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_slave_info_data_size(const struct net_device *dev)
 369 {
 370         struct net_device *master_dev;
 371         const struct rtnl_link_ops *ops;
 372
 373         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 374         if (!master_dev)
 375                 return 0;
 376         ops = master_dev->rtnl_link_ops;
 377         if (!ops || !ops->get_slave_size)
 378                 return 0;
 379         /* IFLA_INFO_SLAVE_DATA + nested data */
 380         return nla_total_size(sizeof(struct nlattr)) +
 381                ops->get_slave_size(master_dev, dev);
 382 }
 383
 384 static size_t rtnl_link_get_size(const struct net_device *dev)
 385 {
 386         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 387         size_t size;
 388
 389         if (!ops)
 390                 return 0;
 391
 392         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 393                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 394
 395         if (ops->get_size)
 396                 /* IFLA_INFO_DATA + nested data */
 397                 size += nla_total_size(sizeof(struct nlattr)) +
 398                         ops->get_size(dev);
 399
 400         if (ops->get_xstats_size)
 401                 /* IFLA_INFO_XSTATS */
 402                 size += nla_total_size(ops->get_xstats_size(dev));
 403
 404         size += rtnl_link_get_slave_info_data_size(dev);
 405
 406         return size;
 407 }
 408
 409 static LIST_HEAD(rtnl_af_ops);
 410
 411 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 412 {
 413         const struct rtnl_af_ops *ops;
 414
 415         list_for_each_entry(ops, &rtnl_af_ops, list) {
 416                 if (ops->family == family)
 417                         return ops;
 418         }
 419
 420         return NULL;
 421 }
 422
 423 /**
 424  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 425  * @ops: struct rtnl_af_ops * to register
 426  *
 427  * Returns 0 on success or a negative error code.
 428  */
 429 void rtnl_af_register(struct rtnl_af_ops *ops)
 430 {
 431         rtnl_lock();
 432         list_add_tail(&ops->list, &rtnl_af_ops);
 433         rtnl_unlock();
 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 bool rtnl_have_link_slave_info(const struct net_device *dev)
 481 {
 482         struct net_device *master_dev;
 483
 484         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 485         if (master_dev && master_dev->rtnl_link_ops)
 486                 return true;
 487         return false;
 488 }
 489
 490 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 491                                      const struct net_device *dev)
 492 {
 493         struct net_device *master_dev;
 494         const struct rtnl_link_ops *ops;
 495         struct nlattr *slave_data;
 496         int err;
 497
 498         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 499         if (!master_dev)
 500                 return 0;
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops)
 503                 return 0;
 504         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 505                 return -EMSGSIZE;
 506         if (ops->fill_slave_info) {
 507                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 508                 if (!slave_data)
 509                         return -EMSGSIZE;
 510                 err = ops->fill_slave_info(skb, master_dev, dev);
 511                 if (err < 0)
 512                         goto err_cancel_slave_data;
 513                 nla_nest_end(skb, slave_data);
 514         }
 515         return 0;
 516
 517 err_cancel_slave_data:
 518         nla_nest_cancel(skb, slave_data);
 519         return err;
 520 }
 521
 522 static int rtnl_link_info_fill(struct sk_buff *skb,
 523                                const struct net_device *dev)
 524 {
 525         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 526         struct nlattr *data;
 527         int err;
 528
 529         if (!ops)
 530                 return 0;
 531         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 532                 return -EMSGSIZE;
 533         if (ops->fill_xstats) {
 534                 err = ops->fill_xstats(skb, dev);
 535                 if (err < 0)
 536                         return err;
 537         }
 538         if (ops->fill_info) {
 539                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 540                 if (data == NULL)
 541                         return -EMSGSIZE;
 542                 err = ops->fill_info(skb, dev);
 543                 if (err < 0)
 544                         goto err_cancel_data;
 545                 nla_nest_end(skb, data);
 546         }
 547         return 0;
 548
 549 err_cancel_data:
 550         nla_nest_cancel(skb, data);
 551         return err;
 552 }
 553
 554 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 555 {
 556         struct nlattr *linkinfo;
 557         int err = -EMSGSIZE;
 558
 559         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 560         if (linkinfo == NULL)
 561                 goto out;
 562
 563         err = rtnl_link_info_fill(skb, dev);
 564         if (err < 0)
 565                 goto err_cancel_link;
 566
 567         err = rtnl_link_slave_info_fill(skb, dev);
 568         if (err < 0)
 569                 goto err_cancel_link;
 570
 571         nla_nest_end(skb, linkinfo);
 572         return 0;
 573
 574 err_cancel_link:
 575         nla_nest_cancel(skb, linkinfo);
 576 out:
 577         return err;
 578 }
 579
 580 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 581 {
 582         struct sock *rtnl = net->rtnl;
 583         int err = 0;
 584
 585         NETLINK_CB(skb).dst_group = group;
 586         if (echo)
 587                 atomic_inc(&skb->users);
 588         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 589         if (echo)
 590                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 591         return err;
 592 }
 593
 594 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 595 {
 596         struct sock *rtnl = net->rtnl;
 597
 598         return nlmsg_unicast(rtnl, skb, pid);
 599 }
 600 EXPORT_SYMBOL(rtnl_unicast);
 601
 602 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 603                  struct nlmsghdr *nlh, gfp_t flags)
 604 {
 605         struct sock *rtnl = net->rtnl;
 606         int report = 0;
 607
 608         if (nlh)
 609                 report = nlmsg_report(nlh);
 610
 611         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 612 }
 613 EXPORT_SYMBOL(rtnl_notify);
 614
 615 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 616 {
 617         struct sock *rtnl = net->rtnl;
 618
 619         netlink_set_err(rtnl, 0, group, error);
 620 }
 621 EXPORT_SYMBOL(rtnl_set_sk_err);
 622
 623 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 624 {
 625         struct nlattr *mx;
 626         int i, valid = 0;
 627
 628         mx = nla_nest_start(skb, RTA_METRICS);
 629         if (mx == NULL)
 630                 return -ENOBUFS;
 631
 632         for (i = 0; i < RTAX_MAX; i++) {
 633                 if (metrics[i]) {
 634                         valid++;
 635                         if (nla_put_u32(skb, i+1, metrics[i]))
 636                                 goto nla_put_failure;
 637                 }
 638         }
 639
 640         if (!valid) {
 641                 nla_nest_cancel(skb, mx);
 642                 return 0;
 643         }
 644
 645         return nla_nest_end(skb, mx);
 646
 647 nla_put_failure:
 648         nla_nest_cancel(skb, mx);
 649         return -EMSGSIZE;
 650 }
 651 EXPORT_SYMBOL(rtnetlink_put_metrics);
 652
 653 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 654                        long expires, u32 error)
 655 {
 656         struct rta_cacheinfo ci = {
 657                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 658                 .rta_used = dst->__use,
 659                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 660                 .rta_error = error,
 661                 .rta_id =  id,
 662         };
 663
 664         if (expires) {
 665                 unsigned long clock;
 666
 667                 clock = jiffies_to_clock_t(abs(expires));
 668                 clock = min_t(unsigned long, clock, INT_MAX);
 669                 ci.rta_expires = (expires > 0) ? clock : -clock;
 670         }
 671         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 672 }
 673 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 674
 675 static void set_operstate(struct net_device *dev, unsigned char transition)
 676 {
 677         unsigned char operstate = dev->operstate;
 678
 679         switch (transition) {
 680         case IF_OPER_UP:
 681                 if ((operstate == IF_OPER_DORMANT ||
 682                      operstate == IF_OPER_UNKNOWN) &&
 683                     !netif_dormant(dev))
 684                         operstate = IF_OPER_UP;
 685                 break;
 686
 687         case IF_OPER_DORMANT:
 688                 if (operstate == IF_OPER_UP ||
 689                     operstate == IF_OPER_UNKNOWN)
 690                         operstate = IF_OPER_DORMANT;
 691                 break;
 692         }
 693
 694         if (dev->operstate != operstate) {
 695                 write_lock_bh(&dev_base_lock);
 696                 dev->operstate = operstate;
 697                 write_unlock_bh(&dev_base_lock);
 698                 netdev_state_change(dev);
 699         }
 700 }
 701
 702 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 703 {
 704         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 705                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 706 }
 707
 708 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 709                                            const struct ifinfomsg *ifm)
 710 {
 711         unsigned int flags = ifm->ifi_flags;
 712
 713         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 714         if (ifm->ifi_change)
 715                 flags = (flags & ifm->ifi_change) |
 716                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 717
 718         return flags;
 719 }
 720
 721 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 722                                  const struct rtnl_link_stats64 *b)
 723 {
 724         a->rx_packets = b->rx_packets;
 725         a->tx_packets = b->tx_packets;
 726         a->rx_bytes = b->rx_bytes;
 727         a->tx_bytes = b->tx_bytes;
 728         a->rx_errors = b->rx_errors;
 729         a->tx_errors = b->tx_errors;
 730         a->rx_dropped = b->rx_dropped;
 731         a->tx_dropped = b->tx_dropped;
 732
 733         a->multicast = b->multicast;
 734         a->collisions = b->collisions;
 735
 736         a->rx_length_errors = b->rx_length_errors;
 737         a->rx_over_errors = b->rx_over_errors;
 738         a->rx_crc_errors = b->rx_crc_errors;
 739         a->rx_frame_errors = b->rx_frame_errors;
 740         a->rx_fifo_errors = b->rx_fifo_errors;
 741         a->rx_missed_errors = b->rx_missed_errors;
 742
 743         a->tx_aborted_errors = b->tx_aborted_errors;
 744         a->tx_carrier_errors = b->tx_carrier_errors;
 745         a->tx_fifo_errors = b->tx_fifo_errors;
 746         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 747         a->tx_window_errors = b->tx_window_errors;
 748
 749         a->rx_compressed = b->rx_compressed;
 750         a->tx_compressed = b->tx_compressed;
 751 }
 752
 753 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 754 {
 755         memcpy(v, b, sizeof(*b));
 756 }
 757
 758 /* All VF info */
 759 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 760                                    u32 ext_filter_mask)
 761 {
 762         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 763             (ext_filter_mask & RTEXT_FILTER_VF)) {
 764                 int num_vfs = dev_num_vf(dev->dev.parent);
 765                 size_t size = nla_total_size(sizeof(struct nlattr));
 766                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 767                 size += num_vfs *
 768                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 769                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 770                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 771                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 772                 return size;
 773         } else
 774                 return 0;
 775 }
 776
 777 static size_t rtnl_port_size(const struct net_device *dev,
 778                              u32 ext_filter_mask)
 779 {
 780         size_t port_size = nla_total_size(4)            /* PORT_VF */
 781                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 782                 + nla_total_size(sizeof(struct ifla_port_vsi))
 783                                                         /* PORT_VSI_TYPE */
 784                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 785                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 786                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 787                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 788         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 789         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 790                 + port_size;
 791         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 792                 + port_size;
 793
 794         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 795             !(ext_filter_mask & RTEXT_FILTER_VF))
 796                 return 0;
 797         if (dev_num_vf(dev->dev.parent))
 798                 return port_self_size + vf_ports_size +
 799                         vf_port_size * dev_num_vf(dev->dev.parent);
 800         else
 801                 return port_self_size;
 802 }
 803
 804 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 805                                      u32 ext_filter_mask)
 806 {
 807         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 808                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 809                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 810                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 811                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 812                + nla_total_size(sizeof(struct rtnl_link_stats))
 813                + nla_total_size(sizeof(struct rtnl_link_stats64))
 814                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 815                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 816                + nla_total_size(4) /* IFLA_TXQLEN */
 817                + nla_total_size(4) /* IFLA_WEIGHT */
 818                + nla_total_size(4) /* IFLA_MTU */
 819                + nla_total_size(4) /* IFLA_LINK */
 820                + nla_total_size(4) /* IFLA_MASTER */
 821                + nla_total_size(1) /* IFLA_CARRIER */
 822                + nla_total_size(4) /* IFLA_PROMISCUITY */
 823                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 824                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 825                + nla_total_size(1) /* IFLA_OPERSTATE */
 826                + nla_total_size(1) /* IFLA_LINKMODE */
 827                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 828                + nla_total_size(ext_filter_mask
 829                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 830                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 831                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 832                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 833                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 834                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 835 }
 836
 837 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 838 {
 839         struct nlattr *vf_ports;
 840         struct nlattr *vf_port;
 841         int vf;
 842         int err;
 843
 844         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 845         if (!vf_ports)
 846                 return -EMSGSIZE;
 847
 848         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 849                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 850                 if (!vf_port)
 851                         goto nla_put_failure;
 852                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 853                         goto nla_put_failure;
 854                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 855                 if (err == -EMSGSIZE)
 856                         goto nla_put_failure;
 857                 if (err) {
 858                         nla_nest_cancel(skb, vf_port);
 859                         continue;
 860                 }
 861                 nla_nest_end(skb, vf_port);
 862         }
 863
 864         nla_nest_end(skb, vf_ports);
 865
 866         return 0;
 867
 868 nla_put_failure:
 869         nla_nest_cancel(skb, vf_ports);
 870         return -EMSGSIZE;
 871 }
 872
 873 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 874 {
 875         struct nlattr *port_self;
 876         int err;
 877
 878         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 879         if (!port_self)
 880                 return -EMSGSIZE;
 881
 882         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 883         if (err) {
 884                 nla_nest_cancel(skb, port_self);
 885                 return (err == -EMSGSIZE) ? err : 0;
 886         }
 887
 888         nla_nest_end(skb, port_self);
 889
 890         return 0;
 891 }
 892
 893 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 894                           u32 ext_filter_mask)
 895 {
 896         int err;
 897
 898         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 899             !(ext_filter_mask & RTEXT_FILTER_VF))
 900                 return 0;
 901
 902         err = rtnl_port_self_fill(skb, dev);
 903         if (err)
 904                 return err;
 905
 906         if (dev_num_vf(dev->dev.parent)) {
 907                 err = rtnl_vf_ports_fill(skb, dev);
 908                 if (err)
 909                         return err;
 910         }
 911
 912         return 0;
 913 }
 914
 915 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 916 {
 917         int err;
 918         struct netdev_phys_port_id ppid;
 919
 920         err = dev_get_phys_port_id(dev, &ppid);
 921         if (err) {
 922                 if (err == -EOPNOTSUPP)
 923                         return 0;
 924                 return err;
 925         }
 926
 927         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 928                 return -EMSGSIZE;
 929
 930         return 0;
 931 }
 932
 933 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 934                             int type, u32 pid, u32 seq, u32 change,
 935                             unsigned int flags, u32 ext_filter_mask)
 936 {
 937         struct ifinfomsg *ifm;
 938         struct nlmsghdr *nlh;
 939         struct rtnl_link_stats64 temp;
 940         const struct rtnl_link_stats64 *stats;
 941         struct nlattr *attr, *af_spec;
 942         struct rtnl_af_ops *af_ops;
 943         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 944
 945         ASSERT_RTNL();
 946         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 947         if (nlh == NULL)
 948                 return -EMSGSIZE;
 949
 950         ifm = nlmsg_data(nlh);
 951         ifm->ifi_family = AF_UNSPEC;
 952         ifm->__ifi_pad = 0;
 953         ifm->ifi_type = dev->type;
 954         ifm->ifi_index = dev->ifindex;
 955         ifm->ifi_flags = dev_get_flags(dev);
 956         ifm->ifi_change = change;
 957
 958         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 959             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 960             nla_put_u8(skb, IFLA_OPERSTATE,
 961                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 962             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 963             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 964             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 965             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 966             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 967 #ifdef CONFIG_RPS
 968             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 969 #endif
 970             (dev->ifindex != dev->iflink &&
 971              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 972             (upper_dev &&
 973              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
 974             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
 975             (dev->qdisc &&
 976              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 977             (dev->ifalias &&
 978              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
 979             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
 980                         atomic_read(&dev->carrier_changes)))
 981                 goto nla_put_failure;
 982
 983         if (1) {
 984                 struct rtnl_link_ifmap map = {
 985                         .mem_start   = dev->mem_start,
 986                         .mem_end     = dev->mem_end,
 987                         .base_addr   = dev->base_addr,
 988                         .irq         = dev->irq,
 989                         .dma         = dev->dma,
 990                         .port        = dev->if_port,
 991                 };
 992                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 993                         goto nla_put_failure;
 994         }
 995
 996         if (dev->addr_len) {
 997                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 998                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 999                         goto nla_put_failure;
1000         }
1001
1002         if (rtnl_phys_port_id_fill(skb, dev))
1003                 goto nla_put_failure;
1004
1005         attr = nla_reserve(skb, IFLA_STATS,
1006                         sizeof(struct rtnl_link_stats));
1007         if (attr == NULL)
1008                 goto nla_put_failure;
1009
1010         stats = dev_get_stats(dev, &temp);
1011         copy_rtnl_link_stats(nla_data(attr), stats);
1012
1013         attr = nla_reserve(skb, IFLA_STATS64,
1014                         sizeof(struct rtnl_link_stats64));
1015         if (attr == NULL)
1016                 goto nla_put_failure;
1017         copy_rtnl_link_stats64(nla_data(attr), stats);
1018
1019         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1020             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1021                 goto nla_put_failure;
1022
1023         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1024             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1025                 int i;
1026
1027                 struct nlattr *vfinfo, *vf;
1028                 int num_vfs = dev_num_vf(dev->dev.parent);
1029
1030                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1031                 if (!vfinfo)
1032                         goto nla_put_failure;
1033                 for (i = 0; i < num_vfs; i++) {
1034                         struct ifla_vf_info ivi;
1035                         struct ifla_vf_mac vf_mac;
1036                         struct ifla_vf_vlan vf_vlan;
1037                         struct ifla_vf_tx_rate vf_tx_rate;
1038                         struct ifla_vf_spoofchk vf_spoofchk;
1039                         struct ifla_vf_link_state vf_linkstate;
1040
1041                         /*
1042                          * Not all SR-IOV capable drivers support the
1043                          * spoofcheck query.  Preset to -1 so the user
1044                          * space tool can detect that the driver didn't
1045                          * report anything.
1046                          */
1047                         ivi.spoofchk = -1;
1048                         memset(ivi.mac, 0, sizeof(ivi.mac));
1049                         /* The default value for VF link state is "auto"
1050                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1051                          */
1052                         ivi.linkstate = 0;
1053                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1054                                 break;
1055                         vf_mac.vf =
1056                                 vf_vlan.vf =
1057                                 vf_tx_rate.vf =
1058                                 vf_spoofchk.vf =
1059                                 vf_linkstate.vf = ivi.vf;
1060
1061                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1062                         vf_vlan.vlan = ivi.vlan;
1063                         vf_vlan.qos = ivi.qos;
1064                         vf_tx_rate.rate = ivi.tx_rate;
1065                         vf_spoofchk.setting = ivi.spoofchk;
1066                         vf_linkstate.link_state = ivi.linkstate;
1067                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1068                         if (!vf) {
1069                                 nla_nest_cancel(skb, vfinfo);
1070                                 goto nla_put_failure;
1071                         }
1072                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1073                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1074                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1075                                     &vf_tx_rate) ||
1076                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1077                                     &vf_spoofchk) ||
1078                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1079                                     &vf_linkstate))
1080                                 goto nla_put_failure;
1081                         nla_nest_end(skb, vf);
1082                 }
1083                 nla_nest_end(skb, vfinfo);
1084         }
1085
1086         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1087                 goto nla_put_failure;
1088
1089         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1090                 if (rtnl_link_fill(skb, dev) < 0)
1091                         goto nla_put_failure;
1092         }
1093
1094         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1095                 goto nla_put_failure;
1096
1097         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1098                 if (af_ops->fill_link_af) {
1099                         struct nlattr *af;
1100                         int err;
1101
1102                         if (!(af = nla_nest_start(skb, af_ops->family)))
1103                                 goto nla_put_failure;
1104
1105                         err = af_ops->fill_link_af(skb, dev);
1106
1107                         /*
1108                          * Caller may return ENODATA to indicate that there
1109                          * was no data to be dumped. This is not an error, it
1110                          * means we should trim the attribute header and
1111                          * continue.
1112                          */
1113                         if (err == -ENODATA)
1114                                 nla_nest_cancel(skb, af);
1115                         else if (err < 0)
1116                                 goto nla_put_failure;
1117
1118                         nla_nest_end(skb, af);
1119                 }
1120         }
1121
1122         nla_nest_end(skb, af_spec);
1123
1124         return nlmsg_end(skb, nlh);
1125
1126 nla_put_failure:
1127         nlmsg_cancel(skb, nlh);
1128         return -EMSGSIZE;
1129 }
1130
1131 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1132         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1133         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1134         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1135         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1136         [IFLA_MTU]              = { .type = NLA_U32 },
1137         [IFLA_LINK]             = { .type = NLA_U32 },
1138         [IFLA_MASTER]           = { .type = NLA_U32 },
1139         [IFLA_CARRIER]          = { .type = NLA_U8 },
1140         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1141         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1142         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1143         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1144         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1145         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1146         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1147         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1148         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1149         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1150         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1151         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1152         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1153         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1154         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1155         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1156         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1157         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1158 };
1159
1160 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1161         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1162         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1163         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1164         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1165 };
1166
1167 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1168         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1169 };
1170
1171 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1172         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1173                                     .len = sizeof(struct ifla_vf_mac) },
1174         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1175                                     .len = sizeof(struct ifla_vf_vlan) },
1176         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1177                                     .len = sizeof(struct ifla_vf_tx_rate) },
1178         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1179                                     .len = sizeof(struct ifla_vf_spoofchk) },
1180 };
1181
1182 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1183         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1184         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1185                                     .len = PORT_PROFILE_MAX },
1186         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1187                                     .len = sizeof(struct ifla_port_vsi)},
1188         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1189                                       .len = PORT_UUID_MAX },
1190         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1191                                     .len = PORT_UUID_MAX },
1192         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1193         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1194 };
1195
1196 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1197 {
1198         struct net *net = sock_net(skb->sk);
1199         int h, s_h;
1200         int idx = 0, s_idx;
1201         struct net_device *dev;
1202         struct hlist_head *head;
1203         struct nlattr *tb[IFLA_MAX+1];
1204         u32 ext_filter_mask = 0;
1205         int err;
1206
1207         s_h = cb->args[0];
1208         s_idx = cb->args[1];
1209
1210         rcu_read_lock();
1211         cb->seq = net->dev_base_seq;
1212
1213         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1214                         ifla_policy) >= 0) {
1215
1216                 if (tb[IFLA_EXT_MASK])
1217                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1218         }
1219
1220         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1221                 idx = 0;
1222                 head = &net->dev_index_head[h];
1223                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1224                         if (idx < s_idx)
1225                                 goto cont;
1226                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1227                                                NETLINK_CB(cb->skb).portid,
1228                                                cb->nlh->nlmsg_seq, 0,
1229                                                NLM_F_MULTI,
1230                                                ext_filter_mask);
1231                         /* If we ran out of room on the first message,
1232                          * we're in trouble
1233                          */
1234                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1235
1236                         if (err <= 0)
1237                                 goto out;
1238
1239                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1240 cont:
1241                         idx++;
1242                 }
1243         }
1244 out:
1245         rcu_read_unlock();
1246         cb->args[1] = idx;
1247         cb->args[0] = h;
1248
1249         return skb->len;
1250 }
1251
1252 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1253 {
1254         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1255 }
1256 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1257
1258 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1259 {
1260         struct net *net;
1261         /* Examine the link attributes and figure out which
1262          * network namespace we are talking about.
1263          */
1264         if (tb[IFLA_NET_NS_PID])
1265                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1266         else if (tb[IFLA_NET_NS_FD])
1267                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1268         else
1269                 net = get_net(src_net);
1270         return net;
1271 }
1272 EXPORT_SYMBOL(rtnl_link_get_net);
1273
1274 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1275 {
1276         if (dev) {
1277                 if (tb[IFLA_ADDRESS] &&
1278                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1279                         return -EINVAL;
1280
1281                 if (tb[IFLA_BROADCAST] &&
1282                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1283                         return -EINVAL;
1284         }
1285
1286         if (tb[IFLA_AF_SPEC]) {
1287                 struct nlattr *af;
1288                 int rem, err;
1289
1290                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1291                         const struct rtnl_af_ops *af_ops;
1292
1293                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1294                                 return -EAFNOSUPPORT;
1295
1296                         if (!af_ops->set_link_af)
1297                                 return -EOPNOTSUPP;
1298
1299                         if (af_ops->validate_link_af) {
1300                                 err = af_ops->validate_link_af(dev, af);
1301                                 if (err < 0)
1302                                         return err;
1303                         }
1304                 }
1305         }
1306
1307         return 0;
1308 }
1309
1310 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1311 {
1312         int rem, err = -EINVAL;
1313         struct nlattr *vf;
1314         const struct net_device_ops *ops = dev->netdev_ops;
1315
1316         nla_for_each_nested(vf, attr, rem) {
1317                 switch (nla_type(vf)) {
1318                 case IFLA_VF_MAC: {
1319                         struct ifla_vf_mac *ivm;
1320                         ivm = nla_data(vf);
1321                         err = -EOPNOTSUPP;
1322                         if (ops->ndo_set_vf_mac)
1323                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1324                                                           ivm->mac);
1325                         break;
1326                 }
1327                 case IFLA_VF_VLAN: {
1328                         struct ifla_vf_vlan *ivv;
1329                         ivv = nla_data(vf);
1330                         err = -EOPNOTSUPP;
1331                         if (ops->ndo_set_vf_vlan)
1332                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1333                                                            ivv->vlan,
1334                                                            ivv->qos);
1335                         break;
1336                 }
1337                 case IFLA_VF_TX_RATE: {
1338                         struct ifla_vf_tx_rate *ivt;
1339                         ivt = nla_data(vf);
1340                         err = -EOPNOTSUPP;
1341                         if (ops->ndo_set_vf_tx_rate)
1342                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1343                                                               ivt->rate);
1344                         break;
1345                 }
1346                 case IFLA_VF_SPOOFCHK: {
1347                         struct ifla_vf_spoofchk *ivs;
1348                         ivs = nla_data(vf);
1349                         err = -EOPNOTSUPP;
1350                         if (ops->ndo_set_vf_spoofchk)
1351                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1352                                                                ivs->setting);
1353                         break;
1354                 }
1355                 case IFLA_VF_LINK_STATE: {
1356                         struct ifla_vf_link_state *ivl;
1357                         ivl = nla_data(vf);
1358                         err = -EOPNOTSUPP;
1359                         if (ops->ndo_set_vf_link_state)
1360                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1361                                                                  ivl->link_state);
1362                         break;
1363                 }
1364                 default:
1365                         err = -EINVAL;
1366                         break;
1367                 }
1368                 if (err)
1369                         break;
1370         }
1371         return err;
1372 }
1373
1374 static int do_set_master(struct net_device *dev, int ifindex)
1375 {
1376         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1377         const struct net_device_ops *ops;
1378         int err;
1379
1380         if (upper_dev) {
1381                 if (upper_dev->ifindex == ifindex)
1382                         return 0;
1383                 ops = upper_dev->netdev_ops;
1384                 if (ops->ndo_del_slave) {
1385                         err = ops->ndo_del_slave(upper_dev, dev);
1386                         if (err)
1387                                 return err;
1388                 } else {
1389                         return -EOPNOTSUPP;
1390                 }
1391         }
1392
1393         if (ifindex) {
1394                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1395                 if (!upper_dev)
1396                         return -EINVAL;
1397                 ops = upper_dev->netdev_ops;
1398                 if (ops->ndo_add_slave) {
1399                         err = ops->ndo_add_slave(upper_dev, dev);
1400                         if (err)
1401                                 return err;
1402                 } else {
1403                         return -EOPNOTSUPP;
1404                 }
1405         }
1406         return 0;
1407 }
1408
1409 static int do_setlink(const struct sk_buff *skb,
1410                       struct net_device *dev, struct ifinfomsg *ifm,
1411                       struct nlattr **tb, char *ifname, int modified)
1412 {
1413         const struct net_device_ops *ops = dev->netdev_ops;
1414         int err;
1415
1416         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1417                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1418                 if (IS_ERR(net)) {
1419                         err = PTR_ERR(net);
1420                         goto errout;
1421                 }
1422                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1423                         err = -EPERM;
1424                         goto errout;
1425                 }
1426                 err = dev_change_net_namespace(dev, net, ifname);
1427                 put_net(net);
1428                 if (err)
1429                         goto errout;
1430                 modified = 1;
1431         }
1432
1433         if (tb[IFLA_MAP]) {
1434                 struct rtnl_link_ifmap *u_map;
1435                 struct ifmap k_map;
1436
1437                 if (!ops->ndo_set_config) {
1438                         err = -EOPNOTSUPP;
1439                         goto errout;
1440                 }
1441
1442                 if (!netif_device_present(dev)) {
1443                         err = -ENODEV;
1444                         goto errout;
1445                 }
1446
1447                 u_map = nla_data(tb[IFLA_MAP]);
1448                 k_map.mem_start = (unsigned long) u_map->mem_start;
1449                 k_map.mem_end = (unsigned long) u_map->mem_end;
1450                 k_map.base_addr = (unsigned short) u_map->base_addr;
1451                 k_map.irq = (unsigned char) u_map->irq;
1452                 k_map.dma = (unsigned char) u_map->dma;
1453                 k_map.port = (unsigned char) u_map->port;
1454
1455                 err = ops->ndo_set_config(dev, &k_map);
1456                 if (err < 0)
1457                         goto errout;
1458
1459                 modified = 1;
1460         }
1461
1462         if (tb[IFLA_ADDRESS]) {
1463                 struct sockaddr *sa;
1464                 int len;
1465
1466                 len = sizeof(sa_family_t) + dev->addr_len;
1467                 sa = kmalloc(len, GFP_KERNEL);
1468                 if (!sa) {
1469                         err = -ENOMEM;
1470                         goto errout;
1471                 }
1472                 sa->sa_family = dev->type;
1473                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1474                        dev->addr_len);
1475                 err = dev_set_mac_address(dev, sa);
1476                 kfree(sa);
1477                 if (err)
1478                         goto errout;
1479                 modified = 1;
1480         }
1481
1482         if (tb[IFLA_MTU]) {
1483                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1484                 if (err < 0)
1485                         goto errout;
1486                 modified = 1;
1487         }
1488
1489         if (tb[IFLA_GROUP]) {
1490                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1491                 modified = 1;
1492         }
1493
1494         /*
1495          * Interface selected by interface index but interface
1496          * name provided implies that a name change has been
1497          * requested.
1498          */
1499         if (ifm->ifi_index > 0 && ifname[0]) {
1500                 err = dev_change_name(dev, ifname);
1501                 if (err < 0)
1502                         goto errout;
1503                 modified = 1;
1504         }
1505
1506         if (tb[IFLA_IFALIAS]) {
1507                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1508                                     nla_len(tb[IFLA_IFALIAS]));
1509                 if (err < 0)
1510                         goto errout;
1511                 modified = 1;
1512         }
1513
1514         if (tb[IFLA_BROADCAST]) {
1515                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1516                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1517         }
1518
1519         if (ifm->ifi_flags || ifm->ifi_change) {
1520                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1521                 if (err < 0)
1522                         goto errout;
1523         }
1524
1525         if (tb[IFLA_MASTER]) {
1526                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1527                 if (err)
1528                         goto errout;
1529                 modified = 1;
1530         }
1531
1532         if (tb[IFLA_CARRIER]) {
1533                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1534                 if (err)
1535                         goto errout;
1536                 modified = 1;
1537         }
1538
1539         if (tb[IFLA_TXQLEN])
1540                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1541
1542         if (tb[IFLA_OPERSTATE])
1543                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1544
1545         if (tb[IFLA_LINKMODE]) {
1546                 write_lock_bh(&dev_base_lock);
1547                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1548                 write_unlock_bh(&dev_base_lock);
1549         }
1550
1551         if (tb[IFLA_VFINFO_LIST]) {
1552                 struct nlattr *attr;
1553                 int rem;
1554                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1555                         if (nla_type(attr) != IFLA_VF_INFO) {
1556                                 err = -EINVAL;
1557                                 goto errout;
1558                         }
1559                         err = do_setvfinfo(dev, attr);
1560                         if (err < 0)
1561                                 goto errout;
1562                         modified = 1;
1563                 }
1564         }
1565         err = 0;
1566
1567         if (tb[IFLA_VF_PORTS]) {
1568                 struct nlattr *port[IFLA_PORT_MAX+1];
1569                 struct nlattr *attr;
1570                 int vf;
1571                 int rem;
1572
1573                 err = -EOPNOTSUPP;
1574                 if (!ops->ndo_set_vf_port)
1575                         goto errout;
1576
1577                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1578                         if (nla_type(attr) != IFLA_VF_PORT)
1579                                 continue;
1580                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1581                                 attr, ifla_port_policy);
1582                         if (err < 0)
1583                                 goto errout;
1584                         if (!port[IFLA_PORT_VF]) {
1585                                 err = -EOPNOTSUPP;
1586                                 goto errout;
1587                         }
1588                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1589                         err = ops->ndo_set_vf_port(dev, vf, port);
1590                         if (err < 0)
1591                                 goto errout;
1592                         modified = 1;
1593                 }
1594         }
1595         err = 0;
1596
1597         if (tb[IFLA_PORT_SELF]) {
1598                 struct nlattr *port[IFLA_PORT_MAX+1];
1599
1600                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1601                         tb[IFLA_PORT_SELF], ifla_port_policy);
1602                 if (err < 0)
1603                         goto errout;
1604
1605                 err = -EOPNOTSUPP;
1606                 if (ops->ndo_set_vf_port)
1607                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1608                 if (err < 0)
1609                         goto errout;
1610                 modified = 1;
1611         }
1612
1613         if (tb[IFLA_AF_SPEC]) {
1614                 struct nlattr *af;
1615                 int rem;
1616
1617                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1618                         const struct rtnl_af_ops *af_ops;
1619
1620                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1621                                 BUG();
1622
1623                         err = af_ops->set_link_af(dev, af);
1624                         if (err < 0)
1625                                 goto errout;
1626
1627                         modified = 1;
1628                 }
1629         }
1630         err = 0;
1631
1632 errout:
1633         if (err < 0 && modified)
1634                 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",
1635                                      dev->name);
1636
1637         return err;
1638 }
1639
1640 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1641 {
1642         struct net *net = sock_net(skb->sk);
1643         struct ifinfomsg *ifm;
1644         struct net_device *dev;
1645         int err;
1646         struct nlattr *tb[IFLA_MAX+1];
1647         char ifname[IFNAMSIZ];
1648
1649         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1650         if (err < 0)
1651                 goto errout;
1652
1653         if (tb[IFLA_IFNAME])
1654                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1655         else
1656                 ifname[0] = '\0';
1657
1658         err = -EINVAL;
1659         ifm = nlmsg_data(nlh);
1660         if (ifm->ifi_index > 0)
1661                 dev = __dev_get_by_index(net, ifm->ifi_index);
1662         else if (tb[IFLA_IFNAME])
1663                 dev = __dev_get_by_name(net, ifname);
1664         else
1665                 goto errout;
1666
1667         if (dev == NULL) {
1668                 err = -ENODEV;
1669                 goto errout;
1670         }
1671
1672         err = validate_linkmsg(dev, tb);
1673         if (err < 0)
1674                 goto errout;
1675
1676         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1677 errout:
1678         return err;
1679 }
1680
1681 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1682 {
1683         struct net *net = sock_net(skb->sk);
1684         const struct rtnl_link_ops *ops;
1685         struct net_device *dev;
1686         struct ifinfomsg *ifm;
1687         char ifname[IFNAMSIZ];
1688         struct nlattr *tb[IFLA_MAX+1];
1689         int err;
1690         LIST_HEAD(list_kill);
1691
1692         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1693         if (err < 0)
1694                 return err;
1695
1696         if (tb[IFLA_IFNAME])
1697                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1698
1699         ifm = nlmsg_data(nlh);
1700         if (ifm->ifi_index > 0)
1701                 dev = __dev_get_by_index(net, ifm->ifi_index);
1702         else if (tb[IFLA_IFNAME])
1703                 dev = __dev_get_by_name(net, ifname);
1704         else
1705                 return -EINVAL;
1706
1707         if (!dev)
1708                 return -ENODEV;
1709
1710         ops = dev->rtnl_link_ops;
1711         if (!ops)
1712                 return -EOPNOTSUPP;
1713
1714         ops->dellink(dev, &list_kill);
1715         unregister_netdevice_many(&list_kill);
1716         list_del(&list_kill);
1717         return 0;
1718 }
1719
1720 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1721 {
1722         unsigned int old_flags;
1723         int err;
1724
1725         old_flags = dev->flags;
1726         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1727                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1728                 if (err < 0)
1729                         return err;
1730         }
1731
1732         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1733
1734         __dev_notify_flags(dev, old_flags, ~0U);
1735         return 0;
1736 }
1737 EXPORT_SYMBOL(rtnl_configure_link);
1738
1739 struct net_device *rtnl_create_link(struct net *net,
1740         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1741 {
1742         int err;
1743         struct net_device *dev;
1744         unsigned int num_tx_queues = 1;
1745         unsigned int num_rx_queues = 1;
1746
1747         if (tb[IFLA_NUM_TX_QUEUES])
1748                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1749         else if (ops->get_num_tx_queues)
1750                 num_tx_queues = ops->get_num_tx_queues();
1751
1752         if (tb[IFLA_NUM_RX_QUEUES])
1753                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1754         else if (ops->get_num_rx_queues)
1755                 num_rx_queues = ops->get_num_rx_queues();
1756
1757         err = -ENOMEM;
1758         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1759                                num_tx_queues, num_rx_queues);
1760         if (!dev)
1761                 goto err;
1762
1763         dev_net_set(dev, net);
1764         dev->rtnl_link_ops = ops;
1765         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1766
1767         if (tb[IFLA_MTU])
1768                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1769         if (tb[IFLA_ADDRESS]) {
1770                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1771                                 nla_len(tb[IFLA_ADDRESS]));
1772                 dev->addr_assign_type = NET_ADDR_SET;
1773         }
1774         if (tb[IFLA_BROADCAST])
1775                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1776                                 nla_len(tb[IFLA_BROADCAST]));
1777         if (tb[IFLA_TXQLEN])
1778                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1779         if (tb[IFLA_OPERSTATE])
1780                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1781         if (tb[IFLA_LINKMODE])
1782                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1783         if (tb[IFLA_GROUP])
1784                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1785
1786         return dev;
1787
1788 err:
1789         return ERR_PTR(err);
1790 }
1791 EXPORT_SYMBOL(rtnl_create_link);
1792
1793 static int rtnl_group_changelink(const struct sk_buff *skb,
1794                 struct net *net, int group,
1795                 struct ifinfomsg *ifm,
1796                 struct nlattr **tb)
1797 {
1798         struct net_device *dev;
1799         int err;
1800
1801         for_each_netdev(net, dev) {
1802                 if (dev->group == group) {
1803                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1804                         if (err < 0)
1805                                 return err;
1806                 }
1807         }
1808
1809         return 0;
1810 }
1811
1812 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1813 {
1814         struct net *net = sock_net(skb->sk);
1815         const struct rtnl_link_ops *ops;
1816         const struct rtnl_link_ops *m_ops = NULL;
1817         struct net_device *dev;
1818         struct net_device *master_dev = NULL;
1819         struct ifinfomsg *ifm;
1820         char kind[MODULE_NAME_LEN];
1821         char ifname[IFNAMSIZ];
1822         struct nlattr *tb[IFLA_MAX+1];
1823         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1824         int err;
1825
1826 #ifdef CONFIG_MODULES
1827 replay:
1828 #endif
1829         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1830         if (err < 0)
1831                 return err;
1832
1833         if (tb[IFLA_IFNAME])
1834                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1835         else
1836                 ifname[0] = '\0';
1837
1838         ifm = nlmsg_data(nlh);
1839         if (ifm->ifi_index > 0)
1840                 dev = __dev_get_by_index(net, ifm->ifi_index);
1841         else {
1842                 if (ifname[0])
1843                         dev = __dev_get_by_name(net, ifname);
1844                 else
1845                         dev = NULL;
1846         }
1847
1848         if (dev) {
1849                 master_dev = netdev_master_upper_dev_get(dev);
1850                 if (master_dev)
1851                         m_ops = master_dev->rtnl_link_ops;
1852         }
1853
1854         err = validate_linkmsg(dev, tb);
1855         if (err < 0)
1856                 return err;
1857
1858         if (tb[IFLA_LINKINFO]) {
1859                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1860                                        tb[IFLA_LINKINFO], ifla_info_policy);
1861                 if (err < 0)
1862                         return err;
1863         } else
1864                 memset(linkinfo, 0, sizeof(linkinfo));
1865
1866         if (linkinfo[IFLA_INFO_KIND]) {
1867                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1868                 ops = rtnl_link_ops_get(kind);
1869         } else {
1870                 kind[0] = '\0';
1871                 ops = NULL;
1872         }
1873
1874         if (1) {
1875                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1876                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1877                 struct nlattr **data = NULL;
1878                 struct nlattr **slave_data = NULL;
1879                 struct net *dest_net;
1880
1881                 if (ops) {
1882                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1883                                 err = nla_parse_nested(attr, ops->maxtype,
1884                                                        linkinfo[IFLA_INFO_DATA],
1885                                                        ops->policy);
1886                                 if (err < 0)
1887                                         return err;
1888                                 data = attr;
1889                         }
1890                         if (ops->validate) {
1891                                 err = ops->validate(tb, data);
1892                                 if (err < 0)
1893                                         return err;
1894                         }
1895                 }
1896
1897                 if (m_ops) {
1898                         if (m_ops->slave_maxtype &&
1899                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
1900                                 err = nla_parse_nested(slave_attr,
1901                                                        m_ops->slave_maxtype,
1902                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
1903                                                        m_ops->slave_policy);
1904                                 if (err < 0)
1905                                         return err;
1906                                 slave_data = slave_attr;
1907                         }
1908                         if (m_ops->slave_validate) {
1909                                 err = m_ops->slave_validate(tb, slave_data);
1910                                 if (err < 0)
1911                                         return err;
1912                         }
1913                 }
1914
1915                 if (dev) {
1916                         int modified = 0;
1917
1918                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1919                                 return -EEXIST;
1920                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1921                                 return -EOPNOTSUPP;
1922
1923                         if (linkinfo[IFLA_INFO_DATA]) {
1924                                 if (!ops || ops != dev->rtnl_link_ops ||
1925                                     !ops->changelink)
1926                                         return -EOPNOTSUPP;
1927
1928                                 err = ops->changelink(dev, tb, data);
1929                                 if (err < 0)
1930                                         return err;
1931                                 modified = 1;
1932                         }
1933
1934                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1935                                 if (!m_ops || !m_ops->slave_changelink)
1936                                         return -EOPNOTSUPP;
1937
1938                                 err = m_ops->slave_changelink(master_dev, dev,
1939                                                               tb, slave_data);
1940                                 if (err < 0)
1941                                         return err;
1942                                 modified = 1;
1943                         }
1944
1945                         return do_setlink(skb, dev, ifm, tb, ifname, modified);
1946                 }
1947
1948                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1949                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1950                                 return rtnl_group_changelink(skb, net,
1951                                                 nla_get_u32(tb[IFLA_GROUP]),
1952                                                 ifm, tb);
1953                         return -ENODEV;
1954                 }
1955
1956                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1957                         return -EOPNOTSUPP;
1958
1959                 if (!ops) {
1960 #ifdef CONFIG_MODULES
1961                         if (kind[0]) {
1962                                 __rtnl_unlock();
1963                                 request_module("rtnl-link-%s", kind);
1964                                 rtnl_lock();
1965                                 ops = rtnl_link_ops_get(kind);
1966                                 if (ops)
1967                                         goto replay;
1968                         }
1969 #endif
1970                         return -EOPNOTSUPP;
1971                 }
1972
1973                 if (!ifname[0])
1974                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1975
1976                 dest_net = rtnl_link_get_net(net, tb);
1977                 if (IS_ERR(dest_net))
1978                         return PTR_ERR(dest_net);
1979
1980                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1981                 if (IS_ERR(dev)) {
1982                         err = PTR_ERR(dev);
1983                         goto out;
1984                 }
1985
1986                 dev->ifindex = ifm->ifi_index;
1987
1988                 if (ops->newlink) {
1989                         err = ops->newlink(net, dev, tb, data);
1990                         /* Drivers should call free_netdev() in ->destructor
1991                          * and unregister it on failure so that device could be
1992                          * finally freed in rtnl_unlock.
1993                          */
1994                         if (err < 0)
1995                                 goto out;
1996                 } else {
1997                         err = register_netdevice(dev);
1998                         if (err < 0) {
1999                                 free_netdev(dev);
2000                                 goto out;
2001                         }
2002                 }
2003                 err = rtnl_configure_link(dev, ifm);
2004                 if (err < 0)
2005                         unregister_netdevice(dev);
2006 out:
2007                 put_net(dest_net);
2008                 return err;
2009         }
2010 }
2011
2012 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2013 {
2014         struct net *net = sock_net(skb->sk);
2015         struct ifinfomsg *ifm;
2016         char ifname[IFNAMSIZ];
2017         struct nlattr *tb[IFLA_MAX+1];
2018         struct net_device *dev = NULL;
2019         struct sk_buff *nskb;
2020         int err;
2021         u32 ext_filter_mask = 0;
2022
2023         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2024         if (err < 0)
2025                 return err;
2026
2027         if (tb[IFLA_IFNAME])
2028                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2029
2030         if (tb[IFLA_EXT_MASK])
2031                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2032
2033         ifm = nlmsg_data(nlh);
2034         if (ifm->ifi_index > 0)
2035                 dev = __dev_get_by_index(net, ifm->ifi_index);
2036         else if (tb[IFLA_IFNAME])
2037                 dev = __dev_get_by_name(net, ifname);
2038         else
2039                 return -EINVAL;
2040
2041         if (dev == NULL)
2042                 return -ENODEV;
2043
2044         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2045         if (nskb == NULL)
2046                 return -ENOBUFS;
2047
2048         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2049                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2050         if (err < 0) {
2051                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2052                 WARN_ON(err == -EMSGSIZE);
2053                 kfree_skb(nskb);
2054         } else
2055                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2056
2057         return err;
2058 }
2059
2060 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2061 {
2062         struct net *net = sock_net(skb->sk);
2063         struct net_device *dev;
2064         struct nlattr *tb[IFLA_MAX+1];
2065         u32 ext_filter_mask = 0;
2066         u16 min_ifinfo_dump_size = 0;
2067
2068         if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2069                         ifla_policy) >= 0) {
2070                 if (tb[IFLA_EXT_MASK])
2071                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2072         }
2073
2074         if (!ext_filter_mask)
2075                 return NLMSG_GOODSIZE;
2076         /*
2077          * traverse the list of net devices and compute the minimum
2078          * buffer size based upon the filter mask.
2079          */
2080         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2081                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2082                                              if_nlmsg_size(dev,
2083                                                            ext_filter_mask));
2084         }
2085
2086         return min_ifinfo_dump_size;
2087 }
2088
2089 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2090 {
2091         int idx;
2092         int s_idx = cb->family;
2093
2094         if (s_idx == 0)
2095                 s_idx = 1;
2096         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2097                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2098                 if (idx < s_idx || idx == PF_PACKET)
2099                         continue;
2100                 if (rtnl_msg_handlers[idx] == NULL ||
2101                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2102                         continue;
2103                 if (idx > s_idx) {
2104                         memset(&cb->args[0], 0, sizeof(cb->args));
2105                         cb->prev_seq = 0;
2106                         cb->seq = 0;
2107                 }
2108                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2109                         break;
2110         }
2111         cb->family = idx;
2112
2113         return skb->len;
2114 }
2115
2116 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2117                   gfp_t flags)
2118 {
2119         struct net *net = dev_net(dev);
2120         struct sk_buff *skb;
2121         int err = -ENOBUFS;
2122         size_t if_info_size;
2123
2124         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2125         if (skb == NULL)
2126                 goto errout;
2127
2128         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2129         if (err < 0) {
2130                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2131                 WARN_ON(err == -EMSGSIZE);
2132                 kfree_skb(skb);
2133                 goto errout;
2134         }
2135         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2136         return;
2137 errout:
2138         if (err < 0)
2139                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2140 }
2141 EXPORT_SYMBOL(rtmsg_ifinfo);
2142
2143 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2144                                    struct net_device *dev,
2145                                    u8 *addr, u32 pid, u32 seq,
2146                                    int type, unsigned int flags,
2147                                    int nlflags)
2148 {
2149         struct nlmsghdr *nlh;
2150         struct ndmsg *ndm;
2151
2152         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2153         if (!nlh)
2154                 return -EMSGSIZE;
2155
2156         ndm = nlmsg_data(nlh);
2157         ndm->ndm_family  = AF_BRIDGE;
2158         ndm->ndm_pad1    = 0;
2159         ndm->ndm_pad2    = 0;
2160         ndm->ndm_flags   = flags;
2161         ndm->ndm_type    = 0;
2162         ndm->ndm_ifindex = dev->ifindex;
2163         ndm->ndm_state   = NUD_PERMANENT;
2164
2165         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2166                 goto nla_put_failure;
2167
2168         return nlmsg_end(skb, nlh);
2169
2170 nla_put_failure:
2171         nlmsg_cancel(skb, nlh);
2172         return -EMSGSIZE;
2173 }
2174
2175 static inline size_t rtnl_fdb_nlmsg_size(void)
2176 {
2177         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2178 }
2179
2180 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2181 {
2182         struct net *net = dev_net(dev);
2183         struct sk_buff *skb;
2184         int err = -ENOBUFS;
2185
2186         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2187         if (!skb)
2188                 goto errout;
2189
2190         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2191         if (err < 0) {
2192                 kfree_skb(skb);
2193                 goto errout;
2194         }
2195
2196         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2197         return;
2198 errout:
2199         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2200 }
2201
2202 /**
2203  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2204  */
2205 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2206                      struct nlattr *tb[],
2207                      struct net_device *dev,
2208                      const unsigned char *addr,
2209                      u16 flags)
2210 {
2211         int err = -EINVAL;
2212
2213         /* If aging addresses are supported device will need to
2214          * implement its own handler for this.
2215          */
2216         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2217                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2218                 return err;
2219         }
2220
2221         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2222                 err = dev_uc_add_excl(dev, addr);
2223         else if (is_multicast_ether_addr(addr))
2224                 err = dev_mc_add_excl(dev, addr);
2225
2226         /* Only return duplicate errors if NLM_F_EXCL is set */
2227         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2228                 err = 0;
2229
2230         return err;
2231 }
2232 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2233
2234 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2235 {
2236         struct net *net = sock_net(skb->sk);
2237         struct ndmsg *ndm;
2238         struct nlattr *tb[NDA_MAX+1];
2239         struct net_device *dev;
2240         u8 *addr;
2241         int err;
2242
2243         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2244         if (err < 0)
2245                 return err;
2246
2247         ndm = nlmsg_data(nlh);
2248         if (ndm->ndm_ifindex == 0) {
2249                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2250                 return -EINVAL;
2251         }
2252
2253         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2254         if (dev == NULL) {
2255                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2256                 return -ENODEV;
2257         }
2258
2259         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2260                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2261                 return -EINVAL;
2262         }
2263
2264         addr = nla_data(tb[NDA_LLADDR]);
2265
2266         err = -EOPNOTSUPP;
2267
2268         /* Support fdb on master device the net/bridge default case */
2269         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2270             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2271                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2272                 const struct net_device_ops *ops = br_dev->netdev_ops;
2273
2274                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2275                 if (err)
2276                         goto out;
2277                 else
2278                         ndm->ndm_flags &= ~NTF_MASTER;
2279         }
2280
2281         /* Embedded bridge, macvlan, and any other device support */
2282         if ((ndm->ndm_flags & NTF_SELF)) {
2283                 if (dev->netdev_ops->ndo_fdb_add)
2284                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2285                                                            nlh->nlmsg_flags);
2286                 else
2287                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2288                                                nlh->nlmsg_flags);
2289
2290                 if (!err) {
2291                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2292                         ndm->ndm_flags &= ~NTF_SELF;
2293                 }
2294         }
2295 out:
2296         return err;
2297 }
2298
2299 /**
2300  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2301  */
2302 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2303                      struct nlattr *tb[],
2304                      struct net_device *dev,
2305                      const unsigned char *addr)
2306 {
2307         int err = -EOPNOTSUPP;
2308
2309         /* If aging addresses are supported device will need to
2310          * implement its own handler for this.
2311          */
2312         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2313                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2314                 return -EINVAL;
2315         }
2316
2317         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2318                 err = dev_uc_del(dev, addr);
2319         else if (is_multicast_ether_addr(addr))
2320                 err = dev_mc_del(dev, addr);
2321         else
2322                 err = -EINVAL;
2323
2324         return err;
2325 }
2326 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2327
2328 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2329 {
2330         struct net *net = sock_net(skb->sk);
2331         struct ndmsg *ndm;
2332         struct nlattr *tb[NDA_MAX+1];
2333         struct net_device *dev;
2334         int err = -EINVAL;
2335         __u8 *addr;
2336
2337         if (!netlink_capable(skb, CAP_NET_ADMIN))
2338                 return -EPERM;
2339
2340         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2341         if (err < 0)
2342                 return err;
2343
2344         ndm = nlmsg_data(nlh);
2345         if (ndm->ndm_ifindex == 0) {
2346                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2347                 return -EINVAL;
2348         }
2349
2350         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2351         if (dev == NULL) {
2352                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2353                 return -ENODEV;
2354         }
2355
2356         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2357                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2358                 return -EINVAL;
2359         }
2360
2361         addr = nla_data(tb[NDA_LLADDR]);
2362
2363         err = -EOPNOTSUPP;
2364
2365         /* Support fdb on master device the net/bridge default case */
2366         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2367             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2368                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2369                 const struct net_device_ops *ops = br_dev->netdev_ops;
2370
2371                 if (ops->ndo_fdb_del)
2372                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2373
2374                 if (err)
2375                         goto out;
2376                 else
2377                         ndm->ndm_flags &= ~NTF_MASTER;
2378         }
2379
2380         /* Embedded bridge, macvlan, and any other device support */
2381         if (ndm->ndm_flags & NTF_SELF) {
2382                 if (dev->netdev_ops->ndo_fdb_del)
2383                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2384                 else
2385                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2386
2387                 if (!err) {
2388                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2389                         ndm->ndm_flags &= ~NTF_SELF;
2390                 }
2391         }
2392 out:
2393         return err;
2394 }
2395
2396 static int nlmsg_populate_fdb(struct sk_buff *skb,
2397                               struct netlink_callback *cb,
2398                               struct net_device *dev,
2399                               int *idx,
2400                               struct netdev_hw_addr_list *list)
2401 {
2402         struct netdev_hw_addr *ha;
2403         int err;
2404         u32 portid, seq;
2405
2406         portid = NETLINK_CB(cb->skb).portid;
2407         seq = cb->nlh->nlmsg_seq;
2408
2409         list_for_each_entry(ha, &list->list, list) {
2410                 if (*idx < cb->args[0])
2411                         goto skip;
2412
2413                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2414                                               portid, seq,
2415                                               RTM_NEWNEIGH, NTF_SELF,
2416                                               NLM_F_MULTI);
2417                 if (err < 0)
2418                         return err;
2419 skip:
2420                 *idx += 1;
2421         }
2422         return 0;
2423 }
2424
2425 /**
2426  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2427  * @nlh: netlink message header
2428  * @dev: netdevice
2429  *
2430  * Default netdevice operation to dump the existing unicast address list.
2431  * Returns number of addresses from list put in skb.
2432  */
2433 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2434                       struct netlink_callback *cb,
2435                       struct net_device *dev,
2436                       int idx)
2437 {
2438         int err;
2439
2440         netif_addr_lock_bh(dev);
2441         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2442         if (err)
2443                 goto out;
2444         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2445 out:
2446         netif_addr_unlock_bh(dev);
2447         return idx;
2448 }
2449 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2450
2451 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2452 {
2453         int idx = 0;
2454         struct net *net = sock_net(skb->sk);
2455         struct net_device *dev;
2456
2457         rcu_read_lock();
2458         for_each_netdev_rcu(net, dev) {
2459                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2460                         struct net_device *br_dev;
2461                         const struct net_device_ops *ops;
2462
2463                         br_dev = netdev_master_upper_dev_get(dev);
2464                         ops = br_dev->netdev_ops;
2465                         if (ops->ndo_fdb_dump)
2466                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2467                 }
2468
2469                 if (dev->netdev_ops->ndo_fdb_dump)
2470                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2471                 else
2472                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2473         }
2474         rcu_read_unlock();
2475
2476         cb->args[0] = idx;
2477         return skb->len;
2478 }
2479
2480 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2481                             struct net_device *dev, u16 mode)
2482 {
2483         struct nlmsghdr *nlh;
2484         struct ifinfomsg *ifm;
2485         struct nlattr *br_afspec;
2486         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2487         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2488
2489         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2490         if (nlh == NULL)
2491                 return -EMSGSIZE;
2492
2493         ifm = nlmsg_data(nlh);
2494         ifm->ifi_family = AF_BRIDGE;
2495         ifm->__ifi_pad = 0;
2496         ifm->ifi_type = dev->type;
2497         ifm->ifi_index = dev->ifindex;
2498         ifm->ifi_flags = dev_get_flags(dev);
2499         ifm->ifi_change = 0;
2500
2501
2502         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2503             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2504             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2505             (br_dev &&
2506              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2507             (dev->addr_len &&
2508              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2509             (dev->ifindex != dev->iflink &&
2510              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2511                 goto nla_put_failure;
2512
2513         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2514         if (!br_afspec)
2515                 goto nla_put_failure;
2516
2517         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2518             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2519                 nla_nest_cancel(skb, br_afspec);
2520                 goto nla_put_failure;
2521         }
2522         nla_nest_end(skb, br_afspec);
2523
2524         return nlmsg_end(skb, nlh);
2525 nla_put_failure:
2526         nlmsg_cancel(skb, nlh);
2527         return -EMSGSIZE;
2528 }
2529 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2530
2531 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2532 {
2533         struct net *net = sock_net(skb->sk);
2534         struct net_device *dev;
2535         int idx = 0;
2536         u32 portid = NETLINK_CB(cb->skb).portid;
2537         u32 seq = cb->nlh->nlmsg_seq;
2538         struct nlattr *extfilt;
2539         u32 filter_mask = 0;
2540
2541         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2542                                   IFLA_EXT_MASK);
2543         if (extfilt)
2544                 filter_mask = nla_get_u32(extfilt);
2545
2546         rcu_read_lock();
2547         for_each_netdev_rcu(net, dev) {
2548                 const struct net_device_ops *ops = dev->netdev_ops;
2549                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2550
2551                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2552                         if (idx >= cb->args[0] &&
2553                             br_dev->netdev_ops->ndo_bridge_getlink(
2554                                     skb, portid, seq, dev, filter_mask) < 0)
2555                                 break;
2556                         idx++;
2557                 }
2558
2559                 if (ops->ndo_bridge_getlink) {
2560                         if (idx >= cb->args[0] &&
2561                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2562                                                     filter_mask) < 0)
2563                                 break;
2564                         idx++;
2565                 }
2566         }
2567         rcu_read_unlock();
2568         cb->args[0] = idx;
2569
2570         return skb->len;
2571 }
2572
2573 static inline size_t bridge_nlmsg_size(void)
2574 {
2575         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2576                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2577                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2578                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2579                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2580                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2581                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2582                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2583                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2584                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2585                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2586 }
2587
2588 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2589 {
2590         struct net *net = dev_net(dev);
2591         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2592         struct sk_buff *skb;
2593         int err = -EOPNOTSUPP;
2594
2595         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2596         if (!skb) {
2597                 err = -ENOMEM;
2598                 goto errout;
2599         }
2600
2601         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2602             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2603                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2604                 if (err < 0)
2605                         goto errout;
2606         }
2607
2608         if ((flags & BRIDGE_FLAGS_SELF) &&
2609             dev->netdev_ops->ndo_bridge_getlink) {
2610                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2611                 if (err < 0)
2612                         goto errout;
2613         }
2614
2615         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2616         return 0;
2617 errout:
2618         WARN_ON(err == -EMSGSIZE);
2619         kfree_skb(skb);
2620         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2621         return err;
2622 }
2623
2624 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2625 {
2626         struct net *net = sock_net(skb->sk);
2627         struct ifinfomsg *ifm;
2628         struct net_device *dev;
2629         struct nlattr *br_spec, *attr = NULL;
2630         int rem, err = -EOPNOTSUPP;
2631         u16 oflags, flags = 0;
2632         bool have_flags = false;
2633
2634         if (nlmsg_len(nlh) < sizeof(*ifm))
2635                 return -EINVAL;
2636
2637         ifm = nlmsg_data(nlh);
2638         if (ifm->ifi_family != AF_BRIDGE)
2639                 return -EPFNOSUPPORT;
2640
2641         dev = __dev_get_by_index(net, ifm->ifi_index);
2642         if (!dev) {
2643                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2644                 return -ENODEV;
2645         }
2646
2647         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2648         if (br_spec) {
2649                 nla_for_each_nested(attr, br_spec, rem) {
2650                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2651                                 have_flags = true;
2652                                 flags = nla_get_u16(attr);
2653                                 break;
2654                         }
2655                 }
2656         }
2657
2658         oflags = flags;
2659
2660         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2661                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2662
2663                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2664                         err = -EOPNOTSUPP;
2665                         goto out;
2666                 }
2667
2668                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2669                 if (err)
2670                         goto out;
2671
2672                 flags &= ~BRIDGE_FLAGS_MASTER;
2673         }
2674
2675         if ((flags & BRIDGE_FLAGS_SELF)) {
2676                 if (!dev->netdev_ops->ndo_bridge_setlink)
2677                         err = -EOPNOTSUPP;
2678                 else
2679                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2680
2681                 if (!err)
2682                         flags &= ~BRIDGE_FLAGS_SELF;
2683         }
2684
2685         if (have_flags)
2686                 memcpy(nla_data(attr), &flags, sizeof(flags));
2687         /* Generate event to notify upper layer of bridge change */
2688         if (!err)
2689                 err = rtnl_bridge_notify(dev, oflags);
2690 out:
2691         return err;
2692 }
2693
2694 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2695 {
2696         struct net *net = sock_net(skb->sk);
2697         struct ifinfomsg *ifm;
2698         struct net_device *dev;
2699         struct nlattr *br_spec, *attr = NULL;
2700         int rem, err = -EOPNOTSUPP;
2701         u16 oflags, flags = 0;
2702         bool have_flags = false;
2703
2704         if (nlmsg_len(nlh) < sizeof(*ifm))
2705                 return -EINVAL;
2706
2707         ifm = nlmsg_data(nlh);
2708         if (ifm->ifi_family != AF_BRIDGE)
2709                 return -EPFNOSUPPORT;
2710
2711         dev = __dev_get_by_index(net, ifm->ifi_index);
2712         if (!dev) {
2713                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2714                 return -ENODEV;
2715         }
2716
2717         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2718         if (br_spec) {
2719                 nla_for_each_nested(attr, br_spec, rem) {
2720                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2721                                 have_flags = true;
2722                                 flags = nla_get_u16(attr);
2723                                 break;
2724                         }
2725                 }
2726         }
2727
2728         oflags = flags;
2729
2730         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2731                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2732
2733                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2734                         err = -EOPNOTSUPP;
2735                         goto out;
2736                 }
2737
2738                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2739                 if (err)
2740                         goto out;
2741
2742                 flags &= ~BRIDGE_FLAGS_MASTER;
2743         }
2744
2745         if ((flags & BRIDGE_FLAGS_SELF)) {
2746                 if (!dev->netdev_ops->ndo_bridge_dellink)
2747                         err = -EOPNOTSUPP;
2748                 else
2749                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2750
2751                 if (!err)
2752                         flags &= ~BRIDGE_FLAGS_SELF;
2753         }
2754
2755         if (have_flags)
2756                 memcpy(nla_data(attr), &flags, sizeof(flags));
2757         /* Generate event to notify upper layer of bridge change */
2758         if (!err)
2759                 err = rtnl_bridge_notify(dev, oflags);
2760 out:
2761         return err;
2762 }
2763
2764 /* Process one rtnetlink message. */
2765
2766 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2767 {
2768         struct net *net = sock_net(skb->sk);
2769         rtnl_doit_func doit;
2770         int sz_idx, kind;
2771         int family;
2772         int type;
2773         int err;
2774
2775         type = nlh->nlmsg_type;
2776         if (type > RTM_MAX)
2777                 return -EOPNOTSUPP;
2778
2779         type -= RTM_BASE;
2780
2781         /* All the messages must have at least 1 byte length */
2782         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2783                 return 0;
2784
2785         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2786         sz_idx = type>>2;
2787         kind = type&3;
2788
2789         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2790                 return -EPERM;
2791
2792         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2793                 struct sock *rtnl;
2794                 rtnl_dumpit_func dumpit;
2795                 rtnl_calcit_func calcit;
2796                 u16 min_dump_alloc = 0;
2797
2798                 dumpit = rtnl_get_dumpit(family, type);
2799                 if (dumpit == NULL)
2800                         return -EOPNOTSUPP;
2801                 calcit = rtnl_get_calcit(family, type);
2802                 if (calcit)
2803                         min_dump_alloc = calcit(skb, nlh);
2804
2805                 __rtnl_unlock();
2806                 rtnl = net->rtnl;
2807                 {
2808                         struct netlink_dump_control c = {
2809                                 .dump           = dumpit,
2810                                 .min_dump_alloc = min_dump_alloc,
2811                         };
2812                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2813                 }
2814                 rtnl_lock();
2815                 return err;
2816         }
2817
2818         doit = rtnl_get_doit(family, type);
2819         if (doit == NULL)
2820                 return -EOPNOTSUPP;
2821
2822         return doit(skb, nlh);
2823 }
2824
2825 static void rtnetlink_rcv(struct sk_buff *skb)
2826 {
2827         rtnl_lock();
2828         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2829         rtnl_unlock();
2830 }
2831
2832 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2833 {
2834         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2835
2836         switch (event) {
2837         case NETDEV_UP:
2838         case NETDEV_DOWN:
2839         case NETDEV_PRE_UP:
2840         case NETDEV_POST_INIT:
2841         case NETDEV_REGISTER:
2842         case NETDEV_CHANGE:
2843         case NETDEV_PRE_TYPE_CHANGE:
2844         case NETDEV_GOING_DOWN:
2845         case NETDEV_UNREGISTER:
2846         case NETDEV_UNREGISTER_FINAL:
2847         case NETDEV_RELEASE:
2848         case NETDEV_JOIN:
2849                 break;
2850         default:
2851                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2852                 break;
2853         }
2854         return NOTIFY_DONE;
2855 }
2856
2857 static struct notifier_block rtnetlink_dev_notifier = {
2858         .notifier_call  = rtnetlink_event,
2859 };
2860
2861
2862 static int __net_init rtnetlink_net_init(struct net *net)
2863 {
2864         struct sock *sk;
2865         struct netlink_kernel_cfg cfg = {
2866                 .groups         = RTNLGRP_MAX,
2867                 .input          = rtnetlink_rcv,
2868                 .cb_mutex       = &rtnl_mutex,
2869                 .flags          = NL_CFG_F_NONROOT_RECV,
2870         };
2871
2872         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2873         if (!sk)
2874                 return -ENOMEM;
2875         net->rtnl = sk;
2876         return 0;
2877 }
2878
2879 static void __net_exit rtnetlink_net_exit(struct net *net)
2880 {
2881         netlink_kernel_release(net->rtnl);
2882         net->rtnl = NULL;
2883 }
2884
2885 static struct pernet_operations rtnetlink_net_ops = {
2886         .init = rtnetlink_net_init,
2887         .exit = rtnetlink_net_exit,
2888 };
2889
2890 void __init rtnetlink_init(void)
2891 {
2892         if (register_pernet_subsys(&rtnetlink_net_ops))
2893                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2894
2895         register_netdevice_notifier(&rtnetlink_dev_notifier);
2896
2897         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2898                       rtnl_dump_ifinfo, rtnl_calcit);
2899         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2900         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2901         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2902
2903         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2904         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2905
2906         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2907         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2908         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2909
2910         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2911         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2912         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2913 }
2914