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