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_spoofchk)) +
 802                          nla_total_size(sizeof(struct ifla_vf_rate)));
 803                 return size;
 804         } else
 805                 return 0;
 806 }
 807
 808 static size_t rtnl_port_size(const struct net_device *dev,
 809                              u32 ext_filter_mask)
 810 {
 811         size_t port_size = nla_total_size(4)            /* PORT_VF */
 812                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 813                 + nla_total_size(sizeof(struct ifla_port_vsi))
 814                                                         /* PORT_VSI_TYPE */
 815                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 816                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 817                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 818                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 819         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 820         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 821                 + port_size;
 822         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 823                 + port_size;
 824
 825         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 826             !(ext_filter_mask & RTEXT_FILTER_VF))
 827                 return 0;
 828         if (dev_num_vf(dev->dev.parent))
 829                 return port_self_size + vf_ports_size +
 830                         vf_port_size * dev_num_vf(dev->dev.parent);
 831         else
 832                 return port_self_size;
 833 }
 834
 835 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 836                                      u32 ext_filter_mask)
 837 {
 838         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 839                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 840                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 841                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 842                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 843                + nla_total_size(sizeof(struct rtnl_link_stats))
 844                + nla_total_size(sizeof(struct rtnl_link_stats64))
 845                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 846                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 847                + nla_total_size(4) /* IFLA_TXQLEN */
 848                + nla_total_size(4) /* IFLA_WEIGHT */
 849                + nla_total_size(4) /* IFLA_MTU */
 850                + nla_total_size(4) /* IFLA_LINK */
 851                + nla_total_size(4) /* IFLA_MASTER */
 852                + nla_total_size(1) /* IFLA_CARRIER */
 853                + nla_total_size(4) /* IFLA_PROMISCUITY */
 854                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 855                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 856                + nla_total_size(1) /* IFLA_OPERSTATE */
 857                + nla_total_size(1) /* IFLA_LINKMODE */
 858                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 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             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1011                         atomic_read(&dev->carrier_changes)))
1012                 goto nla_put_failure;
1013
1014         if (1) {
1015                 struct rtnl_link_ifmap map = {
1016                         .mem_start   = dev->mem_start,
1017                         .mem_end     = dev->mem_end,
1018                         .base_addr   = dev->base_addr,
1019                         .irq         = dev->irq,
1020                         .dma         = dev->dma,
1021                         .port        = dev->if_port,
1022                 };
1023                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1024                         goto nla_put_failure;
1025         }
1026
1027         if (dev->addr_len) {
1028                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1029                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1030                         goto nla_put_failure;
1031         }
1032
1033         if (rtnl_phys_port_id_fill(skb, dev))
1034                 goto nla_put_failure;
1035
1036         attr = nla_reserve(skb, IFLA_STATS,
1037                         sizeof(struct rtnl_link_stats));
1038         if (attr == NULL)
1039                 goto nla_put_failure;
1040
1041         stats = dev_get_stats(dev, &temp);
1042         copy_rtnl_link_stats(nla_data(attr), stats);
1043
1044         attr = nla_reserve(skb, IFLA_STATS64,
1045                         sizeof(struct rtnl_link_stats64));
1046         if (attr == NULL)
1047                 goto nla_put_failure;
1048         copy_rtnl_link_stats64(nla_data(attr), stats);
1049
1050         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1051             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1052                 goto nla_put_failure;
1053
1054         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1055             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1056                 int i;
1057
1058                 struct nlattr *vfinfo, *vf;
1059                 int num_vfs = dev_num_vf(dev->dev.parent);
1060
1061                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1062                 if (!vfinfo)
1063                         goto nla_put_failure;
1064                 for (i = 0; i < num_vfs; i++) {
1065                         struct ifla_vf_info ivi;
1066                         struct ifla_vf_mac vf_mac;
1067                         struct ifla_vf_vlan vf_vlan;
1068                         struct ifla_vf_rate vf_rate;
1069                         struct ifla_vf_tx_rate vf_tx_rate;
1070                         struct ifla_vf_spoofchk vf_spoofchk;
1071                         struct ifla_vf_link_state vf_linkstate;
1072
1073                         /*
1074                          * Not all SR-IOV capable drivers support the
1075                          * spoofcheck query.  Preset to -1 so the user
1076                          * space tool can detect that the driver didn't
1077                          * report anything.
1078                          */
1079                         ivi.spoofchk = -1;
1080                         memset(ivi.mac, 0, sizeof(ivi.mac));
1081                         /* The default value for VF link state is "auto"
1082                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1083                          */
1084                         ivi.linkstate = 0;
1085                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1086                                 break;
1087                         vf_mac.vf =
1088                                 vf_vlan.vf =
1089                                 vf_rate.vf =
1090                                 vf_tx_rate.vf =
1091                                 vf_spoofchk.vf =
1092                                 vf_linkstate.vf = ivi.vf;
1093
1094                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1095                         vf_vlan.vlan = ivi.vlan;
1096                         vf_vlan.qos = ivi.qos;
1097                         vf_tx_rate.rate = ivi.max_tx_rate;
1098                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1099                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1100                         vf_spoofchk.setting = ivi.spoofchk;
1101                         vf_linkstate.link_state = ivi.linkstate;
1102                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1103                         if (!vf) {
1104                                 nla_nest_cancel(skb, vfinfo);
1105                                 goto nla_put_failure;
1106                         }
1107                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1108                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1109                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1110                                     &vf_rate) ||
1111                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1112                                     &vf_tx_rate) ||
1113                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1114                                     &vf_spoofchk) ||
1115                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1116                                     &vf_linkstate))
1117                                 goto nla_put_failure;
1118                         nla_nest_end(skb, vf);
1119                 }
1120                 nla_nest_end(skb, vfinfo);
1121         }
1122
1123         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1124                 goto nla_put_failure;
1125
1126         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1127                 if (rtnl_link_fill(skb, dev) < 0)
1128                         goto nla_put_failure;
1129         }
1130
1131         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1132                 goto nla_put_failure;
1133
1134         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1135                 if (af_ops->fill_link_af) {
1136                         struct nlattr *af;
1137                         int err;
1138
1139                         if (!(af = nla_nest_start(skb, af_ops->family)))
1140                                 goto nla_put_failure;
1141
1142                         err = af_ops->fill_link_af(skb, dev);
1143
1144                         /*
1145                          * Caller may return ENODATA to indicate that there
1146                          * was no data to be dumped. This is not an error, it
1147                          * means we should trim the attribute header and
1148                          * continue.
1149                          */
1150                         if (err == -ENODATA)
1151                                 nla_nest_cancel(skb, af);
1152                         else if (err < 0)
1153                                 goto nla_put_failure;
1154
1155                         nla_nest_end(skb, af);
1156                 }
1157         }
1158
1159         nla_nest_end(skb, af_spec);
1160
1161         return nlmsg_end(skb, nlh);
1162
1163 nla_put_failure:
1164         nlmsg_cancel(skb, nlh);
1165         return -EMSGSIZE;
1166 }
1167
1168 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1169         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1170         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1171         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1172         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1173         [IFLA_MTU]              = { .type = NLA_U32 },
1174         [IFLA_LINK]             = { .type = NLA_U32 },
1175         [IFLA_MASTER]           = { .type = NLA_U32 },
1176         [IFLA_CARRIER]          = { .type = NLA_U8 },
1177         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1178         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1179         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1180         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1181         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1182         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1183         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1184         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1185         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1186         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1187         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1188         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1189         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1190         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1191         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1192         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1193         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1194         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1195 };
1196
1197 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1198         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1199         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1200         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1201         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1202 };
1203
1204 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1205         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1206 };
1207
1208 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1209         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1210                                     .len = sizeof(struct ifla_vf_mac) },
1211         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1212                                     .len = sizeof(struct ifla_vf_vlan) },
1213         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1214                                     .len = sizeof(struct ifla_vf_tx_rate) },
1215         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1216                                     .len = sizeof(struct ifla_vf_spoofchk) },
1217         [IFLA_VF_RATE]          = { .type = NLA_BINARY,
1218                                     .len = sizeof(struct ifla_vf_rate) },
1219         [IFLA_VF_LINK_STATE]    = { .type = NLA_BINARY,
1220                                     .len = sizeof(struct ifla_vf_link_state) },
1221 };
1222
1223 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1224         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1225         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1226                                     .len = PORT_PROFILE_MAX },
1227         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1228                                     .len = sizeof(struct ifla_port_vsi)},
1229         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1230                                       .len = PORT_UUID_MAX },
1231         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1232                                     .len = PORT_UUID_MAX },
1233         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1234         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1235 };
1236
1237 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1238 {
1239         struct net *net = sock_net(skb->sk);
1240         int h, s_h;
1241         int idx = 0, s_idx;
1242         struct net_device *dev;
1243         struct hlist_head *head;
1244         struct nlattr *tb[IFLA_MAX+1];
1245         u32 ext_filter_mask = 0;
1246         int err;
1247         int hdrlen;
1248
1249         s_h = cb->args[0];
1250         s_idx = cb->args[1];
1251
1252         rcu_read_lock();
1253         cb->seq = net->dev_base_seq;
1254
1255         /* A hack to preserve kernel<->userspace interface.
1256          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1257          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1258          * what iproute2 < v3.9.0 used.
1259          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1260          * attribute, its netlink message is shorter than struct ifinfomsg.
1261          */
1262         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1263                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1264
1265         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1266
1267                 if (tb[IFLA_EXT_MASK])
1268                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1269         }
1270
1271         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1272                 idx = 0;
1273                 head = &net->dev_index_head[h];
1274                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1275                         if (idx < s_idx)
1276                                 goto cont;
1277                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1278                                                NETLINK_CB(cb->skb).portid,
1279                                                cb->nlh->nlmsg_seq, 0,
1280                                                NLM_F_MULTI,
1281                                                ext_filter_mask);
1282                         /* If we ran out of room on the first message,
1283                          * we're in trouble
1284                          */
1285                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1286
1287                         if (err <= 0)
1288                                 goto out;
1289
1290                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1291 cont:
1292                         idx++;
1293                 }
1294         }
1295 out:
1296         rcu_read_unlock();
1297         cb->args[1] = idx;
1298         cb->args[0] = h;
1299
1300         return skb->len;
1301 }
1302
1303 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1304 {
1305         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1306 }
1307 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1308
1309 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1310 {
1311         struct net *net;
1312         /* Examine the link attributes and figure out which
1313          * network namespace we are talking about.
1314          */
1315         if (tb[IFLA_NET_NS_PID])
1316                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1317         else if (tb[IFLA_NET_NS_FD])
1318                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1319         else
1320                 net = get_net(src_net);
1321         return net;
1322 }
1323 EXPORT_SYMBOL(rtnl_link_get_net);
1324
1325 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1326 {
1327         if (dev) {
1328                 if (tb[IFLA_ADDRESS] &&
1329                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1330                         return -EINVAL;
1331
1332                 if (tb[IFLA_BROADCAST] &&
1333                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1334                         return -EINVAL;
1335         }
1336
1337         if (tb[IFLA_AF_SPEC]) {
1338                 struct nlattr *af;
1339                 int rem, err;
1340
1341                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1342                         const struct rtnl_af_ops *af_ops;
1343
1344                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1345                                 return -EAFNOSUPPORT;
1346
1347                         if (!af_ops->set_link_af)
1348                                 return -EOPNOTSUPP;
1349
1350                         if (af_ops->validate_link_af) {
1351                                 err = af_ops->validate_link_af(dev, af);
1352                                 if (err < 0)
1353                                         return err;
1354                         }
1355                 }
1356         }
1357
1358         return 0;
1359 }
1360
1361 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1362 {
1363         int rem, err = -EINVAL;
1364         struct nlattr *vf;
1365         const struct net_device_ops *ops = dev->netdev_ops;
1366
1367         nla_for_each_nested(vf, attr, rem) {
1368                 switch (nla_type(vf)) {
1369                 case IFLA_VF_MAC: {
1370                         struct ifla_vf_mac *ivm;
1371                         ivm = nla_data(vf);
1372                         err = -EOPNOTSUPP;
1373                         if (ops->ndo_set_vf_mac)
1374                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1375                                                           ivm->mac);
1376                         break;
1377                 }
1378                 case IFLA_VF_VLAN: {
1379                         struct ifla_vf_vlan *ivv;
1380                         ivv = nla_data(vf);
1381                         err = -EOPNOTSUPP;
1382                         if (ops->ndo_set_vf_vlan)
1383                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1384                                                            ivv->vlan,
1385                                                            ivv->qos);
1386                         break;
1387                 }
1388                 case IFLA_VF_TX_RATE: {
1389                         struct ifla_vf_tx_rate *ivt;
1390                         struct ifla_vf_info ivf;
1391                         ivt = nla_data(vf);
1392                         err = -EOPNOTSUPP;
1393                         if (ops->ndo_get_vf_config)
1394                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1395                                                              &ivf);
1396                         if (err)
1397                                 break;
1398                         err = -EOPNOTSUPP;
1399                         if (ops->ndo_set_vf_rate)
1400                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1401                                                            ivf.min_tx_rate,
1402                                                            ivt->rate);
1403                         break;
1404                 }
1405                 case IFLA_VF_RATE: {
1406                         struct ifla_vf_rate *ivt;
1407                         ivt = nla_data(vf);
1408                         err = -EOPNOTSUPP;
1409                         if (ops->ndo_set_vf_rate)
1410                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1411                                                            ivt->min_tx_rate,
1412                                                            ivt->max_tx_rate);
1413                         break;
1414                 }
1415                 case IFLA_VF_SPOOFCHK: {
1416                         struct ifla_vf_spoofchk *ivs;
1417                         ivs = nla_data(vf);
1418                         err = -EOPNOTSUPP;
1419                         if (ops->ndo_set_vf_spoofchk)
1420                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1421                                                                ivs->setting);
1422                         break;
1423                 }
1424                 case IFLA_VF_LINK_STATE: {
1425                         struct ifla_vf_link_state *ivl;
1426                         ivl = nla_data(vf);
1427                         err = -EOPNOTSUPP;
1428                         if (ops->ndo_set_vf_link_state)
1429                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1430                                                                  ivl->link_state);
1431                         break;
1432                 }
1433                 default:
1434                         err = -EINVAL;
1435                         break;
1436                 }
1437                 if (err)
1438                         break;
1439         }
1440         return err;
1441 }
1442
1443 static int do_set_master(struct net_device *dev, int ifindex)
1444 {
1445         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1446         const struct net_device_ops *ops;
1447         int err;
1448
1449         if (upper_dev) {
1450                 if (upper_dev->ifindex == ifindex)
1451                         return 0;
1452                 ops = upper_dev->netdev_ops;
1453                 if (ops->ndo_del_slave) {
1454                         err = ops->ndo_del_slave(upper_dev, dev);
1455                         if (err)
1456                                 return err;
1457                 } else {
1458                         return -EOPNOTSUPP;
1459                 }
1460         }
1461
1462         if (ifindex) {
1463                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1464                 if (!upper_dev)
1465                         return -EINVAL;
1466                 ops = upper_dev->netdev_ops;
1467                 if (ops->ndo_add_slave) {
1468                         err = ops->ndo_add_slave(upper_dev, dev);
1469                         if (err)
1470                                 return err;
1471                 } else {
1472                         return -EOPNOTSUPP;
1473                 }
1474         }
1475         return 0;
1476 }
1477
1478 static int do_setlink(const struct sk_buff *skb,
1479                       struct net_device *dev, struct ifinfomsg *ifm,
1480                       struct nlattr **tb, char *ifname, int modified)
1481 {
1482         const struct net_device_ops *ops = dev->netdev_ops;
1483         int err;
1484
1485         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1486                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1487                 if (IS_ERR(net)) {
1488                         err = PTR_ERR(net);
1489                         goto errout;
1490                 }
1491                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1492                         err = -EPERM;
1493                         goto errout;
1494                 }
1495                 err = dev_change_net_namespace(dev, net, ifname);
1496                 put_net(net);
1497                 if (err)
1498                         goto errout;
1499                 modified = 1;
1500         }
1501
1502         if (tb[IFLA_MAP]) {
1503                 struct rtnl_link_ifmap *u_map;
1504                 struct ifmap k_map;
1505
1506                 if (!ops->ndo_set_config) {
1507                         err = -EOPNOTSUPP;
1508                         goto errout;
1509                 }
1510
1511                 if (!netif_device_present(dev)) {
1512                         err = -ENODEV;
1513                         goto errout;
1514                 }
1515
1516                 u_map = nla_data(tb[IFLA_MAP]);
1517                 k_map.mem_start = (unsigned long) u_map->mem_start;
1518                 k_map.mem_end = (unsigned long) u_map->mem_end;
1519                 k_map.base_addr = (unsigned short) u_map->base_addr;
1520                 k_map.irq = (unsigned char) u_map->irq;
1521                 k_map.dma = (unsigned char) u_map->dma;
1522                 k_map.port = (unsigned char) u_map->port;
1523
1524                 err = ops->ndo_set_config(dev, &k_map);
1525                 if (err < 0)
1526                         goto errout;
1527
1528                 modified = 1;
1529         }
1530
1531         if (tb[IFLA_ADDRESS]) {
1532                 struct sockaddr *sa;
1533                 int len;
1534
1535                 len = sizeof(sa_family_t) + dev->addr_len;
1536                 sa = kmalloc(len, GFP_KERNEL);
1537                 if (!sa) {
1538                         err = -ENOMEM;
1539                         goto errout;
1540                 }
1541                 sa->sa_family = dev->type;
1542                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1543                        dev->addr_len);
1544                 err = dev_set_mac_address(dev, sa);
1545                 kfree(sa);
1546                 if (err)
1547                         goto errout;
1548                 modified = 1;
1549         }
1550
1551         if (tb[IFLA_MTU]) {
1552                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1553                 if (err < 0)
1554                         goto errout;
1555                 modified = 1;
1556         }
1557
1558         if (tb[IFLA_GROUP]) {
1559                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1560                 modified = 1;
1561         }
1562
1563         /*
1564          * Interface selected by interface index but interface
1565          * name provided implies that a name change has been
1566          * requested.
1567          */
1568         if (ifm->ifi_index > 0 && ifname[0]) {
1569                 err = dev_change_name(dev, ifname);
1570                 if (err < 0)
1571                         goto errout;
1572                 modified = 1;
1573         }
1574
1575         if (tb[IFLA_IFALIAS]) {
1576                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1577                                     nla_len(tb[IFLA_IFALIAS]));
1578                 if (err < 0)
1579                         goto errout;
1580                 modified = 1;
1581         }
1582
1583         if (tb[IFLA_BROADCAST]) {
1584                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1585                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1586         }
1587
1588         if (ifm->ifi_flags || ifm->ifi_change) {
1589                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1590                 if (err < 0)
1591                         goto errout;
1592         }
1593
1594         if (tb[IFLA_MASTER]) {
1595                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1596                 if (err)
1597                         goto errout;
1598                 modified = 1;
1599         }
1600
1601         if (tb[IFLA_CARRIER]) {
1602                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1603                 if (err)
1604                         goto errout;
1605                 modified = 1;
1606         }
1607
1608         if (tb[IFLA_TXQLEN])
1609                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1610
1611         if (tb[IFLA_OPERSTATE])
1612                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1613
1614         if (tb[IFLA_LINKMODE]) {
1615                 write_lock_bh(&dev_base_lock);
1616                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1617                 write_unlock_bh(&dev_base_lock);
1618         }
1619
1620         if (tb[IFLA_VFINFO_LIST]) {
1621                 struct nlattr *attr;
1622                 int rem;
1623                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1624                         if (nla_type(attr) != IFLA_VF_INFO) {
1625                                 err = -EINVAL;
1626                                 goto errout;
1627                         }
1628                         err = do_setvfinfo(dev, attr);
1629                         if (err < 0)
1630                                 goto errout;
1631                         modified = 1;
1632                 }
1633         }
1634         err = 0;
1635
1636         if (tb[IFLA_VF_PORTS]) {
1637                 struct nlattr *port[IFLA_PORT_MAX+1];
1638                 struct nlattr *attr;
1639                 int vf;
1640                 int rem;
1641
1642                 err = -EOPNOTSUPP;
1643                 if (!ops->ndo_set_vf_port)
1644                         goto errout;
1645
1646                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1647                         if (nla_type(attr) != IFLA_VF_PORT)
1648                                 continue;
1649                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1650                                 attr, ifla_port_policy);
1651                         if (err < 0)
1652                                 goto errout;
1653                         if (!port[IFLA_PORT_VF]) {
1654                                 err = -EOPNOTSUPP;
1655                                 goto errout;
1656                         }
1657                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1658                         err = ops->ndo_set_vf_port(dev, vf, port);
1659                         if (err < 0)
1660                                 goto errout;
1661                         modified = 1;
1662                 }
1663         }
1664         err = 0;
1665
1666         if (tb[IFLA_PORT_SELF]) {
1667                 struct nlattr *port[IFLA_PORT_MAX+1];
1668
1669                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1670                         tb[IFLA_PORT_SELF], ifla_port_policy);
1671                 if (err < 0)
1672                         goto errout;
1673
1674                 err = -EOPNOTSUPP;
1675                 if (ops->ndo_set_vf_port)
1676                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1677                 if (err < 0)
1678                         goto errout;
1679                 modified = 1;
1680         }
1681
1682         if (tb[IFLA_AF_SPEC]) {
1683                 struct nlattr *af;
1684                 int rem;
1685
1686                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1687                         const struct rtnl_af_ops *af_ops;
1688
1689                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1690                                 BUG();
1691
1692                         err = af_ops->set_link_af(dev, af);
1693                         if (err < 0)
1694                                 goto errout;
1695
1696                         modified = 1;
1697                 }
1698         }
1699         err = 0;
1700
1701 errout:
1702         if (err < 0 && modified)
1703                 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",
1704                                      dev->name);
1705
1706         return err;
1707 }
1708
1709 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1710 {
1711         struct net *net = sock_net(skb->sk);
1712         struct ifinfomsg *ifm;
1713         struct net_device *dev;
1714         int err;
1715         struct nlattr *tb[IFLA_MAX+1];
1716         char ifname[IFNAMSIZ];
1717
1718         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1719         if (err < 0)
1720                 goto errout;
1721
1722         if (tb[IFLA_IFNAME])
1723                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1724         else
1725                 ifname[0] = '\0';
1726
1727         err = -EINVAL;
1728         ifm = nlmsg_data(nlh);
1729         if (ifm->ifi_index > 0)
1730                 dev = __dev_get_by_index(net, ifm->ifi_index);
1731         else if (tb[IFLA_IFNAME])
1732                 dev = __dev_get_by_name(net, ifname);
1733         else
1734                 goto errout;
1735
1736         if (dev == NULL) {
1737                 err = -ENODEV;
1738                 goto errout;
1739         }
1740
1741         err = validate_linkmsg(dev, tb);
1742         if (err < 0)
1743                 goto errout;
1744
1745         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1746 errout:
1747         return err;
1748 }
1749
1750 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1751 {
1752         struct net *net = sock_net(skb->sk);
1753         const struct rtnl_link_ops *ops;
1754         struct net_device *dev;
1755         struct ifinfomsg *ifm;
1756         char ifname[IFNAMSIZ];
1757         struct nlattr *tb[IFLA_MAX+1];
1758         int err;
1759         LIST_HEAD(list_kill);
1760
1761         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1762         if (err < 0)
1763                 return err;
1764
1765         if (tb[IFLA_IFNAME])
1766                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1767
1768         ifm = nlmsg_data(nlh);
1769         if (ifm->ifi_index > 0)
1770                 dev = __dev_get_by_index(net, ifm->ifi_index);
1771         else if (tb[IFLA_IFNAME])
1772                 dev = __dev_get_by_name(net, ifname);
1773         else
1774                 return -EINVAL;
1775
1776         if (!dev)
1777                 return -ENODEV;
1778
1779         ops = dev->rtnl_link_ops;
1780         if (!ops)
1781                 return -EOPNOTSUPP;
1782
1783         ops->dellink(dev, &list_kill);
1784         unregister_netdevice_many(&list_kill);
1785         return 0;
1786 }
1787
1788 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1789 {
1790         unsigned int old_flags;
1791         int err;
1792
1793         old_flags = dev->flags;
1794         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1795                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1796                 if (err < 0)
1797                         return err;
1798         }
1799
1800         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1801
1802         __dev_notify_flags(dev, old_flags, ~0U);
1803         return 0;
1804 }
1805 EXPORT_SYMBOL(rtnl_configure_link);
1806
1807 struct net_device *rtnl_create_link(struct net *net,
1808         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1809 {
1810         int err;
1811         struct net_device *dev;
1812         unsigned int num_tx_queues = 1;
1813         unsigned int num_rx_queues = 1;
1814
1815         if (tb[IFLA_NUM_TX_QUEUES])
1816                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1817         else if (ops->get_num_tx_queues)
1818                 num_tx_queues = ops->get_num_tx_queues();
1819
1820         if (tb[IFLA_NUM_RX_QUEUES])
1821                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1822         else if (ops->get_num_rx_queues)
1823                 num_rx_queues = ops->get_num_rx_queues();
1824
1825         err = -ENOMEM;
1826         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1827                                num_tx_queues, num_rx_queues);
1828         if (!dev)
1829                 goto err;
1830
1831         dev_net_set(dev, net);
1832         dev->rtnl_link_ops = ops;
1833         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1834
1835         if (tb[IFLA_MTU])
1836                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1837         if (tb[IFLA_ADDRESS]) {
1838                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1839                                 nla_len(tb[IFLA_ADDRESS]));
1840                 dev->addr_assign_type = NET_ADDR_SET;
1841         }
1842         if (tb[IFLA_BROADCAST])
1843                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1844                                 nla_len(tb[IFLA_BROADCAST]));
1845         if (tb[IFLA_TXQLEN])
1846                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1847         if (tb[IFLA_OPERSTATE])
1848                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1849         if (tb[IFLA_LINKMODE])
1850                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1851         if (tb[IFLA_GROUP])
1852                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1853
1854         return dev;
1855
1856 err:
1857         return ERR_PTR(err);
1858 }
1859 EXPORT_SYMBOL(rtnl_create_link);
1860
1861 static int rtnl_group_changelink(const struct sk_buff *skb,
1862                 struct net *net, int group,
1863                 struct ifinfomsg *ifm,
1864                 struct nlattr **tb)
1865 {
1866         struct net_device *dev;
1867         int err;
1868
1869         for_each_netdev(net, dev) {
1870                 if (dev->group == group) {
1871                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1872                         if (err < 0)
1873                                 return err;
1874                 }
1875         }
1876
1877         return 0;
1878 }
1879
1880 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1881 {
1882         struct net *net = sock_net(skb->sk);
1883         const struct rtnl_link_ops *ops;
1884         const struct rtnl_link_ops *m_ops = NULL;
1885         struct net_device *dev;
1886         struct net_device *master_dev = NULL;
1887         struct ifinfomsg *ifm;
1888         char kind[MODULE_NAME_LEN];
1889         char ifname[IFNAMSIZ];
1890         struct nlattr *tb[IFLA_MAX+1];
1891         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1892         int err;
1893
1894 #ifdef CONFIG_MODULES
1895 replay:
1896 #endif
1897         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1898         if (err < 0)
1899                 return err;
1900
1901         if (tb[IFLA_IFNAME])
1902                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1903         else
1904                 ifname[0] = '\0';
1905
1906         ifm = nlmsg_data(nlh);
1907         if (ifm->ifi_index > 0)
1908                 dev = __dev_get_by_index(net, ifm->ifi_index);
1909         else {
1910                 if (ifname[0])
1911                         dev = __dev_get_by_name(net, ifname);
1912                 else
1913                         dev = NULL;
1914         }
1915
1916         if (dev) {
1917                 master_dev = netdev_master_upper_dev_get(dev);
1918                 if (master_dev)
1919                         m_ops = master_dev->rtnl_link_ops;
1920         }
1921
1922         err = validate_linkmsg(dev, tb);
1923         if (err < 0)
1924                 return err;
1925
1926         if (tb[IFLA_LINKINFO]) {
1927                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1928                                        tb[IFLA_LINKINFO], ifla_info_policy);
1929                 if (err < 0)
1930                         return err;
1931         } else
1932                 memset(linkinfo, 0, sizeof(linkinfo));
1933
1934         if (linkinfo[IFLA_INFO_KIND]) {
1935                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1936                 ops = rtnl_link_ops_get(kind);
1937         } else {
1938                 kind[0] = '\0';
1939                 ops = NULL;
1940         }
1941
1942         if (1) {
1943                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1944                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1945                 struct nlattr **data = NULL;
1946                 struct nlattr **slave_data = NULL;
1947                 struct net *dest_net;
1948
1949                 if (ops) {
1950                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1951                                 err = nla_parse_nested(attr, ops->maxtype,
1952                                                        linkinfo[IFLA_INFO_DATA],
1953                                                        ops->policy);
1954                                 if (err < 0)
1955                                         return err;
1956                                 data = attr;
1957                         }
1958                         if (ops->validate) {
1959                                 err = ops->validate(tb, data);
1960                                 if (err < 0)
1961                                         return err;
1962                         }
1963                 }
1964
1965                 if (m_ops) {
1966                         if (m_ops->slave_maxtype &&
1967                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
1968                                 err = nla_parse_nested(slave_attr,
1969                                                        m_ops->slave_maxtype,
1970                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
1971                                                        m_ops->slave_policy);
1972                                 if (err < 0)
1973                                         return err;
1974                                 slave_data = slave_attr;
1975                         }
1976                         if (m_ops->slave_validate) {
1977                                 err = m_ops->slave_validate(tb, slave_data);
1978                                 if (err < 0)
1979                                         return err;
1980                         }
1981                 }
1982
1983                 if (dev) {
1984                         int modified = 0;
1985
1986                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1987                                 return -EEXIST;
1988                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1989                                 return -EOPNOTSUPP;
1990
1991                         if (linkinfo[IFLA_INFO_DATA]) {
1992                                 if (!ops || ops != dev->rtnl_link_ops ||
1993                                     !ops->changelink)
1994                                         return -EOPNOTSUPP;
1995
1996                                 err = ops->changelink(dev, tb, data);
1997                                 if (err < 0)
1998                                         return err;
1999                                 modified = 1;
2000                         }
2001
2002                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2003                                 if (!m_ops || !m_ops->slave_changelink)
2004                                         return -EOPNOTSUPP;
2005
2006                                 err = m_ops->slave_changelink(master_dev, dev,
2007                                                               tb, slave_data);
2008                                 if (err < 0)
2009                                         return err;
2010                                 modified = 1;
2011                         }
2012
2013                         return do_setlink(skb, dev, ifm, tb, ifname, modified);
2014                 }
2015
2016                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2017                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2018                                 return rtnl_group_changelink(skb, net,
2019                                                 nla_get_u32(tb[IFLA_GROUP]),
2020                                                 ifm, tb);
2021                         return -ENODEV;
2022                 }
2023
2024                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2025                         return -EOPNOTSUPP;
2026
2027                 if (!ops) {
2028 #ifdef CONFIG_MODULES
2029                         if (kind[0]) {
2030                                 __rtnl_unlock();
2031                                 request_module("rtnl-link-%s", kind);
2032                                 rtnl_lock();
2033                                 ops = rtnl_link_ops_get(kind);
2034                                 if (ops)
2035                                         goto replay;
2036                         }
2037 #endif
2038                         return -EOPNOTSUPP;
2039                 }
2040
2041                 if (!ifname[0])
2042                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2043
2044                 dest_net = rtnl_link_get_net(net, tb);
2045                 if (IS_ERR(dest_net))
2046                         return PTR_ERR(dest_net);
2047
2048                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
2049                 if (IS_ERR(dev)) {
2050                         err = PTR_ERR(dev);
2051                         goto out;
2052                 }
2053
2054                 dev->ifindex = ifm->ifi_index;
2055
2056                 if (ops->newlink) {
2057                         err = ops->newlink(net, dev, tb, data);
2058                         /* Drivers should call free_netdev() in ->destructor
2059                          * and unregister it on failure after registration
2060                          * so that device could be finally freed in rtnl_unlock.
2061                          */
2062                         if (err < 0) {
2063                                 /* If device is not registered at all, free it now */
2064                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2065                                         free_netdev(dev);
2066                                 goto out;
2067                         }
2068                 } else {
2069                         err = register_netdevice(dev);
2070                         if (err < 0) {
2071                                 free_netdev(dev);
2072                                 goto out;
2073                         }
2074                 }
2075                 err = rtnl_configure_link(dev, ifm);
2076                 if (err < 0)
2077                         unregister_netdevice(dev);
2078 out:
2079                 put_net(dest_net);
2080                 return err;
2081         }
2082 }
2083
2084 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2085 {
2086         struct net *net = sock_net(skb->sk);
2087         struct ifinfomsg *ifm;
2088         char ifname[IFNAMSIZ];
2089         struct nlattr *tb[IFLA_MAX+1];
2090         struct net_device *dev = NULL;
2091         struct sk_buff *nskb;
2092         int err;
2093         u32 ext_filter_mask = 0;
2094
2095         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2096         if (err < 0)
2097                 return err;
2098
2099         if (tb[IFLA_IFNAME])
2100                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2101
2102         if (tb[IFLA_EXT_MASK])
2103                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2104
2105         ifm = nlmsg_data(nlh);
2106         if (ifm->ifi_index > 0)
2107                 dev = __dev_get_by_index(net, ifm->ifi_index);
2108         else if (tb[IFLA_IFNAME])
2109                 dev = __dev_get_by_name(net, ifname);
2110         else
2111                 return -EINVAL;
2112
2113         if (dev == NULL)
2114                 return -ENODEV;
2115
2116         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2117         if (nskb == NULL)
2118                 return -ENOBUFS;
2119
2120         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2121                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2122         if (err < 0) {
2123                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2124                 WARN_ON(err == -EMSGSIZE);
2125                 kfree_skb(nskb);
2126         } else
2127                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2128
2129         return err;
2130 }
2131
2132 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2133 {
2134         struct net *net = sock_net(skb->sk);
2135         struct net_device *dev;
2136         struct nlattr *tb[IFLA_MAX+1];
2137         u32 ext_filter_mask = 0;
2138         u16 min_ifinfo_dump_size = 0;
2139         int hdrlen;
2140
2141         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2142         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2143                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2144
2145         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2146                 if (tb[IFLA_EXT_MASK])
2147                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2148         }
2149
2150         if (!ext_filter_mask)
2151                 return NLMSG_GOODSIZE;
2152         /*
2153          * traverse the list of net devices and compute the minimum
2154          * buffer size based upon the filter mask.
2155          */
2156         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2157                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2158                                              if_nlmsg_size(dev,
2159                                                            ext_filter_mask));
2160         }
2161
2162         return min_ifinfo_dump_size;
2163 }
2164
2165 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2166 {
2167         int idx;
2168         int s_idx = cb->family;
2169
2170         if (s_idx == 0)
2171                 s_idx = 1;
2172         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2173                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2174                 if (idx < s_idx || idx == PF_PACKET)
2175                         continue;
2176                 if (rtnl_msg_handlers[idx] == NULL ||
2177                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2178                         continue;
2179                 if (idx > s_idx) {
2180                         memset(&cb->args[0], 0, sizeof(cb->args));
2181                         cb->prev_seq = 0;
2182                         cb->seq = 0;
2183                 }
2184                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2185                         break;
2186         }
2187         cb->family = idx;
2188
2189         return skb->len;
2190 }
2191
2192 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2193                   gfp_t flags)
2194 {
2195         struct net *net = dev_net(dev);
2196         struct sk_buff *skb;
2197         int err = -ENOBUFS;
2198         size_t if_info_size;
2199
2200         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2201         if (skb == NULL)
2202                 goto errout;
2203
2204         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2205         if (err < 0) {
2206                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2207                 WARN_ON(err == -EMSGSIZE);
2208                 kfree_skb(skb);
2209                 goto errout;
2210         }
2211         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2212         return;
2213 errout:
2214         if (err < 0)
2215                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2216 }
2217 EXPORT_SYMBOL(rtmsg_ifinfo);
2218
2219 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2220                                    struct net_device *dev,
2221                                    u8 *addr, u32 pid, u32 seq,
2222                                    int type, unsigned int flags,
2223                                    int nlflags)
2224 {
2225         struct nlmsghdr *nlh;
2226         struct ndmsg *ndm;
2227
2228         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2229         if (!nlh)
2230                 return -EMSGSIZE;
2231
2232         ndm = nlmsg_data(nlh);
2233         ndm->ndm_family  = AF_BRIDGE;
2234         ndm->ndm_pad1    = 0;
2235         ndm->ndm_pad2    = 0;
2236         ndm->ndm_flags   = flags;
2237         ndm->ndm_type    = 0;
2238         ndm->ndm_ifindex = dev->ifindex;
2239         ndm->ndm_state   = NUD_PERMANENT;
2240
2241         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2242                 goto nla_put_failure;
2243
2244         return nlmsg_end(skb, nlh);
2245
2246 nla_put_failure:
2247         nlmsg_cancel(skb, nlh);
2248         return -EMSGSIZE;
2249 }
2250
2251 static inline size_t rtnl_fdb_nlmsg_size(void)
2252 {
2253         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2254 }
2255
2256 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2257 {
2258         struct net *net = dev_net(dev);
2259         struct sk_buff *skb;
2260         int err = -ENOBUFS;
2261
2262         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2263         if (!skb)
2264                 goto errout;
2265
2266         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2267         if (err < 0) {
2268                 kfree_skb(skb);
2269                 goto errout;
2270         }
2271
2272         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2273         return;
2274 errout:
2275         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2276 }
2277
2278 /**
2279  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2280  */
2281 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2282                      struct nlattr *tb[],
2283                      struct net_device *dev,
2284                      const unsigned char *addr,
2285                      u16 flags)
2286 {
2287         int err = -EINVAL;
2288
2289         /* If aging addresses are supported device will need to
2290          * implement its own handler for this.
2291          */
2292         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2293                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2294                 return err;
2295         }
2296
2297         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2298                 err = dev_uc_add_excl(dev, addr);
2299         else if (is_multicast_ether_addr(addr))
2300                 err = dev_mc_add_excl(dev, addr);
2301
2302         /* Only return duplicate errors if NLM_F_EXCL is set */
2303         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2304                 err = 0;
2305
2306         return err;
2307 }
2308 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2309
2310 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2311 {
2312         struct net *net = sock_net(skb->sk);
2313         struct ndmsg *ndm;
2314         struct nlattr *tb[NDA_MAX+1];
2315         struct net_device *dev;
2316         u8 *addr;
2317         int err;
2318
2319         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2320         if (err < 0)
2321                 return err;
2322
2323         ndm = nlmsg_data(nlh);
2324         if (ndm->ndm_ifindex == 0) {
2325                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2326                 return -EINVAL;
2327         }
2328
2329         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2330         if (dev == NULL) {
2331                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2332                 return -ENODEV;
2333         }
2334
2335         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2336                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2337                 return -EINVAL;
2338         }
2339
2340         addr = nla_data(tb[NDA_LLADDR]);
2341
2342         err = -EOPNOTSUPP;
2343
2344         /* Support fdb on master device the net/bridge default case */
2345         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2346             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2347                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2348                 const struct net_device_ops *ops = br_dev->netdev_ops;
2349
2350                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2351                 if (err)
2352                         goto out;
2353                 else
2354                         ndm->ndm_flags &= ~NTF_MASTER;
2355         }
2356
2357         /* Embedded bridge, macvlan, and any other device support */
2358         if ((ndm->ndm_flags & NTF_SELF)) {
2359                 if (dev->netdev_ops->ndo_fdb_add)
2360                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2361                                                            nlh->nlmsg_flags);
2362                 else
2363                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2364                                                nlh->nlmsg_flags);
2365
2366                 if (!err) {
2367                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2368                         ndm->ndm_flags &= ~NTF_SELF;
2369                 }
2370         }
2371 out:
2372         return err;
2373 }
2374
2375 /**
2376  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2377  */
2378 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2379                      struct nlattr *tb[],
2380                      struct net_device *dev,
2381                      const unsigned char *addr)
2382 {
2383         int err = -EOPNOTSUPP;
2384
2385         /* If aging addresses are supported device will need to
2386          * implement its own handler for this.
2387          */
2388         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2389                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2390                 return -EINVAL;
2391         }
2392
2393         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2394                 err = dev_uc_del(dev, addr);
2395         else if (is_multicast_ether_addr(addr))
2396                 err = dev_mc_del(dev, addr);
2397         else
2398                 err = -EINVAL;
2399
2400         return err;
2401 }
2402 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2403
2404 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2405 {
2406         struct net *net = sock_net(skb->sk);
2407         struct ndmsg *ndm;
2408         struct nlattr *tb[NDA_MAX+1];
2409         struct net_device *dev;
2410         int err = -EINVAL;
2411         __u8 *addr;
2412
2413         if (!netlink_capable(skb, CAP_NET_ADMIN))
2414                 return -EPERM;
2415
2416         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2417         if (err < 0)
2418                 return err;
2419
2420         ndm = nlmsg_data(nlh);
2421         if (ndm->ndm_ifindex == 0) {
2422                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2423                 return -EINVAL;
2424         }
2425
2426         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2427         if (dev == NULL) {
2428                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2429                 return -ENODEV;
2430         }
2431
2432         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2433                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2434                 return -EINVAL;
2435         }
2436
2437         addr = nla_data(tb[NDA_LLADDR]);
2438
2439         err = -EOPNOTSUPP;
2440
2441         /* Support fdb on master device the net/bridge default case */
2442         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2443             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2444                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2445                 const struct net_device_ops *ops = br_dev->netdev_ops;
2446
2447                 if (ops->ndo_fdb_del)
2448                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2449
2450                 if (err)
2451                         goto out;
2452                 else
2453                         ndm->ndm_flags &= ~NTF_MASTER;
2454         }
2455
2456         /* Embedded bridge, macvlan, and any other device support */
2457         if (ndm->ndm_flags & NTF_SELF) {
2458                 if (dev->netdev_ops->ndo_fdb_del)
2459                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2460                 else
2461                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2462
2463                 if (!err) {
2464                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2465                         ndm->ndm_flags &= ~NTF_SELF;
2466                 }
2467         }
2468 out:
2469         return err;
2470 }
2471
2472 static int nlmsg_populate_fdb(struct sk_buff *skb,
2473                               struct netlink_callback *cb,
2474                               struct net_device *dev,
2475                               int *idx,
2476                               struct netdev_hw_addr_list *list)
2477 {
2478         struct netdev_hw_addr *ha;
2479         int err;
2480         u32 portid, seq;
2481
2482         portid = NETLINK_CB(cb->skb).portid;
2483         seq = cb->nlh->nlmsg_seq;
2484
2485         list_for_each_entry(ha, &list->list, list) {
2486                 if (*idx < cb->args[0])
2487                         goto skip;
2488
2489                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2490                                               portid, seq,
2491                                               RTM_NEWNEIGH, NTF_SELF,
2492                                               NLM_F_MULTI);
2493                 if (err < 0)
2494                         return err;
2495 skip:
2496                 *idx += 1;
2497         }
2498         return 0;
2499 }
2500
2501 /**
2502  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2503  * @nlh: netlink message header
2504  * @dev: netdevice
2505  *
2506  * Default netdevice operation to dump the existing unicast address list.
2507  * Returns number of addresses from list put in skb.
2508  */
2509 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2510                       struct netlink_callback *cb,
2511                       struct net_device *dev,
2512                       int idx)
2513 {
2514         int err;
2515
2516         netif_addr_lock_bh(dev);
2517         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2518         if (err)
2519                 goto out;
2520         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2521 out:
2522         netif_addr_unlock_bh(dev);
2523         return idx;
2524 }
2525 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2526
2527 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2528 {
2529         int idx = 0;
2530         struct net *net = sock_net(skb->sk);
2531         struct net_device *dev;
2532
2533         rcu_read_lock();
2534         for_each_netdev_rcu(net, dev) {
2535                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2536                         struct net_device *br_dev;
2537                         const struct net_device_ops *ops;
2538
2539                         br_dev = netdev_master_upper_dev_get(dev);
2540                         ops = br_dev->netdev_ops;
2541                         if (ops->ndo_fdb_dump)
2542                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2543                 }
2544
2545                 if (dev->netdev_ops->ndo_fdb_dump)
2546                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2547                 else
2548                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2549         }
2550         rcu_read_unlock();
2551
2552         cb->args[0] = idx;
2553         return skb->len;
2554 }
2555
2556 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2557                             struct net_device *dev, u16 mode)
2558 {
2559         struct nlmsghdr *nlh;
2560         struct ifinfomsg *ifm;
2561         struct nlattr *br_afspec;
2562         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2563         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2564
2565         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2566         if (nlh == NULL)
2567                 return -EMSGSIZE;
2568
2569         ifm = nlmsg_data(nlh);
2570         ifm->ifi_family = AF_BRIDGE;
2571         ifm->__ifi_pad = 0;
2572         ifm->ifi_type = dev->type;
2573         ifm->ifi_index = dev->ifindex;
2574         ifm->ifi_flags = dev_get_flags(dev);
2575         ifm->ifi_change = 0;
2576
2577
2578         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2579             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2580             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2581             (br_dev &&
2582              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2583             (dev->addr_len &&
2584              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2585             (dev->ifindex != dev->iflink &&
2586              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2587                 goto nla_put_failure;
2588
2589         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2590         if (!br_afspec)
2591                 goto nla_put_failure;
2592
2593         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2594             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2595                 nla_nest_cancel(skb, br_afspec);
2596                 goto nla_put_failure;
2597         }
2598         nla_nest_end(skb, br_afspec);
2599
2600         return nlmsg_end(skb, nlh);
2601 nla_put_failure:
2602         nlmsg_cancel(skb, nlh);
2603         return -EMSGSIZE;
2604 }
2605 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2606
2607 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2608 {
2609         struct net *net = sock_net(skb->sk);
2610         struct net_device *dev;
2611         int idx = 0;
2612         u32 portid = NETLINK_CB(cb->skb).portid;
2613         u32 seq = cb->nlh->nlmsg_seq;
2614         struct nlattr *extfilt;
2615         u32 filter_mask = 0;
2616
2617         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2618                                   IFLA_EXT_MASK);
2619         if (extfilt)
2620                 filter_mask = nla_get_u32(extfilt);
2621
2622         rcu_read_lock();
2623         for_each_netdev_rcu(net, dev) {
2624                 const struct net_device_ops *ops = dev->netdev_ops;
2625                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2626
2627                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2628                         if (idx >= cb->args[0] &&
2629                             br_dev->netdev_ops->ndo_bridge_getlink(
2630                                     skb, portid, seq, dev, filter_mask) < 0)
2631                                 break;
2632                         idx++;
2633                 }
2634
2635                 if (ops->ndo_bridge_getlink) {
2636                         if (idx >= cb->args[0] &&
2637                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2638                                                     filter_mask) < 0)
2639                                 break;
2640                         idx++;
2641                 }
2642         }
2643         rcu_read_unlock();
2644         cb->args[0] = idx;
2645
2646         return skb->len;
2647 }
2648
2649 static inline size_t bridge_nlmsg_size(void)
2650 {
2651         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2652                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2653                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2654                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2655                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2656                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2657                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2658                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2659                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2660                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2661                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2662 }
2663
2664 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2665 {
2666         struct net *net = dev_net(dev);
2667         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2668         struct sk_buff *skb;
2669         int err = -EOPNOTSUPP;
2670
2671         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2672         if (!skb) {
2673                 err = -ENOMEM;
2674                 goto errout;
2675         }
2676
2677         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2678             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2679                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2680                 if (err < 0)
2681                         goto errout;
2682         }
2683
2684         if ((flags & BRIDGE_FLAGS_SELF) &&
2685             dev->netdev_ops->ndo_bridge_getlink) {
2686                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2687                 if (err < 0)
2688                         goto errout;
2689         }
2690
2691         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2692         return 0;
2693 errout:
2694         WARN_ON(err == -EMSGSIZE);
2695         kfree_skb(skb);
2696         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2697         return err;
2698 }
2699
2700 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2701 {
2702         struct net *net = sock_net(skb->sk);
2703         struct ifinfomsg *ifm;
2704         struct net_device *dev;
2705         struct nlattr *br_spec, *attr = NULL;
2706         int rem, err = -EOPNOTSUPP;
2707         u16 oflags, flags = 0;
2708         bool have_flags = false;
2709
2710         if (nlmsg_len(nlh) < sizeof(*ifm))
2711                 return -EINVAL;
2712
2713         ifm = nlmsg_data(nlh);
2714         if (ifm->ifi_family != AF_BRIDGE)
2715                 return -EPFNOSUPPORT;
2716
2717         dev = __dev_get_by_index(net, ifm->ifi_index);
2718         if (!dev) {
2719                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2720                 return -ENODEV;
2721         }
2722
2723         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2724         if (br_spec) {
2725                 nla_for_each_nested(attr, br_spec, rem) {
2726                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2727                                 have_flags = true;
2728                                 flags = nla_get_u16(attr);
2729                                 break;
2730                         }
2731                 }
2732         }
2733
2734         oflags = flags;
2735
2736         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2737                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2738
2739                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2740                         err = -EOPNOTSUPP;
2741                         goto out;
2742                 }
2743
2744                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2745                 if (err)
2746                         goto out;
2747
2748                 flags &= ~BRIDGE_FLAGS_MASTER;
2749         }
2750
2751         if ((flags & BRIDGE_FLAGS_SELF)) {
2752                 if (!dev->netdev_ops->ndo_bridge_setlink)
2753                         err = -EOPNOTSUPP;
2754                 else
2755                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2756
2757                 if (!err)
2758                         flags &= ~BRIDGE_FLAGS_SELF;
2759         }
2760
2761         if (have_flags)
2762                 memcpy(nla_data(attr), &flags, sizeof(flags));
2763         /* Generate event to notify upper layer of bridge change */
2764         if (!err)
2765                 err = rtnl_bridge_notify(dev, oflags);
2766 out:
2767         return err;
2768 }
2769
2770 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2771 {
2772         struct net *net = sock_net(skb->sk);
2773         struct ifinfomsg *ifm;
2774         struct net_device *dev;
2775         struct nlattr *br_spec, *attr = NULL;
2776         int rem, err = -EOPNOTSUPP;
2777         u16 oflags, flags = 0;
2778         bool have_flags = false;
2779
2780         if (nlmsg_len(nlh) < sizeof(*ifm))
2781                 return -EINVAL;
2782
2783         ifm = nlmsg_data(nlh);
2784         if (ifm->ifi_family != AF_BRIDGE)
2785                 return -EPFNOSUPPORT;
2786
2787         dev = __dev_get_by_index(net, ifm->ifi_index);
2788         if (!dev) {
2789                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2790                 return -ENODEV;
2791         }
2792
2793         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2794         if (br_spec) {
2795                 nla_for_each_nested(attr, br_spec, rem) {
2796                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2797                                 have_flags = true;
2798                                 flags = nla_get_u16(attr);
2799                                 break;
2800                         }
2801                 }
2802         }
2803
2804         oflags = flags;
2805
2806         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2807                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2808
2809                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2810                         err = -EOPNOTSUPP;
2811                         goto out;
2812                 }
2813
2814                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2815                 if (err)
2816                         goto out;
2817
2818                 flags &= ~BRIDGE_FLAGS_MASTER;
2819         }
2820
2821         if ((flags & BRIDGE_FLAGS_SELF)) {
2822                 if (!dev->netdev_ops->ndo_bridge_dellink)
2823                         err = -EOPNOTSUPP;
2824                 else
2825                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2826
2827                 if (!err)
2828                         flags &= ~BRIDGE_FLAGS_SELF;
2829         }
2830
2831         if (have_flags)
2832                 memcpy(nla_data(attr), &flags, sizeof(flags));
2833         /* Generate event to notify upper layer of bridge change */
2834         if (!err)
2835                 err = rtnl_bridge_notify(dev, oflags);
2836 out:
2837         return err;
2838 }
2839
2840 /* Process one rtnetlink message. */
2841
2842 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2843 {
2844         struct net *net = sock_net(skb->sk);
2845         rtnl_doit_func doit;
2846         int sz_idx, kind;
2847         int family;
2848         int type;
2849         int err;
2850
2851         type = nlh->nlmsg_type;
2852         if (type > RTM_MAX)
2853                 return -EOPNOTSUPP;
2854
2855         type -= RTM_BASE;
2856
2857         /* All the messages must have at least 1 byte length */
2858         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2859                 return 0;
2860
2861         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2862         sz_idx = type>>2;
2863         kind = type&3;
2864
2865         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2866                 return -EPERM;
2867
2868         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2869                 struct sock *rtnl;
2870                 rtnl_dumpit_func dumpit;
2871                 rtnl_calcit_func calcit;
2872                 u16 min_dump_alloc = 0;
2873
2874                 dumpit = rtnl_get_dumpit(family, type);
2875                 if (dumpit == NULL)
2876                         return -EOPNOTSUPP;
2877                 calcit = rtnl_get_calcit(family, type);
2878                 if (calcit)
2879                         min_dump_alloc = calcit(skb, nlh);
2880
2881                 __rtnl_unlock();
2882                 rtnl = net->rtnl;
2883                 {
2884                         struct netlink_dump_control c = {
2885                                 .dump           = dumpit,
2886                                 .min_dump_alloc = min_dump_alloc,
2887                         };
2888                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2889                 }
2890                 rtnl_lock();
2891                 return err;
2892         }
2893
2894         doit = rtnl_get_doit(family, type);
2895         if (doit == NULL)
2896                 return -EOPNOTSUPP;
2897
2898         return doit(skb, nlh);
2899 }
2900
2901 static void rtnetlink_rcv(struct sk_buff *skb)
2902 {
2903         rtnl_lock();
2904         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2905         rtnl_unlock();
2906 }
2907
2908 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2909 {
2910         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2911
2912         switch (event) {
2913         case NETDEV_UP:
2914         case NETDEV_DOWN:
2915         case NETDEV_PRE_UP:
2916         case NETDEV_POST_INIT:
2917         case NETDEV_REGISTER:
2918         case NETDEV_CHANGE:
2919         case NETDEV_PRE_TYPE_CHANGE:
2920         case NETDEV_GOING_DOWN:
2921         case NETDEV_UNREGISTER:
2922         case NETDEV_UNREGISTER_FINAL:
2923         case NETDEV_RELEASE:
2924         case NETDEV_JOIN:
2925                 break;
2926         default:
2927                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2928                 break;
2929         }
2930         return NOTIFY_DONE;
2931 }
2932
2933 static struct notifier_block rtnetlink_dev_notifier = {
2934         .notifier_call  = rtnetlink_event,
2935 };
2936
2937
2938 static int __net_init rtnetlink_net_init(struct net *net)
2939 {
2940         struct sock *sk;
2941         struct netlink_kernel_cfg cfg = {
2942                 .groups         = RTNLGRP_MAX,
2943                 .input          = rtnetlink_rcv,
2944                 .cb_mutex       = &rtnl_mutex,
2945                 .flags          = NL_CFG_F_NONROOT_RECV,
2946         };
2947
2948         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2949         if (!sk)
2950                 return -ENOMEM;
2951         net->rtnl = sk;
2952         return 0;
2953 }
2954
2955 static void __net_exit rtnetlink_net_exit(struct net *net)
2956 {
2957         netlink_kernel_release(net->rtnl);
2958         net->rtnl = NULL;
2959 }
2960
2961 static struct pernet_operations rtnetlink_net_ops = {
2962         .init = rtnetlink_net_init,
2963         .exit = rtnetlink_net_exit,
2964 };
2965
2966 void __init rtnetlink_init(void)
2967 {
2968         if (register_pernet_subsys(&rtnetlink_net_ops))
2969                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2970
2971         register_netdevice_notifier(&rtnetlink_dev_notifier);
2972
2973         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2974                       rtnl_dump_ifinfo, rtnl_calcit);
2975         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2976         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2977         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2978
2979         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2980         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2981
2982         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2983         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2984         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2985
2986         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2987         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2988         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2989 }
2990