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