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 /* Return with the rtnl_lock held when there are no network
 357  * devices unregistering in any network namespace.
 358  */
 359 static void rtnl_lock_unregistering_all(void)
 360 {
 361         struct net *net;
 362         bool unregistering;
 363         DEFINE_WAIT(wait);
 364
 365         for (;;) {
 366                 prepare_to_wait(&netdev_unregistering_wq, &wait,
 367                                 TASK_UNINTERRUPTIBLE);
 368                 unregistering = false;
 369                 rtnl_lock();
 370                 for_each_net(net) {
 371                         if (net->dev_unreg_count > 0) {
 372                                 unregistering = true;
 373                                 break;
 374                         }
 375                 }
 376                 if (!unregistering)
 377                         break;
 378                 __rtnl_unlock();
 379                 schedule();
 380         }
 381         finish_wait(&netdev_unregistering_wq, &wait);
 382 }
 383
 384 /**
 385  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 386  * @ops: struct rtnl_link_ops * to unregister
 387  */
 388 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 389 {
 390         /* Close the race with cleanup_net() */
 391         mutex_lock(&net_mutex);
 392         rtnl_lock_unregistering_all();
 393         __rtnl_link_unregister(ops);
 394         rtnl_unlock();
 395         mutex_unlock(&net_mutex);
 396 }
 397 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 398
 399 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 400 {
 401         struct net_device *master_dev;
 402         const struct rtnl_link_ops *ops;
 403
 404         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 405         if (!master_dev)
 406                 return 0;
 407         ops = master_dev->rtnl_link_ops;
 408         if (!ops || !ops->get_slave_size)
 409                 return 0;
 410         /* IFLA_INFO_SLAVE_DATA + nested data */
 411         return nla_total_size(sizeof(struct nlattr)) +
 412                ops->get_slave_size(master_dev, dev);
 413 }
 414
 415 static size_t rtnl_link_get_size(const struct net_device *dev)
 416 {
 417         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 418         size_t size;
 419
 420         if (!ops)
 421                 return 0;
 422
 423         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 424                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 425
 426         if (ops->get_size)
 427                 /* IFLA_INFO_DATA + nested data */
 428                 size += nla_total_size(sizeof(struct nlattr)) +
 429                         ops->get_size(dev);
 430
 431         if (ops->get_xstats_size)
 432                 /* IFLA_INFO_XSTATS */
 433                 size += nla_total_size(ops->get_xstats_size(dev));
 434
 435         size += rtnl_link_get_slave_info_data_size(dev);
 436
 437         return size;
 438 }
 439
 440 static LIST_HEAD(rtnl_af_ops);
 441
 442 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 443 {
 444         const struct rtnl_af_ops *ops;
 445
 446         list_for_each_entry(ops, &rtnl_af_ops, list) {
 447                 if (ops->family == family)
 448                         return ops;
 449         }
 450
 451         return NULL;
 452 }
 453
 454 /**
 455  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 456  * @ops: struct rtnl_af_ops * to register
 457  *
 458  * Returns 0 on success or a negative error code.
 459  */
 460 void rtnl_af_register(struct rtnl_af_ops *ops)
 461 {
 462         rtnl_lock();
 463         list_add_tail(&ops->list, &rtnl_af_ops);
 464         rtnl_unlock();
 465 }
 466 EXPORT_SYMBOL_GPL(rtnl_af_register);
 467
 468 /**
 469  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 470  * @ops: struct rtnl_af_ops * to unregister
 471  *
 472  * The caller must hold the rtnl_mutex.
 473  */
 474 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 475 {
 476         list_del(&ops->list);
 477 }
 478 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 479
 480 /**
 481  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 482  * @ops: struct rtnl_af_ops * to unregister
 483  */
 484 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 485 {
 486         rtnl_lock();
 487         __rtnl_af_unregister(ops);
 488         rtnl_unlock();
 489 }
 490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 491
 492 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 493 {
 494         struct rtnl_af_ops *af_ops;
 495         size_t size;
 496
 497         /* IFLA_AF_SPEC */
 498         size = nla_total_size(sizeof(struct nlattr));
 499
 500         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 501                 if (af_ops->get_link_af_size) {
 502                         /* AF_* + nested data */
 503                         size += nla_total_size(sizeof(struct nlattr)) +
 504                                 af_ops->get_link_af_size(dev);
 505                 }
 506         }
 507
 508         return size;
 509 }
 510
 511 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 512 {
 513         struct net_device *master_dev;
 514
 515         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 516         if (master_dev && master_dev->rtnl_link_ops)
 517                 return true;
 518         return false;
 519 }
 520
 521 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 522                                      const struct net_device *dev)
 523 {
 524         struct net_device *master_dev;
 525         const struct rtnl_link_ops *ops;
 526         struct nlattr *slave_data;
 527         int err;
 528
 529         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 530         if (!master_dev)
 531                 return 0;
 532         ops = master_dev->rtnl_link_ops;
 533         if (!ops)
 534                 return 0;
 535         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 536                 return -EMSGSIZE;
 537         if (ops->fill_slave_info) {
 538                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 539                 if (!slave_data)
 540                         return -EMSGSIZE;
 541                 err = ops->fill_slave_info(skb, master_dev, dev);
 542                 if (err < 0)
 543                         goto err_cancel_slave_data;
 544                 nla_nest_end(skb, slave_data);
 545         }
 546         return 0;
 547
 548 err_cancel_slave_data:
 549         nla_nest_cancel(skb, slave_data);
 550         return err;
 551 }
 552
 553 static int rtnl_link_info_fill(struct sk_buff *skb,
 554                                const struct net_device *dev)
 555 {
 556         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 557         struct nlattr *data;
 558         int err;
 559
 560         if (!ops)
 561                 return 0;
 562         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 563                 return -EMSGSIZE;
 564         if (ops->fill_xstats) {
 565                 err = ops->fill_xstats(skb, dev);
 566                 if (err < 0)
 567                         return err;
 568         }
 569         if (ops->fill_info) {
 570                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 571                 if (data == NULL)
 572                         return -EMSGSIZE;
 573                 err = ops->fill_info(skb, dev);
 574                 if (err < 0)
 575                         goto err_cancel_data;
 576                 nla_nest_end(skb, data);
 577         }
 578         return 0;
 579
 580 err_cancel_data:
 581         nla_nest_cancel(skb, data);
 582         return err;
 583 }
 584
 585 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 586 {
 587         struct nlattr *linkinfo;
 588         int err = -EMSGSIZE;
 589
 590         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 591         if (linkinfo == NULL)
 592                 goto out;
 593
 594         err = rtnl_link_info_fill(skb, dev);
 595         if (err < 0)
 596                 goto err_cancel_link;
 597
 598         err = rtnl_link_slave_info_fill(skb, dev);
 599         if (err < 0)
 600                 goto err_cancel_link;
 601
 602         nla_nest_end(skb, linkinfo);
 603         return 0;
 604
 605 err_cancel_link:
 606         nla_nest_cancel(skb, linkinfo);
 607 out:
 608         return err;
 609 }
 610
 611 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 612 {
 613         struct sock *rtnl = net->rtnl;
 614         int err = 0;
 615
 616         NETLINK_CB(skb).dst_group = group;
 617         if (echo)
 618                 atomic_inc(&skb->users);
 619         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 620         if (echo)
 621                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 622         return err;
 623 }
 624
 625 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 626 {
 627         struct sock *rtnl = net->rtnl;
 628
 629         return nlmsg_unicast(rtnl, skb, pid);
 630 }
 631 EXPORT_SYMBOL(rtnl_unicast);
 632
 633 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 634                  struct nlmsghdr *nlh, gfp_t flags)
 635 {
 636         struct sock *rtnl = net->rtnl;
 637         int report = 0;
 638
 639         if (nlh)
 640                 report = nlmsg_report(nlh);
 641
 642         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 643 }
 644 EXPORT_SYMBOL(rtnl_notify);
 645
 646 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 647 {
 648         struct sock *rtnl = net->rtnl;
 649
 650         netlink_set_err(rtnl, 0, group, error);
 651 }
 652 EXPORT_SYMBOL(rtnl_set_sk_err);
 653
 654 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 655 {
 656         struct nlattr *mx;
 657         int i, valid = 0;
 658
 659         mx = nla_nest_start(skb, RTA_METRICS);
 660         if (mx == NULL)
 661                 return -ENOBUFS;
 662
 663         for (i = 0; i < RTAX_MAX; i++) {
 664                 if (metrics[i]) {
 665                         valid++;
 666                         if (nla_put_u32(skb, i+1, metrics[i]))
 667                                 goto nla_put_failure;
 668                 }
 669         }
 670
 671         if (!valid) {
 672                 nla_nest_cancel(skb, mx);
 673                 return 0;
 674         }
 675
 676         return nla_nest_end(skb, mx);
 677
 678 nla_put_failure:
 679         nla_nest_cancel(skb, mx);
 680         return -EMSGSIZE;
 681 }
 682 EXPORT_SYMBOL(rtnetlink_put_metrics);
 683
 684 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 685                        long expires, u32 error)
 686 {
 687         struct rta_cacheinfo ci = {
 688                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 689                 .rta_used = dst->__use,
 690                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 691                 .rta_error = error,
 692                 .rta_id =  id,
 693         };
 694
 695         if (expires) {
 696                 unsigned long clock;
 697
 698                 clock = jiffies_to_clock_t(abs(expires));
 699                 clock = min_t(unsigned long, clock, INT_MAX);
 700                 ci.rta_expires = (expires > 0) ? clock : -clock;
 701         }
 702         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 703 }
 704 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 705
 706 static void set_operstate(struct net_device *dev, unsigned char transition)
 707 {
 708         unsigned char operstate = dev->operstate;
 709
 710         switch (transition) {
 711         case IF_OPER_UP:
 712                 if ((operstate == IF_OPER_DORMANT ||
 713                      operstate == IF_OPER_UNKNOWN) &&
 714                     !netif_dormant(dev))
 715                         operstate = IF_OPER_UP;
 716                 break;
 717
 718         case IF_OPER_DORMANT:
 719                 if (operstate == IF_OPER_UP ||
 720                     operstate == IF_OPER_UNKNOWN)
 721                         operstate = IF_OPER_DORMANT;
 722                 break;
 723         }
 724
 725         if (dev->operstate != operstate) {
 726                 write_lock_bh(&dev_base_lock);
 727                 dev->operstate = operstate;
 728                 write_unlock_bh(&dev_base_lock);
 729                 netdev_state_change(dev);
 730         }
 731 }
 732
 733 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 734 {
 735         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 736                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 737 }
 738
 739 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 740                                            const struct ifinfomsg *ifm)
 741 {
 742         unsigned int flags = ifm->ifi_flags;
 743
 744         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 745         if (ifm->ifi_change)
 746                 flags = (flags & ifm->ifi_change) |
 747                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 748
 749         return flags;
 750 }
 751
 752 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 753                                  const struct rtnl_link_stats64 *b)
 754 {
 755         a->rx_packets = b->rx_packets;
 756         a->tx_packets = b->tx_packets;
 757         a->rx_bytes = b->rx_bytes;
 758         a->tx_bytes = b->tx_bytes;
 759         a->rx_errors = b->rx_errors;
 760         a->tx_errors = b->tx_errors;
 761         a->rx_dropped = b->rx_dropped;
 762         a->tx_dropped = b->tx_dropped;
 763
 764         a->multicast = b->multicast;
 765         a->collisions = b->collisions;
 766
 767         a->rx_length_errors = b->rx_length_errors;
 768         a->rx_over_errors = b->rx_over_errors;
 769         a->rx_crc_errors = b->rx_crc_errors;
 770         a->rx_frame_errors = b->rx_frame_errors;
 771         a->rx_fifo_errors = b->rx_fifo_errors;
 772         a->rx_missed_errors = b->rx_missed_errors;
 773
 774         a->tx_aborted_errors = b->tx_aborted_errors;
 775         a->tx_carrier_errors = b->tx_carrier_errors;
 776         a->tx_fifo_errors = b->tx_fifo_errors;
 777         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 778         a->tx_window_errors = b->tx_window_errors;
 779
 780         a->rx_compressed = b->rx_compressed;
 781         a->tx_compressed = b->tx_compressed;
 782 }
 783
 784 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 785 {
 786         memcpy(v, b, sizeof(*b));
 787 }
 788
 789 /* All VF info */
 790 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 791                                    u32 ext_filter_mask)
 792 {
 793         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 794             (ext_filter_mask & RTEXT_FILTER_VF)) {
 795                 int num_vfs = dev_num_vf(dev->dev.parent);
 796                 size_t size = nla_total_size(sizeof(struct nlattr));
 797                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 798                 size += num_vfs *
 799                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 800                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 801                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 802                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 803                          nla_total_size(sizeof(struct ifla_vf_link_state)));
 804                 return size;
 805         } else
 806                 return 0;
 807 }
 808
 809 static size_t rtnl_port_size(const struct net_device *dev,
 810                              u32 ext_filter_mask)
 811 {
 812         size_t port_size = nla_total_size(4)            /* PORT_VF */
 813                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 814                 + nla_total_size(sizeof(struct ifla_port_vsi))
 815                                                         /* PORT_VSI_TYPE */
 816                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 817                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 818                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 819                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 820         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 821         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 822                 + port_size;
 823         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 824                 + port_size;
 825
 826         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 827             !(ext_filter_mask & RTEXT_FILTER_VF))
 828                 return 0;
 829         if (dev_num_vf(dev->dev.parent))
 830                 return port_self_size + vf_ports_size +
 831                         vf_port_size * dev_num_vf(dev->dev.parent);
 832         else
 833                 return port_self_size;
 834 }
 835
 836 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 837                                      u32 ext_filter_mask)
 838 {
 839         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 840                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 841                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 842                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 843                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 844                + nla_total_size(sizeof(struct rtnl_link_stats))
 845                + nla_total_size(sizeof(struct rtnl_link_stats64))
 846                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 847                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 848                + nla_total_size(4) /* IFLA_TXQLEN */
 849                + nla_total_size(4) /* IFLA_WEIGHT */
 850                + nla_total_size(4) /* IFLA_MTU */
 851                + nla_total_size(4) /* IFLA_LINK */
 852                + nla_total_size(4) /* IFLA_MASTER */
 853                + nla_total_size(1) /* IFLA_CARRIER */
 854                + nla_total_size(4) /* IFLA_PROMISCUITY */
 855                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 856                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 857                + nla_total_size(1) /* IFLA_OPERSTATE */
 858                + nla_total_size(1) /* IFLA_LINKMODE */
 859                + nla_total_size(ext_filter_mask
 860                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 861                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 862                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 863                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 864                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 865                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 866 }
 867
 868 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 869 {
 870         struct nlattr *vf_ports;
 871         struct nlattr *vf_port;
 872         int vf;
 873         int err;
 874
 875         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 876         if (!vf_ports)
 877                 return -EMSGSIZE;
 878
 879         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 880                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 881                 if (!vf_port)
 882                         goto nla_put_failure;
 883                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 884                         goto nla_put_failure;
 885                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 886                 if (err == -EMSGSIZE)
 887                         goto nla_put_failure;
 888                 if (err) {
 889                         nla_nest_cancel(skb, vf_port);
 890                         continue;
 891                 }
 892                 nla_nest_end(skb, vf_port);
 893         }
 894
 895         nla_nest_end(skb, vf_ports);
 896
 897         return 0;
 898
 899 nla_put_failure:
 900         nla_nest_cancel(skb, vf_ports);
 901         return -EMSGSIZE;
 902 }
 903
 904 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 905 {
 906         struct nlattr *port_self;
 907         int err;
 908
 909         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 910         if (!port_self)
 911                 return -EMSGSIZE;
 912
 913         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 914         if (err) {
 915                 nla_nest_cancel(skb, port_self);
 916                 return (err == -EMSGSIZE) ? err : 0;
 917         }
 918
 919         nla_nest_end(skb, port_self);
 920
 921         return 0;
 922 }
 923
 924 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 925                           u32 ext_filter_mask)
 926 {
 927         int err;
 928
 929         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 930             !(ext_filter_mask & RTEXT_FILTER_VF))
 931                 return 0;
 932
 933         err = rtnl_port_self_fill(skb, dev);
 934         if (err)
 935                 return err;
 936
 937         if (dev_num_vf(dev->dev.parent)) {
 938                 err = rtnl_vf_ports_fill(skb, dev);
 939                 if (err)
 940                         return err;
 941         }
 942
 943         return 0;
 944 }
 945
 946 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 947 {
 948         int err;
 949         struct netdev_phys_port_id ppid;
 950
 951         err = dev_get_phys_port_id(dev, &ppid);
 952         if (err) {
 953                 if (err == -EOPNOTSUPP)
 954                         return 0;
 955                 return err;
 956         }
 957
 958         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 959                 return -EMSGSIZE;
 960
 961         return 0;
 962 }
 963
 964 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 965                             int type, u32 pid, u32 seq, u32 change,
 966                             unsigned int flags, u32 ext_filter_mask)
 967 {
 968         struct ifinfomsg *ifm;
 969         struct nlmsghdr *nlh;
 970         struct rtnl_link_stats64 temp;
 971         const struct rtnl_link_stats64 *stats;
 972         struct nlattr *attr, *af_spec;
 973         struct rtnl_af_ops *af_ops;
 974         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 975
 976         ASSERT_RTNL();
 977         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 978         if (nlh == NULL)
 979                 return -EMSGSIZE;
 980
 981         ifm = nlmsg_data(nlh);
 982         ifm->ifi_family = AF_UNSPEC;
 983         ifm->__ifi_pad = 0;
 984         ifm->ifi_type = dev->type;
 985         ifm->ifi_index = dev->ifindex;
 986         ifm->ifi_flags = dev_get_flags(dev);
 987         ifm->ifi_change = change;
 988
 989         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 990             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 991             nla_put_u8(skb, IFLA_OPERSTATE,
 992                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 993             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 994             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 995             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 996             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 997             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 998 #ifdef CONFIG_RPS
 999             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1000 #endif
1001             (dev->ifindex != dev->iflink &&
1002              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1003             (upper_dev &&
1004              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1005             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1006             (dev->qdisc &&
1007              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1008             (dev->ifalias &&
1009              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
1010                 goto nla_put_failure;
1011
1012         if (1) {
1013                 struct rtnl_link_ifmap map = {
1014                         .mem_start   = dev->mem_start,
1015                         .mem_end     = dev->mem_end,
1016                         .base_addr   = dev->base_addr,
1017                         .irq         = dev->irq,
1018                         .dma         = dev->dma,
1019                         .port        = dev->if_port,
1020                 };
1021                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1022                         goto nla_put_failure;
1023         }
1024
1025         if (dev->addr_len) {
1026                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1027                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1028                         goto nla_put_failure;
1029         }
1030
1031         if (rtnl_phys_port_id_fill(skb, dev))
1032                 goto nla_put_failure;
1033
1034         attr = nla_reserve(skb, IFLA_STATS,
1035                         sizeof(struct rtnl_link_stats));
1036         if (attr == NULL)
1037                 goto nla_put_failure;
1038
1039         stats = dev_get_stats(dev, &temp);
1040         copy_rtnl_link_stats(nla_data(attr), stats);
1041
1042         attr = nla_reserve(skb, IFLA_STATS64,
1043                         sizeof(struct rtnl_link_stats64));
1044         if (attr == NULL)
1045                 goto nla_put_failure;
1046         copy_rtnl_link_stats64(nla_data(attr), stats);
1047
1048         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1049             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1050                 goto nla_put_failure;
1051
1052         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1053             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1054                 int i;
1055
1056                 struct nlattr *vfinfo, *vf;
1057                 int num_vfs = dev_num_vf(dev->dev.parent);
1058
1059                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1060                 if (!vfinfo)
1061                         goto nla_put_failure;
1062                 for (i = 0; i < num_vfs; i++) {
1063                         struct ifla_vf_info ivi;
1064                         struct ifla_vf_mac vf_mac;
1065                         struct ifla_vf_vlan vf_vlan;
1066                         struct ifla_vf_tx_rate vf_tx_rate;
1067                         struct ifla_vf_spoofchk vf_spoofchk;
1068                         struct ifla_vf_link_state vf_linkstate;
1069
1070                         /*
1071                          * Not all SR-IOV capable drivers support the
1072                          * spoofcheck query.  Preset to -1 so the user
1073                          * space tool can detect that the driver didn't
1074                          * report anything.
1075                          */
1076                         ivi.spoofchk = -1;
1077                         memset(ivi.mac, 0, sizeof(ivi.mac));
1078                         /* The default value for VF link state is "auto"
1079                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1080                          */
1081                         ivi.linkstate = 0;
1082                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1083                                 break;
1084                         vf_mac.vf =
1085                                 vf_vlan.vf =
1086                                 vf_tx_rate.vf =
1087                                 vf_spoofchk.vf =
1088                                 vf_linkstate.vf = ivi.vf;
1089
1090                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1091                         vf_vlan.vlan = ivi.vlan;
1092                         vf_vlan.qos = ivi.qos;
1093                         vf_tx_rate.rate = ivi.tx_rate;
1094                         vf_spoofchk.setting = ivi.spoofchk;
1095                         vf_linkstate.link_state = ivi.linkstate;
1096                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1097                         if (!vf) {
1098                                 nla_nest_cancel(skb, vfinfo);
1099                                 goto nla_put_failure;
1100                         }
1101                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1102                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1103                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1104                                     &vf_tx_rate) ||
1105                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1106                                     &vf_spoofchk) ||
1107                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1108                                     &vf_linkstate))
1109                                 goto nla_put_failure;
1110                         nla_nest_end(skb, vf);
1111                 }
1112                 nla_nest_end(skb, vfinfo);
1113         }
1114
1115         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1116                 goto nla_put_failure;
1117
1118         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1119                 if (rtnl_link_fill(skb, dev) < 0)
1120                         goto nla_put_failure;
1121         }
1122
1123         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1124                 goto nla_put_failure;
1125
1126         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1127                 if (af_ops->fill_link_af) {
1128                         struct nlattr *af;
1129                         int err;
1130
1131                         if (!(af = nla_nest_start(skb, af_ops->family)))
1132                                 goto nla_put_failure;
1133
1134                         err = af_ops->fill_link_af(skb, dev);
1135
1136                         /*
1137                          * Caller may return ENODATA to indicate that there
1138                          * was no data to be dumped. This is not an error, it
1139                          * means we should trim the attribute header and
1140                          * continue.
1141                          */
1142                         if (err == -ENODATA)
1143                                 nla_nest_cancel(skb, af);
1144                         else if (err < 0)
1145                                 goto nla_put_failure;
1146
1147                         nla_nest_end(skb, af);
1148                 }
1149         }
1150
1151         nla_nest_end(skb, af_spec);
1152
1153         return nlmsg_end(skb, nlh);
1154
1155 nla_put_failure:
1156         nlmsg_cancel(skb, nlh);
1157         return -EMSGSIZE;
1158 }
1159
1160 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1161 {
1162         struct net *net = sock_net(skb->sk);
1163         int h, s_h;
1164         int idx = 0, s_idx;
1165         struct net_device *dev;
1166         struct hlist_head *head;
1167         struct nlattr *tb[IFLA_MAX+1];
1168         u32 ext_filter_mask = 0;
1169         int err;
1170         int hdrlen;
1171
1172         s_h = cb->args[0];
1173         s_idx = cb->args[1];
1174
1175         rcu_read_lock();
1176         cb->seq = net->dev_base_seq;
1177
1178         /* A hack to preserve kernel<->userspace interface.
1179          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1180          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1181          * what iproute2 < v3.9.0 used.
1182          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1183          * attribute, its netlink message is shorter than struct ifinfomsg.
1184          */
1185         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1186                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1187
1188         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1189
1190                 if (tb[IFLA_EXT_MASK])
1191                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1192         }
1193
1194         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1195                 idx = 0;
1196                 head = &net->dev_index_head[h];
1197                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1198                         if (idx < s_idx)
1199                                 goto cont;
1200                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1201                                                NETLINK_CB(cb->skb).portid,
1202                                                cb->nlh->nlmsg_seq, 0,
1203                                                NLM_F_MULTI,
1204                                                ext_filter_mask);
1205                         /* If we ran out of room on the first message,
1206                          * we're in trouble
1207                          */
1208                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1209
1210                         if (err <= 0)
1211                                 goto out;
1212
1213                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1214 cont:
1215                         idx++;
1216                 }
1217         }
1218 out:
1219         rcu_read_unlock();
1220         cb->args[1] = idx;
1221         cb->args[0] = h;
1222
1223         return skb->len;
1224 }
1225
1226 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1227         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1228         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1229         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1230         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1231         [IFLA_MTU]              = { .type = NLA_U32 },
1232         [IFLA_LINK]             = { .type = NLA_U32 },
1233         [IFLA_MASTER]           = { .type = NLA_U32 },
1234         [IFLA_CARRIER]          = { .type = NLA_U8 },
1235         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1236         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1237         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1238         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1239         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1240         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1241         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1242         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1243         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1244         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1245         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1246         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1247         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1248         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1249         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1250         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1251         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1252 };
1253 EXPORT_SYMBOL(ifla_policy);
1254
1255 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1256         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1257         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1258         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1259         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1260 };
1261
1262 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1263         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1264 };
1265
1266 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1267         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1268                                     .len = sizeof(struct ifla_vf_mac) },
1269         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1270                                     .len = sizeof(struct ifla_vf_vlan) },
1271         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1272                                     .len = sizeof(struct ifla_vf_tx_rate) },
1273         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1274                                     .len = sizeof(struct ifla_vf_spoofchk) },
1275 };
1276
1277 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1278         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1279         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1280                                     .len = PORT_PROFILE_MAX },
1281         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1282                                     .len = sizeof(struct ifla_port_vsi)},
1283         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1284                                       .len = PORT_UUID_MAX },
1285         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1286                                     .len = PORT_UUID_MAX },
1287         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1288         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1289 };
1290
1291 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1292 {
1293         struct net *net;
1294         /* Examine the link attributes and figure out which
1295          * network namespace we are talking about.
1296          */
1297         if (tb[IFLA_NET_NS_PID])
1298                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1299         else if (tb[IFLA_NET_NS_FD])
1300                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1301         else
1302                 net = get_net(src_net);
1303         return net;
1304 }
1305 EXPORT_SYMBOL(rtnl_link_get_net);
1306
1307 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1308 {
1309         if (dev) {
1310                 if (tb[IFLA_ADDRESS] &&
1311                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1312                         return -EINVAL;
1313
1314                 if (tb[IFLA_BROADCAST] &&
1315                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1316                         return -EINVAL;
1317         }
1318
1319         if (tb[IFLA_AF_SPEC]) {
1320                 struct nlattr *af;
1321                 int rem, err;
1322
1323                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1324                         const struct rtnl_af_ops *af_ops;
1325
1326                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1327                                 return -EAFNOSUPPORT;
1328
1329                         if (!af_ops->set_link_af)
1330                                 return -EOPNOTSUPP;
1331
1332                         if (af_ops->validate_link_af) {
1333                                 err = af_ops->validate_link_af(dev, af);
1334                                 if (err < 0)
1335                                         return err;
1336                         }
1337                 }
1338         }
1339
1340         return 0;
1341 }
1342
1343 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1344 {
1345         int rem, err = -EINVAL;
1346         struct nlattr *vf;
1347         const struct net_device_ops *ops = dev->netdev_ops;
1348
1349         nla_for_each_nested(vf, attr, rem) {
1350                 switch (nla_type(vf)) {
1351                 case IFLA_VF_MAC: {
1352                         struct ifla_vf_mac *ivm;
1353                         ivm = nla_data(vf);
1354                         err = -EOPNOTSUPP;
1355                         if (ops->ndo_set_vf_mac)
1356                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1357                                                           ivm->mac);
1358                         break;
1359                 }
1360                 case IFLA_VF_VLAN: {
1361                         struct ifla_vf_vlan *ivv;
1362                         ivv = nla_data(vf);
1363                         err = -EOPNOTSUPP;
1364                         if (ops->ndo_set_vf_vlan)
1365                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1366                                                            ivv->vlan,
1367                                                            ivv->qos);
1368                         break;
1369                 }
1370                 case IFLA_VF_TX_RATE: {
1371                         struct ifla_vf_tx_rate *ivt;
1372                         ivt = nla_data(vf);
1373                         err = -EOPNOTSUPP;
1374                         if (ops->ndo_set_vf_tx_rate)
1375                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1376                                                               ivt->rate);
1377                         break;
1378                 }
1379                 case IFLA_VF_SPOOFCHK: {
1380                         struct ifla_vf_spoofchk *ivs;
1381                         ivs = nla_data(vf);
1382                         err = -EOPNOTSUPP;
1383                         if (ops->ndo_set_vf_spoofchk)
1384                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1385                                                                ivs->setting);
1386                         break;
1387                 }
1388                 case IFLA_VF_LINK_STATE: {
1389                         struct ifla_vf_link_state *ivl;
1390                         ivl = nla_data(vf);
1391                         err = -EOPNOTSUPP;
1392                         if (ops->ndo_set_vf_link_state)
1393                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1394                                                                  ivl->link_state);
1395                         break;
1396                 }
1397                 default:
1398                         err = -EINVAL;
1399                         break;
1400                 }
1401                 if (err)
1402                         break;
1403         }
1404         return err;
1405 }
1406
1407 static int do_set_master(struct net_device *dev, int ifindex)
1408 {
1409         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1410         const struct net_device_ops *ops;
1411         int err;
1412
1413         if (upper_dev) {
1414                 if (upper_dev->ifindex == ifindex)
1415                         return 0;
1416                 ops = upper_dev->netdev_ops;
1417                 if (ops->ndo_del_slave) {
1418                         err = ops->ndo_del_slave(upper_dev, dev);
1419                         if (err)
1420                                 return err;
1421                 } else {
1422                         return -EOPNOTSUPP;
1423                 }
1424         }
1425
1426         if (ifindex) {
1427                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1428                 if (!upper_dev)
1429                         return -EINVAL;
1430                 ops = upper_dev->netdev_ops;
1431                 if (ops->ndo_add_slave) {
1432                         err = ops->ndo_add_slave(upper_dev, dev);
1433                         if (err)
1434                                 return err;
1435                 } else {
1436                         return -EOPNOTSUPP;
1437                 }
1438         }
1439         return 0;
1440 }
1441
1442 static int do_setlink(const struct sk_buff *skb,
1443                       struct net_device *dev, struct ifinfomsg *ifm,
1444                       struct nlattr **tb, char *ifname, int modified)
1445 {
1446         const struct net_device_ops *ops = dev->netdev_ops;
1447         int err;
1448
1449         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1450                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1451                 if (IS_ERR(net)) {
1452                         err = PTR_ERR(net);
1453                         goto errout;
1454                 }
1455                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1456                         put_net(net);
1457                         err = -EPERM;
1458                         goto errout;
1459                 }
1460                 err = dev_change_net_namespace(dev, net, ifname);
1461                 put_net(net);
1462                 if (err)
1463                         goto errout;
1464                 modified = 1;
1465         }
1466
1467         if (tb[IFLA_MAP]) {
1468                 struct rtnl_link_ifmap *u_map;
1469                 struct ifmap k_map;
1470
1471                 if (!ops->ndo_set_config) {
1472                         err = -EOPNOTSUPP;
1473                         goto errout;
1474                 }
1475
1476                 if (!netif_device_present(dev)) {
1477                         err = -ENODEV;
1478                         goto errout;
1479                 }
1480
1481                 u_map = nla_data(tb[IFLA_MAP]);
1482                 k_map.mem_start = (unsigned long) u_map->mem_start;
1483                 k_map.mem_end = (unsigned long) u_map->mem_end;
1484                 k_map.base_addr = (unsigned short) u_map->base_addr;
1485                 k_map.irq = (unsigned char) u_map->irq;
1486                 k_map.dma = (unsigned char) u_map->dma;
1487                 k_map.port = (unsigned char) u_map->port;
1488
1489                 err = ops->ndo_set_config(dev, &k_map);
1490                 if (err < 0)
1491                         goto errout;
1492
1493                 modified = 1;
1494         }
1495
1496         if (tb[IFLA_ADDRESS]) {
1497                 struct sockaddr *sa;
1498                 int len;
1499
1500                 len = sizeof(sa_family_t) + dev->addr_len;
1501                 sa = kmalloc(len, GFP_KERNEL);
1502                 if (!sa) {
1503                         err = -ENOMEM;
1504                         goto errout;
1505                 }
1506                 sa->sa_family = dev->type;
1507                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1508                        dev->addr_len);
1509                 err = dev_set_mac_address(dev, sa);
1510                 kfree(sa);
1511                 if (err)
1512                         goto errout;
1513                 modified = 1;
1514         }
1515
1516         if (tb[IFLA_MTU]) {
1517                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1518                 if (err < 0)
1519                         goto errout;
1520                 modified = 1;
1521         }
1522
1523         if (tb[IFLA_GROUP]) {
1524                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1525                 modified = 1;
1526         }
1527
1528         /*
1529          * Interface selected by interface index but interface
1530          * name provided implies that a name change has been
1531          * requested.
1532          */
1533         if (ifm->ifi_index > 0 && ifname[0]) {
1534                 err = dev_change_name(dev, ifname);
1535                 if (err < 0)
1536                         goto errout;
1537                 modified = 1;
1538         }
1539
1540         if (tb[IFLA_IFALIAS]) {
1541                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1542                                     nla_len(tb[IFLA_IFALIAS]));
1543                 if (err < 0)
1544                         goto errout;
1545                 modified = 1;
1546         }
1547
1548         if (tb[IFLA_BROADCAST]) {
1549                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1550                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1551         }
1552
1553         if (ifm->ifi_flags || ifm->ifi_change) {
1554                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1555                 if (err < 0)
1556                         goto errout;
1557         }
1558
1559         if (tb[IFLA_MASTER]) {
1560                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1561                 if (err)
1562                         goto errout;
1563                 modified = 1;
1564         }
1565
1566         if (tb[IFLA_CARRIER]) {
1567                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1568                 if (err)
1569                         goto errout;
1570                 modified = 1;
1571         }
1572
1573         if (tb[IFLA_TXQLEN])
1574                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1575
1576         if (tb[IFLA_OPERSTATE])
1577                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1578
1579         if (tb[IFLA_LINKMODE]) {
1580                 write_lock_bh(&dev_base_lock);
1581                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1582                 write_unlock_bh(&dev_base_lock);
1583         }
1584
1585         if (tb[IFLA_VFINFO_LIST]) {
1586                 struct nlattr *attr;
1587                 int rem;
1588                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1589                         if (nla_type(attr) != IFLA_VF_INFO) {
1590                                 err = -EINVAL;
1591                                 goto errout;
1592                         }
1593                         err = do_setvfinfo(dev, attr);
1594                         if (err < 0)
1595                                 goto errout;
1596                         modified = 1;
1597                 }
1598         }
1599         err = 0;
1600
1601         if (tb[IFLA_VF_PORTS]) {
1602                 struct nlattr *port[IFLA_PORT_MAX+1];
1603                 struct nlattr *attr;
1604                 int vf;
1605                 int rem;
1606
1607                 err = -EOPNOTSUPP;
1608                 if (!ops->ndo_set_vf_port)
1609                         goto errout;
1610
1611                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1612                         if (nla_type(attr) != IFLA_VF_PORT)
1613                                 continue;
1614                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1615                                 attr, ifla_port_policy);
1616                         if (err < 0)
1617                                 goto errout;
1618                         if (!port[IFLA_PORT_VF]) {
1619                                 err = -EOPNOTSUPP;
1620                                 goto errout;
1621                         }
1622                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1623                         err = ops->ndo_set_vf_port(dev, vf, port);
1624                         if (err < 0)
1625                                 goto errout;
1626                         modified = 1;
1627                 }
1628         }
1629         err = 0;
1630
1631         if (tb[IFLA_PORT_SELF]) {
1632                 struct nlattr *port[IFLA_PORT_MAX+1];
1633
1634                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1635                         tb[IFLA_PORT_SELF], ifla_port_policy);
1636                 if (err < 0)
1637                         goto errout;
1638
1639                 err = -EOPNOTSUPP;
1640                 if (ops->ndo_set_vf_port)
1641                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1642                 if (err < 0)
1643                         goto errout;
1644                 modified = 1;
1645         }
1646
1647         if (tb[IFLA_AF_SPEC]) {
1648                 struct nlattr *af;
1649                 int rem;
1650
1651                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1652                         const struct rtnl_af_ops *af_ops;
1653
1654                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1655                                 BUG();
1656
1657                         err = af_ops->set_link_af(dev, af);
1658                         if (err < 0)
1659                                 goto errout;
1660
1661                         modified = 1;
1662                 }
1663         }
1664         err = 0;
1665
1666 errout:
1667         if (err < 0 && modified)
1668                 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",
1669                                      dev->name);
1670
1671         return err;
1672 }
1673
1674 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1675 {
1676         struct net *net = sock_net(skb->sk);
1677         struct ifinfomsg *ifm;
1678         struct net_device *dev;
1679         int err;
1680         struct nlattr *tb[IFLA_MAX+1];
1681         char ifname[IFNAMSIZ];
1682
1683         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1684         if (err < 0)
1685                 goto errout;
1686
1687         if (tb[IFLA_IFNAME])
1688                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1689         else
1690                 ifname[0] = '\0';
1691
1692         err = -EINVAL;
1693         ifm = nlmsg_data(nlh);
1694         if (ifm->ifi_index > 0)
1695                 dev = __dev_get_by_index(net, ifm->ifi_index);
1696         else if (tb[IFLA_IFNAME])
1697                 dev = __dev_get_by_name(net, ifname);
1698         else
1699                 goto errout;
1700
1701         if (dev == NULL) {
1702                 err = -ENODEV;
1703                 goto errout;
1704         }
1705
1706         err = validate_linkmsg(dev, tb);
1707         if (err < 0)
1708                 goto errout;
1709
1710         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1711 errout:
1712         return err;
1713 }
1714
1715 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1716 {
1717         struct net *net = sock_net(skb->sk);
1718         const struct rtnl_link_ops *ops;
1719         struct net_device *dev;
1720         struct ifinfomsg *ifm;
1721         char ifname[IFNAMSIZ];
1722         struct nlattr *tb[IFLA_MAX+1];
1723         int err;
1724         LIST_HEAD(list_kill);
1725
1726         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1727         if (err < 0)
1728                 return err;
1729
1730         if (tb[IFLA_IFNAME])
1731                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1732
1733         ifm = nlmsg_data(nlh);
1734         if (ifm->ifi_index > 0)
1735                 dev = __dev_get_by_index(net, ifm->ifi_index);
1736         else if (tb[IFLA_IFNAME])
1737                 dev = __dev_get_by_name(net, ifname);
1738         else
1739                 return -EINVAL;
1740
1741         if (!dev)
1742                 return -ENODEV;
1743
1744         ops = dev->rtnl_link_ops;
1745         if (!ops)
1746                 return -EOPNOTSUPP;
1747
1748         ops->dellink(dev, &list_kill);
1749         unregister_netdevice_many(&list_kill);
1750         return 0;
1751 }
1752
1753 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1754 {
1755         unsigned int old_flags;
1756         int err;
1757
1758         old_flags = dev->flags;
1759         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1760                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1761                 if (err < 0)
1762                         return err;
1763         }
1764
1765         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1766
1767         __dev_notify_flags(dev, old_flags, ~0U);
1768         return 0;
1769 }
1770 EXPORT_SYMBOL(rtnl_configure_link);
1771
1772 struct net_device *rtnl_create_link(struct net *net,
1773         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1774 {
1775         int err;
1776         struct net_device *dev;
1777         unsigned int num_tx_queues = 1;
1778         unsigned int num_rx_queues = 1;
1779
1780         if (tb[IFLA_NUM_TX_QUEUES])
1781                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1782         else if (ops->get_num_tx_queues)
1783                 num_tx_queues = ops->get_num_tx_queues();
1784
1785         if (tb[IFLA_NUM_RX_QUEUES])
1786                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1787         else if (ops->get_num_rx_queues)
1788                 num_rx_queues = ops->get_num_rx_queues();
1789
1790         err = -ENOMEM;
1791         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1792                                num_tx_queues, num_rx_queues);
1793         if (!dev)
1794                 goto err;
1795
1796         dev_net_set(dev, net);
1797         dev->rtnl_link_ops = ops;
1798         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1799
1800         if (tb[IFLA_MTU])
1801                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1802         if (tb[IFLA_ADDRESS]) {
1803                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1804                                 nla_len(tb[IFLA_ADDRESS]));
1805                 dev->addr_assign_type = NET_ADDR_SET;
1806         }
1807         if (tb[IFLA_BROADCAST])
1808                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1809                                 nla_len(tb[IFLA_BROADCAST]));
1810         if (tb[IFLA_TXQLEN])
1811                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1812         if (tb[IFLA_OPERSTATE])
1813                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1814         if (tb[IFLA_LINKMODE])
1815                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1816         if (tb[IFLA_GROUP])
1817                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1818
1819         return dev;
1820
1821 err:
1822         return ERR_PTR(err);
1823 }
1824 EXPORT_SYMBOL(rtnl_create_link);
1825
1826 static int rtnl_group_changelink(const struct sk_buff *skb,
1827                 struct net *net, int group,
1828                 struct ifinfomsg *ifm,
1829                 struct nlattr **tb)
1830 {
1831         struct net_device *dev;
1832         int err;
1833
1834         for_each_netdev(net, dev) {
1835                 if (dev->group == group) {
1836                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1837                         if (err < 0)
1838                                 return err;
1839                 }
1840         }
1841
1842         return 0;
1843 }
1844
1845 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1846 {
1847         struct net *net = sock_net(skb->sk);
1848         const struct rtnl_link_ops *ops;
1849         const struct rtnl_link_ops *m_ops = NULL;
1850         struct net_device *dev;
1851         struct net_device *master_dev = NULL;
1852         struct ifinfomsg *ifm;
1853         char kind[MODULE_NAME_LEN];
1854         char ifname[IFNAMSIZ];
1855         struct nlattr *tb[IFLA_MAX+1];
1856         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1857         int err;
1858
1859 #ifdef CONFIG_MODULES
1860 replay:
1861 #endif
1862         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1863         if (err < 0)
1864                 return err;
1865
1866         if (tb[IFLA_IFNAME])
1867                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1868         else
1869                 ifname[0] = '\0';
1870
1871         ifm = nlmsg_data(nlh);
1872         if (ifm->ifi_index > 0)
1873                 dev = __dev_get_by_index(net, ifm->ifi_index);
1874         else {
1875                 if (ifname[0])
1876                         dev = __dev_get_by_name(net, ifname);
1877                 else
1878                         dev = NULL;
1879         }
1880
1881         if (dev) {
1882                 master_dev = netdev_master_upper_dev_get(dev);
1883                 if (master_dev)
1884                         m_ops = master_dev->rtnl_link_ops;
1885         }
1886
1887         err = validate_linkmsg(dev, tb);
1888         if (err < 0)
1889                 return err;
1890
1891         if (tb[IFLA_LINKINFO]) {
1892                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1893                                        tb[IFLA_LINKINFO], ifla_info_policy);
1894                 if (err < 0)
1895                         return err;
1896         } else
1897                 memset(linkinfo, 0, sizeof(linkinfo));
1898
1899         if (linkinfo[IFLA_INFO_KIND]) {
1900                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1901                 ops = rtnl_link_ops_get(kind);
1902         } else {
1903                 kind[0] = '\0';
1904                 ops = NULL;
1905         }
1906
1907         if (1) {
1908                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1909                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1910                 struct nlattr **data = NULL;
1911                 struct nlattr **slave_data = NULL;
1912                 struct net *dest_net;
1913
1914                 if (ops) {
1915                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1916                                 err = nla_parse_nested(attr, ops->maxtype,
1917                                                        linkinfo[IFLA_INFO_DATA],
1918                                                        ops->policy);
1919                                 if (err < 0)
1920                                         return err;
1921                                 data = attr;
1922                         }
1923                         if (ops->validate) {
1924                                 err = ops->validate(tb, data);
1925                                 if (err < 0)
1926                                         return err;
1927                         }
1928                 }
1929
1930                 if (m_ops) {
1931                         if (m_ops->slave_maxtype &&
1932                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
1933                                 err = nla_parse_nested(slave_attr,
1934                                                        m_ops->slave_maxtype,
1935                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
1936                                                        m_ops->slave_policy);
1937                                 if (err < 0)
1938                                         return err;
1939                                 slave_data = slave_attr;
1940                         }
1941                         if (m_ops->slave_validate) {
1942                                 err = m_ops->slave_validate(tb, slave_data);
1943                                 if (err < 0)
1944                                         return err;
1945                         }
1946                 }
1947
1948                 if (dev) {
1949                         int modified = 0;
1950
1951                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1952                                 return -EEXIST;
1953                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1954                                 return -EOPNOTSUPP;
1955
1956                         if (linkinfo[IFLA_INFO_DATA]) {
1957                                 if (!ops || ops != dev->rtnl_link_ops ||
1958                                     !ops->changelink)
1959                                         return -EOPNOTSUPP;
1960
1961                                 err = ops->changelink(dev, tb, data);
1962                                 if (err < 0)
1963                                         return err;
1964                                 modified = 1;
1965                         }
1966
1967                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1968                                 if (!m_ops || !m_ops->slave_changelink)
1969                                         return -EOPNOTSUPP;
1970
1971                                 err = m_ops->slave_changelink(master_dev, dev,
1972                                                               tb, slave_data);
1973                                 if (err < 0)
1974                                         return err;
1975                                 modified = 1;
1976                         }
1977
1978                         return do_setlink(skb, dev, ifm, tb, ifname, modified);
1979                 }
1980
1981                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1982                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1983                                 return rtnl_group_changelink(skb, net,
1984                                                 nla_get_u32(tb[IFLA_GROUP]),
1985                                                 ifm, tb);
1986                         return -ENODEV;
1987                 }
1988
1989                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1990                         return -EOPNOTSUPP;
1991
1992                 if (!ops) {
1993 #ifdef CONFIG_MODULES
1994                         if (kind[0]) {
1995                                 __rtnl_unlock();
1996                                 request_module("rtnl-link-%s", kind);
1997                                 rtnl_lock();
1998                                 ops = rtnl_link_ops_get(kind);
1999                                 if (ops)
2000                                         goto replay;
2001                         }
2002 #endif
2003                         return -EOPNOTSUPP;
2004                 }
2005
2006                 if (!ifname[0])
2007                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2008
2009                 dest_net = rtnl_link_get_net(net, tb);
2010                 if (IS_ERR(dest_net))
2011                         return PTR_ERR(dest_net);
2012
2013                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
2014                 if (IS_ERR(dev)) {
2015                         err = PTR_ERR(dev);
2016                         goto out;
2017                 }
2018
2019                 dev->ifindex = ifm->ifi_index;
2020
2021                 if (ops->newlink) {
2022                         err = ops->newlink(net, dev, tb, data);
2023                         /* Drivers should call free_netdev() in ->destructor
2024                          * and unregister it on failure so that device could be
2025                          * finally freed in rtnl_unlock.
2026                          */
2027                         if (err < 0)
2028                                 goto out;
2029                 } else {
2030                         err = register_netdevice(dev);
2031                         if (err < 0) {
2032                                 free_netdev(dev);
2033                                 goto out;
2034                         }
2035                 }
2036                 err = rtnl_configure_link(dev, ifm);
2037                 if (err < 0)
2038                         unregister_netdevice(dev);
2039 out:
2040                 put_net(dest_net);
2041                 return err;
2042         }
2043 }
2044
2045 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2046 {
2047         struct net *net = sock_net(skb->sk);
2048         struct ifinfomsg *ifm;
2049         char ifname[IFNAMSIZ];
2050         struct nlattr *tb[IFLA_MAX+1];
2051         struct net_device *dev = NULL;
2052         struct sk_buff *nskb;
2053         int err;
2054         u32 ext_filter_mask = 0;
2055
2056         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2057         if (err < 0)
2058                 return err;
2059
2060         if (tb[IFLA_IFNAME])
2061                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2062
2063         if (tb[IFLA_EXT_MASK])
2064                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2065
2066         ifm = nlmsg_data(nlh);
2067         if (ifm->ifi_index > 0)
2068                 dev = __dev_get_by_index(net, ifm->ifi_index);
2069         else if (tb[IFLA_IFNAME])
2070                 dev = __dev_get_by_name(net, ifname);
2071         else
2072                 return -EINVAL;
2073
2074         if (dev == NULL)
2075                 return -ENODEV;
2076
2077         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2078         if (nskb == NULL)
2079                 return -ENOBUFS;
2080
2081         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2082                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2083         if (err < 0) {
2084                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2085                 WARN_ON(err == -EMSGSIZE);
2086                 kfree_skb(nskb);
2087         } else
2088                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2089
2090         return err;
2091 }
2092
2093 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2094 {
2095         struct net *net = sock_net(skb->sk);
2096         struct net_device *dev;
2097         struct nlattr *tb[IFLA_MAX+1];
2098         u32 ext_filter_mask = 0;
2099         u16 min_ifinfo_dump_size = 0;
2100         int hdrlen;
2101
2102         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2103         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2104                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2105
2106         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2107                 if (tb[IFLA_EXT_MASK])
2108                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2109         }
2110
2111         if (!ext_filter_mask)
2112                 return NLMSG_GOODSIZE;
2113         /*
2114          * traverse the list of net devices and compute the minimum
2115          * buffer size based upon the filter mask.
2116          */
2117         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2118                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2119                                              if_nlmsg_size(dev,
2120                                                            ext_filter_mask));
2121         }
2122
2123         return min_ifinfo_dump_size;
2124 }
2125
2126 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2127 {
2128         int idx;
2129         int s_idx = cb->family;
2130
2131         if (s_idx == 0)
2132                 s_idx = 1;
2133         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2134                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2135                 if (idx < s_idx || idx == PF_PACKET)
2136                         continue;
2137                 if (rtnl_msg_handlers[idx] == NULL ||
2138                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2139                         continue;
2140                 if (idx > s_idx) {
2141                         memset(&cb->args[0], 0, sizeof(cb->args));
2142                         cb->prev_seq = 0;
2143                         cb->seq = 0;
2144                 }
2145                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2146                         break;
2147         }
2148         cb->family = idx;
2149
2150         return skb->len;
2151 }
2152
2153 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2154                   gfp_t flags)
2155 {
2156         struct net *net = dev_net(dev);
2157         struct sk_buff *skb;
2158         int err = -ENOBUFS;
2159         size_t if_info_size;
2160
2161         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2162         if (skb == NULL)
2163                 goto errout;
2164
2165         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2166         if (err < 0) {
2167                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2168                 WARN_ON(err == -EMSGSIZE);
2169                 kfree_skb(skb);
2170                 goto errout;
2171         }
2172         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2173         return;
2174 errout:
2175         if (err < 0)
2176                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2177 }
2178 EXPORT_SYMBOL(rtmsg_ifinfo);
2179
2180 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2181                                    struct net_device *dev,
2182                                    u8 *addr, u32 pid, u32 seq,
2183                                    int type, unsigned int flags,
2184                                    int nlflags)
2185 {
2186         struct nlmsghdr *nlh;
2187         struct ndmsg *ndm;
2188
2189         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2190         if (!nlh)
2191                 return -EMSGSIZE;
2192
2193         ndm = nlmsg_data(nlh);
2194         ndm->ndm_family  = AF_BRIDGE;
2195         ndm->ndm_pad1    = 0;
2196         ndm->ndm_pad2    = 0;
2197         ndm->ndm_flags   = flags;
2198         ndm->ndm_type    = 0;
2199         ndm->ndm_ifindex = dev->ifindex;
2200         ndm->ndm_state   = NUD_PERMANENT;
2201
2202         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2203                 goto nla_put_failure;
2204
2205         return nlmsg_end(skb, nlh);
2206
2207 nla_put_failure:
2208         nlmsg_cancel(skb, nlh);
2209         return -EMSGSIZE;
2210 }
2211
2212 static inline size_t rtnl_fdb_nlmsg_size(void)
2213 {
2214         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2215 }
2216
2217 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2218 {
2219         struct net *net = dev_net(dev);
2220         struct sk_buff *skb;
2221         int err = -ENOBUFS;
2222
2223         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2224         if (!skb)
2225                 goto errout;
2226
2227         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2228         if (err < 0) {
2229                 kfree_skb(skb);
2230                 goto errout;
2231         }
2232
2233         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2234         return;
2235 errout:
2236         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2237 }
2238
2239 /**
2240  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2241  */
2242 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2243                      struct nlattr *tb[],
2244                      struct net_device *dev,
2245                      const unsigned char *addr,
2246                      u16 flags)
2247 {
2248         int err = -EINVAL;
2249
2250         /* If aging addresses are supported device will need to
2251          * implement its own handler for this.
2252          */
2253         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2254                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2255                 return err;
2256         }
2257
2258         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2259                 err = dev_uc_add_excl(dev, addr);
2260         else if (is_multicast_ether_addr(addr))
2261                 err = dev_mc_add_excl(dev, addr);
2262
2263         /* Only return duplicate errors if NLM_F_EXCL is set */
2264         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2265                 err = 0;
2266
2267         return err;
2268 }
2269 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2270
2271 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2272 {
2273         struct net *net = sock_net(skb->sk);
2274         struct ndmsg *ndm;
2275         struct nlattr *tb[NDA_MAX+1];
2276         struct net_device *dev;
2277         u8 *addr;
2278         int err;
2279
2280         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2281         if (err < 0)
2282                 return err;
2283
2284         ndm = nlmsg_data(nlh);
2285         if (ndm->ndm_ifindex == 0) {
2286                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2287                 return -EINVAL;
2288         }
2289
2290         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2291         if (dev == NULL) {
2292                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2293                 return -ENODEV;
2294         }
2295
2296         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2297                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2298                 return -EINVAL;
2299         }
2300
2301         addr = nla_data(tb[NDA_LLADDR]);
2302
2303         err = -EOPNOTSUPP;
2304
2305         /* Support fdb on master device the net/bridge default case */
2306         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2307             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2308                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2309                 const struct net_device_ops *ops = br_dev->netdev_ops;
2310
2311                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2312                 if (err)
2313                         goto out;
2314                 else
2315                         ndm->ndm_flags &= ~NTF_MASTER;
2316         }
2317
2318         /* Embedded bridge, macvlan, and any other device support */
2319         if ((ndm->ndm_flags & NTF_SELF)) {
2320                 if (dev->netdev_ops->ndo_fdb_add)
2321                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2322                                                            nlh->nlmsg_flags);
2323                 else
2324                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2325                                                nlh->nlmsg_flags);
2326
2327                 if (!err) {
2328                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2329                         ndm->ndm_flags &= ~NTF_SELF;
2330                 }
2331         }
2332 out:
2333         return err;
2334 }
2335
2336 /**
2337  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2338  */
2339 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2340                      struct nlattr *tb[],
2341                      struct net_device *dev,
2342                      const unsigned char *addr)
2343 {
2344         int err = -EOPNOTSUPP;
2345
2346         /* If aging addresses are supported device will need to
2347          * implement its own handler for this.
2348          */
2349         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2350                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2351                 return -EINVAL;
2352         }
2353
2354         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2355                 err = dev_uc_del(dev, addr);
2356         else if (is_multicast_ether_addr(addr))
2357                 err = dev_mc_del(dev, addr);
2358         else
2359                 err = -EINVAL;
2360
2361         return err;
2362 }
2363 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2364
2365 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2366 {
2367         struct net *net = sock_net(skb->sk);
2368         struct ndmsg *ndm;
2369         struct nlattr *tb[NDA_MAX+1];
2370         struct net_device *dev;
2371         int err = -EINVAL;
2372         __u8 *addr;
2373
2374         if (!netlink_capable(skb, CAP_NET_ADMIN))
2375                 return -EPERM;
2376
2377         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2378         if (err < 0)
2379                 return err;
2380
2381         ndm = nlmsg_data(nlh);
2382         if (ndm->ndm_ifindex == 0) {
2383                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2384                 return -EINVAL;
2385         }
2386
2387         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2388         if (dev == NULL) {
2389                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2390                 return -ENODEV;
2391         }
2392
2393         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2394                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2395                 return -EINVAL;
2396         }
2397
2398         addr = nla_data(tb[NDA_LLADDR]);
2399
2400         err = -EOPNOTSUPP;
2401
2402         /* Support fdb on master device the net/bridge default case */
2403         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2404             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2405                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2406                 const struct net_device_ops *ops = br_dev->netdev_ops;
2407
2408                 if (ops->ndo_fdb_del)
2409                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2410
2411                 if (err)
2412                         goto out;
2413                 else
2414                         ndm->ndm_flags &= ~NTF_MASTER;
2415         }
2416
2417         /* Embedded bridge, macvlan, and any other device support */
2418         if (ndm->ndm_flags & NTF_SELF) {
2419                 if (dev->netdev_ops->ndo_fdb_del)
2420                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2421                 else
2422                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2423
2424                 if (!err) {
2425                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2426                         ndm->ndm_flags &= ~NTF_SELF;
2427                 }
2428         }
2429 out:
2430         return err;
2431 }
2432
2433 static int nlmsg_populate_fdb(struct sk_buff *skb,
2434                               struct netlink_callback *cb,
2435                               struct net_device *dev,
2436                               int *idx,
2437                               struct netdev_hw_addr_list *list)
2438 {
2439         struct netdev_hw_addr *ha;
2440         int err;
2441         u32 portid, seq;
2442
2443         portid = NETLINK_CB(cb->skb).portid;
2444         seq = cb->nlh->nlmsg_seq;
2445
2446         list_for_each_entry(ha, &list->list, list) {
2447                 if (*idx < cb->args[0])
2448                         goto skip;
2449
2450                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2451                                               portid, seq,
2452                                               RTM_NEWNEIGH, NTF_SELF,
2453                                               NLM_F_MULTI);
2454                 if (err < 0)
2455                         return err;
2456 skip:
2457                 *idx += 1;
2458         }
2459         return 0;
2460 }
2461
2462 /**
2463  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2464  * @nlh: netlink message header
2465  * @dev: netdevice
2466  *
2467  * Default netdevice operation to dump the existing unicast address list.
2468  * Returns number of addresses from list put in skb.
2469  */
2470 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2471                       struct netlink_callback *cb,
2472                       struct net_device *dev,
2473                       int idx)
2474 {
2475         int err;
2476
2477         netif_addr_lock_bh(dev);
2478         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2479         if (err)
2480                 goto out;
2481         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2482 out:
2483         netif_addr_unlock_bh(dev);
2484         return idx;
2485 }
2486 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2487
2488 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2489 {
2490         int idx = 0;
2491         struct net *net = sock_net(skb->sk);
2492         struct net_device *dev;
2493
2494         rcu_read_lock();
2495         for_each_netdev_rcu(net, dev) {
2496                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2497                         struct net_device *br_dev;
2498                         const struct net_device_ops *ops;
2499
2500                         br_dev = netdev_master_upper_dev_get(dev);
2501                         ops = br_dev->netdev_ops;
2502                         if (ops->ndo_fdb_dump)
2503                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2504                 }
2505
2506                 if (dev->netdev_ops->ndo_fdb_dump)
2507                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2508                 else
2509                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2510         }
2511         rcu_read_unlock();
2512
2513         cb->args[0] = idx;
2514         return skb->len;
2515 }
2516
2517 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2518                             struct net_device *dev, u16 mode)
2519 {
2520         struct nlmsghdr *nlh;
2521         struct ifinfomsg *ifm;
2522         struct nlattr *br_afspec;
2523         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2524         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2525
2526         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2527         if (nlh == NULL)
2528                 return -EMSGSIZE;
2529
2530         ifm = nlmsg_data(nlh);
2531         ifm->ifi_family = AF_BRIDGE;
2532         ifm->__ifi_pad = 0;
2533         ifm->ifi_type = dev->type;
2534         ifm->ifi_index = dev->ifindex;
2535         ifm->ifi_flags = dev_get_flags(dev);
2536         ifm->ifi_change = 0;
2537
2538
2539         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2540             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2541             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2542             (br_dev &&
2543              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2544             (dev->addr_len &&
2545              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2546             (dev->ifindex != dev->iflink &&
2547              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2548                 goto nla_put_failure;
2549
2550         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2551         if (!br_afspec)
2552                 goto nla_put_failure;
2553
2554         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2555             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2556                 nla_nest_cancel(skb, br_afspec);
2557                 goto nla_put_failure;
2558         }
2559         nla_nest_end(skb, br_afspec);
2560
2561         return nlmsg_end(skb, nlh);
2562 nla_put_failure:
2563         nlmsg_cancel(skb, nlh);
2564         return -EMSGSIZE;
2565 }
2566 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2567
2568 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2569 {
2570         struct net *net = sock_net(skb->sk);
2571         struct net_device *dev;
2572         int idx = 0;
2573         u32 portid = NETLINK_CB(cb->skb).portid;
2574         u32 seq = cb->nlh->nlmsg_seq;
2575         struct nlattr *extfilt;
2576         u32 filter_mask = 0;
2577
2578         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2579                                   IFLA_EXT_MASK);
2580         if (extfilt)
2581                 filter_mask = nla_get_u32(extfilt);
2582
2583         rcu_read_lock();
2584         for_each_netdev_rcu(net, dev) {
2585                 const struct net_device_ops *ops = dev->netdev_ops;
2586                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2587
2588                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2589                         if (idx >= cb->args[0] &&
2590                             br_dev->netdev_ops->ndo_bridge_getlink(
2591                                     skb, portid, seq, dev, filter_mask) < 0)
2592                                 break;
2593                         idx++;
2594                 }
2595
2596                 if (ops->ndo_bridge_getlink) {
2597                         if (idx >= cb->args[0] &&
2598                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2599                                                     filter_mask) < 0)
2600                                 break;
2601                         idx++;
2602                 }
2603         }
2604         rcu_read_unlock();
2605         cb->args[0] = idx;
2606
2607         return skb->len;
2608 }
2609
2610 static inline size_t bridge_nlmsg_size(void)
2611 {
2612         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2613                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2614                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2615                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2616                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2617                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2618                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2619                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2620                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2621                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2622                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2623 }
2624
2625 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2626 {
2627         struct net *net = dev_net(dev);
2628         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2629         struct sk_buff *skb;
2630         int err = -EOPNOTSUPP;
2631
2632         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2633         if (!skb) {
2634                 err = -ENOMEM;
2635                 goto errout;
2636         }
2637
2638         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2639             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2640                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2641                 if (err < 0)
2642                         goto errout;
2643         }
2644
2645         if ((flags & BRIDGE_FLAGS_SELF) &&
2646             dev->netdev_ops->ndo_bridge_getlink) {
2647                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2648                 if (err < 0)
2649                         goto errout;
2650         }
2651
2652         if (!skb->len)
2653                 goto errout;
2654
2655         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2656         return 0;
2657 errout:
2658         WARN_ON(err == -EMSGSIZE);
2659         kfree_skb(skb);
2660         if (err)
2661                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2662         return err;
2663 }
2664
2665 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2666 {
2667         struct net *net = sock_net(skb->sk);
2668         struct ifinfomsg *ifm;
2669         struct net_device *dev;
2670         struct nlattr *br_spec, *attr = NULL;
2671         int rem, err = -EOPNOTSUPP;
2672         u16 oflags, flags = 0;
2673         bool have_flags = false;
2674
2675         if (nlmsg_len(nlh) < sizeof(*ifm))
2676                 return -EINVAL;
2677
2678         ifm = nlmsg_data(nlh);
2679         if (ifm->ifi_family != AF_BRIDGE)
2680                 return -EPFNOSUPPORT;
2681
2682         dev = __dev_get_by_index(net, ifm->ifi_index);
2683         if (!dev) {
2684                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2685                 return -ENODEV;
2686         }
2687
2688         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2689         if (br_spec) {
2690                 nla_for_each_nested(attr, br_spec, rem) {
2691                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2692                                 have_flags = true;
2693                                 flags = nla_get_u16(attr);
2694                                 break;
2695                         }
2696                 }
2697         }
2698
2699         oflags = flags;
2700
2701         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2702                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2703
2704                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2705                         err = -EOPNOTSUPP;
2706                         goto out;
2707                 }
2708
2709                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2710                 if (err)
2711                         goto out;
2712
2713                 flags &= ~BRIDGE_FLAGS_MASTER;
2714         }
2715
2716         if ((flags & BRIDGE_FLAGS_SELF)) {
2717                 if (!dev->netdev_ops->ndo_bridge_setlink)
2718                         err = -EOPNOTSUPP;
2719                 else
2720                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2721
2722                 if (!err)
2723                         flags &= ~BRIDGE_FLAGS_SELF;
2724         }
2725
2726         if (have_flags)
2727                 memcpy(nla_data(attr), &flags, sizeof(flags));
2728         /* Generate event to notify upper layer of bridge change */
2729         if (!err)
2730                 err = rtnl_bridge_notify(dev, oflags);
2731 out:
2732         return err;
2733 }
2734
2735 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2736 {
2737         struct net *net = sock_net(skb->sk);
2738         struct ifinfomsg *ifm;
2739         struct net_device *dev;
2740         struct nlattr *br_spec, *attr = NULL;
2741         int rem, err = -EOPNOTSUPP;
2742         u16 oflags, flags = 0;
2743         bool have_flags = false;
2744
2745         if (nlmsg_len(nlh) < sizeof(*ifm))
2746                 return -EINVAL;
2747
2748         ifm = nlmsg_data(nlh);
2749         if (ifm->ifi_family != AF_BRIDGE)
2750                 return -EPFNOSUPPORT;
2751
2752         dev = __dev_get_by_index(net, ifm->ifi_index);
2753         if (!dev) {
2754                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2755                 return -ENODEV;
2756         }
2757
2758         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2759         if (br_spec) {
2760                 nla_for_each_nested(attr, br_spec, rem) {
2761                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2762                                 have_flags = true;
2763                                 flags = nla_get_u16(attr);
2764                                 break;
2765                         }
2766                 }
2767         }
2768
2769         oflags = flags;
2770
2771         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2772                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2773
2774                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2775                         err = -EOPNOTSUPP;
2776                         goto out;
2777                 }
2778
2779                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2780                 if (err)
2781                         goto out;
2782
2783                 flags &= ~BRIDGE_FLAGS_MASTER;
2784         }
2785
2786         if ((flags & BRIDGE_FLAGS_SELF)) {
2787                 if (!dev->netdev_ops->ndo_bridge_dellink)
2788                         err = -EOPNOTSUPP;
2789                 else
2790                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2791
2792                 if (!err)
2793                         flags &= ~BRIDGE_FLAGS_SELF;
2794         }
2795
2796         if (have_flags)
2797                 memcpy(nla_data(attr), &flags, sizeof(flags));
2798         /* Generate event to notify upper layer of bridge change */
2799         if (!err)
2800                 err = rtnl_bridge_notify(dev, oflags);
2801 out:
2802         return err;
2803 }
2804
2805 /* Process one rtnetlink message. */
2806
2807 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2808 {
2809         struct net *net = sock_net(skb->sk);
2810         rtnl_doit_func doit;
2811         int sz_idx, kind;
2812         int family;
2813         int type;
2814         int err;
2815
2816         type = nlh->nlmsg_type;
2817         if (type > RTM_MAX)
2818                 return -EOPNOTSUPP;
2819
2820         type -= RTM_BASE;
2821
2822         /* All the messages must have at least 1 byte length */
2823         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2824                 return 0;
2825
2826         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2827         sz_idx = type>>2;
2828         kind = type&3;
2829
2830         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2831                 return -EPERM;
2832
2833         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2834                 struct sock *rtnl;
2835                 rtnl_dumpit_func dumpit;
2836                 rtnl_calcit_func calcit;
2837                 u16 min_dump_alloc = 0;
2838
2839                 dumpit = rtnl_get_dumpit(family, type);
2840                 if (dumpit == NULL)
2841                         return -EOPNOTSUPP;
2842                 calcit = rtnl_get_calcit(family, type);
2843                 if (calcit)
2844                         min_dump_alloc = calcit(skb, nlh);
2845
2846                 __rtnl_unlock();
2847                 rtnl = net->rtnl;
2848                 {
2849                         struct netlink_dump_control c = {
2850                                 .dump           = dumpit,
2851                                 .min_dump_alloc = min_dump_alloc,
2852                         };
2853                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2854                 }
2855                 rtnl_lock();
2856                 return err;
2857         }
2858
2859         doit = rtnl_get_doit(family, type);
2860         if (doit == NULL)
2861                 return -EOPNOTSUPP;
2862
2863         return doit(skb, nlh);
2864 }
2865
2866 static void rtnetlink_rcv(struct sk_buff *skb)
2867 {
2868         rtnl_lock();
2869         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2870         rtnl_unlock();
2871 }
2872
2873 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2874 {
2875         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2876
2877         switch (event) {
2878         case NETDEV_UP:
2879         case NETDEV_DOWN:
2880         case NETDEV_PRE_UP:
2881         case NETDEV_POST_INIT:
2882         case NETDEV_REGISTER:
2883         case NETDEV_CHANGE:
2884         case NETDEV_PRE_TYPE_CHANGE:
2885         case NETDEV_GOING_DOWN:
2886         case NETDEV_UNREGISTER:
2887         case NETDEV_UNREGISTER_FINAL:
2888         case NETDEV_RELEASE:
2889         case NETDEV_JOIN:
2890                 break;
2891         default:
2892                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2893                 break;
2894         }
2895         return NOTIFY_DONE;
2896 }
2897
2898 static struct notifier_block rtnetlink_dev_notifier = {
2899         .notifier_call  = rtnetlink_event,
2900 };
2901
2902
2903 static int __net_init rtnetlink_net_init(struct net *net)
2904 {
2905         struct sock *sk;
2906         struct netlink_kernel_cfg cfg = {
2907                 .groups         = RTNLGRP_MAX,
2908                 .input          = rtnetlink_rcv,
2909                 .cb_mutex       = &rtnl_mutex,
2910                 .flags          = NL_CFG_F_NONROOT_RECV,
2911         };
2912
2913         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2914         if (!sk)
2915                 return -ENOMEM;
2916         net->rtnl = sk;
2917         return 0;
2918 }
2919
2920 static void __net_exit rtnetlink_net_exit(struct net *net)
2921 {
2922         netlink_kernel_release(net->rtnl);
2923         net->rtnl = NULL;
2924 }
2925
2926 static struct pernet_operations rtnetlink_net_ops = {
2927         .init = rtnetlink_net_init,
2928         .exit = rtnetlink_net_exit,
2929 };
2930
2931 void __init rtnetlink_init(void)
2932 {
2933         if (register_pernet_subsys(&rtnetlink_net_ops))
2934                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2935
2936         register_netdevice_notifier(&rtnetlink_dev_notifier);
2937
2938         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2939                       rtnl_dump_ifinfo, rtnl_calcit);
2940         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2941         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2942         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2943
2944         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2945         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2946
2947         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2948         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2949         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2950
2951         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2952         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2953         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2954 }
2955