net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 int lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256
 257         return 0;
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_unregister);
 260
 261 /**
 262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 263  * @protocol : Protocol family or PF_UNSPEC
 264  *
 265  * Identical to calling rtnl_unregster() for all registered message types
 266  * of a certain protocol family.
 267  */
 268 void rtnl_unregister_all(int protocol)
 269 {
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271
 272         kfree(rtnl_msg_handlers[protocol]);
 273         rtnl_msg_handlers[protocol] = NULL;
 274 }
 275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 276
 277 static LIST_HEAD(link_ops);
 278
 279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 280 {
 281         const struct rtnl_link_ops *ops;
 282
 283         list_for_each_entry(ops, &link_ops, list) {
 284                 if (!strcmp(ops->kind, kind))
 285                         return ops;
 286         }
 287         return NULL;
 288 }
 289
 290 /**
 291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 292  * @ops: struct rtnl_link_ops * to register
 293  *
 294  * The caller must hold the rtnl_mutex. This function should be used
 295  * by drivers that create devices during module initialization. It
 296  * must be called before registering the devices.
 297  *
 298  * Returns 0 on success or a negative error code.
 299  */
 300 int __rtnl_link_register(struct rtnl_link_ops *ops)
 301 {
 302         if (rtnl_link_ops_get(ops->kind))
 303                 return -EEXIST;
 304
 305         /* The check for setup is here because if ops
 306          * does not have that filled up, it is not possible
 307          * to use the ops for creating device. So do not
 308          * fill up dellink as well. That disables rtnl_dellink.
 309          */
 310         if (ops->setup && !ops->dellink)
 311                 ops->dellink = unregister_netdevice_queue;
 312
 313         list_add_tail(&ops->list, &link_ops);
 314         return 0;
 315 }
 316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 317
 318 /**
 319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 320  * @ops: struct rtnl_link_ops * to register
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_link_register(struct rtnl_link_ops *ops)
 325 {
 326         int err;
 327
 328         rtnl_lock();
 329         err = __rtnl_link_register(ops);
 330         rtnl_unlock();
 331         return err;
 332 }
 333 EXPORT_SYMBOL_GPL(rtnl_link_register);
 334
 335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 336 {
 337         struct net_device *dev;
 338         LIST_HEAD(list_kill);
 339
 340         for_each_netdev(net, dev) {
 341                 if (dev->rtnl_link_ops == ops)
 342                         ops->dellink(dev, &list_kill);
 343         }
 344         unregister_netdevice_many(&list_kill);
 345 }
 346
 347 /**
 348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 349  * @ops: struct rtnl_link_ops * to unregister
 350  *
 351  * The caller must hold the rtnl_mutex.
 352  */
 353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 354 {
 355         struct net *net;
 356
 357         for_each_net(net) {
 358                 __rtnl_kill_links(net, ops);
 359         }
 360         list_del(&ops->list);
 361 }
 362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 363
 364 /* Return with the rtnl_lock held when there are no network
 365  * devices unregistering in any network namespace.
 366  */
 367 static void rtnl_lock_unregistering_all(void)
 368 {
 369         struct net *net;
 370         bool unregistering;
 371         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 372
 373         add_wait_queue(&netdev_unregistering_wq, &wait);
 374         for (;;) {
 375                 unregistering = false;
 376                 rtnl_lock();
 377                 for_each_net(net) {
 378                         if (net->dev_unreg_count > 0) {
 379                                 unregistering = true;
 380                                 break;
 381                         }
 382                 }
 383                 if (!unregistering)
 384                         break;
 385                 __rtnl_unlock();
 386
 387                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 388         }
 389         remove_wait_queue(&netdev_unregistering_wq, &wait);
 390 }
 391
 392 /**
 393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 394  * @ops: struct rtnl_link_ops * to unregister
 395  */
 396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 397 {
 398         /* Close the race with cleanup_net() */
 399         mutex_lock(&net_mutex);
 400         rtnl_lock_unregistering_all();
 401         __rtnl_link_unregister(ops);
 402         rtnl_unlock();
 403         mutex_unlock(&net_mutex);
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 406
 407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 408 {
 409         struct net_device *master_dev;
 410         const struct rtnl_link_ops *ops;
 411
 412         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 413         if (!master_dev)
 414                 return 0;
 415         ops = master_dev->rtnl_link_ops;
 416         if (!ops || !ops->get_slave_size)
 417                 return 0;
 418         /* IFLA_INFO_SLAVE_DATA + nested data */
 419         return nla_total_size(sizeof(struct nlattr)) +
 420                ops->get_slave_size(master_dev, dev);
 421 }
 422
 423 static size_t rtnl_link_get_size(const struct net_device *dev)
 424 {
 425         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 426         size_t size;
 427
 428         if (!ops)
 429                 return 0;
 430
 431         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 432                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 433
 434         if (ops->get_size)
 435                 /* IFLA_INFO_DATA + nested data */
 436                 size += nla_total_size(sizeof(struct nlattr)) +
 437                         ops->get_size(dev);
 438
 439         if (ops->get_xstats_size)
 440                 /* IFLA_INFO_XSTATS */
 441                 size += nla_total_size(ops->get_xstats_size(dev));
 442
 443         size += rtnl_link_get_slave_info_data_size(dev);
 444
 445         return size;
 446 }
 447
 448 static LIST_HEAD(rtnl_af_ops);
 449
 450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 451 {
 452         const struct rtnl_af_ops *ops;
 453
 454         list_for_each_entry(ops, &rtnl_af_ops, list) {
 455                 if (ops->family == family)
 456                         return ops;
 457         }
 458
 459         return NULL;
 460 }
 461
 462 /**
 463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 464  * @ops: struct rtnl_af_ops * to register
 465  *
 466  * Returns 0 on success or a negative error code.
 467  */
 468 void rtnl_af_register(struct rtnl_af_ops *ops)
 469 {
 470         rtnl_lock();
 471         list_add_tail(&ops->list, &rtnl_af_ops);
 472         rtnl_unlock();
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_af_register);
 475
 476 /**
 477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 478  * @ops: struct rtnl_af_ops * to unregister
 479  *
 480  * The caller must hold the rtnl_mutex.
 481  */
 482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         list_del(&ops->list);
 485 }
 486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 487
 488 /**
 489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 490  * @ops: struct rtnl_af_ops * to unregister
 491  */
 492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 493 {
 494         rtnl_lock();
 495         __rtnl_af_unregister(ops);
 496         rtnl_unlock();
 497 }
 498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 499
 500 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 501 {
 502         struct rtnl_af_ops *af_ops;
 503         size_t size;
 504
 505         /* IFLA_AF_SPEC */
 506         size = nla_total_size(sizeof(struct nlattr));
 507
 508         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 509                 if (af_ops->get_link_af_size) {
 510                         /* AF_* + nested data */
 511                         size += nla_total_size(sizeof(struct nlattr)) +
 512                                 af_ops->get_link_af_size(dev);
 513                 }
 514         }
 515
 516         return size;
 517 }
 518
 519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 520 {
 521         struct net_device *master_dev;
 522
 523         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 524         if (master_dev && master_dev->rtnl_link_ops)
 525                 return true;
 526         return false;
 527 }
 528
 529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 530                                      const struct net_device *dev)
 531 {
 532         struct net_device *master_dev;
 533         const struct rtnl_link_ops *ops;
 534         struct nlattr *slave_data;
 535         int err;
 536
 537         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 538         if (!master_dev)
 539                 return 0;
 540         ops = master_dev->rtnl_link_ops;
 541         if (!ops)
 542                 return 0;
 543         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 544                 return -EMSGSIZE;
 545         if (ops->fill_slave_info) {
 546                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 547                 if (!slave_data)
 548                         return -EMSGSIZE;
 549                 err = ops->fill_slave_info(skb, master_dev, dev);
 550                 if (err < 0)
 551                         goto err_cancel_slave_data;
 552                 nla_nest_end(skb, slave_data);
 553         }
 554         return 0;
 555
 556 err_cancel_slave_data:
 557         nla_nest_cancel(skb, slave_data);
 558         return err;
 559 }
 560
 561 static int rtnl_link_info_fill(struct sk_buff *skb,
 562                                const struct net_device *dev)
 563 {
 564         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 565         struct nlattr *data;
 566         int err;
 567
 568         if (!ops)
 569                 return 0;
 570         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 571                 return -EMSGSIZE;
 572         if (ops->fill_xstats) {
 573                 err = ops->fill_xstats(skb, dev);
 574                 if (err < 0)
 575                         return err;
 576         }
 577         if (ops->fill_info) {
 578                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 579                 if (data == NULL)
 580                         return -EMSGSIZE;
 581                 err = ops->fill_info(skb, dev);
 582                 if (err < 0)
 583                         goto err_cancel_data;
 584                 nla_nest_end(skb, data);
 585         }
 586         return 0;
 587
 588 err_cancel_data:
 589         nla_nest_cancel(skb, data);
 590         return err;
 591 }
 592
 593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 594 {
 595         struct nlattr *linkinfo;
 596         int err = -EMSGSIZE;
 597
 598         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 599         if (linkinfo == NULL)
 600                 goto out;
 601
 602         err = rtnl_link_info_fill(skb, dev);
 603         if (err < 0)
 604                 goto err_cancel_link;
 605
 606         err = rtnl_link_slave_info_fill(skb, dev);
 607         if (err < 0)
 608                 goto err_cancel_link;
 609
 610         nla_nest_end(skb, linkinfo);
 611         return 0;
 612
 613 err_cancel_link:
 614         nla_nest_cancel(skb, linkinfo);
 615 out:
 616         return err;
 617 }
 618
 619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 620 {
 621         struct sock *rtnl = net->rtnl;
 622         int err = 0;
 623
 624         NETLINK_CB(skb).dst_group = group;
 625         if (echo)
 626                 atomic_inc(&skb->users);
 627         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 628         if (echo)
 629                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 630         return err;
 631 }
 632
 633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 634 {
 635         struct sock *rtnl = net->rtnl;
 636
 637         return nlmsg_unicast(rtnl, skb, pid);
 638 }
 639 EXPORT_SYMBOL(rtnl_unicast);
 640
 641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 642                  struct nlmsghdr *nlh, gfp_t flags)
 643 {
 644         struct sock *rtnl = net->rtnl;
 645         int report = 0;
 646
 647         if (nlh)
 648                 report = nlmsg_report(nlh);
 649
 650         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 651 }
 652 EXPORT_SYMBOL(rtnl_notify);
 653
 654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 655 {
 656         struct sock *rtnl = net->rtnl;
 657
 658         netlink_set_err(rtnl, 0, group, error);
 659 }
 660 EXPORT_SYMBOL(rtnl_set_sk_err);
 661
 662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 663 {
 664         struct nlattr *mx;
 665         int i, valid = 0;
 666
 667         mx = nla_nest_start(skb, RTA_METRICS);
 668         if (mx == NULL)
 669                 return -ENOBUFS;
 670
 671         for (i = 0; i < RTAX_MAX; i++) {
 672                 if (metrics[i]) {
 673                         if (i == RTAX_CC_ALGO - 1) {
 674                                 char tmp[TCP_CA_NAME_MAX], *name;
 675
 676                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 677                                 if (!name)
 678                                         continue;
 679                                 if (nla_put_string(skb, i + 1, name))
 680                                         goto nla_put_failure;
 681                         } else {
 682                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 683                                         goto nla_put_failure;
 684                         }
 685                         valid++;
 686                 }
 687         }
 688
 689         if (!valid) {
 690                 nla_nest_cancel(skb, mx);
 691                 return 0;
 692         }
 693
 694         return nla_nest_end(skb, mx);
 695
 696 nla_put_failure:
 697         nla_nest_cancel(skb, mx);
 698         return -EMSGSIZE;
 699 }
 700 EXPORT_SYMBOL(rtnetlink_put_metrics);
 701
 702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 703                        long expires, u32 error)
 704 {
 705         struct rta_cacheinfo ci = {
 706                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 707                 .rta_used = dst->__use,
 708                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 709                 .rta_error = error,
 710                 .rta_id =  id,
 711         };
 712
 713         if (expires) {
 714                 unsigned long clock;
 715
 716                 clock = jiffies_to_clock_t(abs(expires));
 717                 clock = min_t(unsigned long, clock, INT_MAX);
 718                 ci.rta_expires = (expires > 0) ? clock : -clock;
 719         }
 720         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 721 }
 722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 723
 724 static void set_operstate(struct net_device *dev, unsigned char transition)
 725 {
 726         unsigned char operstate = dev->operstate;
 727
 728         switch (transition) {
 729         case IF_OPER_UP:
 730                 if ((operstate == IF_OPER_DORMANT ||
 731                      operstate == IF_OPER_UNKNOWN) &&
 732                     !netif_dormant(dev))
 733                         operstate = IF_OPER_UP;
 734                 break;
 735
 736         case IF_OPER_DORMANT:
 737                 if (operstate == IF_OPER_UP ||
 738                     operstate == IF_OPER_UNKNOWN)
 739                         operstate = IF_OPER_DORMANT;
 740                 break;
 741         }
 742
 743         if (dev->operstate != operstate) {
 744                 write_lock_bh(&dev_base_lock);
 745                 dev->operstate = operstate;
 746                 write_unlock_bh(&dev_base_lock);
 747                 netdev_state_change(dev);
 748         }
 749 }
 750
 751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 752 {
 753         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 754                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 755 }
 756
 757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 758                                            const struct ifinfomsg *ifm)
 759 {
 760         unsigned int flags = ifm->ifi_flags;
 761
 762         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 763         if (ifm->ifi_change)
 764                 flags = (flags & ifm->ifi_change) |
 765                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 766
 767         return flags;
 768 }
 769
 770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 771                                  const struct rtnl_link_stats64 *b)
 772 {
 773         a->rx_packets = b->rx_packets;
 774         a->tx_packets = b->tx_packets;
 775         a->rx_bytes = b->rx_bytes;
 776         a->tx_bytes = b->tx_bytes;
 777         a->rx_errors = b->rx_errors;
 778         a->tx_errors = b->tx_errors;
 779         a->rx_dropped = b->rx_dropped;
 780         a->tx_dropped = b->tx_dropped;
 781
 782         a->multicast = b->multicast;
 783         a->collisions = b->collisions;
 784
 785         a->rx_length_errors = b->rx_length_errors;
 786         a->rx_over_errors = b->rx_over_errors;
 787         a->rx_crc_errors = b->rx_crc_errors;
 788         a->rx_frame_errors = b->rx_frame_errors;
 789         a->rx_fifo_errors = b->rx_fifo_errors;
 790         a->rx_missed_errors = b->rx_missed_errors;
 791
 792         a->tx_aborted_errors = b->tx_aborted_errors;
 793         a->tx_carrier_errors = b->tx_carrier_errors;
 794         a->tx_fifo_errors = b->tx_fifo_errors;
 795         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 796         a->tx_window_errors = b->tx_window_errors;
 797
 798         a->rx_compressed = b->rx_compressed;
 799         a->tx_compressed = b->tx_compressed;
 800 }
 801
 802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 803 {
 804         memcpy(v, b, sizeof(*b));
 805 }
 806
 807 /* All VF info */
 808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 809                                    u32 ext_filter_mask)
 810 {
 811         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 812             (ext_filter_mask & RTEXT_FILTER_VF)) {
 813                 int num_vfs = dev_num_vf(dev->dev.parent);
 814                 size_t size = nla_total_size(sizeof(struct nlattr));
 815                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 816                 size += num_vfs *
 817                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 818                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 819                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 820                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 821                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 822                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)));
 823                 return size;
 824         } else
 825                 return 0;
 826 }
 827
 828 static size_t rtnl_port_size(const struct net_device *dev,
 829                              u32 ext_filter_mask)
 830 {
 831         size_t port_size = nla_total_size(4)            /* PORT_VF */
 832                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 833                 + nla_total_size(sizeof(struct ifla_port_vsi))
 834                                                         /* PORT_VSI_TYPE */
 835                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 836                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 837                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 838                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 839         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 840         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 841                 + port_size;
 842         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 843                 + port_size;
 844
 845         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 846             !(ext_filter_mask & RTEXT_FILTER_VF))
 847                 return 0;
 848         if (dev_num_vf(dev->dev.parent))
 849                 return port_self_size + vf_ports_size +
 850                         vf_port_size * dev_num_vf(dev->dev.parent);
 851         else
 852                 return port_self_size;
 853 }
 854
 855 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 856                                      u32 ext_filter_mask)
 857 {
 858         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 859                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 860                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 861                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 862                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 863                + nla_total_size(sizeof(struct rtnl_link_stats))
 864                + nla_total_size(sizeof(struct rtnl_link_stats64))
 865                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 866                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 867                + nla_total_size(4) /* IFLA_TXQLEN */
 868                + nla_total_size(4) /* IFLA_WEIGHT */
 869                + nla_total_size(4) /* IFLA_MTU */
 870                + nla_total_size(4) /* IFLA_LINK */
 871                + nla_total_size(4) /* IFLA_MASTER */
 872                + nla_total_size(1) /* IFLA_CARRIER */
 873                + nla_total_size(4) /* IFLA_PROMISCUITY */
 874                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 875                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 876                + nla_total_size(1) /* IFLA_OPERSTATE */
 877                + nla_total_size(1) /* IFLA_LINKMODE */
 878                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 879                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 880                + nla_total_size(ext_filter_mask
 881                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 882                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 883                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 884                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 885                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 886                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 887                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 888 }
 889
 890 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 891 {
 892         struct nlattr *vf_ports;
 893         struct nlattr *vf_port;
 894         int vf;
 895         int err;
 896
 897         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 898         if (!vf_ports)
 899                 return -EMSGSIZE;
 900
 901         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 902                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 903                 if (!vf_port)
 904                         goto nla_put_failure;
 905                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 906                         goto nla_put_failure;
 907                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 908                 if (err == -EMSGSIZE)
 909                         goto nla_put_failure;
 910                 if (err) {
 911                         nla_nest_cancel(skb, vf_port);
 912                         continue;
 913                 }
 914                 nla_nest_end(skb, vf_port);
 915         }
 916
 917         nla_nest_end(skb, vf_ports);
 918
 919         return 0;
 920
 921 nla_put_failure:
 922         nla_nest_cancel(skb, vf_ports);
 923         return -EMSGSIZE;
 924 }
 925
 926 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 927 {
 928         struct nlattr *port_self;
 929         int err;
 930
 931         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 932         if (!port_self)
 933                 return -EMSGSIZE;
 934
 935         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 936         if (err) {
 937                 nla_nest_cancel(skb, port_self);
 938                 return (err == -EMSGSIZE) ? err : 0;
 939         }
 940
 941         nla_nest_end(skb, port_self);
 942
 943         return 0;
 944 }
 945
 946 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 947                           u32 ext_filter_mask)
 948 {
 949         int err;
 950
 951         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 952             !(ext_filter_mask & RTEXT_FILTER_VF))
 953                 return 0;
 954
 955         err = rtnl_port_self_fill(skb, dev);
 956         if (err)
 957                 return err;
 958
 959         if (dev_num_vf(dev->dev.parent)) {
 960                 err = rtnl_vf_ports_fill(skb, dev);
 961                 if (err)
 962                         return err;
 963         }
 964
 965         return 0;
 966 }
 967
 968 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 969 {
 970         int err;
 971         struct netdev_phys_item_id ppid;
 972
 973         err = dev_get_phys_port_id(dev, &ppid);
 974         if (err) {
 975                 if (err == -EOPNOTSUPP)
 976                         return 0;
 977                 return err;
 978         }
 979
 980         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 981                 return -EMSGSIZE;
 982
 983         return 0;
 984 }
 985
 986 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
 987 {
 988         char name[IFNAMSIZ];
 989         int err;
 990
 991         err = dev_get_phys_port_name(dev, name, sizeof(name));
 992         if (err) {
 993                 if (err == -EOPNOTSUPP)
 994                         return 0;
 995                 return err;
 996         }
 997
 998         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
 999                 return -EMSGSIZE;
1000
1001         return 0;
1002 }
1003
1004 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1005 {
1006         int err;
1007         struct switchdev_attr attr = {
1008                 .id = SWITCHDEV_ATTR_PORT_PARENT_ID,
1009                 .flags = SWITCHDEV_F_NO_RECURSE,
1010         };
1011
1012         err = switchdev_port_attr_get(dev, &attr);
1013         if (err) {
1014                 if (err == -EOPNOTSUPP)
1015                         return 0;
1016                 return err;
1017         }
1018
1019         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1020                     attr.u.ppid.id))
1021                 return -EMSGSIZE;
1022
1023         return 0;
1024 }
1025
1026 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1027                             int type, u32 pid, u32 seq, u32 change,
1028                             unsigned int flags, u32 ext_filter_mask)
1029 {
1030         struct ifinfomsg *ifm;
1031         struct nlmsghdr *nlh;
1032         struct rtnl_link_stats64 temp;
1033         const struct rtnl_link_stats64 *stats;
1034         struct nlattr *attr, *af_spec;
1035         struct rtnl_af_ops *af_ops;
1036         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1037
1038         ASSERT_RTNL();
1039         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1040         if (nlh == NULL)
1041                 return -EMSGSIZE;
1042
1043         ifm = nlmsg_data(nlh);
1044         ifm->ifi_family = AF_UNSPEC;
1045         ifm->__ifi_pad = 0;
1046         ifm->ifi_type = dev->type;
1047         ifm->ifi_index = dev->ifindex;
1048         ifm->ifi_flags = dev_get_flags(dev);
1049         ifm->ifi_change = change;
1050
1051         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1052             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1053             nla_put_u8(skb, IFLA_OPERSTATE,
1054                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1055             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1056             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1057             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1058             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1059             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1060 #ifdef CONFIG_RPS
1061             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1062 #endif
1063             (dev->ifindex != dev_get_iflink(dev) &&
1064              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1065             (upper_dev &&
1066              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1067             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1068             (dev->qdisc &&
1069              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1070             (dev->ifalias &&
1071              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1072             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1073                         atomic_read(&dev->carrier_changes)))
1074                 goto nla_put_failure;
1075
1076         if (1) {
1077                 struct rtnl_link_ifmap map = {
1078                         .mem_start   = dev->mem_start,
1079                         .mem_end     = dev->mem_end,
1080                         .base_addr   = dev->base_addr,
1081                         .irq         = dev->irq,
1082                         .dma         = dev->dma,
1083                         .port        = dev->if_port,
1084                 };
1085                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1086                         goto nla_put_failure;
1087         }
1088
1089         if (dev->addr_len) {
1090                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1091                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1092                         goto nla_put_failure;
1093         }
1094
1095         if (rtnl_phys_port_id_fill(skb, dev))
1096                 goto nla_put_failure;
1097
1098         if (rtnl_phys_port_name_fill(skb, dev))
1099                 goto nla_put_failure;
1100
1101         if (rtnl_phys_switch_id_fill(skb, dev))
1102                 goto nla_put_failure;
1103
1104         attr = nla_reserve(skb, IFLA_STATS,
1105                         sizeof(struct rtnl_link_stats));
1106         if (attr == NULL)
1107                 goto nla_put_failure;
1108
1109         stats = dev_get_stats(dev, &temp);
1110         copy_rtnl_link_stats(nla_data(attr), stats);
1111
1112         attr = nla_reserve(skb, IFLA_STATS64,
1113                         sizeof(struct rtnl_link_stats64));
1114         if (attr == NULL)
1115                 goto nla_put_failure;
1116         copy_rtnl_link_stats64(nla_data(attr), stats);
1117
1118         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1119             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1120                 goto nla_put_failure;
1121
1122         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1123             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1124                 int i;
1125
1126                 struct nlattr *vfinfo, *vf;
1127                 int num_vfs = dev_num_vf(dev->dev.parent);
1128
1129                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1130                 if (!vfinfo)
1131                         goto nla_put_failure;
1132                 for (i = 0; i < num_vfs; i++) {
1133                         struct ifla_vf_info ivi;
1134                         struct ifla_vf_mac vf_mac;
1135                         struct ifla_vf_vlan vf_vlan;
1136                         struct ifla_vf_rate vf_rate;
1137                         struct ifla_vf_tx_rate vf_tx_rate;
1138                         struct ifla_vf_spoofchk vf_spoofchk;
1139                         struct ifla_vf_link_state vf_linkstate;
1140                         struct ifla_vf_rss_query_en vf_rss_query_en;
1141
1142                         /*
1143                          * Not all SR-IOV capable drivers support the
1144                          * spoofcheck and "RSS query enable" query.  Preset to
1145                          * -1 so the user space tool can detect that the driver
1146                          * didn't report anything.
1147                          */
1148                         ivi.spoofchk = -1;
1149                         ivi.rss_query_en = -1;
1150                         memset(ivi.mac, 0, sizeof(ivi.mac));
1151                         /* The default value for VF link state is "auto"
1152                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1153                          */
1154                         ivi.linkstate = 0;
1155                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1156                                 break;
1157                         vf_mac.vf =
1158                                 vf_vlan.vf =
1159                                 vf_rate.vf =
1160                                 vf_tx_rate.vf =
1161                                 vf_spoofchk.vf =
1162                                 vf_linkstate.vf =
1163                                 vf_rss_query_en.vf = ivi.vf;
1164
1165                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1166                         vf_vlan.vlan = ivi.vlan;
1167                         vf_vlan.qos = ivi.qos;
1168                         vf_tx_rate.rate = ivi.max_tx_rate;
1169                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1170                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1171                         vf_spoofchk.setting = ivi.spoofchk;
1172                         vf_linkstate.link_state = ivi.linkstate;
1173                         vf_rss_query_en.setting = ivi.rss_query_en;
1174                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1175                         if (!vf) {
1176                                 nla_nest_cancel(skb, vfinfo);
1177                                 goto nla_put_failure;
1178                         }
1179                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1182                                     &vf_rate) ||
1183                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1184                                     &vf_tx_rate) ||
1185                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1186                                     &vf_spoofchk) ||
1187                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1188                                     &vf_linkstate) ||
1189                             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190                                     sizeof(vf_rss_query_en),
1191                                     &vf_rss_query_en))
1192                                 goto nla_put_failure;
1193                         nla_nest_end(skb, vf);
1194                 }
1195                 nla_nest_end(skb, vfinfo);
1196         }
1197
1198         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1199                 goto nla_put_failure;
1200
1201         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1202                 if (rtnl_link_fill(skb, dev) < 0)
1203                         goto nla_put_failure;
1204         }
1205
1206         if (dev->rtnl_link_ops &&
1207             dev->rtnl_link_ops->get_link_net) {
1208                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1209
1210                 if (!net_eq(dev_net(dev), link_net)) {
1211                         int id = peernet2id_alloc(dev_net(dev), link_net);
1212
1213                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1214                                 goto nla_put_failure;
1215                 }
1216         }
1217
1218         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1219                 goto nla_put_failure;
1220
1221         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1222                 if (af_ops->fill_link_af) {
1223                         struct nlattr *af;
1224                         int err;
1225
1226                         if (!(af = nla_nest_start(skb, af_ops->family)))
1227                                 goto nla_put_failure;
1228
1229                         err = af_ops->fill_link_af(skb, dev);
1230
1231                         /*
1232                          * Caller may return ENODATA to indicate that there
1233                          * was no data to be dumped. This is not an error, it
1234                          * means we should trim the attribute header and
1235                          * continue.
1236                          */
1237                         if (err == -ENODATA)
1238                                 nla_nest_cancel(skb, af);
1239                         else if (err < 0)
1240                                 goto nla_put_failure;
1241
1242                         nla_nest_end(skb, af);
1243                 }
1244         }
1245
1246         nla_nest_end(skb, af_spec);
1247
1248         nlmsg_end(skb, nlh);
1249         return 0;
1250
1251 nla_put_failure:
1252         nlmsg_cancel(skb, nlh);
1253         return -EMSGSIZE;
1254 }
1255
1256 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1257         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1258         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1259         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1260         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1261         [IFLA_MTU]              = { .type = NLA_U32 },
1262         [IFLA_LINK]             = { .type = NLA_U32 },
1263         [IFLA_MASTER]           = { .type = NLA_U32 },
1264         [IFLA_CARRIER]          = { .type = NLA_U8 },
1265         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1266         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1267         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1268         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1269         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1270         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1271         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1272         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1273         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1274         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1275         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1276         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1277         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1278         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1279         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1280         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1281         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1282         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1283         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1284         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1285 };
1286
1287 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1288         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1289         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1290         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1291         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1292 };
1293
1294 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1295         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1296 };
1297
1298 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1299         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1300         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1301         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1302         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1303         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1304         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1305         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1306 };
1307
1308 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1309         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1310         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1311                                     .len = PORT_PROFILE_MAX },
1312         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1313                                     .len = sizeof(struct ifla_port_vsi)},
1314         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1315                                       .len = PORT_UUID_MAX },
1316         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1317                                     .len = PORT_UUID_MAX },
1318         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1319         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1320 };
1321
1322 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1323 {
1324         struct net *net = sock_net(skb->sk);
1325         int h, s_h;
1326         int idx = 0, s_idx;
1327         struct net_device *dev;
1328         struct hlist_head *head;
1329         struct nlattr *tb[IFLA_MAX+1];
1330         u32 ext_filter_mask = 0;
1331         int err;
1332         int hdrlen;
1333
1334         s_h = cb->args[0];
1335         s_idx = cb->args[1];
1336
1337         cb->seq = net->dev_base_seq;
1338
1339         /* A hack to preserve kernel<->userspace interface.
1340          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1341          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1342          * what iproute2 < v3.9.0 used.
1343          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1344          * attribute, its netlink message is shorter than struct ifinfomsg.
1345          */
1346         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1347                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1348
1349         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1350
1351                 if (tb[IFLA_EXT_MASK])
1352                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1353         }
1354
1355         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1356                 idx = 0;
1357                 head = &net->dev_index_head[h];
1358                 hlist_for_each_entry(dev, head, index_hlist) {
1359                         if (idx < s_idx)
1360                                 goto cont;
1361                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1362                                                NETLINK_CB(cb->skb).portid,
1363                                                cb->nlh->nlmsg_seq, 0,
1364                                                NLM_F_MULTI,
1365                                                ext_filter_mask);
1366                         /* If we ran out of room on the first message,
1367                          * we're in trouble
1368                          */
1369                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1370
1371                         if (err < 0)
1372                                 goto out;
1373
1374                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1375 cont:
1376                         idx++;
1377                 }
1378         }
1379 out:
1380         cb->args[1] = idx;
1381         cb->args[0] = h;
1382
1383         return skb->len;
1384 }
1385
1386 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1387 {
1388         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1389 }
1390 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1391
1392 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1393 {
1394         struct net *net;
1395         /* Examine the link attributes and figure out which
1396          * network namespace we are talking about.
1397          */
1398         if (tb[IFLA_NET_NS_PID])
1399                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1400         else if (tb[IFLA_NET_NS_FD])
1401                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1402         else
1403                 net = get_net(src_net);
1404         return net;
1405 }
1406 EXPORT_SYMBOL(rtnl_link_get_net);
1407
1408 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1409 {
1410         if (dev) {
1411                 if (tb[IFLA_ADDRESS] &&
1412                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1413                         return -EINVAL;
1414
1415                 if (tb[IFLA_BROADCAST] &&
1416                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1417                         return -EINVAL;
1418         }
1419
1420         if (tb[IFLA_AF_SPEC]) {
1421                 struct nlattr *af;
1422                 int rem, err;
1423
1424                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1425                         const struct rtnl_af_ops *af_ops;
1426
1427                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1428                                 return -EAFNOSUPPORT;
1429
1430                         if (!af_ops->set_link_af)
1431                                 return -EOPNOTSUPP;
1432
1433                         if (af_ops->validate_link_af) {
1434                                 err = af_ops->validate_link_af(dev, af);
1435                                 if (err < 0)
1436                                         return err;
1437                         }
1438                 }
1439         }
1440
1441         return 0;
1442 }
1443
1444 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1445 {
1446         int rem, err = -EINVAL;
1447         struct nlattr *vf;
1448         const struct net_device_ops *ops = dev->netdev_ops;
1449
1450         nla_for_each_nested(vf, attr, rem) {
1451                 switch (nla_type(vf)) {
1452                 case IFLA_VF_MAC: {
1453                         struct ifla_vf_mac *ivm;
1454                         ivm = nla_data(vf);
1455                         err = -EOPNOTSUPP;
1456                         if (ops->ndo_set_vf_mac)
1457                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1458                                                           ivm->mac);
1459                         break;
1460                 }
1461                 case IFLA_VF_VLAN: {
1462                         struct ifla_vf_vlan *ivv;
1463                         ivv = nla_data(vf);
1464                         err = -EOPNOTSUPP;
1465                         if (ops->ndo_set_vf_vlan)
1466                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1467                                                            ivv->vlan,
1468                                                            ivv->qos);
1469                         break;
1470                 }
1471                 case IFLA_VF_TX_RATE: {
1472                         struct ifla_vf_tx_rate *ivt;
1473                         struct ifla_vf_info ivf;
1474                         ivt = nla_data(vf);
1475                         err = -EOPNOTSUPP;
1476                         if (ops->ndo_get_vf_config)
1477                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1478                                                              &ivf);
1479                         if (err)
1480                                 break;
1481                         err = -EOPNOTSUPP;
1482                         if (ops->ndo_set_vf_rate)
1483                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1484                                                            ivf.min_tx_rate,
1485                                                            ivt->rate);
1486                         break;
1487                 }
1488                 case IFLA_VF_RATE: {
1489                         struct ifla_vf_rate *ivt;
1490                         ivt = nla_data(vf);
1491                         err = -EOPNOTSUPP;
1492                         if (ops->ndo_set_vf_rate)
1493                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1494                                                            ivt->min_tx_rate,
1495                                                            ivt->max_tx_rate);
1496                         break;
1497                 }
1498                 case IFLA_VF_SPOOFCHK: {
1499                         struct ifla_vf_spoofchk *ivs;
1500                         ivs = nla_data(vf);
1501                         err = -EOPNOTSUPP;
1502                         if (ops->ndo_set_vf_spoofchk)
1503                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1504                                                                ivs->setting);
1505                         break;
1506                 }
1507                 case IFLA_VF_LINK_STATE: {
1508                         struct ifla_vf_link_state *ivl;
1509                         ivl = nla_data(vf);
1510                         err = -EOPNOTSUPP;
1511                         if (ops->ndo_set_vf_link_state)
1512                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1513                                                                  ivl->link_state);
1514                         break;
1515                 }
1516                 case IFLA_VF_RSS_QUERY_EN: {
1517                         struct ifla_vf_rss_query_en *ivrssq_en;
1518
1519                         ivrssq_en = nla_data(vf);
1520                         err = -EOPNOTSUPP;
1521                         if (ops->ndo_set_vf_rss_query_en)
1522                                 err = ops->ndo_set_vf_rss_query_en(dev,
1523                                                             ivrssq_en->vf,
1524                                                             ivrssq_en->setting);
1525                         break;
1526                 }
1527                 default:
1528                         err = -EINVAL;
1529                         break;
1530                 }
1531                 if (err)
1532                         break;
1533         }
1534         return err;
1535 }
1536
1537 static int do_set_master(struct net_device *dev, int ifindex)
1538 {
1539         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1540         const struct net_device_ops *ops;
1541         int err;
1542
1543         if (upper_dev) {
1544                 if (upper_dev->ifindex == ifindex)
1545                         return 0;
1546                 ops = upper_dev->netdev_ops;
1547                 if (ops->ndo_del_slave) {
1548                         err = ops->ndo_del_slave(upper_dev, dev);
1549                         if (err)
1550                                 return err;
1551                 } else {
1552                         return -EOPNOTSUPP;
1553                 }
1554         }
1555
1556         if (ifindex) {
1557                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1558                 if (!upper_dev)
1559                         return -EINVAL;
1560                 ops = upper_dev->netdev_ops;
1561                 if (ops->ndo_add_slave) {
1562                         err = ops->ndo_add_slave(upper_dev, dev);
1563                         if (err)
1564                                 return err;
1565                 } else {
1566                         return -EOPNOTSUPP;
1567                 }
1568         }
1569         return 0;
1570 }
1571
1572 #define DO_SETLINK_MODIFIED     0x01
1573 /* notify flag means notify + modified. */
1574 #define DO_SETLINK_NOTIFY       0x03
1575 static int do_setlink(const struct sk_buff *skb,
1576                       struct net_device *dev, struct ifinfomsg *ifm,
1577                       struct nlattr **tb, char *ifname, int status)
1578 {
1579         const struct net_device_ops *ops = dev->netdev_ops;
1580         int err;
1581
1582         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1583                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1584                 if (IS_ERR(net)) {
1585                         err = PTR_ERR(net);
1586                         goto errout;
1587                 }
1588                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1589                         put_net(net);
1590                         err = -EPERM;
1591                         goto errout;
1592                 }
1593                 err = dev_change_net_namespace(dev, net, ifname);
1594                 put_net(net);
1595                 if (err)
1596                         goto errout;
1597                 status |= DO_SETLINK_MODIFIED;
1598         }
1599
1600         if (tb[IFLA_MAP]) {
1601                 struct rtnl_link_ifmap *u_map;
1602                 struct ifmap k_map;
1603
1604                 if (!ops->ndo_set_config) {
1605                         err = -EOPNOTSUPP;
1606                         goto errout;
1607                 }
1608
1609                 if (!netif_device_present(dev)) {
1610                         err = -ENODEV;
1611                         goto errout;
1612                 }
1613
1614                 u_map = nla_data(tb[IFLA_MAP]);
1615                 k_map.mem_start = (unsigned long) u_map->mem_start;
1616                 k_map.mem_end = (unsigned long) u_map->mem_end;
1617                 k_map.base_addr = (unsigned short) u_map->base_addr;
1618                 k_map.irq = (unsigned char) u_map->irq;
1619                 k_map.dma = (unsigned char) u_map->dma;
1620                 k_map.port = (unsigned char) u_map->port;
1621
1622                 err = ops->ndo_set_config(dev, &k_map);
1623                 if (err < 0)
1624                         goto errout;
1625
1626                 status |= DO_SETLINK_NOTIFY;
1627         }
1628
1629         if (tb[IFLA_ADDRESS]) {
1630                 struct sockaddr *sa;
1631                 int len;
1632
1633                 len = sizeof(sa_family_t) + dev->addr_len;
1634                 sa = kmalloc(len, GFP_KERNEL);
1635                 if (!sa) {
1636                         err = -ENOMEM;
1637                         goto errout;
1638                 }
1639                 sa->sa_family = dev->type;
1640                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1641                        dev->addr_len);
1642                 err = dev_set_mac_address(dev, sa);
1643                 kfree(sa);
1644                 if (err)
1645                         goto errout;
1646                 status |= DO_SETLINK_MODIFIED;
1647         }
1648
1649         if (tb[IFLA_MTU]) {
1650                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1651                 if (err < 0)
1652                         goto errout;
1653                 status |= DO_SETLINK_MODIFIED;
1654         }
1655
1656         if (tb[IFLA_GROUP]) {
1657                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1658                 status |= DO_SETLINK_NOTIFY;
1659         }
1660
1661         /*
1662          * Interface selected by interface index but interface
1663          * name provided implies that a name change has been
1664          * requested.
1665          */
1666         if (ifm->ifi_index > 0 && ifname[0]) {
1667                 err = dev_change_name(dev, ifname);
1668                 if (err < 0)
1669                         goto errout;
1670                 status |= DO_SETLINK_MODIFIED;
1671         }
1672
1673         if (tb[IFLA_IFALIAS]) {
1674                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1675                                     nla_len(tb[IFLA_IFALIAS]));
1676                 if (err < 0)
1677                         goto errout;
1678                 status |= DO_SETLINK_NOTIFY;
1679         }
1680
1681         if (tb[IFLA_BROADCAST]) {
1682                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1683                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1684         }
1685
1686         if (ifm->ifi_flags || ifm->ifi_change) {
1687                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1688                 if (err < 0)
1689                         goto errout;
1690         }
1691
1692         if (tb[IFLA_MASTER]) {
1693                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1694                 if (err)
1695                         goto errout;
1696                 status |= DO_SETLINK_MODIFIED;
1697         }
1698
1699         if (tb[IFLA_CARRIER]) {
1700                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1701                 if (err)
1702                         goto errout;
1703                 status |= DO_SETLINK_MODIFIED;
1704         }
1705
1706         if (tb[IFLA_TXQLEN]) {
1707                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1708
1709                 if (dev->tx_queue_len ^ value)
1710                         status |= DO_SETLINK_NOTIFY;
1711
1712                 dev->tx_queue_len = value;
1713         }
1714
1715         if (tb[IFLA_OPERSTATE])
1716                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1717
1718         if (tb[IFLA_LINKMODE]) {
1719                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1720
1721                 write_lock_bh(&dev_base_lock);
1722                 if (dev->link_mode ^ value)
1723                         status |= DO_SETLINK_NOTIFY;
1724                 dev->link_mode = value;
1725                 write_unlock_bh(&dev_base_lock);
1726         }
1727
1728         if (tb[IFLA_VFINFO_LIST]) {
1729                 struct nlattr *attr;
1730                 int rem;
1731                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1732                         if (nla_type(attr) != IFLA_VF_INFO) {
1733                                 err = -EINVAL;
1734                                 goto errout;
1735                         }
1736                         err = do_setvfinfo(dev, attr);
1737                         if (err < 0)
1738                                 goto errout;
1739                         status |= DO_SETLINK_NOTIFY;
1740                 }
1741         }
1742         err = 0;
1743
1744         if (tb[IFLA_VF_PORTS]) {
1745                 struct nlattr *port[IFLA_PORT_MAX+1];
1746                 struct nlattr *attr;
1747                 int vf;
1748                 int rem;
1749
1750                 err = -EOPNOTSUPP;
1751                 if (!ops->ndo_set_vf_port)
1752                         goto errout;
1753
1754                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1755                         if (nla_type(attr) != IFLA_VF_PORT)
1756                                 continue;
1757                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1758                                 attr, ifla_port_policy);
1759                         if (err < 0)
1760                                 goto errout;
1761                         if (!port[IFLA_PORT_VF]) {
1762                                 err = -EOPNOTSUPP;
1763                                 goto errout;
1764                         }
1765                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1766                         err = ops->ndo_set_vf_port(dev, vf, port);
1767                         if (err < 0)
1768                                 goto errout;
1769                         status |= DO_SETLINK_NOTIFY;
1770                 }
1771         }
1772         err = 0;
1773
1774         if (tb[IFLA_PORT_SELF]) {
1775                 struct nlattr *port[IFLA_PORT_MAX+1];
1776
1777                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1778                         tb[IFLA_PORT_SELF], ifla_port_policy);
1779                 if (err < 0)
1780                         goto errout;
1781
1782                 err = -EOPNOTSUPP;
1783                 if (ops->ndo_set_vf_port)
1784                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1785                 if (err < 0)
1786                         goto errout;
1787                 status |= DO_SETLINK_NOTIFY;
1788         }
1789
1790         if (tb[IFLA_AF_SPEC]) {
1791                 struct nlattr *af;
1792                 int rem;
1793
1794                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1795                         const struct rtnl_af_ops *af_ops;
1796
1797                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1798                                 BUG();
1799
1800                         err = af_ops->set_link_af(dev, af);
1801                         if (err < 0)
1802                                 goto errout;
1803
1804                         status |= DO_SETLINK_NOTIFY;
1805                 }
1806         }
1807         err = 0;
1808
1809 errout:
1810         if (status & DO_SETLINK_MODIFIED) {
1811                 if (status & DO_SETLINK_NOTIFY)
1812                         netdev_state_change(dev);
1813
1814                 if (err < 0)
1815                         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",
1816                                              dev->name);
1817         }
1818
1819         return err;
1820 }
1821
1822 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1823 {
1824         struct net *net = sock_net(skb->sk);
1825         struct ifinfomsg *ifm;
1826         struct net_device *dev;
1827         int err;
1828         struct nlattr *tb[IFLA_MAX+1];
1829         char ifname[IFNAMSIZ];
1830
1831         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1832         if (err < 0)
1833                 goto errout;
1834
1835         if (tb[IFLA_IFNAME])
1836                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1837         else
1838                 ifname[0] = '\0';
1839
1840         err = -EINVAL;
1841         ifm = nlmsg_data(nlh);
1842         if (ifm->ifi_index > 0)
1843                 dev = __dev_get_by_index(net, ifm->ifi_index);
1844         else if (tb[IFLA_IFNAME])
1845                 dev = __dev_get_by_name(net, ifname);
1846         else
1847                 goto errout;
1848
1849         if (dev == NULL) {
1850                 err = -ENODEV;
1851                 goto errout;
1852         }
1853
1854         err = validate_linkmsg(dev, tb);
1855         if (err < 0)
1856                 goto errout;
1857
1858         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1859 errout:
1860         return err;
1861 }
1862
1863 static int rtnl_group_dellink(const struct net *net, int group)
1864 {
1865         struct net_device *dev, *aux;
1866         LIST_HEAD(list_kill);
1867         bool found = false;
1868
1869         if (!group)
1870                 return -EPERM;
1871
1872         for_each_netdev(net, dev) {
1873                 if (dev->group == group) {
1874                         const struct rtnl_link_ops *ops;
1875
1876                         found = true;
1877                         ops = dev->rtnl_link_ops;
1878                         if (!ops || !ops->dellink)
1879                                 return -EOPNOTSUPP;
1880                 }
1881         }
1882
1883         if (!found)
1884                 return -ENODEV;
1885
1886         for_each_netdev_safe(net, dev, aux) {
1887                 if (dev->group == group) {
1888                         const struct rtnl_link_ops *ops;
1889
1890                         ops = dev->rtnl_link_ops;
1891                         ops->dellink(dev, &list_kill);
1892                 }
1893         }
1894         unregister_netdevice_many(&list_kill);
1895
1896         return 0;
1897 }
1898
1899 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1900 {
1901         struct net *net = sock_net(skb->sk);
1902         const struct rtnl_link_ops *ops;
1903         struct net_device *dev;
1904         struct ifinfomsg *ifm;
1905         char ifname[IFNAMSIZ];
1906         struct nlattr *tb[IFLA_MAX+1];
1907         int err;
1908         LIST_HEAD(list_kill);
1909
1910         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1911         if (err < 0)
1912                 return err;
1913
1914         if (tb[IFLA_IFNAME])
1915                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1916
1917         ifm = nlmsg_data(nlh);
1918         if (ifm->ifi_index > 0)
1919                 dev = __dev_get_by_index(net, ifm->ifi_index);
1920         else if (tb[IFLA_IFNAME])
1921                 dev = __dev_get_by_name(net, ifname);
1922         else if (tb[IFLA_GROUP])
1923                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1924         else
1925                 return -EINVAL;
1926
1927         if (!dev)
1928                 return -ENODEV;
1929
1930         ops = dev->rtnl_link_ops;
1931         if (!ops || !ops->dellink)
1932                 return -EOPNOTSUPP;
1933
1934         ops->dellink(dev, &list_kill);
1935         unregister_netdevice_many(&list_kill);
1936         return 0;
1937 }
1938
1939 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1940 {
1941         unsigned int old_flags;
1942         int err;
1943
1944         old_flags = dev->flags;
1945         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1946                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1947                 if (err < 0)
1948                         return err;
1949         }
1950
1951         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1952
1953         __dev_notify_flags(dev, old_flags, ~0U);
1954         return 0;
1955 }
1956 EXPORT_SYMBOL(rtnl_configure_link);
1957
1958 struct net_device *rtnl_create_link(struct net *net,
1959         const char *ifname, unsigned char name_assign_type,
1960         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1961 {
1962         int err;
1963         struct net_device *dev;
1964         unsigned int num_tx_queues = 1;
1965         unsigned int num_rx_queues = 1;
1966
1967         if (tb[IFLA_NUM_TX_QUEUES])
1968                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1969         else if (ops->get_num_tx_queues)
1970                 num_tx_queues = ops->get_num_tx_queues();
1971
1972         if (tb[IFLA_NUM_RX_QUEUES])
1973                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1974         else if (ops->get_num_rx_queues)
1975                 num_rx_queues = ops->get_num_rx_queues();
1976
1977         err = -ENOMEM;
1978         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1979                                ops->setup, num_tx_queues, num_rx_queues);
1980         if (!dev)
1981                 goto err;
1982
1983         dev_net_set(dev, net);
1984         dev->rtnl_link_ops = ops;
1985         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1986
1987         if (tb[IFLA_MTU])
1988                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1989         if (tb[IFLA_ADDRESS]) {
1990                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1991                                 nla_len(tb[IFLA_ADDRESS]));
1992                 dev->addr_assign_type = NET_ADDR_SET;
1993         }
1994         if (tb[IFLA_BROADCAST])
1995                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1996                                 nla_len(tb[IFLA_BROADCAST]));
1997         if (tb[IFLA_TXQLEN])
1998                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1999         if (tb[IFLA_OPERSTATE])
2000                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2001         if (tb[IFLA_LINKMODE])
2002                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2003         if (tb[IFLA_GROUP])
2004                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2005
2006         return dev;
2007
2008 err:
2009         return ERR_PTR(err);
2010 }
2011 EXPORT_SYMBOL(rtnl_create_link);
2012
2013 static int rtnl_group_changelink(const struct sk_buff *skb,
2014                 struct net *net, int group,
2015                 struct ifinfomsg *ifm,
2016                 struct nlattr **tb)
2017 {
2018         struct net_device *dev, *aux;
2019         int err;
2020
2021         for_each_netdev_safe(net, dev, aux) {
2022                 if (dev->group == group) {
2023                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2024                         if (err < 0)
2025                                 return err;
2026                 }
2027         }
2028
2029         return 0;
2030 }
2031
2032 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2033 {
2034         struct net *net = sock_net(skb->sk);
2035         const struct rtnl_link_ops *ops;
2036         const struct rtnl_link_ops *m_ops = NULL;
2037         struct net_device *dev;
2038         struct net_device *master_dev = NULL;
2039         struct ifinfomsg *ifm;
2040         char kind[MODULE_NAME_LEN];
2041         char ifname[IFNAMSIZ];
2042         struct nlattr *tb[IFLA_MAX+1];
2043         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2044         unsigned char name_assign_type = NET_NAME_USER;
2045         int err;
2046
2047 #ifdef CONFIG_MODULES
2048 replay:
2049 #endif
2050         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2051         if (err < 0)
2052                 return err;
2053
2054         if (tb[IFLA_IFNAME])
2055                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2056         else
2057                 ifname[0] = '\0';
2058
2059         ifm = nlmsg_data(nlh);
2060         if (ifm->ifi_index > 0)
2061                 dev = __dev_get_by_index(net, ifm->ifi_index);
2062         else {
2063                 if (ifname[0])
2064                         dev = __dev_get_by_name(net, ifname);
2065                 else
2066                         dev = NULL;
2067         }
2068
2069         if (dev) {
2070                 master_dev = netdev_master_upper_dev_get(dev);
2071                 if (master_dev)
2072                         m_ops = master_dev->rtnl_link_ops;
2073         }
2074
2075         err = validate_linkmsg(dev, tb);
2076         if (err < 0)
2077                 return err;
2078
2079         if (tb[IFLA_LINKINFO]) {
2080                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2081                                        tb[IFLA_LINKINFO], ifla_info_policy);
2082                 if (err < 0)
2083                         return err;
2084         } else
2085                 memset(linkinfo, 0, sizeof(linkinfo));
2086
2087         if (linkinfo[IFLA_INFO_KIND]) {
2088                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2089                 ops = rtnl_link_ops_get(kind);
2090         } else {
2091                 kind[0] = '\0';
2092                 ops = NULL;
2093         }
2094
2095         if (1) {
2096                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2097                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2098                 struct nlattr **data = NULL;
2099                 struct nlattr **slave_data = NULL;
2100                 struct net *dest_net, *link_net = NULL;
2101
2102                 if (ops) {
2103                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2104                                 err = nla_parse_nested(attr, ops->maxtype,
2105                                                        linkinfo[IFLA_INFO_DATA],
2106                                                        ops->policy);
2107                                 if (err < 0)
2108                                         return err;
2109                                 data = attr;
2110                         }
2111                         if (ops->validate) {
2112                                 err = ops->validate(tb, data);
2113                                 if (err < 0)
2114                                         return err;
2115                         }
2116                 }
2117
2118                 if (m_ops) {
2119                         if (m_ops->slave_maxtype &&
2120                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2121                                 err = nla_parse_nested(slave_attr,
2122                                                        m_ops->slave_maxtype,
2123                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2124                                                        m_ops->slave_policy);
2125                                 if (err < 0)
2126                                         return err;
2127                                 slave_data = slave_attr;
2128                         }
2129                         if (m_ops->slave_validate) {
2130                                 err = m_ops->slave_validate(tb, slave_data);
2131                                 if (err < 0)
2132                                         return err;
2133                         }
2134                 }
2135
2136                 if (dev) {
2137                         int status = 0;
2138
2139                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2140                                 return -EEXIST;
2141                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2142                                 return -EOPNOTSUPP;
2143
2144                         if (linkinfo[IFLA_INFO_DATA]) {
2145                                 if (!ops || ops != dev->rtnl_link_ops ||
2146                                     !ops->changelink)
2147                                         return -EOPNOTSUPP;
2148
2149                                 err = ops->changelink(dev, tb, data);
2150                                 if (err < 0)
2151                                         return err;
2152                                 status |= DO_SETLINK_NOTIFY;
2153                         }
2154
2155                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2156                                 if (!m_ops || !m_ops->slave_changelink)
2157                                         return -EOPNOTSUPP;
2158
2159                                 err = m_ops->slave_changelink(master_dev, dev,
2160                                                               tb, slave_data);
2161                                 if (err < 0)
2162                                         return err;
2163                                 status |= DO_SETLINK_NOTIFY;
2164                         }
2165
2166                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2167                 }
2168
2169                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2170                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2171                                 return rtnl_group_changelink(skb, net,
2172                                                 nla_get_u32(tb[IFLA_GROUP]),
2173                                                 ifm, tb);
2174                         return -ENODEV;
2175                 }
2176
2177                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2178                         return -EOPNOTSUPP;
2179
2180                 if (!ops) {
2181 #ifdef CONFIG_MODULES
2182                         if (kind[0]) {
2183                                 __rtnl_unlock();
2184                                 request_module("rtnl-link-%s", kind);
2185                                 rtnl_lock();
2186                                 ops = rtnl_link_ops_get(kind);
2187                                 if (ops)
2188                                         goto replay;
2189                         }
2190 #endif
2191                         return -EOPNOTSUPP;
2192                 }
2193
2194                 if (!ops->setup)
2195                         return -EOPNOTSUPP;
2196
2197                 if (!ifname[0]) {
2198                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2199                         name_assign_type = NET_NAME_ENUM;
2200                 }
2201
2202                 dest_net = rtnl_link_get_net(net, tb);
2203                 if (IS_ERR(dest_net))
2204                         return PTR_ERR(dest_net);
2205
2206                 err = -EPERM;
2207                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2208                         goto out;
2209
2210                 if (tb[IFLA_LINK_NETNSID]) {
2211                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2212
2213                         link_net = get_net_ns_by_id(dest_net, id);
2214                         if (!link_net) {
2215                                 err =  -EINVAL;
2216                                 goto out;
2217                         }
2218                         err = -EPERM;
2219                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2220                                 goto out;
2221                 }
2222
2223                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2224                                        name_assign_type, ops, tb);
2225                 if (IS_ERR(dev)) {
2226                         err = PTR_ERR(dev);
2227                         goto out;
2228                 }
2229
2230                 dev->ifindex = ifm->ifi_index;
2231
2232                 if (ops->newlink) {
2233                         err = ops->newlink(link_net ? : net, dev, tb, data);
2234                         /* Drivers should call free_netdev() in ->destructor
2235                          * and unregister it on failure after registration
2236                          * so that device could be finally freed in rtnl_unlock.
2237                          */
2238                         if (err < 0) {
2239                                 /* If device is not registered at all, free it now */
2240                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2241                                         free_netdev(dev);
2242                                 goto out;
2243                         }
2244                 } else {
2245                         err = register_netdevice(dev);
2246                         if (err < 0) {
2247                                 free_netdev(dev);
2248                                 goto out;
2249                         }
2250                 }
2251                 err = rtnl_configure_link(dev, ifm);
2252                 if (err < 0)
2253                         goto out_unregister;
2254                 if (link_net) {
2255                         err = dev_change_net_namespace(dev, dest_net, ifname);
2256                         if (err < 0)
2257                                 goto out_unregister;
2258                 }
2259 out:
2260                 if (link_net)
2261                         put_net(link_net);
2262                 put_net(dest_net);
2263                 return err;
2264 out_unregister:
2265                 if (ops->newlink) {
2266                         LIST_HEAD(list_kill);
2267
2268                         ops->dellink(dev, &list_kill);
2269                         unregister_netdevice_many(&list_kill);
2270                 } else {
2271                         unregister_netdevice(dev);
2272                 }
2273                 goto out;
2274         }
2275 }
2276
2277 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2278 {
2279         struct net *net = sock_net(skb->sk);
2280         struct ifinfomsg *ifm;
2281         char ifname[IFNAMSIZ];
2282         struct nlattr *tb[IFLA_MAX+1];
2283         struct net_device *dev = NULL;
2284         struct sk_buff *nskb;
2285         int err;
2286         u32 ext_filter_mask = 0;
2287
2288         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2289         if (err < 0)
2290                 return err;
2291
2292         if (tb[IFLA_IFNAME])
2293                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2294
2295         if (tb[IFLA_EXT_MASK])
2296                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2297
2298         ifm = nlmsg_data(nlh);
2299         if (ifm->ifi_index > 0)
2300                 dev = __dev_get_by_index(net, ifm->ifi_index);
2301         else if (tb[IFLA_IFNAME])
2302                 dev = __dev_get_by_name(net, ifname);
2303         else
2304                 return -EINVAL;
2305
2306         if (dev == NULL)
2307                 return -ENODEV;
2308
2309         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2310         if (nskb == NULL)
2311                 return -ENOBUFS;
2312
2313         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2314                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2315         if (err < 0) {
2316                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2317                 WARN_ON(err == -EMSGSIZE);
2318                 kfree_skb(nskb);
2319         } else
2320                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2321
2322         return err;
2323 }
2324
2325 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2326 {
2327         struct net *net = sock_net(skb->sk);
2328         struct net_device *dev;
2329         struct nlattr *tb[IFLA_MAX+1];
2330         u32 ext_filter_mask = 0;
2331         u16 min_ifinfo_dump_size = 0;
2332         int hdrlen;
2333
2334         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2335         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2336                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2337
2338         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2339                 if (tb[IFLA_EXT_MASK])
2340                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2341         }
2342
2343         if (!ext_filter_mask)
2344                 return NLMSG_GOODSIZE;
2345         /*
2346          * traverse the list of net devices and compute the minimum
2347          * buffer size based upon the filter mask.
2348          */
2349         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2350                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2351                                              if_nlmsg_size(dev,
2352                                                            ext_filter_mask));
2353         }
2354
2355         return min_ifinfo_dump_size;
2356 }
2357
2358 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2359 {
2360         int idx;
2361         int s_idx = cb->family;
2362
2363         if (s_idx == 0)
2364                 s_idx = 1;
2365         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2366                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2367                 if (idx < s_idx || idx == PF_PACKET)
2368                         continue;
2369                 if (rtnl_msg_handlers[idx] == NULL ||
2370                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2371                         continue;
2372                 if (idx > s_idx) {
2373                         memset(&cb->args[0], 0, sizeof(cb->args));
2374                         cb->prev_seq = 0;
2375                         cb->seq = 0;
2376                 }
2377                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2378                         break;
2379         }
2380         cb->family = idx;
2381
2382         return skb->len;
2383 }
2384
2385 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2386                                        unsigned int change, gfp_t flags)
2387 {
2388         struct net *net = dev_net(dev);
2389         struct sk_buff *skb;
2390         int err = -ENOBUFS;
2391         size_t if_info_size;
2392
2393         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2394         if (skb == NULL)
2395                 goto errout;
2396
2397         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2398         if (err < 0) {
2399                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2400                 WARN_ON(err == -EMSGSIZE);
2401                 kfree_skb(skb);
2402                 goto errout;
2403         }
2404         return skb;
2405 errout:
2406         if (err < 0)
2407                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2408         return NULL;
2409 }
2410
2411 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2412 {
2413         struct net *net = dev_net(dev);
2414
2415         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2416 }
2417
2418 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2419                   gfp_t flags)
2420 {
2421         struct sk_buff *skb;
2422
2423         if (dev->reg_state != NETREG_REGISTERED)
2424                 return;
2425
2426         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2427         if (skb)
2428                 rtmsg_ifinfo_send(skb, dev, flags);
2429 }
2430 EXPORT_SYMBOL(rtmsg_ifinfo);
2431
2432 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2433                                    struct net_device *dev,
2434                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2435                                    int type, unsigned int flags,
2436                                    int nlflags)
2437 {
2438         struct nlmsghdr *nlh;
2439         struct ndmsg *ndm;
2440
2441         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2442         if (!nlh)
2443                 return -EMSGSIZE;
2444
2445         ndm = nlmsg_data(nlh);
2446         ndm->ndm_family  = AF_BRIDGE;
2447         ndm->ndm_pad1    = 0;
2448         ndm->ndm_pad2    = 0;
2449         ndm->ndm_flags   = flags;
2450         ndm->ndm_type    = 0;
2451         ndm->ndm_ifindex = dev->ifindex;
2452         ndm->ndm_state   = NUD_PERMANENT;
2453
2454         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2455                 goto nla_put_failure;
2456         if (vid)
2457                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2458                         goto nla_put_failure;
2459
2460         nlmsg_end(skb, nlh);
2461         return 0;
2462
2463 nla_put_failure:
2464         nlmsg_cancel(skb, nlh);
2465         return -EMSGSIZE;
2466 }
2467
2468 static inline size_t rtnl_fdb_nlmsg_size(void)
2469 {
2470         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2471 }
2472
2473 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2474 {
2475         struct net *net = dev_net(dev);
2476         struct sk_buff *skb;
2477         int err = -ENOBUFS;
2478
2479         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2480         if (!skb)
2481                 goto errout;
2482
2483         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2484                                       0, 0, type, NTF_SELF, 0);
2485         if (err < 0) {
2486                 kfree_skb(skb);
2487                 goto errout;
2488         }
2489
2490         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2491         return;
2492 errout:
2493         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2494 }
2495
2496 /**
2497  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2498  */
2499 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2500                      struct nlattr *tb[],
2501                      struct net_device *dev,
2502                      const unsigned char *addr, u16 vid,
2503                      u16 flags)
2504 {
2505         int err = -EINVAL;
2506
2507         /* If aging addresses are supported device will need to
2508          * implement its own handler for this.
2509          */
2510         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2511                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2512                 return err;
2513         }
2514
2515         if (vid) {
2516                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2517                 return err;
2518         }
2519
2520         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2521                 err = dev_uc_add_excl(dev, addr);
2522         else if (is_multicast_ether_addr(addr))
2523                 err = dev_mc_add_excl(dev, addr);
2524
2525         /* Only return duplicate errors if NLM_F_EXCL is set */
2526         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2527                 err = 0;
2528
2529         return err;
2530 }
2531 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2532
2533 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2534 {
2535         u16 vid = 0;
2536
2537         if (vlan_attr) {
2538                 if (nla_len(vlan_attr) != sizeof(u16)) {
2539                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2540                         return -EINVAL;
2541                 }
2542
2543                 vid = nla_get_u16(vlan_attr);
2544
2545                 if (!vid || vid >= VLAN_VID_MASK) {
2546                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2547                                 vid);
2548                         return -EINVAL;
2549                 }
2550         }
2551         *p_vid = vid;
2552         return 0;
2553 }
2554
2555 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2556 {
2557         struct net *net = sock_net(skb->sk);
2558         struct ndmsg *ndm;
2559         struct nlattr *tb[NDA_MAX+1];
2560         struct net_device *dev;
2561         u8 *addr;
2562         u16 vid;
2563         int err;
2564
2565         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2566         if (err < 0)
2567                 return err;
2568
2569         ndm = nlmsg_data(nlh);
2570         if (ndm->ndm_ifindex == 0) {
2571                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2572                 return -EINVAL;
2573         }
2574
2575         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2576         if (dev == NULL) {
2577                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2578                 return -ENODEV;
2579         }
2580
2581         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2582                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2583                 return -EINVAL;
2584         }
2585
2586         addr = nla_data(tb[NDA_LLADDR]);
2587
2588         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2589         if (err)
2590                 return err;
2591
2592         err = -EOPNOTSUPP;
2593
2594         /* Support fdb on master device the net/bridge default case */
2595         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2596             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2597                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2598                 const struct net_device_ops *ops = br_dev->netdev_ops;
2599
2600                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2601                                        nlh->nlmsg_flags);
2602                 if (err)
2603                         goto out;
2604                 else
2605                         ndm->ndm_flags &= ~NTF_MASTER;
2606         }
2607
2608         /* Embedded bridge, macvlan, and any other device support */
2609         if ((ndm->ndm_flags & NTF_SELF)) {
2610                 if (dev->netdev_ops->ndo_fdb_add)
2611                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2612                                                            vid,
2613                                                            nlh->nlmsg_flags);
2614                 else
2615                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2616                                                nlh->nlmsg_flags);
2617
2618                 if (!err) {
2619                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2620                         ndm->ndm_flags &= ~NTF_SELF;
2621                 }
2622         }
2623 out:
2624         return err;
2625 }
2626
2627 /**
2628  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2629  */
2630 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2631                      struct nlattr *tb[],
2632                      struct net_device *dev,
2633                      const unsigned char *addr, u16 vid)
2634 {
2635         int err = -EINVAL;
2636
2637         /* If aging addresses are supported device will need to
2638          * implement its own handler for this.
2639          */
2640         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2641                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2642                 return err;
2643         }
2644
2645         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2646                 err = dev_uc_del(dev, addr);
2647         else if (is_multicast_ether_addr(addr))
2648                 err = dev_mc_del(dev, addr);
2649
2650         return err;
2651 }
2652 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2653
2654 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2655 {
2656         struct net *net = sock_net(skb->sk);
2657         struct ndmsg *ndm;
2658         struct nlattr *tb[NDA_MAX+1];
2659         struct net_device *dev;
2660         int err = -EINVAL;
2661         __u8 *addr;
2662         u16 vid;
2663
2664         if (!netlink_capable(skb, CAP_NET_ADMIN))
2665                 return -EPERM;
2666
2667         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2668         if (err < 0)
2669                 return err;
2670
2671         ndm = nlmsg_data(nlh);
2672         if (ndm->ndm_ifindex == 0) {
2673                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2674                 return -EINVAL;
2675         }
2676
2677         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2678         if (dev == NULL) {
2679                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2680                 return -ENODEV;
2681         }
2682
2683         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2684                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2685                 return -EINVAL;
2686         }
2687
2688         addr = nla_data(tb[NDA_LLADDR]);
2689
2690         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2691         if (err)
2692                 return err;
2693
2694         err = -EOPNOTSUPP;
2695
2696         /* Support fdb on master device the net/bridge default case */
2697         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2698             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2699                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2700                 const struct net_device_ops *ops = br_dev->netdev_ops;
2701
2702                 if (ops->ndo_fdb_del)
2703                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2704
2705                 if (err)
2706                         goto out;
2707                 else
2708                         ndm->ndm_flags &= ~NTF_MASTER;
2709         }
2710
2711         /* Embedded bridge, macvlan, and any other device support */
2712         if (ndm->ndm_flags & NTF_SELF) {
2713                 if (dev->netdev_ops->ndo_fdb_del)
2714                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2715                                                            vid);
2716                 else
2717                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2718
2719                 if (!err) {
2720                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2721                         ndm->ndm_flags &= ~NTF_SELF;
2722                 }
2723         }
2724 out:
2725         return err;
2726 }
2727
2728 static int nlmsg_populate_fdb(struct sk_buff *skb,
2729                               struct netlink_callback *cb,
2730                               struct net_device *dev,
2731                               int *idx,
2732                               struct netdev_hw_addr_list *list)
2733 {
2734         struct netdev_hw_addr *ha;
2735         int err;
2736         u32 portid, seq;
2737
2738         portid = NETLINK_CB(cb->skb).portid;
2739         seq = cb->nlh->nlmsg_seq;
2740
2741         list_for_each_entry(ha, &list->list, list) {
2742                 if (*idx < cb->args[0])
2743                         goto skip;
2744
2745                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2746                                               portid, seq,
2747                                               RTM_NEWNEIGH, NTF_SELF,
2748                                               NLM_F_MULTI);
2749                 if (err < 0)
2750                         return err;
2751 skip:
2752                 *idx += 1;
2753         }
2754         return 0;
2755 }
2756
2757 /**
2758  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2759  * @nlh: netlink message header
2760  * @dev: netdevice
2761  *
2762  * Default netdevice operation to dump the existing unicast address list.
2763  * Returns number of addresses from list put in skb.
2764  */
2765 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2766                       struct netlink_callback *cb,
2767                       struct net_device *dev,
2768                       struct net_device *filter_dev,
2769                       int idx)
2770 {
2771         int err;
2772
2773         netif_addr_lock_bh(dev);
2774         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2775         if (err)
2776                 goto out;
2777         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2778 out:
2779         netif_addr_unlock_bh(dev);
2780         return idx;
2781 }
2782 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2783
2784 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2785 {
2786         struct net_device *dev;
2787         struct nlattr *tb[IFLA_MAX+1];
2788         struct net_device *br_dev = NULL;
2789         const struct net_device_ops *ops = NULL;
2790         const struct net_device_ops *cops = NULL;
2791         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2792         struct net *net = sock_net(skb->sk);
2793         int brport_idx = 0;
2794         int br_idx = 0;
2795         int idx = 0;
2796
2797         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2798                         ifla_policy) == 0) {
2799                 if (tb[IFLA_MASTER])
2800                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2801         }
2802
2803         brport_idx = ifm->ifi_index;
2804
2805         if (br_idx) {
2806                 br_dev = __dev_get_by_index(net, br_idx);
2807                 if (!br_dev)
2808                         return -ENODEV;
2809
2810                 ops = br_dev->netdev_ops;
2811         }
2812
2813         for_each_netdev(net, dev) {
2814                 if (brport_idx && (dev->ifindex != brport_idx))
2815                         continue;
2816
2817                 if (!br_idx) { /* user did not specify a specific bridge */
2818                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2819                                 br_dev = netdev_master_upper_dev_get(dev);
2820                                 cops = br_dev->netdev_ops;
2821                         }
2822
2823                 } else {
2824                         if (dev != br_dev &&
2825                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2826                                 continue;
2827
2828                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2829                             !(dev->priv_flags & IFF_EBRIDGE))
2830                                 continue;
2831
2832                         cops = ops;
2833                 }
2834
2835                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2836                         if (cops && cops->ndo_fdb_dump)
2837                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2838                                                          idx);
2839                 }
2840
2841                 if (dev->netdev_ops->ndo_fdb_dump)
2842                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2843                                                             idx);
2844                 else
2845                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2846
2847                 cops = NULL;
2848         }
2849
2850         cb->args[0] = idx;
2851         return skb->len;
2852 }
2853
2854 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2855                                unsigned int attrnum, unsigned int flag)
2856 {
2857         if (mask & flag)
2858                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2859         return 0;
2860 }
2861
2862 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2863                             struct net_device *dev, u16 mode,
2864                             u32 flags, u32 mask, int nlflags)
2865 {
2866         struct nlmsghdr *nlh;
2867         struct ifinfomsg *ifm;
2868         struct nlattr *br_afspec;
2869         struct nlattr *protinfo;
2870         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2871         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2872
2873         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2874         if (nlh == NULL)
2875                 return -EMSGSIZE;
2876
2877         ifm = nlmsg_data(nlh);
2878         ifm->ifi_family = AF_BRIDGE;
2879         ifm->__ifi_pad = 0;
2880         ifm->ifi_type = dev->type;
2881         ifm->ifi_index = dev->ifindex;
2882         ifm->ifi_flags = dev_get_flags(dev);
2883         ifm->ifi_change = 0;
2884
2885
2886         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2887             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2888             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2889             (br_dev &&
2890              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2891             (dev->addr_len &&
2892              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2893             (dev->ifindex != dev_get_iflink(dev) &&
2894              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2895                 goto nla_put_failure;
2896
2897         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2898         if (!br_afspec)
2899                 goto nla_put_failure;
2900
2901         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2902                 nla_nest_cancel(skb, br_afspec);
2903                 goto nla_put_failure;
2904         }
2905
2906         if (mode != BRIDGE_MODE_UNDEF) {
2907                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2908                         nla_nest_cancel(skb, br_afspec);
2909                         goto nla_put_failure;
2910                 }
2911         }
2912         nla_nest_end(skb, br_afspec);
2913
2914         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2915         if (!protinfo)
2916                 goto nla_put_failure;
2917
2918         if (brport_nla_put_flag(skb, flags, mask,
2919                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2920             brport_nla_put_flag(skb, flags, mask,
2921                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2922             brport_nla_put_flag(skb, flags, mask,
2923                                 IFLA_BRPORT_FAST_LEAVE,
2924                                 BR_MULTICAST_FAST_LEAVE) ||
2925             brport_nla_put_flag(skb, flags, mask,
2926                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2927             brport_nla_put_flag(skb, flags, mask,
2928                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2929             brport_nla_put_flag(skb, flags, mask,
2930                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2931             brport_nla_put_flag(skb, flags, mask,
2932                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2933             brport_nla_put_flag(skb, flags, mask,
2934                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2935                 nla_nest_cancel(skb, protinfo);
2936                 goto nla_put_failure;
2937         }
2938
2939         nla_nest_end(skb, protinfo);
2940
2941         nlmsg_end(skb, nlh);
2942         return 0;
2943 nla_put_failure:
2944         nlmsg_cancel(skb, nlh);
2945         return -EMSGSIZE;
2946 }
2947 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2948
2949 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2950 {
2951         struct net *net = sock_net(skb->sk);
2952         struct net_device *dev;
2953         int idx = 0;
2954         u32 portid = NETLINK_CB(cb->skb).portid;
2955         u32 seq = cb->nlh->nlmsg_seq;
2956         u32 filter_mask = 0;
2957
2958         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2959                 struct nlattr *extfilt;
2960
2961                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2962                                           IFLA_EXT_MASK);
2963                 if (extfilt) {
2964                         if (nla_len(extfilt) < sizeof(filter_mask))
2965                                 return -EINVAL;
2966
2967                         filter_mask = nla_get_u32(extfilt);
2968                 }
2969         }
2970
2971         rcu_read_lock();
2972         for_each_netdev_rcu(net, dev) {
2973                 const struct net_device_ops *ops = dev->netdev_ops;
2974                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2975
2976                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2977                         if (idx >= cb->args[0] &&
2978                             br_dev->netdev_ops->ndo_bridge_getlink(
2979                                     skb, portid, seq, dev, filter_mask,
2980                                     NLM_F_MULTI) < 0)
2981                                 break;
2982                         idx++;
2983                 }
2984
2985                 if (ops->ndo_bridge_getlink) {
2986                         if (idx >= cb->args[0] &&
2987                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2988                                                     filter_mask,
2989                                                     NLM_F_MULTI) < 0)
2990                                 break;
2991                         idx++;
2992                 }
2993         }
2994         rcu_read_unlock();
2995         cb->args[0] = idx;
2996
2997         return skb->len;
2998 }
2999
3000 static inline size_t bridge_nlmsg_size(void)
3001 {
3002         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3003                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3004                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3005                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3006                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3007                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3008                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3009                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3010                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3011                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3012                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3013 }
3014
3015 static int rtnl_bridge_notify(struct net_device *dev)
3016 {
3017         struct net *net = dev_net(dev);
3018         struct sk_buff *skb;
3019         int err = -EOPNOTSUPP;
3020
3021         if (!dev->netdev_ops->ndo_bridge_getlink)
3022                 return 0;
3023
3024         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3025         if (!skb) {
3026                 err = -ENOMEM;
3027                 goto errout;
3028         }
3029
3030         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3031         if (err < 0)
3032                 goto errout;
3033
3034         if (!skb->len)
3035                 goto errout;
3036
3037         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3038         return 0;
3039 errout:
3040         WARN_ON(err == -EMSGSIZE);
3041         kfree_skb(skb);
3042         if (err)
3043                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3044         return err;
3045 }
3046
3047 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3048 {
3049         struct net *net = sock_net(skb->sk);
3050         struct ifinfomsg *ifm;
3051         struct net_device *dev;
3052         struct nlattr *br_spec, *attr = NULL;
3053         int rem, err = -EOPNOTSUPP;
3054         u16 flags = 0;
3055         bool have_flags = false;
3056
3057         if (nlmsg_len(nlh) < sizeof(*ifm))
3058                 return -EINVAL;
3059
3060         ifm = nlmsg_data(nlh);
3061         if (ifm->ifi_family != AF_BRIDGE)
3062                 return -EPFNOSUPPORT;
3063
3064         dev = __dev_get_by_index(net, ifm->ifi_index);
3065         if (!dev) {
3066                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3067                 return -ENODEV;
3068         }
3069
3070         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3071         if (br_spec) {
3072                 nla_for_each_nested(attr, br_spec, rem) {
3073                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3074                                 if (nla_len(attr) < sizeof(flags))
3075                                         return -EINVAL;
3076
3077                                 have_flags = true;
3078                                 flags = nla_get_u16(attr);
3079                                 break;
3080                         }
3081                 }
3082         }
3083
3084         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3085                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3086
3087                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3088                         err = -EOPNOTSUPP;
3089                         goto out;
3090                 }
3091
3092                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3093                 if (err)
3094                         goto out;
3095
3096                 flags &= ~BRIDGE_FLAGS_MASTER;
3097         }
3098
3099         if ((flags & BRIDGE_FLAGS_SELF)) {
3100                 if (!dev->netdev_ops->ndo_bridge_setlink)
3101                         err = -EOPNOTSUPP;
3102                 else
3103                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3104                                                                   flags);
3105                 if (!err) {
3106                         flags &= ~BRIDGE_FLAGS_SELF;
3107
3108                         /* Generate event to notify upper layer of bridge
3109                          * change
3110                          */
3111                         err = rtnl_bridge_notify(dev);
3112                 }
3113         }
3114
3115         if (have_flags)
3116                 memcpy(nla_data(attr), &flags, sizeof(flags));
3117 out:
3118         return err;
3119 }
3120
3121 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3122 {
3123         struct net *net = sock_net(skb->sk);
3124         struct ifinfomsg *ifm;
3125         struct net_device *dev;
3126         struct nlattr *br_spec, *attr = NULL;
3127         int rem, err = -EOPNOTSUPP;
3128         u16 flags = 0;
3129         bool have_flags = false;
3130
3131         if (nlmsg_len(nlh) < sizeof(*ifm))
3132                 return -EINVAL;
3133
3134         ifm = nlmsg_data(nlh);
3135         if (ifm->ifi_family != AF_BRIDGE)
3136                 return -EPFNOSUPPORT;
3137
3138         dev = __dev_get_by_index(net, ifm->ifi_index);
3139         if (!dev) {
3140                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3141                 return -ENODEV;
3142         }
3143
3144         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3145         if (br_spec) {
3146                 nla_for_each_nested(attr, br_spec, rem) {
3147                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3148                                 if (nla_len(attr) < sizeof(flags))
3149                                         return -EINVAL;
3150
3151                                 have_flags = true;
3152                                 flags = nla_get_u16(attr);
3153                                 break;
3154                         }
3155                 }
3156         }
3157
3158         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3159                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3160
3161                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3162                         err = -EOPNOTSUPP;
3163                         goto out;
3164                 }
3165
3166                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3167                 if (err)
3168                         goto out;
3169
3170                 flags &= ~BRIDGE_FLAGS_MASTER;
3171         }
3172
3173         if ((flags & BRIDGE_FLAGS_SELF)) {
3174                 if (!dev->netdev_ops->ndo_bridge_dellink)
3175                         err = -EOPNOTSUPP;
3176                 else
3177                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3178                                                                   flags);
3179
3180                 if (!err) {
3181                         flags &= ~BRIDGE_FLAGS_SELF;
3182
3183                         /* Generate event to notify upper layer of bridge
3184                          * change
3185                          */
3186                         err = rtnl_bridge_notify(dev);
3187                 }
3188         }
3189
3190         if (have_flags)
3191                 memcpy(nla_data(attr), &flags, sizeof(flags));
3192 out:
3193         return err;
3194 }
3195
3196 /* Process one rtnetlink message. */
3197
3198 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3199 {
3200         struct net *net = sock_net(skb->sk);
3201         rtnl_doit_func doit;
3202         int sz_idx, kind;
3203         int family;
3204         int type;
3205         int err;
3206
3207         type = nlh->nlmsg_type;
3208         if (type > RTM_MAX)
3209                 return -EOPNOTSUPP;
3210
3211         type -= RTM_BASE;
3212
3213         /* All the messages must have at least 1 byte length */
3214         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3215                 return 0;
3216
3217         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3218         sz_idx = type>>2;
3219         kind = type&3;
3220
3221         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3222                 return -EPERM;
3223
3224         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3225                 struct sock *rtnl;
3226                 rtnl_dumpit_func dumpit;
3227                 rtnl_calcit_func calcit;
3228                 u16 min_dump_alloc = 0;
3229
3230                 dumpit = rtnl_get_dumpit(family, type);
3231                 if (dumpit == NULL)
3232                         return -EOPNOTSUPP;
3233                 calcit = rtnl_get_calcit(family, type);
3234                 if (calcit)
3235                         min_dump_alloc = calcit(skb, nlh);
3236
3237                 __rtnl_unlock();
3238                 rtnl = net->rtnl;
3239                 {
3240                         struct netlink_dump_control c = {
3241                                 .dump           = dumpit,
3242                                 .min_dump_alloc = min_dump_alloc,
3243                         };
3244                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3245                 }
3246                 rtnl_lock();
3247                 return err;
3248         }
3249
3250         doit = rtnl_get_doit(family, type);
3251         if (doit == NULL)
3252                 return -EOPNOTSUPP;
3253
3254         return doit(skb, nlh);
3255 }
3256
3257 static void rtnetlink_rcv(struct sk_buff *skb)
3258 {
3259         rtnl_lock();
3260         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3261         rtnl_unlock();
3262 }
3263
3264 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3265 {
3266         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3267
3268         switch (event) {
3269         case NETDEV_UP:
3270         case NETDEV_DOWN:
3271         case NETDEV_PRE_UP:
3272         case NETDEV_POST_INIT:
3273         case NETDEV_REGISTER:
3274         case NETDEV_CHANGE:
3275         case NETDEV_PRE_TYPE_CHANGE:
3276         case NETDEV_GOING_DOWN:
3277         case NETDEV_UNREGISTER:
3278         case NETDEV_UNREGISTER_FINAL:
3279         case NETDEV_RELEASE:
3280         case NETDEV_JOIN:
3281         case NETDEV_BONDING_INFO:
3282                 break;
3283         default:
3284                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3285                 break;
3286         }
3287         return NOTIFY_DONE;
3288 }
3289
3290 static struct notifier_block rtnetlink_dev_notifier = {
3291         .notifier_call  = rtnetlink_event,
3292 };
3293
3294
3295 static int __net_init rtnetlink_net_init(struct net *net)
3296 {
3297         struct sock *sk;
3298         struct netlink_kernel_cfg cfg = {
3299                 .groups         = RTNLGRP_MAX,
3300                 .input          = rtnetlink_rcv,
3301                 .cb_mutex       = &rtnl_mutex,
3302                 .flags          = NL_CFG_F_NONROOT_RECV,
3303         };
3304
3305         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3306         if (!sk)
3307                 return -ENOMEM;
3308         net->rtnl = sk;
3309         return 0;
3310 }
3311
3312 static void __net_exit rtnetlink_net_exit(struct net *net)
3313 {
3314         netlink_kernel_release(net->rtnl);
3315         net->rtnl = NULL;
3316 }
3317
3318 static struct pernet_operations rtnetlink_net_ops = {
3319         .init = rtnetlink_net_init,
3320         .exit = rtnetlink_net_exit,
3321 };
3322
3323 void __init rtnetlink_init(void)
3324 {
3325         if (register_pernet_subsys(&rtnetlink_net_ops))
3326                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3327
3328         register_netdevice_notifier(&rtnetlink_dev_notifier);
3329
3330         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3331                       rtnl_dump_ifinfo, rtnl_calcit);
3332         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3333         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3334         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3335
3336         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3337         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3338
3339         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3340         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3341         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3342
3343         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3344         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3345         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3346 }
3347