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.ppid.id_len, attr.ppid.id))
1020                 return -EMSGSIZE;
1021
1022         return 0;
1023 }
1024
1025 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1026                             int type, u32 pid, u32 seq, u32 change,
1027                             unsigned int flags, u32 ext_filter_mask)
1028 {
1029         struct ifinfomsg *ifm;
1030         struct nlmsghdr *nlh;
1031         struct rtnl_link_stats64 temp;
1032         const struct rtnl_link_stats64 *stats;
1033         struct nlattr *attr, *af_spec;
1034         struct rtnl_af_ops *af_ops;
1035         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1036
1037         ASSERT_RTNL();
1038         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1039         if (nlh == NULL)
1040                 return -EMSGSIZE;
1041
1042         ifm = nlmsg_data(nlh);
1043         ifm->ifi_family = AF_UNSPEC;
1044         ifm->__ifi_pad = 0;
1045         ifm->ifi_type = dev->type;
1046         ifm->ifi_index = dev->ifindex;
1047         ifm->ifi_flags = dev_get_flags(dev);
1048         ifm->ifi_change = change;
1049
1050         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1051             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1052             nla_put_u8(skb, IFLA_OPERSTATE,
1053                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1054             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1055             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1056             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1057             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1058             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1059 #ifdef CONFIG_RPS
1060             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1061 #endif
1062             (dev->ifindex != dev_get_iflink(dev) &&
1063              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1064             (upper_dev &&
1065              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1066             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1067             (dev->qdisc &&
1068              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1069             (dev->ifalias &&
1070              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1071             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1072                         atomic_read(&dev->carrier_changes)))
1073                 goto nla_put_failure;
1074
1075         if (1) {
1076                 struct rtnl_link_ifmap map = {
1077                         .mem_start   = dev->mem_start,
1078                         .mem_end     = dev->mem_end,
1079                         .base_addr   = dev->base_addr,
1080                         .irq         = dev->irq,
1081                         .dma         = dev->dma,
1082                         .port        = dev->if_port,
1083                 };
1084                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1085                         goto nla_put_failure;
1086         }
1087
1088         if (dev->addr_len) {
1089                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1090                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1091                         goto nla_put_failure;
1092         }
1093
1094         if (rtnl_phys_port_id_fill(skb, dev))
1095                 goto nla_put_failure;
1096
1097         if (rtnl_phys_port_name_fill(skb, dev))
1098                 goto nla_put_failure;
1099
1100         if (rtnl_phys_switch_id_fill(skb, dev))
1101                 goto nla_put_failure;
1102
1103         attr = nla_reserve(skb, IFLA_STATS,
1104                         sizeof(struct rtnl_link_stats));
1105         if (attr == NULL)
1106                 goto nla_put_failure;
1107
1108         stats = dev_get_stats(dev, &temp);
1109         copy_rtnl_link_stats(nla_data(attr), stats);
1110
1111         attr = nla_reserve(skb, IFLA_STATS64,
1112                         sizeof(struct rtnl_link_stats64));
1113         if (attr == NULL)
1114                 goto nla_put_failure;
1115         copy_rtnl_link_stats64(nla_data(attr), stats);
1116
1117         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1118             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1119                 goto nla_put_failure;
1120
1121         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1122             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1123                 int i;
1124
1125                 struct nlattr *vfinfo, *vf;
1126                 int num_vfs = dev_num_vf(dev->dev.parent);
1127
1128                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1129                 if (!vfinfo)
1130                         goto nla_put_failure;
1131                 for (i = 0; i < num_vfs; i++) {
1132                         struct ifla_vf_info ivi;
1133                         struct ifla_vf_mac vf_mac;
1134                         struct ifla_vf_vlan vf_vlan;
1135                         struct ifla_vf_rate vf_rate;
1136                         struct ifla_vf_tx_rate vf_tx_rate;
1137                         struct ifla_vf_spoofchk vf_spoofchk;
1138                         struct ifla_vf_link_state vf_linkstate;
1139                         struct ifla_vf_rss_query_en vf_rss_query_en;
1140
1141                         /*
1142                          * Not all SR-IOV capable drivers support the
1143                          * spoofcheck and "RSS query enable" query.  Preset to
1144                          * -1 so the user space tool can detect that the driver
1145                          * didn't report anything.
1146                          */
1147                         ivi.spoofchk = -1;
1148                         ivi.rss_query_en = -1;
1149                         memset(ivi.mac, 0, sizeof(ivi.mac));
1150                         /* The default value for VF link state is "auto"
1151                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1152                          */
1153                         ivi.linkstate = 0;
1154                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1155                                 break;
1156                         vf_mac.vf =
1157                                 vf_vlan.vf =
1158                                 vf_rate.vf =
1159                                 vf_tx_rate.vf =
1160                                 vf_spoofchk.vf =
1161                                 vf_linkstate.vf =
1162                                 vf_rss_query_en.vf = ivi.vf;
1163
1164                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1165                         vf_vlan.vlan = ivi.vlan;
1166                         vf_vlan.qos = ivi.qos;
1167                         vf_tx_rate.rate = ivi.max_tx_rate;
1168                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1169                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1170                         vf_spoofchk.setting = ivi.spoofchk;
1171                         vf_linkstate.link_state = ivi.linkstate;
1172                         vf_rss_query_en.setting = ivi.rss_query_en;
1173                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1174                         if (!vf) {
1175                                 nla_nest_cancel(skb, vfinfo);
1176                                 goto nla_put_failure;
1177                         }
1178                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1179                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1180                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1181                                     &vf_rate) ||
1182                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1183                                     &vf_tx_rate) ||
1184                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1185                                     &vf_spoofchk) ||
1186                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1187                                     &vf_linkstate) ||
1188                             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1189                                     sizeof(vf_rss_query_en),
1190                                     &vf_rss_query_en))
1191                                 goto nla_put_failure;
1192                         nla_nest_end(skb, vf);
1193                 }
1194                 nla_nest_end(skb, vfinfo);
1195         }
1196
1197         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1198                 goto nla_put_failure;
1199
1200         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1201                 if (rtnl_link_fill(skb, dev) < 0)
1202                         goto nla_put_failure;
1203         }
1204
1205         if (dev->rtnl_link_ops &&
1206             dev->rtnl_link_ops->get_link_net) {
1207                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1208
1209                 if (!net_eq(dev_net(dev), link_net)) {
1210                         int id = peernet2id_alloc(dev_net(dev), link_net);
1211
1212                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1213                                 goto nla_put_failure;
1214                 }
1215         }
1216
1217         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1218                 goto nla_put_failure;
1219
1220         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1221                 if (af_ops->fill_link_af) {
1222                         struct nlattr *af;
1223                         int err;
1224
1225                         if (!(af = nla_nest_start(skb, af_ops->family)))
1226                                 goto nla_put_failure;
1227
1228                         err = af_ops->fill_link_af(skb, dev);
1229
1230                         /*
1231                          * Caller may return ENODATA to indicate that there
1232                          * was no data to be dumped. This is not an error, it
1233                          * means we should trim the attribute header and
1234                          * continue.
1235                          */
1236                         if (err == -ENODATA)
1237                                 nla_nest_cancel(skb, af);
1238                         else if (err < 0)
1239                                 goto nla_put_failure;
1240
1241                         nla_nest_end(skb, af);
1242                 }
1243         }
1244
1245         nla_nest_end(skb, af_spec);
1246
1247         nlmsg_end(skb, nlh);
1248         return 0;
1249
1250 nla_put_failure:
1251         nlmsg_cancel(skb, nlh);
1252         return -EMSGSIZE;
1253 }
1254
1255 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1256         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1257         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1258         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1259         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1260         [IFLA_MTU]              = { .type = NLA_U32 },
1261         [IFLA_LINK]             = { .type = NLA_U32 },
1262         [IFLA_MASTER]           = { .type = NLA_U32 },
1263         [IFLA_CARRIER]          = { .type = NLA_U8 },
1264         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1265         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1266         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1267         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1268         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1269         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1270         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1271         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1272         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1273         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1274         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1275         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1276         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1277         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1278         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1279         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1280         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1281         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1282         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1283         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1284 };
1285
1286 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1287         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1288         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1289         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1290         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1291 };
1292
1293 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1294         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1295 };
1296
1297 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1298         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1299         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1300         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1301         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1302         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1303         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1304         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1305 };
1306
1307 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1308         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1309         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1310                                     .len = PORT_PROFILE_MAX },
1311         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1312                                     .len = sizeof(struct ifla_port_vsi)},
1313         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1314                                       .len = PORT_UUID_MAX },
1315         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1316                                     .len = PORT_UUID_MAX },
1317         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1318         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1319 };
1320
1321 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1322 {
1323         struct net *net = sock_net(skb->sk);
1324         int h, s_h;
1325         int idx = 0, s_idx;
1326         struct net_device *dev;
1327         struct hlist_head *head;
1328         struct nlattr *tb[IFLA_MAX+1];
1329         u32 ext_filter_mask = 0;
1330         int err;
1331         int hdrlen;
1332
1333         s_h = cb->args[0];
1334         s_idx = cb->args[1];
1335
1336         cb->seq = net->dev_base_seq;
1337
1338         /* A hack to preserve kernel<->userspace interface.
1339          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1340          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1341          * what iproute2 < v3.9.0 used.
1342          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1343          * attribute, its netlink message is shorter than struct ifinfomsg.
1344          */
1345         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1346                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1347
1348         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1349
1350                 if (tb[IFLA_EXT_MASK])
1351                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1352         }
1353
1354         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1355                 idx = 0;
1356                 head = &net->dev_index_head[h];
1357                 hlist_for_each_entry(dev, head, index_hlist) {
1358                         if (idx < s_idx)
1359                                 goto cont;
1360                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1361                                                NETLINK_CB(cb->skb).portid,
1362                                                cb->nlh->nlmsg_seq, 0,
1363                                                NLM_F_MULTI,
1364                                                ext_filter_mask);
1365                         /* If we ran out of room on the first message,
1366                          * we're in trouble
1367                          */
1368                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1369
1370                         if (err < 0)
1371                                 goto out;
1372
1373                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1374 cont:
1375                         idx++;
1376                 }
1377         }
1378 out:
1379         cb->args[1] = idx;
1380         cb->args[0] = h;
1381
1382         return skb->len;
1383 }
1384
1385 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1386 {
1387         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1388 }
1389 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1390
1391 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1392 {
1393         struct net *net;
1394         /* Examine the link attributes and figure out which
1395          * network namespace we are talking about.
1396          */
1397         if (tb[IFLA_NET_NS_PID])
1398                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1399         else if (tb[IFLA_NET_NS_FD])
1400                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1401         else
1402                 net = get_net(src_net);
1403         return net;
1404 }
1405 EXPORT_SYMBOL(rtnl_link_get_net);
1406
1407 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1408 {
1409         if (dev) {
1410                 if (tb[IFLA_ADDRESS] &&
1411                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1412                         return -EINVAL;
1413
1414                 if (tb[IFLA_BROADCAST] &&
1415                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1416                         return -EINVAL;
1417         }
1418
1419         if (tb[IFLA_AF_SPEC]) {
1420                 struct nlattr *af;
1421                 int rem, err;
1422
1423                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1424                         const struct rtnl_af_ops *af_ops;
1425
1426                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1427                                 return -EAFNOSUPPORT;
1428
1429                         if (!af_ops->set_link_af)
1430                                 return -EOPNOTSUPP;
1431
1432                         if (af_ops->validate_link_af) {
1433                                 err = af_ops->validate_link_af(dev, af);
1434                                 if (err < 0)
1435                                         return err;
1436                         }
1437                 }
1438         }
1439
1440         return 0;
1441 }
1442
1443 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1444 {
1445         int rem, err = -EINVAL;
1446         struct nlattr *vf;
1447         const struct net_device_ops *ops = dev->netdev_ops;
1448
1449         nla_for_each_nested(vf, attr, rem) {
1450                 switch (nla_type(vf)) {
1451                 case IFLA_VF_MAC: {
1452                         struct ifla_vf_mac *ivm;
1453                         ivm = nla_data(vf);
1454                         err = -EOPNOTSUPP;
1455                         if (ops->ndo_set_vf_mac)
1456                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1457                                                           ivm->mac);
1458                         break;
1459                 }
1460                 case IFLA_VF_VLAN: {
1461                         struct ifla_vf_vlan *ivv;
1462                         ivv = nla_data(vf);
1463                         err = -EOPNOTSUPP;
1464                         if (ops->ndo_set_vf_vlan)
1465                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1466                                                            ivv->vlan,
1467                                                            ivv->qos);
1468                         break;
1469                 }
1470                 case IFLA_VF_TX_RATE: {
1471                         struct ifla_vf_tx_rate *ivt;
1472                         struct ifla_vf_info ivf;
1473                         ivt = nla_data(vf);
1474                         err = -EOPNOTSUPP;
1475                         if (ops->ndo_get_vf_config)
1476                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1477                                                              &ivf);
1478                         if (err)
1479                                 break;
1480                         err = -EOPNOTSUPP;
1481                         if (ops->ndo_set_vf_rate)
1482                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1483                                                            ivf.min_tx_rate,
1484                                                            ivt->rate);
1485                         break;
1486                 }
1487                 case IFLA_VF_RATE: {
1488                         struct ifla_vf_rate *ivt;
1489                         ivt = nla_data(vf);
1490                         err = -EOPNOTSUPP;
1491                         if (ops->ndo_set_vf_rate)
1492                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1493                                                            ivt->min_tx_rate,
1494                                                            ivt->max_tx_rate);
1495                         break;
1496                 }
1497                 case IFLA_VF_SPOOFCHK: {
1498                         struct ifla_vf_spoofchk *ivs;
1499                         ivs = nla_data(vf);
1500                         err = -EOPNOTSUPP;
1501                         if (ops->ndo_set_vf_spoofchk)
1502                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1503                                                                ivs->setting);
1504                         break;
1505                 }
1506                 case IFLA_VF_LINK_STATE: {
1507                         struct ifla_vf_link_state *ivl;
1508                         ivl = nla_data(vf);
1509                         err = -EOPNOTSUPP;
1510                         if (ops->ndo_set_vf_link_state)
1511                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1512                                                                  ivl->link_state);
1513                         break;
1514                 }
1515                 case IFLA_VF_RSS_QUERY_EN: {
1516                         struct ifla_vf_rss_query_en *ivrssq_en;
1517
1518                         ivrssq_en = nla_data(vf);
1519                         err = -EOPNOTSUPP;
1520                         if (ops->ndo_set_vf_rss_query_en)
1521                                 err = ops->ndo_set_vf_rss_query_en(dev,
1522                                                             ivrssq_en->vf,
1523                                                             ivrssq_en->setting);
1524                         break;
1525                 }
1526                 default:
1527                         err = -EINVAL;
1528                         break;
1529                 }
1530                 if (err)
1531                         break;
1532         }
1533         return err;
1534 }
1535
1536 static int do_set_master(struct net_device *dev, int ifindex)
1537 {
1538         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1539         const struct net_device_ops *ops;
1540         int err;
1541
1542         if (upper_dev) {
1543                 if (upper_dev->ifindex == ifindex)
1544                         return 0;
1545                 ops = upper_dev->netdev_ops;
1546                 if (ops->ndo_del_slave) {
1547                         err = ops->ndo_del_slave(upper_dev, dev);
1548                         if (err)
1549                                 return err;
1550                 } else {
1551                         return -EOPNOTSUPP;
1552                 }
1553         }
1554
1555         if (ifindex) {
1556                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1557                 if (!upper_dev)
1558                         return -EINVAL;
1559                 ops = upper_dev->netdev_ops;
1560                 if (ops->ndo_add_slave) {
1561                         err = ops->ndo_add_slave(upper_dev, dev);
1562                         if (err)
1563                                 return err;
1564                 } else {
1565                         return -EOPNOTSUPP;
1566                 }
1567         }
1568         return 0;
1569 }
1570
1571 #define DO_SETLINK_MODIFIED     0x01
1572 /* notify flag means notify + modified. */
1573 #define DO_SETLINK_NOTIFY       0x03
1574 static int do_setlink(const struct sk_buff *skb,
1575                       struct net_device *dev, struct ifinfomsg *ifm,
1576                       struct nlattr **tb, char *ifname, int status)
1577 {
1578         const struct net_device_ops *ops = dev->netdev_ops;
1579         int err;
1580
1581         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1582                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1583                 if (IS_ERR(net)) {
1584                         err = PTR_ERR(net);
1585                         goto errout;
1586                 }
1587                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1588                         put_net(net);
1589                         err = -EPERM;
1590                         goto errout;
1591                 }
1592                 err = dev_change_net_namespace(dev, net, ifname);
1593                 put_net(net);
1594                 if (err)
1595                         goto errout;
1596                 status |= DO_SETLINK_MODIFIED;
1597         }
1598
1599         if (tb[IFLA_MAP]) {
1600                 struct rtnl_link_ifmap *u_map;
1601                 struct ifmap k_map;
1602
1603                 if (!ops->ndo_set_config) {
1604                         err = -EOPNOTSUPP;
1605                         goto errout;
1606                 }
1607
1608                 if (!netif_device_present(dev)) {
1609                         err = -ENODEV;
1610                         goto errout;
1611                 }
1612
1613                 u_map = nla_data(tb[IFLA_MAP]);
1614                 k_map.mem_start = (unsigned long) u_map->mem_start;
1615                 k_map.mem_end = (unsigned long) u_map->mem_end;
1616                 k_map.base_addr = (unsigned short) u_map->base_addr;
1617                 k_map.irq = (unsigned char) u_map->irq;
1618                 k_map.dma = (unsigned char) u_map->dma;
1619                 k_map.port = (unsigned char) u_map->port;
1620
1621                 err = ops->ndo_set_config(dev, &k_map);
1622                 if (err < 0)
1623                         goto errout;
1624
1625                 status |= DO_SETLINK_NOTIFY;
1626         }
1627
1628         if (tb[IFLA_ADDRESS]) {
1629                 struct sockaddr *sa;
1630                 int len;
1631
1632                 len = sizeof(sa_family_t) + dev->addr_len;
1633                 sa = kmalloc(len, GFP_KERNEL);
1634                 if (!sa) {
1635                         err = -ENOMEM;
1636                         goto errout;
1637                 }
1638                 sa->sa_family = dev->type;
1639                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1640                        dev->addr_len);
1641                 err = dev_set_mac_address(dev, sa);
1642                 kfree(sa);
1643                 if (err)
1644                         goto errout;
1645                 status |= DO_SETLINK_MODIFIED;
1646         }
1647
1648         if (tb[IFLA_MTU]) {
1649                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1650                 if (err < 0)
1651                         goto errout;
1652                 status |= DO_SETLINK_MODIFIED;
1653         }
1654
1655         if (tb[IFLA_GROUP]) {
1656                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1657                 status |= DO_SETLINK_NOTIFY;
1658         }
1659
1660         /*
1661          * Interface selected by interface index but interface
1662          * name provided implies that a name change has been
1663          * requested.
1664          */
1665         if (ifm->ifi_index > 0 && ifname[0]) {
1666                 err = dev_change_name(dev, ifname);
1667                 if (err < 0)
1668                         goto errout;
1669                 status |= DO_SETLINK_MODIFIED;
1670         }
1671
1672         if (tb[IFLA_IFALIAS]) {
1673                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1674                                     nla_len(tb[IFLA_IFALIAS]));
1675                 if (err < 0)
1676                         goto errout;
1677                 status |= DO_SETLINK_NOTIFY;
1678         }
1679
1680         if (tb[IFLA_BROADCAST]) {
1681                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1682                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1683         }
1684
1685         if (ifm->ifi_flags || ifm->ifi_change) {
1686                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1687                 if (err < 0)
1688                         goto errout;
1689         }
1690
1691         if (tb[IFLA_MASTER]) {
1692                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1693                 if (err)
1694                         goto errout;
1695                 status |= DO_SETLINK_MODIFIED;
1696         }
1697
1698         if (tb[IFLA_CARRIER]) {
1699                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1700                 if (err)
1701                         goto errout;
1702                 status |= DO_SETLINK_MODIFIED;
1703         }
1704
1705         if (tb[IFLA_TXQLEN]) {
1706                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1707
1708                 if (dev->tx_queue_len ^ value)
1709                         status |= DO_SETLINK_NOTIFY;
1710
1711                 dev->tx_queue_len = value;
1712         }
1713
1714         if (tb[IFLA_OPERSTATE])
1715                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1716
1717         if (tb[IFLA_LINKMODE]) {
1718                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1719
1720                 write_lock_bh(&dev_base_lock);
1721                 if (dev->link_mode ^ value)
1722                         status |= DO_SETLINK_NOTIFY;
1723                 dev->link_mode = value;
1724                 write_unlock_bh(&dev_base_lock);
1725         }
1726
1727         if (tb[IFLA_VFINFO_LIST]) {
1728                 struct nlattr *attr;
1729                 int rem;
1730                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1731                         if (nla_type(attr) != IFLA_VF_INFO) {
1732                                 err = -EINVAL;
1733                                 goto errout;
1734                         }
1735                         err = do_setvfinfo(dev, attr);
1736                         if (err < 0)
1737                                 goto errout;
1738                         status |= DO_SETLINK_NOTIFY;
1739                 }
1740         }
1741         err = 0;
1742
1743         if (tb[IFLA_VF_PORTS]) {
1744                 struct nlattr *port[IFLA_PORT_MAX+1];
1745                 struct nlattr *attr;
1746                 int vf;
1747                 int rem;
1748
1749                 err = -EOPNOTSUPP;
1750                 if (!ops->ndo_set_vf_port)
1751                         goto errout;
1752
1753                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1754                         if (nla_type(attr) != IFLA_VF_PORT)
1755                                 continue;
1756                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1757                                 attr, ifla_port_policy);
1758                         if (err < 0)
1759                                 goto errout;
1760                         if (!port[IFLA_PORT_VF]) {
1761                                 err = -EOPNOTSUPP;
1762                                 goto errout;
1763                         }
1764                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1765                         err = ops->ndo_set_vf_port(dev, vf, port);
1766                         if (err < 0)
1767                                 goto errout;
1768                         status |= DO_SETLINK_NOTIFY;
1769                 }
1770         }
1771         err = 0;
1772
1773         if (tb[IFLA_PORT_SELF]) {
1774                 struct nlattr *port[IFLA_PORT_MAX+1];
1775
1776                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1777                         tb[IFLA_PORT_SELF], ifla_port_policy);
1778                 if (err < 0)
1779                         goto errout;
1780
1781                 err = -EOPNOTSUPP;
1782                 if (ops->ndo_set_vf_port)
1783                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1784                 if (err < 0)
1785                         goto errout;
1786                 status |= DO_SETLINK_NOTIFY;
1787         }
1788
1789         if (tb[IFLA_AF_SPEC]) {
1790                 struct nlattr *af;
1791                 int rem;
1792
1793                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1794                         const struct rtnl_af_ops *af_ops;
1795
1796                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1797                                 BUG();
1798
1799                         err = af_ops->set_link_af(dev, af);
1800                         if (err < 0)
1801                                 goto errout;
1802
1803                         status |= DO_SETLINK_NOTIFY;
1804                 }
1805         }
1806         err = 0;
1807
1808 errout:
1809         if (status & DO_SETLINK_MODIFIED) {
1810                 if (status & DO_SETLINK_NOTIFY)
1811                         netdev_state_change(dev);
1812
1813                 if (err < 0)
1814                         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",
1815                                              dev->name);
1816         }
1817
1818         return err;
1819 }
1820
1821 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1822 {
1823         struct net *net = sock_net(skb->sk);
1824         struct ifinfomsg *ifm;
1825         struct net_device *dev;
1826         int err;
1827         struct nlattr *tb[IFLA_MAX+1];
1828         char ifname[IFNAMSIZ];
1829
1830         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1831         if (err < 0)
1832                 goto errout;
1833
1834         if (tb[IFLA_IFNAME])
1835                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1836         else
1837                 ifname[0] = '\0';
1838
1839         err = -EINVAL;
1840         ifm = nlmsg_data(nlh);
1841         if (ifm->ifi_index > 0)
1842                 dev = __dev_get_by_index(net, ifm->ifi_index);
1843         else if (tb[IFLA_IFNAME])
1844                 dev = __dev_get_by_name(net, ifname);
1845         else
1846                 goto errout;
1847
1848         if (dev == NULL) {
1849                 err = -ENODEV;
1850                 goto errout;
1851         }
1852
1853         err = validate_linkmsg(dev, tb);
1854         if (err < 0)
1855                 goto errout;
1856
1857         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1858 errout:
1859         return err;
1860 }
1861
1862 static int rtnl_group_dellink(const struct net *net, int group)
1863 {
1864         struct net_device *dev, *aux;
1865         LIST_HEAD(list_kill);
1866         bool found = false;
1867
1868         if (!group)
1869                 return -EPERM;
1870
1871         for_each_netdev(net, dev) {
1872                 if (dev->group == group) {
1873                         const struct rtnl_link_ops *ops;
1874
1875                         found = true;
1876                         ops = dev->rtnl_link_ops;
1877                         if (!ops || !ops->dellink)
1878                                 return -EOPNOTSUPP;
1879                 }
1880         }
1881
1882         if (!found)
1883                 return -ENODEV;
1884
1885         for_each_netdev_safe(net, dev, aux) {
1886                 if (dev->group == group) {
1887                         const struct rtnl_link_ops *ops;
1888
1889                         ops = dev->rtnl_link_ops;
1890                         ops->dellink(dev, &list_kill);
1891                 }
1892         }
1893         unregister_netdevice_many(&list_kill);
1894
1895         return 0;
1896 }
1897
1898 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1899 {
1900         struct net *net = sock_net(skb->sk);
1901         const struct rtnl_link_ops *ops;
1902         struct net_device *dev;
1903         struct ifinfomsg *ifm;
1904         char ifname[IFNAMSIZ];
1905         struct nlattr *tb[IFLA_MAX+1];
1906         int err;
1907         LIST_HEAD(list_kill);
1908
1909         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1910         if (err < 0)
1911                 return err;
1912
1913         if (tb[IFLA_IFNAME])
1914                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1915
1916         ifm = nlmsg_data(nlh);
1917         if (ifm->ifi_index > 0)
1918                 dev = __dev_get_by_index(net, ifm->ifi_index);
1919         else if (tb[IFLA_IFNAME])
1920                 dev = __dev_get_by_name(net, ifname);
1921         else if (tb[IFLA_GROUP])
1922                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1923         else
1924                 return -EINVAL;
1925
1926         if (!dev)
1927                 return -ENODEV;
1928
1929         ops = dev->rtnl_link_ops;
1930         if (!ops || !ops->dellink)
1931                 return -EOPNOTSUPP;
1932
1933         ops->dellink(dev, &list_kill);
1934         unregister_netdevice_many(&list_kill);
1935         return 0;
1936 }
1937
1938 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1939 {
1940         unsigned int old_flags;
1941         int err;
1942
1943         old_flags = dev->flags;
1944         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1945                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1946                 if (err < 0)
1947                         return err;
1948         }
1949
1950         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1951
1952         __dev_notify_flags(dev, old_flags, ~0U);
1953         return 0;
1954 }
1955 EXPORT_SYMBOL(rtnl_configure_link);
1956
1957 struct net_device *rtnl_create_link(struct net *net,
1958         const char *ifname, unsigned char name_assign_type,
1959         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1960 {
1961         int err;
1962         struct net_device *dev;
1963         unsigned int num_tx_queues = 1;
1964         unsigned int num_rx_queues = 1;
1965
1966         if (tb[IFLA_NUM_TX_QUEUES])
1967                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1968         else if (ops->get_num_tx_queues)
1969                 num_tx_queues = ops->get_num_tx_queues();
1970
1971         if (tb[IFLA_NUM_RX_QUEUES])
1972                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1973         else if (ops->get_num_rx_queues)
1974                 num_rx_queues = ops->get_num_rx_queues();
1975
1976         err = -ENOMEM;
1977         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1978                                ops->setup, num_tx_queues, num_rx_queues);
1979         if (!dev)
1980                 goto err;
1981
1982         dev_net_set(dev, net);
1983         dev->rtnl_link_ops = ops;
1984         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1985
1986         if (tb[IFLA_MTU])
1987                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1988         if (tb[IFLA_ADDRESS]) {
1989                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1990                                 nla_len(tb[IFLA_ADDRESS]));
1991                 dev->addr_assign_type = NET_ADDR_SET;
1992         }
1993         if (tb[IFLA_BROADCAST])
1994                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1995                                 nla_len(tb[IFLA_BROADCAST]));
1996         if (tb[IFLA_TXQLEN])
1997                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1998         if (tb[IFLA_OPERSTATE])
1999                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2000         if (tb[IFLA_LINKMODE])
2001                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2002         if (tb[IFLA_GROUP])
2003                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2004
2005         return dev;
2006
2007 err:
2008         return ERR_PTR(err);
2009 }
2010 EXPORT_SYMBOL(rtnl_create_link);
2011
2012 static int rtnl_group_changelink(const struct sk_buff *skb,
2013                 struct net *net, int group,
2014                 struct ifinfomsg *ifm,
2015                 struct nlattr **tb)
2016 {
2017         struct net_device *dev, *aux;
2018         int err;
2019
2020         for_each_netdev_safe(net, dev, aux) {
2021                 if (dev->group == group) {
2022                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2023                         if (err < 0)
2024                                 return err;
2025                 }
2026         }
2027
2028         return 0;
2029 }
2030
2031 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2032 {
2033         struct net *net = sock_net(skb->sk);
2034         const struct rtnl_link_ops *ops;
2035         const struct rtnl_link_ops *m_ops = NULL;
2036         struct net_device *dev;
2037         struct net_device *master_dev = NULL;
2038         struct ifinfomsg *ifm;
2039         char kind[MODULE_NAME_LEN];
2040         char ifname[IFNAMSIZ];
2041         struct nlattr *tb[IFLA_MAX+1];
2042         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2043         unsigned char name_assign_type = NET_NAME_USER;
2044         int err;
2045
2046 #ifdef CONFIG_MODULES
2047 replay:
2048 #endif
2049         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2050         if (err < 0)
2051                 return err;
2052
2053         if (tb[IFLA_IFNAME])
2054                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2055         else
2056                 ifname[0] = '\0';
2057
2058         ifm = nlmsg_data(nlh);
2059         if (ifm->ifi_index > 0)
2060                 dev = __dev_get_by_index(net, ifm->ifi_index);
2061         else {
2062                 if (ifname[0])
2063                         dev = __dev_get_by_name(net, ifname);
2064                 else
2065                         dev = NULL;
2066         }
2067
2068         if (dev) {
2069                 master_dev = netdev_master_upper_dev_get(dev);
2070                 if (master_dev)
2071                         m_ops = master_dev->rtnl_link_ops;
2072         }
2073
2074         err = validate_linkmsg(dev, tb);
2075         if (err < 0)
2076                 return err;
2077
2078         if (tb[IFLA_LINKINFO]) {
2079                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2080                                        tb[IFLA_LINKINFO], ifla_info_policy);
2081                 if (err < 0)
2082                         return err;
2083         } else
2084                 memset(linkinfo, 0, sizeof(linkinfo));
2085
2086         if (linkinfo[IFLA_INFO_KIND]) {
2087                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2088                 ops = rtnl_link_ops_get(kind);
2089         } else {
2090                 kind[0] = '\0';
2091                 ops = NULL;
2092         }
2093
2094         if (1) {
2095                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2096                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2097                 struct nlattr **data = NULL;
2098                 struct nlattr **slave_data = NULL;
2099                 struct net *dest_net, *link_net = NULL;
2100
2101                 if (ops) {
2102                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2103                                 err = nla_parse_nested(attr, ops->maxtype,
2104                                                        linkinfo[IFLA_INFO_DATA],
2105                                                        ops->policy);
2106                                 if (err < 0)
2107                                         return err;
2108                                 data = attr;
2109                         }
2110                         if (ops->validate) {
2111                                 err = ops->validate(tb, data);
2112                                 if (err < 0)
2113                                         return err;
2114                         }
2115                 }
2116
2117                 if (m_ops) {
2118                         if (m_ops->slave_maxtype &&
2119                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2120                                 err = nla_parse_nested(slave_attr,
2121                                                        m_ops->slave_maxtype,
2122                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2123                                                        m_ops->slave_policy);
2124                                 if (err < 0)
2125                                         return err;
2126                                 slave_data = slave_attr;
2127                         }
2128                         if (m_ops->slave_validate) {
2129                                 err = m_ops->slave_validate(tb, slave_data);
2130                                 if (err < 0)
2131                                         return err;
2132                         }
2133                 }
2134
2135                 if (dev) {
2136                         int status = 0;
2137
2138                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2139                                 return -EEXIST;
2140                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2141                                 return -EOPNOTSUPP;
2142
2143                         if (linkinfo[IFLA_INFO_DATA]) {
2144                                 if (!ops || ops != dev->rtnl_link_ops ||
2145                                     !ops->changelink)
2146                                         return -EOPNOTSUPP;
2147
2148                                 err = ops->changelink(dev, tb, data);
2149                                 if (err < 0)
2150                                         return err;
2151                                 status |= DO_SETLINK_NOTIFY;
2152                         }
2153
2154                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2155                                 if (!m_ops || !m_ops->slave_changelink)
2156                                         return -EOPNOTSUPP;
2157
2158                                 err = m_ops->slave_changelink(master_dev, dev,
2159                                                               tb, slave_data);
2160                                 if (err < 0)
2161                                         return err;
2162                                 status |= DO_SETLINK_NOTIFY;
2163                         }
2164
2165                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2166                 }
2167
2168                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2169                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2170                                 return rtnl_group_changelink(skb, net,
2171                                                 nla_get_u32(tb[IFLA_GROUP]),
2172                                                 ifm, tb);
2173                         return -ENODEV;
2174                 }
2175
2176                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2177                         return -EOPNOTSUPP;
2178
2179                 if (!ops) {
2180 #ifdef CONFIG_MODULES
2181                         if (kind[0]) {
2182                                 __rtnl_unlock();
2183                                 request_module("rtnl-link-%s", kind);
2184                                 rtnl_lock();
2185                                 ops = rtnl_link_ops_get(kind);
2186                                 if (ops)
2187                                         goto replay;
2188                         }
2189 #endif
2190                         return -EOPNOTSUPP;
2191                 }
2192
2193                 if (!ops->setup)
2194                         return -EOPNOTSUPP;
2195
2196                 if (!ifname[0]) {
2197                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2198                         name_assign_type = NET_NAME_ENUM;
2199                 }
2200
2201                 dest_net = rtnl_link_get_net(net, tb);
2202                 if (IS_ERR(dest_net))
2203                         return PTR_ERR(dest_net);
2204
2205                 err = -EPERM;
2206                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2207                         goto out;
2208
2209                 if (tb[IFLA_LINK_NETNSID]) {
2210                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2211
2212                         link_net = get_net_ns_by_id(dest_net, id);
2213                         if (!link_net) {
2214                                 err =  -EINVAL;
2215                                 goto out;
2216                         }
2217                         err = -EPERM;
2218                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2219                                 goto out;
2220                 }
2221
2222                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2223                                        name_assign_type, ops, tb);
2224                 if (IS_ERR(dev)) {
2225                         err = PTR_ERR(dev);
2226                         goto out;
2227                 }
2228
2229                 dev->ifindex = ifm->ifi_index;
2230
2231                 if (ops->newlink) {
2232                         err = ops->newlink(link_net ? : net, dev, tb, data);
2233                         /* Drivers should call free_netdev() in ->destructor
2234                          * and unregister it on failure after registration
2235                          * so that device could be finally freed in rtnl_unlock.
2236                          */
2237                         if (err < 0) {
2238                                 /* If device is not registered at all, free it now */
2239                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2240                                         free_netdev(dev);
2241                                 goto out;
2242                         }
2243                 } else {
2244                         err = register_netdevice(dev);
2245                         if (err < 0) {
2246                                 free_netdev(dev);
2247                                 goto out;
2248                         }
2249                 }
2250                 err = rtnl_configure_link(dev, ifm);
2251                 if (err < 0)
2252                         goto out_unregister;
2253                 if (link_net) {
2254                         err = dev_change_net_namespace(dev, dest_net, ifname);
2255                         if (err < 0)
2256                                 goto out_unregister;
2257                 }
2258 out:
2259                 if (link_net)
2260                         put_net(link_net);
2261                 put_net(dest_net);
2262                 return err;
2263 out_unregister:
2264                 if (ops->newlink) {
2265                         LIST_HEAD(list_kill);
2266
2267                         ops->dellink(dev, &list_kill);
2268                         unregister_netdevice_many(&list_kill);
2269                 } else {
2270                         unregister_netdevice(dev);
2271                 }
2272                 goto out;
2273         }
2274 }
2275
2276 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2277 {
2278         struct net *net = sock_net(skb->sk);
2279         struct ifinfomsg *ifm;
2280         char ifname[IFNAMSIZ];
2281         struct nlattr *tb[IFLA_MAX+1];
2282         struct net_device *dev = NULL;
2283         struct sk_buff *nskb;
2284         int err;
2285         u32 ext_filter_mask = 0;
2286
2287         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2288         if (err < 0)
2289                 return err;
2290
2291         if (tb[IFLA_IFNAME])
2292                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2293
2294         if (tb[IFLA_EXT_MASK])
2295                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2296
2297         ifm = nlmsg_data(nlh);
2298         if (ifm->ifi_index > 0)
2299                 dev = __dev_get_by_index(net, ifm->ifi_index);
2300         else if (tb[IFLA_IFNAME])
2301                 dev = __dev_get_by_name(net, ifname);
2302         else
2303                 return -EINVAL;
2304
2305         if (dev == NULL)
2306                 return -ENODEV;
2307
2308         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2309         if (nskb == NULL)
2310                 return -ENOBUFS;
2311
2312         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2313                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2314         if (err < 0) {
2315                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2316                 WARN_ON(err == -EMSGSIZE);
2317                 kfree_skb(nskb);
2318         } else
2319                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2320
2321         return err;
2322 }
2323
2324 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2325 {
2326         struct net *net = sock_net(skb->sk);
2327         struct net_device *dev;
2328         struct nlattr *tb[IFLA_MAX+1];
2329         u32 ext_filter_mask = 0;
2330         u16 min_ifinfo_dump_size = 0;
2331         int hdrlen;
2332
2333         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2334         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2335                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2336
2337         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2338                 if (tb[IFLA_EXT_MASK])
2339                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2340         }
2341
2342         if (!ext_filter_mask)
2343                 return NLMSG_GOODSIZE;
2344         /*
2345          * traverse the list of net devices and compute the minimum
2346          * buffer size based upon the filter mask.
2347          */
2348         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2349                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2350                                              if_nlmsg_size(dev,
2351                                                            ext_filter_mask));
2352         }
2353
2354         return min_ifinfo_dump_size;
2355 }
2356
2357 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2358 {
2359         int idx;
2360         int s_idx = cb->family;
2361
2362         if (s_idx == 0)
2363                 s_idx = 1;
2364         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2365                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2366                 if (idx < s_idx || idx == PF_PACKET)
2367                         continue;
2368                 if (rtnl_msg_handlers[idx] == NULL ||
2369                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2370                         continue;
2371                 if (idx > s_idx) {
2372                         memset(&cb->args[0], 0, sizeof(cb->args));
2373                         cb->prev_seq = 0;
2374                         cb->seq = 0;
2375                 }
2376                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2377                         break;
2378         }
2379         cb->family = idx;
2380
2381         return skb->len;
2382 }
2383
2384 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2385                                        unsigned int change, gfp_t flags)
2386 {
2387         struct net *net = dev_net(dev);
2388         struct sk_buff *skb;
2389         int err = -ENOBUFS;
2390         size_t if_info_size;
2391
2392         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2393         if (skb == NULL)
2394                 goto errout;
2395
2396         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2397         if (err < 0) {
2398                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2399                 WARN_ON(err == -EMSGSIZE);
2400                 kfree_skb(skb);
2401                 goto errout;
2402         }
2403         return skb;
2404 errout:
2405         if (err < 0)
2406                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2407         return NULL;
2408 }
2409
2410 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2411 {
2412         struct net *net = dev_net(dev);
2413
2414         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2415 }
2416
2417 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2418                   gfp_t flags)
2419 {
2420         struct sk_buff *skb;
2421
2422         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2423         if (skb)
2424                 rtmsg_ifinfo_send(skb, dev, flags);
2425 }
2426 EXPORT_SYMBOL(rtmsg_ifinfo);
2427
2428 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2429                                    struct net_device *dev,
2430                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2431                                    int type, unsigned int flags,
2432                                    int nlflags)
2433 {
2434         struct nlmsghdr *nlh;
2435         struct ndmsg *ndm;
2436
2437         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2438         if (!nlh)
2439                 return -EMSGSIZE;
2440
2441         ndm = nlmsg_data(nlh);
2442         ndm->ndm_family  = AF_BRIDGE;
2443         ndm->ndm_pad1    = 0;
2444         ndm->ndm_pad2    = 0;
2445         ndm->ndm_flags   = flags;
2446         ndm->ndm_type    = 0;
2447         ndm->ndm_ifindex = dev->ifindex;
2448         ndm->ndm_state   = NUD_PERMANENT;
2449
2450         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2451                 goto nla_put_failure;
2452         if (vid)
2453                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2454                         goto nla_put_failure;
2455
2456         nlmsg_end(skb, nlh);
2457         return 0;
2458
2459 nla_put_failure:
2460         nlmsg_cancel(skb, nlh);
2461         return -EMSGSIZE;
2462 }
2463
2464 static inline size_t rtnl_fdb_nlmsg_size(void)
2465 {
2466         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2467 }
2468
2469 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2470 {
2471         struct net *net = dev_net(dev);
2472         struct sk_buff *skb;
2473         int err = -ENOBUFS;
2474
2475         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2476         if (!skb)
2477                 goto errout;
2478
2479         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2480                                       0, 0, type, NTF_SELF, 0);
2481         if (err < 0) {
2482                 kfree_skb(skb);
2483                 goto errout;
2484         }
2485
2486         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2487         return;
2488 errout:
2489         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2490 }
2491
2492 /**
2493  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2494  */
2495 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2496                      struct nlattr *tb[],
2497                      struct net_device *dev,
2498                      const unsigned char *addr, u16 vid,
2499                      u16 flags)
2500 {
2501         int err = -EINVAL;
2502
2503         /* If aging addresses are supported device will need to
2504          * implement its own handler for this.
2505          */
2506         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2507                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2508                 return err;
2509         }
2510
2511         if (vid) {
2512                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2513                 return err;
2514         }
2515
2516         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2517                 err = dev_uc_add_excl(dev, addr);
2518         else if (is_multicast_ether_addr(addr))
2519                 err = dev_mc_add_excl(dev, addr);
2520
2521         /* Only return duplicate errors if NLM_F_EXCL is set */
2522         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2523                 err = 0;
2524
2525         return err;
2526 }
2527 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2528
2529 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2530 {
2531         u16 vid = 0;
2532
2533         if (vlan_attr) {
2534                 if (nla_len(vlan_attr) != sizeof(u16)) {
2535                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2536                         return -EINVAL;
2537                 }
2538
2539                 vid = nla_get_u16(vlan_attr);
2540
2541                 if (!vid || vid >= VLAN_VID_MASK) {
2542                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2543                                 vid);
2544                         return -EINVAL;
2545                 }
2546         }
2547         *p_vid = vid;
2548         return 0;
2549 }
2550
2551 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2552 {
2553         struct net *net = sock_net(skb->sk);
2554         struct ndmsg *ndm;
2555         struct nlattr *tb[NDA_MAX+1];
2556         struct net_device *dev;
2557         u8 *addr;
2558         u16 vid;
2559         int err;
2560
2561         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2562         if (err < 0)
2563                 return err;
2564
2565         ndm = nlmsg_data(nlh);
2566         if (ndm->ndm_ifindex == 0) {
2567                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2568                 return -EINVAL;
2569         }
2570
2571         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2572         if (dev == NULL) {
2573                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2574                 return -ENODEV;
2575         }
2576
2577         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2578                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2579                 return -EINVAL;
2580         }
2581
2582         addr = nla_data(tb[NDA_LLADDR]);
2583
2584         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2585         if (err)
2586                 return err;
2587
2588         err = -EOPNOTSUPP;
2589
2590         /* Support fdb on master device the net/bridge default case */
2591         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2592             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2593                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2594                 const struct net_device_ops *ops = br_dev->netdev_ops;
2595
2596                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2597                                        nlh->nlmsg_flags);
2598                 if (err)
2599                         goto out;
2600                 else
2601                         ndm->ndm_flags &= ~NTF_MASTER;
2602         }
2603
2604         /* Embedded bridge, macvlan, and any other device support */
2605         if ((ndm->ndm_flags & NTF_SELF)) {
2606                 if (dev->netdev_ops->ndo_fdb_add)
2607                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2608                                                            vid,
2609                                                            nlh->nlmsg_flags);
2610                 else
2611                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2612                                                nlh->nlmsg_flags);
2613
2614                 if (!err) {
2615                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2616                         ndm->ndm_flags &= ~NTF_SELF;
2617                 }
2618         }
2619 out:
2620         return err;
2621 }
2622
2623 /**
2624  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2625  */
2626 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2627                      struct nlattr *tb[],
2628                      struct net_device *dev,
2629                      const unsigned char *addr, u16 vid)
2630 {
2631         int err = -EINVAL;
2632
2633         /* If aging addresses are supported device will need to
2634          * implement its own handler for this.
2635          */
2636         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2637                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2638                 return err;
2639         }
2640
2641         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2642                 err = dev_uc_del(dev, addr);
2643         else if (is_multicast_ether_addr(addr))
2644                 err = dev_mc_del(dev, addr);
2645
2646         return err;
2647 }
2648 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2649
2650 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2651 {
2652         struct net *net = sock_net(skb->sk);
2653         struct ndmsg *ndm;
2654         struct nlattr *tb[NDA_MAX+1];
2655         struct net_device *dev;
2656         int err = -EINVAL;
2657         __u8 *addr;
2658         u16 vid;
2659
2660         if (!netlink_capable(skb, CAP_NET_ADMIN))
2661                 return -EPERM;
2662
2663         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2664         if (err < 0)
2665                 return err;
2666
2667         ndm = nlmsg_data(nlh);
2668         if (ndm->ndm_ifindex == 0) {
2669                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2670                 return -EINVAL;
2671         }
2672
2673         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2674         if (dev == NULL) {
2675                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2676                 return -ENODEV;
2677         }
2678
2679         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2680                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2681                 return -EINVAL;
2682         }
2683
2684         addr = nla_data(tb[NDA_LLADDR]);
2685
2686         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2687         if (err)
2688                 return err;
2689
2690         err = -EOPNOTSUPP;
2691
2692         /* Support fdb on master device the net/bridge default case */
2693         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2694             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2695                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2696                 const struct net_device_ops *ops = br_dev->netdev_ops;
2697
2698                 if (ops->ndo_fdb_del)
2699                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2700
2701                 if (err)
2702                         goto out;
2703                 else
2704                         ndm->ndm_flags &= ~NTF_MASTER;
2705         }
2706
2707         /* Embedded bridge, macvlan, and any other device support */
2708         if (ndm->ndm_flags & NTF_SELF) {
2709                 if (dev->netdev_ops->ndo_fdb_del)
2710                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2711                                                            vid);
2712                 else
2713                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2714
2715                 if (!err) {
2716                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2717                         ndm->ndm_flags &= ~NTF_SELF;
2718                 }
2719         }
2720 out:
2721         return err;
2722 }
2723
2724 static int nlmsg_populate_fdb(struct sk_buff *skb,
2725                               struct netlink_callback *cb,
2726                               struct net_device *dev,
2727                               int *idx,
2728                               struct netdev_hw_addr_list *list)
2729 {
2730         struct netdev_hw_addr *ha;
2731         int err;
2732         u32 portid, seq;
2733
2734         portid = NETLINK_CB(cb->skb).portid;
2735         seq = cb->nlh->nlmsg_seq;
2736
2737         list_for_each_entry(ha, &list->list, list) {
2738                 if (*idx < cb->args[0])
2739                         goto skip;
2740
2741                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2742                                               portid, seq,
2743                                               RTM_NEWNEIGH, NTF_SELF,
2744                                               NLM_F_MULTI);
2745                 if (err < 0)
2746                         return err;
2747 skip:
2748                 *idx += 1;
2749         }
2750         return 0;
2751 }
2752
2753 /**
2754  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2755  * @nlh: netlink message header
2756  * @dev: netdevice
2757  *
2758  * Default netdevice operation to dump the existing unicast address list.
2759  * Returns number of addresses from list put in skb.
2760  */
2761 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2762                       struct netlink_callback *cb,
2763                       struct net_device *dev,
2764                       struct net_device *filter_dev,
2765                       int idx)
2766 {
2767         int err;
2768
2769         netif_addr_lock_bh(dev);
2770         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2771         if (err)
2772                 goto out;
2773         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2774 out:
2775         netif_addr_unlock_bh(dev);
2776         return idx;
2777 }
2778 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2779
2780 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2781 {
2782         struct net_device *dev;
2783         struct nlattr *tb[IFLA_MAX+1];
2784         struct net_device *br_dev = NULL;
2785         const struct net_device_ops *ops = NULL;
2786         const struct net_device_ops *cops = NULL;
2787         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2788         struct net *net = sock_net(skb->sk);
2789         int brport_idx = 0;
2790         int br_idx = 0;
2791         int idx = 0;
2792
2793         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2794                         ifla_policy) == 0) {
2795                 if (tb[IFLA_MASTER])
2796                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2797         }
2798
2799         brport_idx = ifm->ifi_index;
2800
2801         if (br_idx) {
2802                 br_dev = __dev_get_by_index(net, br_idx);
2803                 if (!br_dev)
2804                         return -ENODEV;
2805
2806                 ops = br_dev->netdev_ops;
2807         }
2808
2809         for_each_netdev(net, dev) {
2810                 if (brport_idx && (dev->ifindex != brport_idx))
2811                         continue;
2812
2813                 if (!br_idx) { /* user did not specify a specific bridge */
2814                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2815                                 br_dev = netdev_master_upper_dev_get(dev);
2816                                 cops = br_dev->netdev_ops;
2817                         }
2818
2819                 } else {
2820                         if (dev != br_dev &&
2821                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2822                                 continue;
2823
2824                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2825                             !(dev->priv_flags & IFF_EBRIDGE))
2826                                 continue;
2827
2828                         cops = ops;
2829                 }
2830
2831                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2832                         if (cops && cops->ndo_fdb_dump)
2833                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2834                                                          idx);
2835                 }
2836
2837                 if (dev->netdev_ops->ndo_fdb_dump)
2838                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2839                                                             idx);
2840                 else
2841                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2842
2843                 cops = NULL;
2844         }
2845
2846         cb->args[0] = idx;
2847         return skb->len;
2848 }
2849
2850 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2851                                unsigned int attrnum, unsigned int flag)
2852 {
2853         if (mask & flag)
2854                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2855         return 0;
2856 }
2857
2858 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2859                             struct net_device *dev, u16 mode,
2860                             u32 flags, u32 mask, int nlflags)
2861 {
2862         struct nlmsghdr *nlh;
2863         struct ifinfomsg *ifm;
2864         struct nlattr *br_afspec;
2865         struct nlattr *protinfo;
2866         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2867         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2868
2869         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2870         if (nlh == NULL)
2871                 return -EMSGSIZE;
2872
2873         ifm = nlmsg_data(nlh);
2874         ifm->ifi_family = AF_BRIDGE;
2875         ifm->__ifi_pad = 0;
2876         ifm->ifi_type = dev->type;
2877         ifm->ifi_index = dev->ifindex;
2878         ifm->ifi_flags = dev_get_flags(dev);
2879         ifm->ifi_change = 0;
2880
2881
2882         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2883             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2884             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2885             (br_dev &&
2886              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2887             (dev->addr_len &&
2888              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2889             (dev->ifindex != dev_get_iflink(dev) &&
2890              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2891                 goto nla_put_failure;
2892
2893         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2894         if (!br_afspec)
2895                 goto nla_put_failure;
2896
2897         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2898                 nla_nest_cancel(skb, br_afspec);
2899                 goto nla_put_failure;
2900         }
2901
2902         if (mode != BRIDGE_MODE_UNDEF) {
2903                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2904                         nla_nest_cancel(skb, br_afspec);
2905                         goto nla_put_failure;
2906                 }
2907         }
2908         nla_nest_end(skb, br_afspec);
2909
2910         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2911         if (!protinfo)
2912                 goto nla_put_failure;
2913
2914         if (brport_nla_put_flag(skb, flags, mask,
2915                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2916             brport_nla_put_flag(skb, flags, mask,
2917                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2918             brport_nla_put_flag(skb, flags, mask,
2919                                 IFLA_BRPORT_FAST_LEAVE,
2920                                 BR_MULTICAST_FAST_LEAVE) ||
2921             brport_nla_put_flag(skb, flags, mask,
2922                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2923             brport_nla_put_flag(skb, flags, mask,
2924                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2925             brport_nla_put_flag(skb, flags, mask,
2926                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2927             brport_nla_put_flag(skb, flags, mask,
2928                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2929             brport_nla_put_flag(skb, flags, mask,
2930                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2931                 nla_nest_cancel(skb, protinfo);
2932                 goto nla_put_failure;
2933         }
2934
2935         nla_nest_end(skb, protinfo);
2936
2937         nlmsg_end(skb, nlh);
2938         return 0;
2939 nla_put_failure:
2940         nlmsg_cancel(skb, nlh);
2941         return -EMSGSIZE;
2942 }
2943 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2944
2945 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2946 {
2947         struct net *net = sock_net(skb->sk);
2948         struct net_device *dev;
2949         int idx = 0;
2950         u32 portid = NETLINK_CB(cb->skb).portid;
2951         u32 seq = cb->nlh->nlmsg_seq;
2952         u32 filter_mask = 0;
2953
2954         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2955                 struct nlattr *extfilt;
2956
2957                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2958                                           IFLA_EXT_MASK);
2959                 if (extfilt) {
2960                         if (nla_len(extfilt) < sizeof(filter_mask))
2961                                 return -EINVAL;
2962
2963                         filter_mask = nla_get_u32(extfilt);
2964                 }
2965         }
2966
2967         rcu_read_lock();
2968         for_each_netdev_rcu(net, dev) {
2969                 const struct net_device_ops *ops = dev->netdev_ops;
2970                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2971
2972                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2973                         if (idx >= cb->args[0] &&
2974                             br_dev->netdev_ops->ndo_bridge_getlink(
2975                                     skb, portid, seq, dev, filter_mask,
2976                                     NLM_F_MULTI) < 0)
2977                                 break;
2978                         idx++;
2979                 }
2980
2981                 if (ops->ndo_bridge_getlink) {
2982                         if (idx >= cb->args[0] &&
2983                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2984                                                     filter_mask,
2985                                                     NLM_F_MULTI) < 0)
2986                                 break;
2987                         idx++;
2988                 }
2989         }
2990         rcu_read_unlock();
2991         cb->args[0] = idx;
2992
2993         return skb->len;
2994 }
2995
2996 static inline size_t bridge_nlmsg_size(void)
2997 {
2998         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2999                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3000                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3001                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3002                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3003                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3004                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3005                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3006                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3007                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3008                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3009 }
3010
3011 static int rtnl_bridge_notify(struct net_device *dev)
3012 {
3013         struct net *net = dev_net(dev);
3014         struct sk_buff *skb;
3015         int err = -EOPNOTSUPP;
3016
3017         if (!dev->netdev_ops->ndo_bridge_getlink)
3018                 return 0;
3019
3020         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3021         if (!skb) {
3022                 err = -ENOMEM;
3023                 goto errout;
3024         }
3025
3026         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3027         if (err < 0)
3028                 goto errout;
3029
3030         if (!skb->len)
3031                 goto errout;
3032
3033         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3034         return 0;
3035 errout:
3036         WARN_ON(err == -EMSGSIZE);
3037         kfree_skb(skb);
3038         if (err)
3039                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3040         return err;
3041 }
3042
3043 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3044 {
3045         struct net *net = sock_net(skb->sk);
3046         struct ifinfomsg *ifm;
3047         struct net_device *dev;
3048         struct nlattr *br_spec, *attr = NULL;
3049         int rem, err = -EOPNOTSUPP;
3050         u16 flags = 0;
3051         bool have_flags = false;
3052
3053         if (nlmsg_len(nlh) < sizeof(*ifm))
3054                 return -EINVAL;
3055
3056         ifm = nlmsg_data(nlh);
3057         if (ifm->ifi_family != AF_BRIDGE)
3058                 return -EPFNOSUPPORT;
3059
3060         dev = __dev_get_by_index(net, ifm->ifi_index);
3061         if (!dev) {
3062                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3063                 return -ENODEV;
3064         }
3065
3066         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3067         if (br_spec) {
3068                 nla_for_each_nested(attr, br_spec, rem) {
3069                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3070                                 if (nla_len(attr) < sizeof(flags))
3071                                         return -EINVAL;
3072
3073                                 have_flags = true;
3074                                 flags = nla_get_u16(attr);
3075                                 break;
3076                         }
3077                 }
3078         }
3079
3080         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3081                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3082
3083                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3084                         err = -EOPNOTSUPP;
3085                         goto out;
3086                 }
3087
3088                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3089                 if (err)
3090                         goto out;
3091
3092                 flags &= ~BRIDGE_FLAGS_MASTER;
3093         }
3094
3095         if ((flags & BRIDGE_FLAGS_SELF)) {
3096                 if (!dev->netdev_ops->ndo_bridge_setlink)
3097                         err = -EOPNOTSUPP;
3098                 else
3099                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3100                                                                   flags);
3101                 if (!err) {
3102                         flags &= ~BRIDGE_FLAGS_SELF;
3103
3104                         /* Generate event to notify upper layer of bridge
3105                          * change
3106                          */
3107                         err = rtnl_bridge_notify(dev);
3108                 }
3109         }
3110
3111         if (have_flags)
3112                 memcpy(nla_data(attr), &flags, sizeof(flags));
3113 out:
3114         return err;
3115 }
3116
3117 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3118 {
3119         struct net *net = sock_net(skb->sk);
3120         struct ifinfomsg *ifm;
3121         struct net_device *dev;
3122         struct nlattr *br_spec, *attr = NULL;
3123         int rem, err = -EOPNOTSUPP;
3124         u16 flags = 0;
3125         bool have_flags = false;
3126
3127         if (nlmsg_len(nlh) < sizeof(*ifm))
3128                 return -EINVAL;
3129
3130         ifm = nlmsg_data(nlh);
3131         if (ifm->ifi_family != AF_BRIDGE)
3132                 return -EPFNOSUPPORT;
3133
3134         dev = __dev_get_by_index(net, ifm->ifi_index);
3135         if (!dev) {
3136                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3137                 return -ENODEV;
3138         }
3139
3140         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3141         if (br_spec) {
3142                 nla_for_each_nested(attr, br_spec, rem) {
3143                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3144                                 if (nla_len(attr) < sizeof(flags))
3145                                         return -EINVAL;
3146
3147                                 have_flags = true;
3148                                 flags = nla_get_u16(attr);
3149                                 break;
3150                         }
3151                 }
3152         }
3153
3154         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3155                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3156
3157                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3158                         err = -EOPNOTSUPP;
3159                         goto out;
3160                 }
3161
3162                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3163                 if (err)
3164                         goto out;
3165
3166                 flags &= ~BRIDGE_FLAGS_MASTER;
3167         }
3168
3169         if ((flags & BRIDGE_FLAGS_SELF)) {
3170                 if (!dev->netdev_ops->ndo_bridge_dellink)
3171                         err = -EOPNOTSUPP;
3172                 else
3173                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3174                                                                   flags);
3175
3176                 if (!err) {
3177                         flags &= ~BRIDGE_FLAGS_SELF;
3178
3179                         /* Generate event to notify upper layer of bridge
3180                          * change
3181                          */
3182                         err = rtnl_bridge_notify(dev);
3183                 }
3184         }
3185
3186         if (have_flags)
3187                 memcpy(nla_data(attr), &flags, sizeof(flags));
3188 out:
3189         return err;
3190 }
3191
3192 /* Process one rtnetlink message. */
3193
3194 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3195 {
3196         struct net *net = sock_net(skb->sk);
3197         rtnl_doit_func doit;
3198         int sz_idx, kind;
3199         int family;
3200         int type;
3201         int err;
3202
3203         type = nlh->nlmsg_type;
3204         if (type > RTM_MAX)
3205                 return -EOPNOTSUPP;
3206
3207         type -= RTM_BASE;
3208
3209         /* All the messages must have at least 1 byte length */
3210         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3211                 return 0;
3212
3213         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3214         sz_idx = type>>2;
3215         kind = type&3;
3216
3217         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3218                 return -EPERM;
3219
3220         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3221                 struct sock *rtnl;
3222                 rtnl_dumpit_func dumpit;
3223                 rtnl_calcit_func calcit;
3224                 u16 min_dump_alloc = 0;
3225
3226                 dumpit = rtnl_get_dumpit(family, type);
3227                 if (dumpit == NULL)
3228                         return -EOPNOTSUPP;
3229                 calcit = rtnl_get_calcit(family, type);
3230                 if (calcit)
3231                         min_dump_alloc = calcit(skb, nlh);
3232
3233                 __rtnl_unlock();
3234                 rtnl = net->rtnl;
3235                 {
3236                         struct netlink_dump_control c = {
3237                                 .dump           = dumpit,
3238                                 .min_dump_alloc = min_dump_alloc,
3239                         };
3240                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3241                 }
3242                 rtnl_lock();
3243                 return err;
3244         }
3245
3246         doit = rtnl_get_doit(family, type);
3247         if (doit == NULL)
3248                 return -EOPNOTSUPP;
3249
3250         return doit(skb, nlh);
3251 }
3252
3253 static void rtnetlink_rcv(struct sk_buff *skb)
3254 {
3255         rtnl_lock();
3256         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3257         rtnl_unlock();
3258 }
3259
3260 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3261 {
3262         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3263
3264         switch (event) {
3265         case NETDEV_UP:
3266         case NETDEV_DOWN:
3267         case NETDEV_PRE_UP:
3268         case NETDEV_POST_INIT:
3269         case NETDEV_REGISTER:
3270         case NETDEV_CHANGE:
3271         case NETDEV_PRE_TYPE_CHANGE:
3272         case NETDEV_GOING_DOWN:
3273         case NETDEV_UNREGISTER:
3274         case NETDEV_UNREGISTER_FINAL:
3275         case NETDEV_RELEASE:
3276         case NETDEV_JOIN:
3277         case NETDEV_BONDING_INFO:
3278                 break;
3279         default:
3280                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3281                 break;
3282         }
3283         return NOTIFY_DONE;
3284 }
3285
3286 static struct notifier_block rtnetlink_dev_notifier = {
3287         .notifier_call  = rtnetlink_event,
3288 };
3289
3290
3291 static int __net_init rtnetlink_net_init(struct net *net)
3292 {
3293         struct sock *sk;
3294         struct netlink_kernel_cfg cfg = {
3295                 .groups         = RTNLGRP_MAX,
3296                 .input          = rtnetlink_rcv,
3297                 .cb_mutex       = &rtnl_mutex,
3298                 .flags          = NL_CFG_F_NONROOT_RECV,
3299         };
3300
3301         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3302         if (!sk)
3303                 return -ENOMEM;
3304         net->rtnl = sk;
3305         return 0;
3306 }
3307
3308 static void __net_exit rtnetlink_net_exit(struct net *net)
3309 {
3310         netlink_kernel_release(net->rtnl);
3311         net->rtnl = NULL;
3312 }
3313
3314 static struct pernet_operations rtnetlink_net_ops = {
3315         .init = rtnetlink_net_init,
3316         .exit = rtnetlink_net_exit,
3317 };
3318
3319 void __init rtnetlink_init(void)
3320 {
3321         if (register_pernet_subsys(&rtnetlink_net_ops))
3322                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3323
3324         register_netdevice_notifier(&rtnetlink_dev_notifier);
3325
3326         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3327                       rtnl_dump_ifinfo, rtnl_calcit);
3328         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3329         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3330         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3331
3332         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3333         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3334
3335         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3336         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3337         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3338
3339         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3340         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3341         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3342 }
3343