR(dev);
2139                         goto out;
2140                 }
2141
2142                 dev->ifindex = ifm->ifi_index;
2143
2144                 if (ops->newlink) {
2145                         err = ops->newlink(link_net ? : net, dev, tb, data);
2146                         /* Drivers should call free_netdev() in ->destructor
2147                          * and unregister it on failure after registration
2148                          * so that device could be finally freed in rtnl_unlock.
2149                          */
2150                         if (err < 0) {
2151                                 /* If device is not registered at all, free it now */
2152                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2153                                         free_netdev(dev);
2154                                 goto out;
2155                         }
2156                 } else {
2157                         err = register_netdevice(dev);
2158                         if (err < 0) {
2159                                 free_netdev(dev);
2160                                 goto out;
2161                         }
2162                 }
2163                 err = rtnl_configure_link(dev, ifm);
2164                 if (err < 0) {
2165                         if (ops->newlink) {
2166                                 LIST_HEAD(list_kill);
2167
2168                                 ops->dellink(dev, &list_kill);
2169                                 unregister_netdevice_many(&list_kill);
2170                         } else {
2171                                 unregister_netdevice(dev);
2172                         }
2173                         goto out;
2174                 }
2175
2176                 if (link_net) {
2177                         err = dev_change_net_namespace(dev, dest_net, ifname);
2178                         if (err < 0)
2179                                 unregister_netdevice(dev);
2180                 }
2181 out:
2182                 if (link_net)
2183                         put_net(link_net);
2184                 put_net(dest_net);
2185                 return err;
2186         }
2187 }
2188
2189 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2190 {
2191         struct net *net = sock_net(skb->sk);
2192         struct ifinfomsg *ifm;
2193         char ifname[IFNAMSIZ];
2194         struct nlattr *tb[IFLA_MAX+1];
2195         struct net_device *dev = NULL;
2196         struct sk_buff *nskb;
2197         int err;
2198         u32 ext_filter_mask = 0;
2199
2200         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2201         if (err < 0)
2202                 return err;
2203
2204         if (tb[IFLA_IFNAME])
2205                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2206
2207         if (tb[IFLA_EXT_MASK])
2208                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2209
2210         ifm = nlmsg_data(nlh);
2211         if (ifm->ifi_index > 0)
2212                 dev = __dev_get_by_index(net, ifm->ifi_index);
2213         else if (tb[IFLA_IFNAME])
2214                 dev = __dev_get_by_name(net, ifname);
2215         else
2216                 return -EINVAL;
2217
2218         if (dev == NULL)
2219                 return -ENODEV;
2220
2221         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2222         if (nskb == NULL)
2223                 return -ENOBUFS;
2224
2225         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2226                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2227         if (err < 0) {
2228                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2229                 WARN_ON(err == -EMSGSIZE);
2230                 kfree_skb(nskb);
2231         } else
2232                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2233
2234         return err;
2235 }
2236
2237 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2238 {
2239         struct net *net = sock_net(skb->sk);
2240         struct net_device *dev;
2241         struct nlattr *tb[IFLA_MAX+1];
2242         u32 ext_filter_mask = 0;
2243         u16 min_ifinfo_dump_size = 0;
2244         int hdrlen;
2245
2246         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2247         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2248                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2249
2250         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2251                 if (tb[IFLA_EXT_MASK])
2252                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2253         }
2254
2255         if (!ext_filter_mask)
2256                 return NLMSG_GOODSIZE;
2257         /*
2258          * traverse the list of net devices and compute the minimum
2259          * buffer size based upon the filter mask.
2260          */
2261         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2262                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2263                                              if_nlmsg_size(dev,
2264                                                            ext_filter_mask));
2265         }
2266
2267         return min_ifinfo_dump_size;
2268 }
2269
2270 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2271 {
2272         int idx;
2273         int s_idx = cb->family;
2274
2275         if (s_idx == 0)
2276                 s_idx = 1;
2277         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2278                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2279                 if (idx < s_idx || idx == PF_PACKET)
2280                         continue;
2281                 if (rtnl_msg_handlers[idx] == NULL ||
2282                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2283                         continue;
2284                 if (idx > s_idx) {
2285                         memset(&cb->args[0], 0, sizeof(cb->args));
2286                         cb->prev_seq = 0;
2287                         cb->seq = 0;
2288                 }
2289                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2290                         break;
2291         }
2292         cb->family = idx;
2293
2294         return skb->len;
2295 }
2296
2297 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2298                                        unsigned int change, gfp_t flags)
2299 {
2300         struct net *net = dev_net(dev);
2301         struct sk_buff *skb;
2302         int err = -ENOBUFS;
2303         size_t if_info_size;
2304
2305         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2306         if (skb == NULL)
2307                 goto errout;
2308
2309         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2310         if (err < 0) {
2311                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2312                 WARN_ON(err == -EMSGSIZE);
2313                 kfree_skb(skb);
2314                 goto errout;
2315         }
2316         return skb;
2317 errout:
2318         if (err < 0)
2319                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2320         return NULL;
2321 }
2322
2323 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2324 {
2325         struct net *net = dev_net(dev);
2326
2327         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2328 }
2329
2330 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2331                   gfp_t flags)
2332 {
2333         struct sk_buff *skb;
2334
2335         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2336         if (skb)
2337                 rtmsg_ifinfo_send(skb, dev, flags);
2338 }
2339 EXPORT_SYMBOL(rtmsg_ifinfo);
2340
2341 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2342                                    struct net_device *dev,
2343                                    u8 *addr, u32 pid, u32 seq,
2344                                    int type, unsigned int flags,
2345                                    int nlflags)
2346 {
2347         struct nlmsghdr *nlh;
2348         struct ndmsg *ndm;
2349
2350         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2351         if (!nlh)
2352                 return -EMSGSIZE;
2353
2354         ndm = nlmsg_data(nlh);
2355         ndm->ndm_family  = AF_BRIDGE;
2356         ndm->ndm_pad1    = 0;
2357         ndm->ndm_pad2    = 0;
2358         ndm->ndm_flags   = flags;
2359         ndm->ndm_type    = 0;
2360         ndm->ndm_ifindex = dev->ifindex;
2361         ndm->ndm_state   = NUD_PERMANENT;
2362
2363         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2364                 goto nla_put_failure;
2365
2366         nlmsg_end(skb, nlh);
2367         return 0;
2368
2369 nla_put_failure:
2370         nlmsg_cancel(skb, nlh);
2371         return -EMSGSIZE;
2372 }
2373
2374 static inline size_t rtnl_fdb_nlmsg_size(void)
2375 {
2376         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2377 }
2378
2379 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2380 {
2381         struct net *net = dev_net(dev);
2382         struct sk_buff *skb;
2383         int err = -ENOBUFS;
2384
2385         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2386         if (!skb)
2387                 goto errout;
2388
2389         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2390         if (err < 0) {
2391                 kfree_skb(skb);
2392                 goto errout;
2393         }
2394
2395         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2396         return;
2397 errout:
2398         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2399 }
2400
2401 /**
2402  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2403  */
2404 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2405                      struct nlattr *tb[],
2406                      struct net_device *dev,
2407                      const unsigned char *addr, u16 vid,
2408                      u16 flags)
2409 {
2410         int err = -EINVAL;
2411
2412         /* If aging addresses are supported device will need to
2413          * implement its own handler for this.
2414          */
2415         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2416                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2417                 return err;
2418         }
2419
2420         if (vid) {
2421                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2422                 return err;
2423         }
2424
2425         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2426                 err = dev_uc_add_excl(dev, addr);
2427         else if (is_multicast_ether_addr(addr))
2428                 err = dev_mc_add_excl(dev, addr);
2429
2430         /* Only return duplicate errors if NLM_F_EXCL is set */
2431         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2432                 err = 0;
2433
2434         return err;
2435 }
2436 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2437
2438 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2439 {
2440         u16 vid = 0;
2441
2442         if (vlan_attr) {
2443                 if (nla_len(vlan_attr) != sizeof(u16)) {
2444                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2445                         return -EINVAL;
2446                 }
2447
2448                 vid = nla_get_u16(vlan_attr);
2449
2450                 if (!vid || vid >= VLAN_VID_MASK) {
2451                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2452                                 vid);
2453                         return -EINVAL;
2454                 }
2455         }
2456         *p_vid = vid;
2457         return 0;
2458 }
2459
2460 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2461 {
2462         struct net *net = sock_net(skb->sk);
2463         struct ndmsg *ndm;
2464         struct nlattr *tb[NDA_MAX+1];
2465         struct net_device *dev;
2466         u8 *addr;
2467         u16 vid;
2468         int err;
2469
2470         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2471         if (err < 0)
2472                 return err;
2473
2474         ndm = nlmsg_data(nlh);
2475         if (ndm->ndm_ifindex == 0) {
2476                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2477                 return -EINVAL;
2478         }
2479
2480         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2481         if (dev == NULL) {
2482                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2483                 return -ENODEV;
2484         }
2485
2486         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2487                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2488                 return -EINVAL;
2489         }
2490
2491         addr = nla_data(tb[NDA_LLADDR]);
2492
2493         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2494         if (err)
2495                 return err;
2496
2497         err = -EOPNOTSUPP;
2498
2499         /* Support fdb on master device the net/bridge default case */
2500         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2501             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2502                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2503                 const struct net_device_ops *ops = br_dev->netdev_ops;
2504
2505                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2506                                        nlh->nlmsg_flags);
2507                 if (err)
2508                         goto out;
2509                 else
2510                         ndm->ndm_flags &= ~NTF_MASTER;
2511         }
2512
2513         /* Embedded bridge, macvlan, and any other device support */
2514         if ((ndm->ndm_flags & NTF_SELF)) {
2515                 if (dev->netdev_ops->ndo_fdb_add)
2516                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2517                                                            vid,
2518                                                            nlh->nlmsg_flags);
2519                 else
2520                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2521                                                nlh->nlmsg_flags);
2522
2523                 if (!err) {
2524                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2525                         ndm->ndm_flags &= ~NTF_SELF;
2526                 }
2527         }
2528 out:
2529         return err;
2530 }
2531
2532 /**
2533  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2534  */
2535 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2536                      struct nlattr *tb[],
2537                      struct net_device *dev,
2538                      const unsigned char *addr, u16 vid)
2539 {
2540         int err = -EINVAL;
2541
2542         /* If aging addresses are supported device will need to
2543          * implement its own handler for this.
2544          */
2545         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2546                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2547                 return err;
2548         }
2549
2550         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2551                 err = dev_uc_del(dev, addr);
2552         else if (is_multicast_ether_addr(addr))
2553                 err = dev_mc_del(dev, addr);
2554
2555         return err;
2556 }
2557 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2558
2559 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2560 {
2561         struct net *net = sock_net(skb->sk);
2562         struct ndmsg *ndm;
2563         struct nlattr *tb[NDA_MAX+1];
2564         struct net_device *dev;
2565         int err = -EINVAL;
2566         __u8 *addr;
2567         u16 vid;
2568
2569         if (!netlink_capable(skb, CAP_NET_ADMIN))
2570                 return -EPERM;
2571
2572         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2573         if (err < 0)
2574                 return err;
2575
2576         ndm = nlmsg_data(nlh);
2577         if (ndm->ndm_ifindex == 0) {
2578                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2579                 return -EINVAL;
2580         }
2581
2582         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2583         if (dev == NULL) {
2584                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2585                 return -ENODEV;
2586         }
2587
2588         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2589                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2590                 return -EINVAL;
2591         }
2592
2593         addr = nla_data(tb[NDA_LLADDR]);
2594
2595         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2596         if (err)
2597                 return err;
2598
2599         err = -EOPNOTSUPP;
2600
2601         /* Support fdb on master device the net/bridge default case */
2602         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2603             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2604                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2605                 const struct net_device_ops *ops = br_dev->netdev_ops;
2606
2607                 if (ops->ndo_fdb_del)
2608                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2609
2610                 if (err)
2611                         goto out;
2612                 else
2613                         ndm->ndm_flags &= ~NTF_MASTER;
2614         }
2615
2616         /* Embedded bridge, macvlan, and any other device support */
2617         if (ndm->ndm_flags & NTF_SELF) {
2618                 if (dev->netdev_ops->ndo_fdb_del)
2619                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2620                                                            vid);
2621                 else
2622                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2623
2624                 if (!err) {
2625                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2626                         ndm->ndm_flags &= ~NTF_SELF;
2627                 }
2628         }
2629 out:
2630         return err;
2631 }
2632
2633 static int nlmsg_populate_fdb(struct sk_buff *skb,
2634                               struct netlink_callback *cb,
2635                               struct net_device *dev,
2636                               int *idx,
2637                               struct netdev_hw_addr_list *list)
2638 {
2639         struct netdev_hw_addr *ha;
2640         int err;
2641         u32 portid, seq;
2642
2643         portid = NETLINK_CB(cb->skb).portid;
2644         seq = cb->nlh->nlmsg_seq;
2645
2646         list_for_each_entry(ha, &list->list, list) {
2647                 if (*idx < cb->args[0])
2648                         goto skip;
2649
2650                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2651                                               portid, seq,
2652                                               RTM_NEWNEIGH, NTF_SELF,
2653                                               NLM_F_MULTI);
2654                 if (err < 0)
2655                         return err;
2656 skip:
2657                 *idx += 1;
2658         }
2659         return 0;
2660 }
2661
2662 /**
2663  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2664  * @nlh: netlink message header
2665  * @dev: netdevice
2666  *
2667  * Default netdevice operation to dump the existing unicast address list.
2668  * Returns number of addresses from list put in skb.
2669  */
2670 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2671                       struct netlink_callback *cb,
2672                       struct net_device *dev,
2673                       struct net_device *filter_dev,
2674                       int idx)
2675 {
2676         int err;
2677
2678         netif_addr_lock_bh(dev);
2679         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2680         if (err)
2681                 goto out;
2682         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2683 out:
2684         netif_addr_unlock_bh(dev);
2685         return idx;
2686 }
2687 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2688
2689 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2690 {
2691         struct net_device *dev;
2692         struct nlattr *tb[IFLA_MAX+1];
2693         struct net_device *bdev = NULL;
2694         struct net_device *br_dev = NULL;
2695         const struct net_device_ops *ops = NULL;
2696         const struct net_device_ops *cops = NULL;
2697         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2698         struct net *net = sock_net(skb->sk);
2699         int brport_idx = 0;
2700         int br_idx = 0;
2701         int idx = 0;
2702
2703         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2704                         ifla_policy) == 0) {
2705                 if (tb[IFLA_MASTER])
2706                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2707         }
2708
2709         brport_idx = ifm->ifi_index;
2710
2711         if (br_idx) {
2712                 br_dev = __dev_get_by_index(net, br_idx);
2713                 if (!br_dev)
2714                         return -ENODEV;
2715
2716                 ops = br_dev->netdev_ops;
2717                 bdev = br_dev;
2718         }
2719
2720         for_each_netdev(net, dev) {
2721                 if (brport_idx && (dev->ifindex != brport_idx))
2722                         continue;
2723
2724                 if (!br_idx) { /* user did not specify a specific bridge */
2725                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2726                                 br_dev = netdev_master_upper_dev_get(dev);
2727                                 cops = br_dev->netdev_ops;
2728                         }
2729
2730                         bdev = dev;
2731                 } else {
2732                         if (dev != br_dev &&
2733                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2734                                 continue;
2735
2736                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2737                             !(dev->priv_flags & IFF_EBRIDGE))
2738                                 continue;
2739
2740                         bdev = br_dev;
2741                         cops = ops;
2742                 }
2743
2744                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2745                         if (cops && cops->ndo_fdb_dump)
2746                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2747                                                          idx);
2748                 }
2749
2750                 if (dev->netdev_ops->ndo_fdb_dump)
2751                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2752                                                             idx);
2753                 else
2754                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2755
2756                 cops = NULL;
2757         }
2758
2759         cb->args[0] = idx;
2760         return skb->len;
2761 }
2762
2763 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2764                                unsigned int attrnum, unsigned int flag)
2765 {
2766         if (mask & flag)
2767                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2768         return 0;
2769 }
2770
2771 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2772                             struct net_device *dev, u16 mode,
2773                             u32 flags, u32 mask)
2774 {
2775         struct nlmsghdr *nlh;
2776         struct ifinfomsg *ifm;
2777         struct nlattr *br_afspec;
2778         struct nlattr *protinfo;
2779         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2780         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2781
2782         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2783         if (nlh == NULL)
2784                 return -EMSGSIZE;
2785
2786         ifm = nlmsg_data(nlh);
2787         ifm->ifi_family = AF_BRIDGE;
2788         ifm->__ifi_pad = 0;
2789         ifm->ifi_type = dev->type;
2790         ifm->ifi_index = dev->ifindex;
2791         ifm->ifi_flags = dev_get_flags(dev);
2792         ifm->ifi_change = 0;
2793
2794
2795         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2796             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2797             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2798             (br_dev &&
2799              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2800             (dev->addr_len &&
2801              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2802             (dev->ifindex != dev->iflink &&
2803              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2804                 goto nla_put_failure;
2805
2806         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2807         if (!br_afspec)
2808                 goto nla_put_failure;
2809
2810         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2811                 nla_nest_cancel(skb, br_afspec);
2812                 goto nla_put_failure;
2813         }
2814
2815         if (mode != BRIDGE_MODE_UNDEF) {
2816                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2817                         nla_nest_cancel(skb, br_afspec);
2818                         goto nla_put_failure;
2819                 }
2820         }
2821         nla_nest_end(skb, br_afspec);
2822
2823         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2824         if (!protinfo)
2825                 goto nla_put_failure;
2826
2827         if (brport_nla_put_flag(skb, flags, mask,
2828                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2829             brport_nla_put_flag(skb, flags, mask,
2830                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2831             brport_nla_put_flag(skb, flags, mask,
2832                                 IFLA_BRPORT_FAST_LEAVE,
2833                                 BR_MULTICAST_FAST_LEAVE) ||
2834             brport_nla_put_flag(skb, flags, mask,
2835                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2836             brport_nla_put_flag(skb, flags, mask,
2837                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2838             brport_nla_put_flag(skb, flags, mask,
2839                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2840             brport_nla_put_flag(skb, flags, mask,
2841                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2842             brport_nla_put_flag(skb, flags, mask,
2843                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2844                 nla_nest_cancel(skb, protinfo);
2845                 goto nla_put_failure;
2846         }
2847
2848         nla_nest_end(skb, protinfo);
2849
2850         nlmsg_end(skb, nlh);
2851         return 0;
2852 nla_put_failure:
2853         nlmsg_cancel(skb, nlh);
2854         return -EMSGSIZE;
2855 }
2856 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2857
2858 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2859 {
2860         struct net *net = sock_net(skb->sk);
2861         struct net_device *dev;
2862         int idx = 0;
2863         u32 portid = NETLINK_CB(cb->skb).portid;
2864         u32 seq = cb->nlh->nlmsg_seq;
2865         u32 filter_mask = 0;
2866
2867         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2868                 struct nlattr *extfilt;
2869
2870                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2871                                           IFLA_EXT_MASK);
2872                 if (extfilt) {
2873                         if (nla_len(extfilt) < sizeof(filter_mask))
2874                                 return -EINVAL;
2875
2876                         filter_mask = nla_get_u32(extfilt);
2877                 }
2878         }
2879
2880         rcu_read_lock();
2881         for_each_netdev_rcu(net, dev) {
2882                 const struct net_device_ops *ops = dev->netdev_ops;
2883                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2884
2885                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2886                         if (idx >= cb->args[0] &&
2887                             br_dev->netdev_ops->ndo_bridge_getlink(
2888                                     skb, portid, seq, dev, filter_mask) < 0)
2889                                 break;
2890                         idx++;
2891                 }
2892
2893                 if (ops->ndo_bridge_getlink) {
2894                         if (idx >= cb->args[0] &&
2895                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2896                                                     filter_mask) < 0)
2897                                 break;
2898                         idx++;
2899                 }
2900         }
2901         rcu_read_unlock();
2902         cb->args[0] = idx;
2903
2904         return skb->len;
2905 }
2906
2907 static inline size_t bridge_nlmsg_size(void)
2908 {
2909         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2910                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2911                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2912                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2913                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2914                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2915                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2916                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2917                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2918                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2919                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2920 }
2921
2922 static int rtnl_bridge_notify(struct net_device *dev)
2923 {
2924         struct net *net = dev_net(dev);
2925         struct sk_buff *skb;
2926         int err = -EOPNOTSUPP;
2927
2928         if (!dev->netdev_ops->ndo_bridge_getlink)
2929                 return 0;
2930
2931         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2932         if (!skb) {
2933                 err = -ENOMEM;
2934                 goto errout;
2935         }
2936
2937         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2938         if (err < 0)
2939                 goto errout;
2940
2941         if (!skb->len)
2942                 goto errout;
2943
2944         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2945         return 0;
2946 errout:
2947         WARN_ON(err == -EMSGSIZE);
2948         kfree_skb(skb);
2949         if (err)
2950                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2951         return err;
2952 }
2953
2954 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2955 {
2956         struct net *net = sock_net(skb->sk);
2957         struct ifinfomsg *ifm;
2958         struct net_device *dev;
2959         struct nlattr *br_spec, *attr = NULL;
2960         int rem, err = -EOPNOTSUPP;
2961         u16 flags = 0;
2962         bool have_flags = false;
2963
2964         if (nlmsg_len(nlh) < sizeof(*ifm))
2965                 return -EINVAL;
2966
2967         ifm = nlmsg_data(nlh);
2968         if (ifm->ifi_family != AF_BRIDGE)
2969                 return -EPFNOSUPPORT;
2970
2971         dev = __dev_get_by_index(net, ifm->ifi_index);
2972         if (!dev) {
2973                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2974                 return -ENODEV;
2975         }
2976
2977         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2978         if (br_spec) {
2979                 nla_for_each_nested(attr, br_spec, rem) {
2980                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2981                                 if (nla_len(attr) < sizeof(flags))
2982                                         return -EINVAL;
2983
2984                                 have_flags = true;
2985                                 flags = nla_get_u16(attr);
2986                                 break;
2987                         }
2988                 }
2989         }
2990
2991         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2992                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2993
2994                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2995                         err = -EOPNOTSUPP;
2996                         goto out;
2997                 }
2998
2999                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3000                 if (err)
3001                         goto out;
3002
3003                 flags &= ~BRIDGE_FLAGS_MASTER;
3004         }
3005
3006         if ((flags & BRIDGE_FLAGS_SELF)) {
3007                 if (!dev->netdev_ops->ndo_bridge_setlink)
3008                         err = -EOPNOTSUPP;
3009                 else
3010                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3011                                                                   flags);
3012                 if (!err) {
3013                         flags &= ~BRIDGE_FLAGS_SELF;
3014
3015                         /* Generate event to notify upper layer of bridge
3016                          * change
3017                          */
3018                         err = rtnl_bridge_notify(dev);
3019                 }
3020         }
3021
3022         if (have_flags)
3023                 memcpy(nla_data(attr), &flags, sizeof(flags));
3024 out:
3025         return err;
3026 }
3027
3028 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3029 {
3030         struct net *net = sock_net(skb->sk);
3031         struct ifinfomsg *ifm;
3032         struct net_device *dev;
3033         struct nlattr *br_spec, *attr = NULL;
3034         int rem, err = -EOPNOTSUPP;
3035         u16 flags = 0;
3036         bool have_flags = false;
3037
3038         if (nlmsg_len(nlh) < sizeof(*ifm))
3039                 return -EINVAL;
3040
3041         ifm = nlmsg_data(nlh);
3042         if (ifm->ifi_family != AF_BRIDGE)
3043                 return -EPFNOSUPPORT;
3044
3045         dev = __dev_get_by_index(net, ifm->ifi_index);
3046         if (!dev) {
3047                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3048                 return -ENODEV;
3049         }
3050
3051         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3052         if (br_spec) {
3053                 nla_for_each_nested(attr, br_spec, rem) {
3054                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3055                                 if (nla_len(attr) < sizeof(flags))
3056                                         return -EINVAL;
3057
3058                                 have_flags = true;
3059                                 flags = nla_get_u16(attr);
3060                                 break;
3061                         }
3062                 }
3063         }
3064
3065         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3066                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3067
3068                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3069                         err = -EOPNOTSUPP;
3070                         goto out;
3071                 }
3072
3073                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3074                 if (err)
3075                         goto out;
3076
3077                 flags &= ~BRIDGE_FLAGS_MASTER;
3078         }
3079
3080         if ((flags & BRIDGE_FLAGS_SELF)) {
3081                 if (!dev->netdev_ops->ndo_bridge_dellink)
3082                         err = -EOPNOTSUPP;
3083                 else
3084                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3085                                                                   flags);
3086
3087                 if (!err) {
3088                         flags &= ~BRIDGE_FLAGS_SELF;
3089
3090                         /* Generate event to notify upper layer of bridge
3091                          * change
3092                          */
3093                         err = rtnl_bridge_notify(dev);
3094                 }
3095         }
3096
3097         if (have_flags)
3098                 memcpy(nla_data(attr), &flags, sizeof(flags));
3099 out:
3100         return err;
3101 }
3102
3103 /* Process one rtnetlink message. */
3104
3105 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3106 {
3107         struct net *net = sock_net(skb->sk);
3108         rtnl_doit_func doit;
3109         int sz_idx, kind;
3110         int family;
3111         int type;
3112         int err;
3113
3114         type = nlh->nlmsg_type;
3115         if (type > RTM_MAX)
3116                 return -EOPNOTSUPP;
3117
3118         type -= RTM_BASE;
3119
3120         /* All the messages must have at least 1 byte length */
3121         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3122                 return 0;
3123
3124         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3125         sz_idx = type>>2;
3126         kind = type&3;
3127
3128         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3129                 return -EPERM;
3130
3131         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3132                 struct sock *rtnl;
3133                 rtnl_dumpit_func dumpit;
3134                 rtnl_calcit_func calcit;
3135                 u16 min_dump_alloc = 0;
3136
3137                 dumpit = rtnl_get_dumpit(family, type);
3138                 if (dumpit == NULL)
3139                         return -EOPNOTSUPP;
3140                 calcit = rtnl_get_calcit(family, type);
3141                 if (calcit)
3142                         min_dump_alloc = calcit(skb, nlh);
3143
3144                 __rtnl_unlock();
3145                 rtnl = net->rtnl;
3146                 {
3147                         struct netlink_dump_control c = {
3148                                 .dump           = dumpit,
3149                                 .min_dump_alloc = min_dump_alloc,
3150                         };
3151                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3152                 }
3153                 rtnl_lock();
3154                 return err;
3155         }
3156
3157         doit = rtnl_get_doit(family, type);
3158         if (doit == NULL)
3159                 return -EOPNOTSUPP;
3160
3161         return doit(skb, nlh);
3162 }
3163
3164 static void rtnetlink_rcv(struct sk_buff *skb)
3165 {
3166         rtnl_lock();
3167         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3168         rtnl_unlock();
3169 }
3170
3171 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3172 {
3173         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3174
3175         switch (event) {
3176         case NETDEV_UP:
3177         case NETDEV_DOWN:
3178         case NETDEV_PRE_UP:
3179         case NETDEV_POST_INIT:
3180         case NETDEV_REGISTER:
3181         case NETDEV_CHANGE:
3182         case NETDEV_PRE_TYPE_CHANGE:
3183         case NETDEV_GOING_DOWN:
3184         case NETDEV_UNREGISTER:
3185         case NETDEV_UNREGISTER_FINAL:
3186         case NETDEV_RELEASE:
3187         case NETDEV_JOIN:
3188         case NETDEV_BONDING_INFO:
3189                 break;
3190         default:
3191                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3192                 break;
3193         }
3194         return NOTIFY_DONE;
3195 }
3196
3197 static struct notifier_block rtnetlink_dev_notifier = {
3198         .notifier_call  = rtnetlink_event,
3199 };
3200
3201
3202 static int __net_init rtnetlink_net_init(struct net *net)
3203 {
3204         struct sock *sk;
3205         struct netlink_kernel_cfg cfg = {
3206                 .groups         = RTNLGRP_MAX,
3207                 .input          = rtnetlink_rcv,
3208                 .cb_mutex       = &rtnl_mutex,
3209                 .flags          = NL_CFG_F_NONROOT_RECV,
3210         };
3211
3212         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3213         if (!sk)
3214                 return -ENOMEM;
3215         net->rtnl = sk;
3216         return 0;
3217 }
3218
3219 static void __net_exit rtnetlink_net_exit(struct net *net)
3220 {
3221         netlink_kernel_release(net->rtnl);
3222         net->rtnl = NULL;
3223 }
3224
3225 static struct pernet_operations rtnetlink_net_ops = {
3226         .init = rtnetlink_net_init,
3227         .exit = rtnetlink_net_exit,
3228 };
3229
3230 void __init rtnetlink_init(void)
3231 {
3232         if (register_pernet_subsys(&rtnetlink_net_ops))
3233                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3234
3235         register_netdevice_notifier(&rtnetlink_dev_notifier);
3236
3237         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3238                       rtnl_dump_ifinfo, rtnl_calcit);
3239         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3240         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3241         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3242
3243         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3244         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3245
3246         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3247         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3248         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3249
3250         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3251         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3252         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3253 }
3254