net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/pci.h>
  40 #include <linux/etherdevice.h>
  41
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 struct rtnl_link {
  58         rtnl_doit_func          doit;
  59         rtnl_dumpit_func        dumpit;
  60         rtnl_calcit_func        calcit;
  61 };
  62
  63 static DEFINE_MUTEX(rtnl_mutex);
  64
  65 void rtnl_lock(void)
  66 {
  67         mutex_lock(&rtnl_mutex);
  68 }
  69 EXPORT_SYMBOL(rtnl_lock);
  70
  71 void __rtnl_unlock(void)
  72 {
  73         mutex_unlock(&rtnl_mutex);
  74 }
  75
  76 void rtnl_unlock(void)
  77 {
  78         /* This fellow will unlock it for us. */
  79         netdev_run_todo();
  80 }
  81 EXPORT_SYMBOL(rtnl_unlock);
  82
  83 int rtnl_trylock(void)
  84 {
  85         return mutex_trylock(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_trylock);
  88
  89 int rtnl_is_locked(void)
  90 {
  91         return mutex_is_locked(&rtnl_mutex);
  92 }
  93 EXPORT_SYMBOL(rtnl_is_locked);
  94
  95 #ifdef CONFIG_PROVE_LOCKING
  96 int lockdep_rtnl_is_held(void)
  97 {
  98         return lockdep_is_held(&rtnl_mutex);
  99 }
 100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105 static inline int rtm_msgindex(int msgtype)
 106 {
 107         int msgindex = msgtype - RTM_BASE;
 108
 109         /*
 110          * msgindex < 0 implies someone tried to register a netlink
 111          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112          * the message type has not been added to linux/rtnetlink.h
 113          */
 114         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116         return msgindex;
 117 }
 118
 119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120 {
 121         struct rtnl_link *tab;
 122
 123         if (protocol <= RTNL_FAMILY_MAX)
 124                 tab = rtnl_msg_handlers[protocol];
 125         else
 126                 tab = NULL;
 127
 128         if (tab == NULL || tab[msgindex].doit == NULL)
 129                 tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131         return tab[msgindex].doit;
 132 }
 133
 134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135 {
 136         struct rtnl_link *tab;
 137
 138         if (protocol <= RTNL_FAMILY_MAX)
 139                 tab = rtnl_msg_handlers[protocol];
 140         else
 141                 tab = NULL;
 142
 143         if (tab == NULL || tab[msgindex].dumpit == NULL)
 144                 tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146         return tab[msgindex].dumpit;
 147 }
 148
 149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150 {
 151         struct rtnl_link *tab;
 152
 153         if (protocol <= RTNL_FAMILY_MAX)
 154                 tab = rtnl_msg_handlers[protocol];
 155         else
 156                 tab = NULL;
 157
 158         if (tab == NULL || tab[msgindex].calcit == NULL)
 159                 tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161         return tab[msgindex].calcit;
 162 }
 163
 164 /**
 165  * __rtnl_register - Register a rtnetlink message type
 166  * @protocol: Protocol family or PF_UNSPEC
 167  * @msgtype: rtnetlink message type
 168  * @doit: Function pointer called for each request message
 169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170  * @calcit: Function pointer to calc size of dump message
 171  *
 172  * Registers the specified function pointers (at least one of them has
 173  * to be non-NULL) to be called whenever a request message for the
 174  * specified protocol family and message type is received.
 175  *
 176  * The special protocol family PF_UNSPEC may be used to define fallback
 177  * function pointers for the case when no entry for the specific protocol
 178  * family exists.
 179  *
 180  * Returns 0 on success or a negative error code.
 181  */
 182 int __rtnl_register(int protocol, int msgtype,
 183                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184                     rtnl_calcit_func calcit)
 185 {
 186         struct rtnl_link *tab;
 187         int msgindex;
 188
 189         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190         msgindex = rtm_msgindex(msgtype);
 191
 192         tab = rtnl_msg_handlers[protocol];
 193         if (tab == NULL) {
 194                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195                 if (tab == NULL)
 196                         return -ENOBUFS;
 197
 198                 rtnl_msg_handlers[protocol] = tab;
 199         }
 200
 201         if (doit)
 202                 tab[msgindex].doit = doit;
 203
 204         if (dumpit)
 205                 tab[msgindex].dumpit = dumpit;
 206
 207         if (calcit)
 208                 tab[msgindex].calcit = calcit;
 209
 210         return 0;
 211 }
 212 EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214 /**
 215  * rtnl_register - Register a rtnetlink message type
 216  *
 217  * Identical to __rtnl_register() but panics on failure. This is useful
 218  * as failure of this function is very unlikely, it can only happen due
 219  * to lack of memory when allocating the chain to store all message
 220  * handlers for a protocol. Meant for use in init functions where lack
 221  * of memory implies no sense in continuing.
 222  */
 223 void rtnl_register(int protocol, int msgtype,
 224                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225                    rtnl_calcit_func calcit)
 226 {
 227         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228                 panic("Unable to register rtnetlink message handler, "
 229                       "protocol = %d, message type = %d\n",
 230                       protocol, msgtype);
 231 }
 232 EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234 /**
 235  * rtnl_unregister - Unregister a rtnetlink message type
 236  * @protocol: Protocol family or PF_UNSPEC
 237  * @msgtype: rtnetlink message type
 238  *
 239  * Returns 0 on success or a negative error code.
 240  */
 241 int rtnl_unregister(int protocol, int msgtype)
 242 {
 243         int msgindex;
 244
 245         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246         msgindex = rtm_msgindex(msgtype);
 247
 248         if (rtnl_msg_handlers[protocol] == NULL)
 249                 return -ENOENT;
 250
 251         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254         return 0;
 255 }
 256 EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258 /**
 259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260  * @protocol : Protocol family or PF_UNSPEC
 261  *
 262  * Identical to calling rtnl_unregster() for all registered message types
 263  * of a certain protocol family.
 264  */
 265 void rtnl_unregister_all(int protocol)
 266 {
 267         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269         kfree(rtnl_msg_handlers[protocol]);
 270         rtnl_msg_handlers[protocol] = NULL;
 271 }
 272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274 static LIST_HEAD(link_ops);
 275
 276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277 {
 278         const struct rtnl_link_ops *ops;
 279
 280         list_for_each_entry(ops, &link_ops, list) {
 281                 if (!strcmp(ops->kind, kind))
 282                         return ops;
 283         }
 284         return NULL;
 285 }
 286
 287 /**
 288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289  * @ops: struct rtnl_link_ops * to register
 290  *
 291  * The caller must hold the rtnl_mutex. This function should be used
 292  * by drivers that create devices during module initialization. It
 293  * must be called before registering the devices.
 294  *
 295  * Returns 0 on success or a negative error code.
 296  */
 297 int __rtnl_link_register(struct rtnl_link_ops *ops)
 298 {
 299         if (rtnl_link_ops_get(ops->kind))
 300                 return -EEXIST;
 301
 302         if (!ops->dellink)
 303                 ops->dellink = unregister_netdevice_queue;
 304
 305         list_add_tail(&ops->list, &link_ops);
 306         return 0;
 307 }
 308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 309
 310 /**
 311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 312  * @ops: struct rtnl_link_ops * to register
 313  *
 314  * Returns 0 on success or a negative error code.
 315  */
 316 int rtnl_link_register(struct rtnl_link_ops *ops)
 317 {
 318         int err;
 319
 320         rtnl_lock();
 321         err = __rtnl_link_register(ops);
 322         rtnl_unlock();
 323         return err;
 324 }
 325 EXPORT_SYMBOL_GPL(rtnl_link_register);
 326
 327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 328 {
 329         struct net_device *dev;
 330         LIST_HEAD(list_kill);
 331
 332         for_each_netdev(net, dev) {
 333                 if (dev->rtnl_link_ops == ops)
 334                         ops->dellink(dev, &list_kill);
 335         }
 336         unregister_netdevice_many(&list_kill);
 337 }
 338
 339 /**
 340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 341  * @ops: struct rtnl_link_ops * to unregister
 342  *
 343  * The caller must hold the rtnl_mutex.
 344  */
 345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 346 {
 347         struct net *net;
 348
 349         for_each_net(net) {
 350                 __rtnl_kill_links(net, ops);
 351         }
 352         list_del(&ops->list);
 353 }
 354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 355
 356 /* Return with the rtnl_lock held when there are no network
 357  * devices unregistering in any network namespace.
 358  */
 359 static void rtnl_lock_unregistering_all(void)
 360 {
 361         struct net *net;
 362         bool unregistering;
 363         DEFINE_WAIT(wait);
 364
 365         for (;;) {
 366                 prepare_to_wait(&netdev_unregistering_wq, &wait,
 367                                 TASK_UNINTERRUPTIBLE);
 368                 unregistering = false;
 369                 rtnl_lock();
 370                 for_each_net(net) {
 371                         if (net->dev_unreg_count > 0) {
 372                                 unregistering = true;
 373                                 break;
 374                         }
 375                 }
 376                 if (!unregistering)
 377                         break;
 378                 __rtnl_unlock();
 379                 schedule();
 380         }
 381         finish_wait(&netdev_unregistering_wq, &wait);
 382 }
 383
 384 /**
 385  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 386  * @ops: struct rtnl_link_ops * to unregister
 387  */
 388 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 389 {
 390         /* Close the race with cleanup_net() */
 391         mutex_lock(&net_mutex);
 392         rtnl_lock_unregistering_all();
 393         __rtnl_link_unregister(ops);
 394         rtnl_unlock();
 395         mutex_unlock(&net_mutex);
 396 }
 397 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 398
 399 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 400 {
 401         struct net_device *master_dev;
 402         const struct rtnl_link_ops *ops;
 403
 404         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 405         if (!master_dev)
 406                 return 0;
 407         ops = master_dev->rtnl_link_ops;
 408         if (!ops || !ops->get_slave_size)
 409                 return 0;
 410         /* IFLA_INFO_SLAVE_DATA + nested data */
 411         return nla_total_size(sizeof(struct nlattr)) +
 412                ops->get_slave_size(master_dev, dev);
 413 }
 414
 415 static size_t rtnl_link_get_size(const struct net_device *dev)
 416 {
 417         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 418         size_t size;
 419
 420         if (!ops)
 421                 return 0;
 422
 423         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 424                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 425
 426         if (ops->get_size)
 427                 /* IFLA_INFO_DATA + nested data */
 428                 size += nla_total_size(sizeof(struct nlattr)) +
 429                         ops->get_size(dev);
 430
 431         if (ops->get_xstats_size)
 432                 /* IFLA_INFO_XSTATS */
 433                 size += nla_total_size(ops->get_xstats_size(dev));
 434
 435         size += rtnl_link_get_slave_info_data_size(dev);
 436
 437         return size;
 438 }
 439
 440 static LIST_HEAD(rtnl_af_ops);
 441
 442 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 443 {
 444         const struct rtnl_af_ops *ops;
 445
 446         list_for_each_entry(ops, &rtnl_af_ops, list) {
 447                 if (ops->family == family)
 448                         return ops;
 449         }
 450
 451         return NULL;
 452 }
 453
 454 /**
 455  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 456  * @ops: struct rtnl_af_ops * to register
 457  *
 458  * Returns 0 on success or a negative error code.
 459  */
 460 void rtnl_af_register(struct rtnl_af_ops *ops)
 461 {
 462         rtnl_lock();
 463         list_add_tail(&ops->list, &rtnl_af_ops);
 464         rtnl_unlock();
 465 }
 466 EXPORT_SYMBOL_GPL(rtnl_af_register);
 467
 468 /**
 469  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 470  * @ops: struct rtnl_af_ops * to unregister
 471  *
 472  * The caller must hold the rtnl_mutex.
 473  */
 474 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 475 {
 476         list_del(&ops->list);
 477 }
 478 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 479
 480 /**
 481  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 482  * @ops: struct rtnl_af_ops * to unregister
 483  */
 484 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 485 {
 486         rtnl_lock();
 487         __rtnl_af_unregister(ops);
 488         rtnl_unlock();
 489 }
 490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 491
 492 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 493 {
 494         struct rtnl_af_ops *af_ops;
 495         size_t size;
 496
 497         /* IFLA_AF_SPEC */
 498         size = nla_total_size(sizeof(struct nlattr));
 499
 500         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 501                 if (af_ops->get_link_af_size) {
 502                         /* AF_* + nested data */
 503                         size += nla_total_size(sizeof(struct nlattr)) +
 504                                 af_ops->get_link_af_size(dev);
 505                 }
 506         }
 507
 508         return size;
 509 }
 510
 511 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 512 {
 513         struct net_device *master_dev;
 514
 515         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 516         if (master_dev && master_dev->rtnl_link_ops)
 517                 return true;
 518         return false;
 519 }
 520
 521 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 522                                      const struct net_device *dev)
 523 {
 524         struct net_device *master_dev;
 525         const struct rtnl_link_ops *ops;
 526         struct nlattr *slave_data;
 527         int err;
 528
 529         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 530         if (!master_dev)
 531                 return 0;
 532         ops = master_dev->rtnl_link_ops;
 533         if (!ops)
 534                 return 0;
 535         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 536                 return -EMSGSIZE;
 537         if (ops->fill_slave_info) {
 538                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 539                 if (!slave_data)
 540                         return -EMSGSIZE;
 541                 err = ops->fill_slave_info(skb, master_dev, dev);
 542                 if (err < 0)
 543                         goto err_cancel_slave_data;
 544                 nla_nest_end(skb, slave_data);
 545         }
 546         return 0;
 547
 548 err_cancel_slave_data:
 549         nla_nest_cancel(skb, slave_data);
 550         return err;
 551 }
 552
 553 static int rtnl_link_info_fill(struct sk_buff *skb,
 554                                const struct net_device *dev)
 555 {
 556         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 557         struct nlattr *data;
 558         int err;
 559
 560         if (!ops)
 561                 return 0;
 562         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 563                 return -EMSGSIZE;
 564         if (ops->fill_xstats) {
 565                 err = ops->fill_xstats(skb, dev);
 566                 if (err < 0)
 567                         return err;
 568         }
 569         if (ops->fill_info) {
 570                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 571                 if (data == NULL)
 572                         return -EMSGSIZE;
 573                 err = ops->fill_info(skb, dev);
 574                 if (err < 0)
 575                         goto err_cancel_data;
 576                 nla_nest_end(skb, data);
 577         }
 578         return 0;
 579
 580 err_cancel_data:
 581         nla_nest_cancel(skb, data);
 582         return err;
 583 }
 584
 585 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 586 {
 587         struct nlattr *linkinfo;
 588         int err = -EMSGSIZE;
 589
 590         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 591         if (linkinfo == NULL)
 592                 goto out;
 593
 594         err = rtnl_link_info_fill(skb, dev);
 595         if (err < 0)
 596                 goto err_cancel_link;
 597
 598         err = rtnl_link_slave_info_fill(skb, dev);
 599         if (err < 0)
 600                 goto err_cancel_link;
 601
 602         nla_nest_end(skb, linkinfo);
 603         return 0;
 604
 605 err_cancel_link:
 606         nla_nest_cancel(skb, linkinfo);
 607 out:
 608         return err;
 609 }
 610
 611 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 612 {
 613         struct sock *rtnl = net->rtnl;
 614         int err = 0;
 615
 616         NETLINK_CB(skb).dst_group = group;
 617         if (echo)
 618                 atomic_inc(&skb->users);
 619         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 620         if (echo)
 621                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 622         return err;
 623 }
 624
 625 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 626 {
 627         struct sock *rtnl = net->rtnl;
 628
 629         return nlmsg_unicast(rtnl, skb, pid);
 630 }
 631 EXPORT_SYMBOL(rtnl_unicast);
 632
 633 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 634                  struct nlmsghdr *nlh, gfp_t flags)
 635 {
 636         struct sock *rtnl = net->rtnl;
 637         int report = 0;
 638
 639         if (nlh)
 640                 report = nlmsg_report(nlh);
 641
 642         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 643 }
 644 EXPORT_SYMBOL(rtnl_notify);
 645
 646 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 647 {
 648         struct sock *rtnl = net->rtnl;
 649
 650         netlink_set_err(rtnl, 0, group, error);
 651 }
 652 EXPORT_SYMBOL(rtnl_set_sk_err);
 653
 654 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 655 {
 656         struct nlattr *mx;
 657         int i, valid = 0;
 658
 659         mx = nla_nest_start(skb, RTA_METRICS);
 660         if (mx == NULL)
 661                 return -ENOBUFS;
 662
 663         for (i = 0; i < RTAX_MAX; i++) {
 664                 if (metrics[i]) {
 665                         valid++;
 666                         if (nla_put_u32(skb, i+1, metrics[i]))
 667                                 goto nla_put_failure;
 668                 }
 669         }
 670
 671         if (!valid) {
 672                 nla_nest_cancel(skb, mx);
 673                 return 0;
 674         }
 675
 676         return nla_nest_end(skb, mx);
 677
 678 nla_put_failure:
 679         nla_nest_cancel(skb, mx);
 680         return -EMSGSIZE;
 681 }
 682 EXPORT_SYMBOL(rtnetlink_put_metrics);
 683
 684 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 685                        long expires, u32 error)
 686 {
 687         struct rta_cacheinfo ci = {
 688                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 689                 .rta_used = dst->__use,
 690                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 691                 .rta_error = error,
 692                 .rta_id =  id,
 693         };
 694
 695         if (expires) {
 696                 unsigned long clock;
 697
 698                 clock = jiffies_to_clock_t(abs(expires));
 699                 clock = min_t(unsigned long, clock, INT_MAX);
 700                 ci.rta_expires = (expires > 0) ? clock : -clock;
 701         }
 702         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 703 }
 704 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 705
 706 static void set_operstate(struct net_device *dev, unsigned char transition)
 707 {
 708         unsigned char operstate = dev->operstate;
 709
 710         switch (transition) {
 711         case IF_OPER_UP:
 712                 if ((operstate == IF_OPER_DORMANT ||
 713                      operstate == IF_OPER_UNKNOWN) &&
 714                     !netif_dormant(dev))
 715                         operstate = IF_OPER_UP;
 716                 break;
 717
 718         case IF_OPER_DORMANT:
 719                 if (operstate == IF_OPER_UP ||
 720                     operstate == IF_OPER_UNKNOWN)
 721                         operstate = IF_OPER_DORMANT;
 722                 break;
 723         }
 724
 725         if (dev->operstate != operstate) {
 726                 write_lock_bh(&dev_base_lock);
 727                 dev->operstate = operstate;
 728                 write_unlock_bh(&dev_base_lock);
 729                 netdev_state_change(dev);
 730         }
 731 }
 732
 733 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 734 {
 735         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 736                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 737 }
 738
 739 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 740                                            const struct ifinfomsg *ifm)
 741 {
 742         unsigned int flags = ifm->ifi_flags;
 743
 744         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 745         if (ifm->ifi_change)
 746                 flags = (flags & ifm->ifi_change) |
 747                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 748
 749         return flags;
 750 }
 751
 752 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 753                                  const struct rtnl_link_stats64 *b)
 754 {
 755         a->rx_packets = b->rx_packets;
 756         a->tx_packets = b->tx_packets;
 757         a->rx_bytes = b->rx_bytes;
 758         a->tx_bytes = b->tx_bytes;
 759         a->rx_errors = b->rx_errors;
 760         a->tx_errors = b->tx_errors;
 761         a->rx_dropped = b->rx_dropped;
 762         a->tx_dropped = b->tx_dropped;
 763
 764         a->multicast = b->multicast;
 765         a->collisions = b->collisions;
 766
 767         a->rx_length_errors = b->rx_length_errors;
 768         a->rx_over_errors = b->rx_over_errors;
 769         a->rx_crc_errors = b->rx_crc_errors;
 770         a->rx_frame_errors = b->rx_frame_errors;
 771         a->rx_fifo_errors = b->rx_fifo_errors;
 772         a->rx_missed_errors = b->rx_missed_errors;
 773
 774         a->tx_aborted_errors = b->tx_aborted_errors;
 775         a->tx_carrier_errors = b->tx_carrier_errors;
 776         a->tx_fifo_errors = b->tx_fifo_errors;
 777         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 778         a->tx_window_errors = b->tx_window_errors;
 779
 780         a->rx_compressed = b->rx_compressed;
 781         a->tx_compressed = b->tx_compressed;
 782 }
 783
 784 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 785 {
 786         memcpy(v, b, sizeof(*b));
 787 }
 788
 789 /* All VF info */
 790 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 791                                    u32 ext_filter_mask)
 792 {
 793         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 794             (ext_filter_mask & RTEXT_FILTER_VF)) {
 795                 int num_vfs = dev_num_vf(dev->dev.parent);
 796                 size_t size = nla_total_size(sizeof(struct nlattr));
 797                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 798                 size += num_vfs *
 799                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 800                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 801                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 802                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 803                 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(ext_filter_mask
 859                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 860                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 861                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 862                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 863                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 864                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 865 }
 866
 867 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 868 {
 869         struct nlattr *vf_ports;
 870         struct nlattr *vf_port;
 871         int vf;
 872         int err;
 873
 874         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 875         if (!vf_ports)
 876                 return -EMSGSIZE;
 877
 878         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 879                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 880                 if (!vf_port)
 881                         goto nla_put_failure;
 882                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 883                         goto nla_put_failure;
 884                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 885                 if (err == -EMSGSIZE)
 886                         goto nla_put_failure;
 887                 if (err) {
 888                         nla_nest_cancel(skb, vf_port);
 889                         continue;
 890                 }
 891                 nla_nest_end(skb, vf_port);
 892         }
 893
 894         nla_nest_end(skb, vf_ports);
 895
 896         return 0;
 897
 898 nla_put_failure:
 899         nla_nest_cancel(skb, vf_ports);
 900         return -EMSGSIZE;
 901 }
 902
 903 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 904 {
 905         struct nlattr *port_self;
 906         int err;
 907
 908         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 909         if (!port_self)
 910                 return -EMSGSIZE;
 911
 912         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 913         if (err) {
 914                 nla_nest_cancel(skb, port_self);
 915                 return (err == -EMSGSIZE) ? err : 0;
 916         }
 917
 918         nla_nest_end(skb, port_self);
 919
 920         return 0;
 921 }
 922
 923 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 924                           u32 ext_filter_mask)
 925 {
 926         int err;
 927
 928         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 929             !(ext_filter_mask & RTEXT_FILTER_VF))
 930                 return 0;
 931
 932         err = rtnl_port_self_fill(skb, dev);
 933         if (err)
 934                 return err;
 935
 936         if (dev_num_vf(dev->dev.parent)) {
 937                 err = rtnl_vf_ports_fill(skb, dev);
 938                 if (err)
 939                         return err;
 940         }
 941
 942         return 0;
 943 }
 944
 945 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 946 {
 947         int err;
 948         struct netdev_phys_port_id ppid;
 949
 950         err = dev_get_phys_port_id(dev, &ppid);
 951         if (err) {
 952                 if (err == -EOPNOTSUPP)
 953                         return 0;
 954                 return err;
 955         }
 956
 957         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 958                 return -EMSGSIZE;
 959
 960         return 0;
 961 }
 962
 963 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 964                             int type, u32 pid, u32 seq, u32 change,
 965                             unsigned int flags, u32 ext_filter_mask)
 966 {
 967         struct ifinfomsg *ifm;
 968         struct nlmsghdr *nlh;
 969         struct rtnl_link_stats64 temp;
 970         const struct rtnl_link_stats64 *stats;
 971         struct nlattr *attr, *af_spec;
 972         struct rtnl_af_ops *af_ops;
 973         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 974
 975         ASSERT_RTNL();
 976         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 977         if (nlh == NULL)
 978                 return -EMSGSIZE;
 979
 980         ifm = nlmsg_data(nlh);
 981         ifm->ifi_family = AF_UNSPEC;
 982         ifm->__ifi_pad = 0;
 983         ifm->ifi_type = dev->type;
 984         ifm->ifi_index = dev->ifindex;
 985         ifm->ifi_flags = dev_get_flags(dev);
 986         ifm->ifi_change = change;
 987
 988         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 989             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 990             nla_put_u8(skb, IFLA_OPERSTATE,
 991                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 992             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 993             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 994             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 995             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 996             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 997 #ifdef CONFIG_RPS
 998             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 999 #endif
1000             (dev->ifindex != dev->iflink &&
1001              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1002             (upper_dev &&
1003              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1004             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1005             (dev->qdisc &&
1006              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1007             (dev->ifalias &&
1008              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
1009                 goto nla_put_failure;
1010
1011         if (1) {
1012                 struct rtnl_link_ifmap map = {
1013                         .mem_start   = dev->mem_start,
1014                         .mem_end     = dev->mem_end,
1015                         .base_addr   = dev->base_addr,
1016                         .irq         = dev->irq,
1017                         .dma         = dev->dma,
1018                         .port        = dev->if_port,
1019                 };
1020                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1021                         goto nla_put_failure;
1022         }
1023
1024         if (dev->addr_len) {
1025                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1026                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1027                         goto nla_put_failure;
1028         }
1029
1030         if (rtnl_phys_port_id_fill(skb, dev))
1031                 goto nla_put_failure;
1032
1033         attr = nla_reserve(skb, IFLA_STATS,
1034                         sizeof(struct rtnl_link_stats));
1035         if (attr == NULL)
1036                 goto nla_put_failure;
1037
1038         stats = dev_get_stats(dev, &temp);
1039         copy_rtnl_link_stats(nla_data(attr), stats);
1040
1041         attr = nla_reserve(skb, IFLA_STATS64,
1042                         sizeof(struct rtnl_link_stats64));
1043         if (attr == NULL)
1044                 goto nla_put_failure;
1045         copy_rtnl_link_stats64(nla_data(attr), stats);
1046
1047         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1048             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1049                 goto nla_put_failure;
1050
1051         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1052             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1053                 int i;
1054
1055                 struct nlattr *vfinfo, *vf;
1056                 int num_vfs = dev_num_vf(dev->dev.parent);
1057
1058                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1059                 if (!vfinfo)
1060                         goto nla_put_failure;
1061                 for (i = 0; i < num_vfs; i++) {
1062                         struct ifla_vf_info ivi;
1063                         struct ifla_vf_mac vf_mac;
1064                         struct ifla_vf_vlan vf_vlan;
1065                         struct ifla_vf_tx_rate vf_tx_rate;
1066                         struct ifla_vf_spoofchk vf_spoofchk;
1067                         struct ifla_vf_link_state vf_linkstate;
1068
1069                         /*
1070                          * Not all SR-IOV capable drivers support the
1071                          * spoofcheck query.  Preset to -1 so the user
1072                          * space tool can detect that the driver didn't
1073                          * report anything.
1074                          */
1075                         ivi.spoofchk = -1;
1076                         memset(ivi.mac, 0, sizeof(ivi.mac));
1077                         /* The default value for VF link state is "auto"
1078                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1079                          */
1080                         ivi.linkstate = 0;
1081                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1082                                 break;
1083                         vf_mac.vf =
1084                                 vf_vlan.vf =
1085                                 vf_tx_rate.vf =
1086                                 vf_spoofchk.vf =
1087                                 vf_linkstate.vf = ivi.vf;
1088
1089                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1090                         vf_vlan.vlan = ivi.vlan;
1091                         vf_vlan.qos = ivi.qos;
1092                         vf_tx_rate.rate = ivi.tx_rate;
1093                         vf_spoofchk.setting = ivi.spoofchk;
1094                         vf_linkstate.link_state = ivi.linkstate;
1095                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1096                         if (!vf) {
1097                                 nla_nest_cancel(skb, vfinfo);
1098                                 goto nla_put_failure;
1099                         }
1100                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1101                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1102                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1103                                     &vf_tx_rate) ||
1104                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1105                                     &vf_spoofchk) ||
1106                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1107                                     &vf_linkstate))
1108                                 goto nla_put_failure;
1109                         nla_nest_end(skb, vf);
1110                 }
1111                 nla_nest_end(skb, vfinfo);
1112         }
1113
1114         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1115                 goto nla_put_failure;
1116
1117         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1118                 if (rtnl_link_fill(skb, dev) < 0)
1119                         goto nla_put_failure;
1120         }
1121
1122         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1123                 goto nla_put_failure;
1124
1125         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1126                 if (af_ops->fill_link_af) {
1127                         struct nlattr *af;
1128                         int err;
1129
1130                         if (!(af = nla_nest_start(skb, af_ops->family)))
1131                                 goto nla_put_failure;
1132
1133                         err = af_ops->fill_link_af(skb, dev);
1134
1135                         /*
1136                          * Caller may return ENODATA to indicate that there
1137                          * was no data to be dumped. This is not an error, it
1138                          * means we should trim the attribute header and
1139                          * continue.
1140                          */
1141                         if (err == -ENODATA)
1142                                 nla_nest_cancel(skb, af);
1143                         else if (err < 0)
1144                                 goto nla_put_failure;
1145
1146                         nla_nest_end(skb, af);
1147                 }
1148         }
1149
1150         nla_nest_end(skb, af_spec);
1151
1152         return nlmsg_end(skb, nlh);
1153
1154 nla_put_failure:
1155         nlmsg_cancel(skb, nlh);
1156         return -EMSGSIZE;
1157 }
1158
1159 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1160 {
1161         struct net *net = sock_net(skb->sk);
1162         int h, s_h;
1163         int idx = 0, s_idx;
1164         struct net_device *dev;
1165         struct hlist_head *head;
1166         struct nlattr *tb[IFLA_MAX+1];
1167         u32 ext_filter_mask = 0;
1168         int err;
1169         int hdrlen;
1170
1171         s_h = cb->args[0];
1172         s_idx = cb->args[1];
1173
1174         rcu_read_lock();
1175         cb->seq = net->dev_base_seq;
1176
1177         /* A hack to preserve kernel<->userspace interface.
1178          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1179          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1180          * what iproute2 < v3.9.0 used.
1181          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1182          * attribute, its netlink message is shorter than struct ifinfomsg.
1183          */
1184         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1185                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1186
1187         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1188
1189                 if (tb[IFLA_EXT_MASK])
1190                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1191         }
1192
1193         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1194                 idx = 0;
1195                 head = &net->dev_index_head[h];
1196                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1197                         if (idx < s_idx)
1198                                 goto cont;
1199                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1200                                                NETLINK_CB(cb->skb).portid,
1201                                                cb->nlh->nlmsg_seq, 0,
1202                                                NLM_F_MULTI,
1203                                                ext_filter_mask);
1204                         /* If we ran out of room on the first message,
1205                          * we're in trouble
1206                          */
1207                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1208
1209                         if (err <= 0)
1210                                 goto out;
1211
1212                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1213 cont:
1214                         idx++;
1215                 }
1216         }
1217 out:
1218         rcu_read_unlock();
1219         cb->args[1] = idx;
1220         cb->args[0] = h;
1221
1222         return skb->len;
1223 }
1224
1225 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1226         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1227         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1228         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1229         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1230         [IFLA_MTU]              = { .type = NLA_U32 },
1231         [IFLA_LINK]             = { .type = NLA_U32 },
1232         [IFLA_MASTER]           = { .type = NLA_U32 },
1233         [IFLA_CARRIER]          = { .type = NLA_U8 },
1234         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1235         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1236         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1237         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1238         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1239         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1240         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1241         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1242         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1243         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1244         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1245         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1246         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1247         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1248         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1249         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1250         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1251 };
1252 EXPORT_SYMBOL(ifla_policy);
1253
1254 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1255         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1256         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1257         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1258         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1259 };
1260
1261 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1262         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1263 };
1264
1265 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1266         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1267                                     .len = sizeof(struct ifla_vf_mac) },
1268         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1269                                     .len = sizeof(struct ifla_vf_vlan) },
1270         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1271                                     .len = sizeof(struct ifla_vf_tx_rate) },
1272         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1273                                     .len = sizeof(struct ifla_vf_spoofchk) },
1274 };
1275
1276 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1277         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1278         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1279                                     .len = PORT_PROFILE_MAX },
1280         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1281                                     .len = sizeof(struct ifla_port_vsi)},
1282         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1283                                       .len = PORT_UUID_MAX },
1284         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1285                                     .len = PORT_UUID_MAX },
1286         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1287         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1288 };
1289
1290 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1291 {
1292         struct net *net;
1293         /* Examine the link attributes and figure out which
1294          * network namespace we are talking about.
1295          */
1296         if (tb[IFLA_NET_NS_PID])
1297                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1298         else if (tb[IFLA_NET_NS_FD])
1299                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1300         else
1301                 net = get_net(src_net);
1302         return net;
1303 }
1304 EXPORT_SYMBOL(rtnl_link_get_net);
1305
1306 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1307 {
1308         if (dev) {
1309                 if (tb[IFLA_ADDRESS] &&
1310                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1311                         return -EINVAL;
1312
1313                 if (tb[IFLA_BROADCAST] &&
1314                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1315                         return -EINVAL;
1316         }
1317
1318         if (tb[IFLA_AF_SPEC]) {
1319                 struct nlattr *af;
1320                 int rem, err;
1321
1322                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1323                         const struct rtnl_af_ops *af_ops;
1324
1325                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1326                                 return -EAFNOSUPPORT;
1327
1328                         if (!af_ops->set_link_af)
1329                                 return -EOPNOTSUPP;
1330
1331                         if (af_ops->validate_link_af) {
1332                                 err = af_ops->validate_link_af(dev, af);
1333                                 if (err < 0)
1334                                         return err;
1335                         }
1336                 }
1337         }
1338
1339         return 0;
1340 }
1341
1342 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1343 {
1344         int rem, err = -EINVAL;
1345         struct nlattr *vf;
1346         const struct net_device_ops *ops = dev->netdev_ops;
1347
1348         nla_for_each_nested(vf, attr, rem) {
1349                 switch (nla_type(vf)) {
1350                 case IFLA_VF_MAC: {
1351                         struct ifla_vf_mac *ivm;
1352                         ivm = nla_data(vf);
1353                         err = -EOPNOTSUPP;
1354                         if (ops->ndo_set_vf_mac)
1355                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1356                                                           ivm->mac);
1357                         break;
1358                 }
1359                 case IFLA_VF_VLAN: {
1360                         struct ifla_vf_vlan *ivv;
1361                         ivv = nla_data(vf);
1362                         err = -EOPNOTSUPP;
1363                         if (ops->ndo_set_vf_vlan)
1364                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1365                                                            ivv->vlan,
1366                                                            ivv->qos);
1367                         break;
1368                 }
1369                 case IFLA_VF_TX_RATE: {
1370                         struct ifla_vf_tx_rate *ivt;
1371                         ivt = nla_data(vf);
1372                         err = -EOPNOTSUPP;
1373                         if (ops->ndo_set_vf_tx_rate)
1374                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1375                                                               ivt->rate);
1376                         break;
1377                 }
1378                 case IFLA_VF_SPOOFCHK: {
1379                         struct ifla_vf_spoofchk *ivs;
1380                         ivs = nla_data(vf);
1381                         err = -EOPNOTSUPP;
1382                         if (ops->ndo_set_vf_spoofchk)
1383                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1384                                                                ivs->setting);
1385                         break;
1386                 }
1387                 case IFLA_VF_LINK_STATE: {
1388                         struct ifla_vf_link_state *ivl;
1389                         ivl = nla_data(vf);
1390                         err = -EOPNOTSUPP;
1391                         if (ops->ndo_set_vf_link_state)
1392                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1393                                                                  ivl->link_state);
1394                         break;
1395                 }
1396                 default:
1397                         err = -EINVAL;
1398                         break;
1399                 }
1400                 if (err)
1401                         break;
1402         }
1403         return err;
1404 }
1405
1406 static int do_set_master(struct net_device *dev, int ifindex)
1407 {
1408         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1409         const struct net_device_ops *ops;
1410         int err;
1411
1412         if (upper_dev) {
1413                 if (upper_dev->ifindex == ifindex)
1414                         return 0;
1415                 ops = upper_dev->netdev_ops;
1416                 if (ops->ndo_del_slave) {
1417                         err = ops->ndo_del_slave(upper_dev, dev);
1418                         if (err)
1419                                 return err;
1420                 } else {
1421                         return -EOPNOTSUPP;
1422                 }
1423         }
1424
1425         if (ifindex) {
1426                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1427                 if (!upper_dev)
1428                         return -EINVAL;
1429                 ops = upper_dev->netdev_ops;
1430                 if (ops->ndo_add_slave) {
1431                         err = ops->ndo_add_slave(upper_dev, dev);
1432                         if (err)
1433                                 return err;
1434                 } else {
1435                         return -EOPNOTSUPP;
1436                 }
1437         }
1438         return 0;
1439 }
1440
1441 static int do_setlink(const struct sk_buff *skb,
1442                       struct net_device *dev, struct ifinfomsg *ifm,
1443                       struct nlattr **tb, char *ifname, int modified)
1444 {
1445         const struct net_device_ops *ops = dev->netdev_ops;
1446         int err;
1447
1448         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1449                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1450                 if (IS_ERR(net)) {
1451                         err = PTR_ERR(net);
1452                         goto errout;
1453                 }
1454                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1455                         err = -EPERM;
1456                         goto errout;
1457                 }
1458                 err = dev_change_net_namespace(dev, net, ifname);
1459                 put_net(net);
1460                 if (err)
1461                         goto errout;
1462                 modified = 1;
1463         }
1464
1465         if (tb[IFLA_MAP]) {
1466                 struct rtnl_link_ifmap *u_map;
1467                 struct ifmap k_map;
1468
1469                 if (!ops->ndo_set_config) {
1470                         err = -EOPNOTSUPP;
1471                         goto errout;
1472                 }
1473
1474                 if (!netif_device_present(dev)) {
1475                         err = -ENODEV;
1476                         goto errout;
1477                 }
1478
1479                 u_map = nla_data(tb[IFLA_MAP]);
1480                 k_map.mem_start = (unsigned long) u_map->mem_start;
1481                 k_map.mem_end = (unsigned long) u_map->mem_end;
1482                 k_map.base_addr = (unsigned short) u_map->base_addr;
1483                 k_map.irq = (unsigned char) u_map->irq;
1484                 k_map.dma = (unsigned char) u_map->dma;
1485                 k_map.port = (unsigned char) u_map->port;
1486
1487                 err = ops->ndo_set_config(dev, &k_map);
1488                 if (err < 0)
1489                         goto errout;
1490
1491                 modified = 1;
1492         }
1493
1494         if (tb[IFLA_ADDRESS]) {
1495                 struct sockaddr *sa;
1496                 int len;
1497
1498                 len = sizeof(sa_family_t) + dev->addr_len;
1499                 sa = kmalloc(len, GFP_KERNEL);
1500                 if (!sa) {
1501                         err = -ENOMEM;
1502                         goto errout;
1503                 }
1504                 sa->sa_family = dev->type;
1505                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1506                        dev->addr_len);
1507                 err = dev_set_mac_address(dev, sa);
1508                 kfree(sa);
1509                 if (err)
1510                         goto errout;
1511                 modified = 1;
1512         }
1513
1514         if (tb[IFLA_MTU]) {
1515                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1516                 if (err < 0)
1517                         goto errout;
1518                 modified = 1;
1519         }
1520
1521         if (tb[IFLA_GROUP]) {
1522                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1523                 modified = 1;
1524         }
1525
1526         /*
1527          * Interface selected by interface index but interface
1528          * name provided implies that a name change has been
1529          * requested.
1530          */
1531         if (ifm->ifi_index > 0 && ifname[0]) {
1532                 err = dev_change_name(dev, ifname);
1533                 if (err < 0)
1534                         goto errout;
1535                 modified = 1;
1536         }
1537
1538         if (tb[IFLA_IFALIAS]) {
1539                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1540                                     nla_len(tb[IFLA_IFALIAS]));
1541                 if (err < 0)
1542                         goto errout;
1543                 modified = 1;
1544         }
1545
1546         if (tb[IFLA_BROADCAST]) {
1547                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1548                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1549         }
1550
1551         if (ifm->ifi_flags || ifm->ifi_change) {
1552                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1553                 if (err < 0)
1554                         goto errout;
1555         }
1556
1557         if (tb[IFLA_MASTER]) {
1558                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1559                 if (err)
1560                         goto errout;
1561                 modified = 1;
1562         }
1563
1564         if (tb[IFLA_CARRIER]) {
1565                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1566                 if (err)
1567                         goto errout;
1568                 modified = 1;
1569         }
1570
1571         if (tb[IFLA_TXQLEN])
1572                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1573
1574         if (tb[IFLA_OPERSTATE])
1575                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1576
1577         if (tb[IFLA_LINKMODE]) {
1578                 write_lock_bh(&dev_base_lock);
1579                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1580                 write_unlock_bh(&dev_base_lock);
1581         }
1582
1583         if (tb[IFLA_VFINFO_LIST]) {
1584                 struct nlattr *attr;
1585                 int rem;
1586                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1587                         if (nla_type(attr) != IFLA_VF_INFO) {
1588                                 err = -EINVAL;
1589                                 goto errout;
1590                         }
1591                         err = do_setvfinfo(dev, attr);
1592                         if (err < 0)
1593                                 goto errout;
1594                         modified = 1;
1595                 }
1596         }
1597         err = 0;
1598
1599         if (tb[IFLA_VF_PORTS]) {
1600                 struct nlattr *port[IFLA_PORT_MAX+1];
1601                 struct nlattr *attr;
1602                 int vf;
1603                 int rem;
1604
1605                 err = -EOPNOTSUPP;
1606                 if (!ops->ndo_set_vf_port)
1607                         goto errout;
1608
1609                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1610                         if (nla_type(attr) != IFLA_VF_PORT)
1611                                 continue;
1612                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1613                                 attr, ifla_port_policy);
1614                         if (err < 0)
1615                                 goto errout;
1616                         if (!port[IFLA_PORT_VF]) {
1617                                 err = -EOPNOTSUPP;
1618                                 goto errout;
1619                         }
1620                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1621                         err = ops->ndo_set_vf_port(dev, vf, port);
1622                         if (err < 0)
1623                                 goto errout;
1624                         modified = 1;
1625                 }
1626         }
1627         err = 0;
1628
1629         if (tb[IFLA_PORT_SELF]) {
1630                 struct nlattr *port[IFLA_PORT_MAX+1];
1631
1632                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1633                         tb[IFLA_PORT_SELF], ifla_port_policy);
1634                 if (err < 0)
1635                         goto errout;
1636
1637                 err = -EOPNOTSUPP;
1638                 if (ops->ndo_set_vf_port)
1639                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1640                 if (err < 0)
1641                         goto errout;
1642                 modified = 1;
1643         }
1644
1645         if (tb[IFLA_AF_SPEC]) {
1646                 struct nlattr *af;
1647                 int rem;
1648
1649                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1650                         const struct rtnl_af_ops *af_ops;
1651
1652                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1653                                 BUG();
1654
1655                         err = af_ops->set_link_af(dev, af);
1656                         if (err < 0)
1657                                 goto errout;
1658
1659                         modified = 1;
1660                 }
1661         }
1662         err = 0;
1663
1664 errout:
1665         if (err < 0 && modified)
1666                 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",
1667                                      dev->name);
1668
1669         return err;
1670 }
1671
1672 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1673 {
1674         struct net *net = sock_net(skb->sk);
1675         struct ifinfomsg *ifm;
1676         struct net_device *dev;
1677         int err;
1678         struct nlattr *tb[IFLA_MAX+1];
1679         char ifname[IFNAMSIZ];
1680
1681         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1682         if (err < 0)
1683                 goto errout;
1684
1685         if (tb[IFLA_IFNAME])
1686                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1687         else
1688                 ifname[0] = '\0';
1689
1690         err = -EINVAL;
1691         ifm = nlmsg_data(nlh);
1692         if (ifm->ifi_index > 0)
1693                 dev = __dev_get_by_index(net, ifm->ifi_index);
1694         else if (tb[IFLA_IFNAME])
1695                 dev = __dev_get_by_name(net, ifname);
1696         else
1697                 goto errout;
1698
1699         if (dev == NULL) {
1700                 err = -ENODEV;
1701                 goto errout;
1702         }
1703
1704         err = validate_linkmsg(dev, tb);
1705         if (err < 0)
1706                 goto errout;
1707
1708         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1709 errout:
1710         return err;
1711 }
1712
1713 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1714 {
1715         struct net *net = sock_net(skb->sk);
1716         const struct rtnl_link_ops *ops;
1717         struct net_device *dev;
1718         struct ifinfomsg *ifm;
1719         char ifname[IFNAMSIZ];
1720         struct nlattr *tb[IFLA_MAX+1];
1721         int err;
1722         LIST_HEAD(list_kill);
1723
1724         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1725         if (err < 0)
1726                 return err;
1727
1728         if (tb[IFLA_IFNAME])
1729                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1730
1731         ifm = nlmsg_data(nlh);
1732         if (ifm->ifi_index > 0)
1733                 dev = __dev_get_by_index(net, ifm->ifi_index);
1734         else if (tb[IFLA_IFNAME])
1735                 dev = __dev_get_by_name(net, ifname);
1736         else
1737                 return -EINVAL;
1738
1739         if (!dev)
1740                 return -ENODEV;
1741
1742         ops = dev->rtnl_link_ops;
1743         if (!ops)
1744                 return -EOPNOTSUPP;
1745
1746         ops->dellink(dev, &list_kill);
1747         unregister_netdevice_many(&list_kill);
1748         return 0;
1749 }
1750
1751 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1752 {
1753         unsigned int old_flags;
1754         int err;
1755
1756         old_flags = dev->flags;
1757         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1758                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1759                 if (err < 0)
1760                         return err;
1761         }
1762
1763         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1764
1765         __dev_notify_flags(dev, old_flags, ~0U);
1766         return 0;
1767 }
1768 EXPORT_SYMBOL(rtnl_configure_link);
1769
1770 struct net_device *rtnl_create_link(struct net *net,
1771         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1772 {
1773         int err;
1774         struct net_device *dev;
1775         unsigned int num_tx_queues = 1;
1776         unsigned int num_rx_queues = 1;
1777
1778         if (tb[IFLA_NUM_TX_QUEUES])
1779                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1780         else if (ops->get_num_tx_queues)
1781                 num_tx_queues = ops->get_num_tx_queues();
1782
1783         if (tb[IFLA_NUM_RX_QUEUES])
1784                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1785         else if (ops->get_num_rx_queues)
1786                 num_rx_queues = ops->get_num_rx_queues();
1787
1788         err = -ENOMEM;
1789         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1790                                num_tx_queues, num_rx_queues);
1791         if (!dev)
1792                 goto err;
1793
1794         dev_net_set(dev, net);
1795         dev->rtnl_link_ops = ops;
1796         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1797
1798         if (tb[IFLA_MTU])
1799                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1800         if (tb[IFLA_ADDRESS]) {
1801                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1802                                 nla_len(tb[IFLA_ADDRESS]));
1803                 dev->addr_assign_type = NET_ADDR_SET;
1804         }
1805         if (tb[IFLA_BROADCAST])
1806                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1807                                 nla_len(tb[IFLA_BROADCAST]));
1808         if (tb[IFLA_TXQLEN])
1809                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1810         if (tb[IFLA_OPERSTATE])
1811                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1812         if (tb[IFLA_LINKMODE])
1813                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1814         if (tb[IFLA_GROUP])
1815                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1816
1817         return dev;
1818
1819 err:
1820         return ERR_PTR(err);
1821 }
1822 EXPORT_SYMBOL(rtnl_create_link);
1823
1824 static int rtnl_group_changelink(const struct sk_buff *skb,
1825                 struct net *net, int group,
1826                 struct ifinfomsg *ifm,
1827                 struct nlattr **tb)
1828 {
1829         struct net_device *dev;
1830         int err;
1831
1832         for_each_netdev(net, dev) {
1833                 if (dev->group == group) {
1834                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1835                         if (err < 0)
1836                                 return err;
1837                 }
1838         }
1839
1840         return 0;
1841 }
1842
1843 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1844 {
1845         struct net *net = sock_net(skb->sk);
1846         const struct rtnl_link_ops *ops;
1847         const struct rtnl_link_ops *m_ops = NULL;
1848         struct net_device *dev;
1849         struct net_device *master_dev = NULL;
1850         struct ifinfomsg *ifm;
1851         char kind[MODULE_NAME_LEN];
1852         char ifname[IFNAMSIZ];
1853         struct nlattr *tb[IFLA_MAX+1];
1854         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1855         int err;
1856
1857 #ifdef CONFIG_MODULES
1858 replay:
1859 #endif
1860         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1861         if (err < 0)
1862                 return err;
1863
1864         if (tb[IFLA_IFNAME])
1865                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1866         else
1867                 ifname[0] = '\0';
1868
1869         ifm = nlmsg_data(nlh);
1870         if (ifm->ifi_index > 0)
1871                 dev = __dev_get_by_index(net, ifm->ifi_index);
1872         else {
1873                 if (ifname[0])
1874                         dev = __dev_get_by_name(net, ifname);
1875                 else
1876                         dev = NULL;
1877         }
1878
1879         if (dev) {
1880                 master_dev = netdev_master_upper_dev_get(dev);
1881                 if (master_dev)
1882                         m_ops = master_dev->rtnl_link_ops;
1883         }
1884
1885         err = validate_linkmsg(dev, tb);
1886         if (err < 0)
1887                 return err;
1888
1889         if (tb[IFLA_LINKINFO]) {
1890                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1891                                        tb[IFLA_LINKINFO], ifla_info_policy);
1892                 if (err < 0)
1893                         return err;
1894         } else
1895                 memset(linkinfo, 0, sizeof(linkinfo));
1896
1897         if (linkinfo[IFLA_INFO_KIND]) {
1898                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1899                 ops = rtnl_link_ops_get(kind);
1900         } else {
1901                 kind[0] = '\0';
1902                 ops = NULL;
1903         }
1904
1905         if (1) {
1906                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1907                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1908                 struct nlattr **data = NULL;
1909                 struct nlattr **slave_data = NULL;
1910                 struct net *dest_net;
1911
1912                 if (ops) {
1913                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1914                                 err = nla_parse_nested(attr, ops->maxtype,
1915                                                        linkinfo[IFLA_INFO_DATA],
1916                                                        ops->policy);
1917                                 if (err < 0)
1918                                         return err;
1919                                 data = attr;
1920                         }
1921                         if (ops->validate) {
1922                                 err = ops->validate(tb, data);
1923                                 if (err < 0)
1924                                         return err;
1925                         }
1926                 }
1927
1928                 if (m_ops) {
1929                         if (m_ops->slave_maxtype &&
1930                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
1931                                 err = nla_parse_nested(slave_attr,
1932                                                        m_ops->slave_maxtype,
1933                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
1934                                                        m_ops->slave_policy);
1935                                 if (err < 0)
1936                                         return err;
1937                                 slave_data = slave_attr;
1938                         }
1939                         if (m_ops->slave_validate) {
1940                                 err = m_ops->slave_validate(tb, slave_data);
1941                                 if (err < 0)
1942                                         return err;
1943                         }
1944                 }
1945
1946                 if (dev) {
1947                         int modified = 0;
1948
1949                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1950                                 return -EEXIST;
1951                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1952                                 return -EOPNOTSUPP;
1953
1954                         if (linkinfo[IFLA_INFO_DATA]) {
1955                                 if (!ops || ops != dev->rtnl_link_ops ||
1956                                     !ops->changelink)
1957                                         return -EOPNOTSUPP;
1958
1959                                 err = ops->changelink(dev, tb, data);
1960                                 if (err < 0)
1961                                         return err;
1962                                 modified = 1;
1963                         }
1964
1965                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1966                                 if (!m_ops || !m_ops->slave_changelink)
1967                                         return -EOPNOTSUPP;
1968
1969                                 err = m_ops->slave_changelink(master_dev, dev,
1970                                                               tb, slave_data);
1971                                 if (err < 0)
1972                                         return err;
1973                                 modified = 1;
1974                         }
1975
1976                         return do_setlink(skb, dev, ifm, tb, ifname, modified);
1977                 }
1978
1979                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1980                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1981                                 return rtnl_group_changelink(skb, net,
1982                                                 nla_get_u32(tb[IFLA_GROUP]),
1983                                                 ifm, tb);
1984                         return -ENODEV;
1985                 }
1986
1987                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1988                         return -EOPNOTSUPP;
1989
1990                 if (!ops) {
1991 #ifdef CONFIG_MODULES
1992                         if (kind[0]) {
1993                                 __rtnl_unlock();
1994                                 request_module("rtnl-link-%s", kind);
1995                                 rtnl_lock();
1996                                 ops = rtnl_link_ops_get(kind);
1997                                 if (ops)
1998                                         goto replay;
1999                         }
2000 #endif
2001                         return -EOPNOTSUPP;
2002                 }
2003
2004                 if (!ifname[0])
2005                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2006
2007                 dest_net = rtnl_link_get_net(net, tb);
2008                 if (IS_ERR(dest_net))
2009                         return PTR_ERR(dest_net);
2010
2011                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
2012                 if (IS_ERR(dev)) {
2013                         err = PTR_ERR(dev);
2014                         goto out;
2015                 }
2016
2017                 dev->ifindex = ifm->ifi_index;
2018
2019                 if (ops->newlink) {
2020                         err = ops->newlink(net, dev, tb, data);
2021                         /* Drivers should call free_netdev() in ->destructor
2022                          * and unregister it on failure so that device could be
2023                          * finally freed in rtnl_unlock.
2024                          */
2025                         if (err < 0)
2026                                 goto out;
2027                 } else {
2028                         err = register_netdevice(dev);
2029                         if (err < 0) {
2030                                 free_netdev(dev);
2031                                 goto out;
2032                         }
2033                 }
2034                 err = rtnl_configure_link(dev, ifm);
2035                 if (err < 0)
2036                         unregister_netdevice(dev);
2037 out:
2038                 put_net(dest_net);
2039                 return err;
2040         }
2041 }
2042
2043 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2044 {
2045         struct net *net = sock_net(skb->sk);
2046         struct ifinfomsg *ifm;
2047         char ifname[IFNAMSIZ];
2048         struct nlattr *tb[IFLA_MAX+1];
2049         struct net_device *dev = NULL;
2050         struct sk_buff *nskb;
2051         int err;
2052         u32 ext_filter_mask = 0;
2053
2054         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2055         if (err < 0)
2056                 return err;
2057
2058         if (tb[IFLA_IFNAME])
2059                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2060
2061         if (tb[IFLA_EXT_MASK])
2062                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2063
2064         ifm = nlmsg_data(nlh);
2065         if (ifm->ifi_index > 0)
2066                 dev = __dev_get_by_index(net, ifm->ifi_index);
2067         else if (tb[IFLA_IFNAME])
2068                 dev = __dev_get_by_name(net, ifname);
2069         else
2070                 return -EINVAL;
2071
2072         if (dev == NULL)
2073                 return -ENODEV;
2074
2075         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2076         if (nskb == NULL)
2077                 return -ENOBUFS;
2078
2079         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2080                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2081         if (err < 0) {
2082                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2083                 WARN_ON(err == -EMSGSIZE);
2084                 kfree_skb(nskb);
2085         } else
2086                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2087
2088         return err;
2089 }
2090
2091 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2092 {
2093         struct net *net = sock_net(skb->sk);
2094         struct net_device *dev;
2095         struct nlattr *tb[IFLA_MAX+1];
2096         u32 ext_filter_mask = 0;
2097         u16 min_ifinfo_dump_size = 0;
2098         int hdrlen;
2099
2100         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2101         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2102                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2103
2104         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2105                 if (tb[IFLA_EXT_MASK])
2106                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2107         }
2108
2109         if (!ext_filter_mask)
2110                 return NLMSG_GOODSIZE;
2111         /*
2112          * traverse the list of net devices and compute the minimum
2113          * buffer size based upon the filter mask.
2114          */
2115         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2116                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2117                                              if_nlmsg_size(dev,
2118                                                            ext_filter_mask));
2119         }
2120
2121         return min_ifinfo_dump_size;
2122 }
2123
2124 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2125 {
2126         int idx;
2127         int s_idx = cb->family;
2128
2129         if (s_idx == 0)
2130                 s_idx = 1;
2131         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2132                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2133                 if (idx < s_idx || idx == PF_PACKET)
2134                         continue;
2135                 if (rtnl_msg_handlers[idx] == NULL ||
2136                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2137                         continue;
2138                 if (idx > s_idx) {
2139                         memset(&cb->args[0], 0, sizeof(cb->args));
2140                         cb->prev_seq = 0;
2141                         cb->seq = 0;
2142                 }
2143                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2144                         break;
2145         }
2146         cb->family = idx;
2147
2148         return skb->len;
2149 }
2150
2151 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2152                   gfp_t flags)
2153 {
2154         struct net *net = dev_net(dev);
2155         struct sk_buff *skb;
2156         int err = -ENOBUFS;
2157         size_t if_info_size;
2158
2159         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2160         if (skb == NULL)
2161                 goto errout;
2162
2163         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2164         if (err < 0) {
2165                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2166                 WARN_ON(err == -EMSGSIZE);
2167                 kfree_skb(skb);
2168                 goto errout;
2169         }
2170         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2171         return;
2172 errout:
2173         if (err < 0)
2174                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2175 }
2176 EXPORT_SYMBOL(rtmsg_ifinfo);
2177
2178 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2179                                    struct net_device *dev,
2180                                    u8 *addr, u32 pid, u32 seq,
2181                                    int type, unsigned int flags,
2182                                    int nlflags)
2183 {
2184         struct nlmsghdr *nlh;
2185         struct ndmsg *ndm;
2186
2187         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2188         if (!nlh)
2189                 return -EMSGSIZE;
2190
2191         ndm = nlmsg_data(nlh);
2192         ndm->ndm_family  = AF_BRIDGE;
2193         ndm->ndm_pad1    = 0;
2194         ndm->ndm_pad2    = 0;
2195         ndm->ndm_flags   = flags;
2196         ndm->ndm_type    = 0;
2197         ndm->ndm_ifindex = dev->ifindex;
2198         ndm->ndm_state   = NUD_PERMANENT;
2199
2200         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2201                 goto nla_put_failure;
2202
2203         return nlmsg_end(skb, nlh);
2204
2205 nla_put_failure:
2206         nlmsg_cancel(skb, nlh);
2207         return -EMSGSIZE;
2208 }
2209
2210 static inline size_t rtnl_fdb_nlmsg_size(void)
2211 {
2212         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2213 }
2214
2215 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2216 {
2217         struct net *net = dev_net(dev);
2218         struct sk_buff *skb;
2219         int err = -ENOBUFS;
2220
2221         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2222         if (!skb)
2223                 goto errout;
2224
2225         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2226         if (err < 0) {
2227                 kfree_skb(skb);
2228                 goto errout;
2229         }
2230
2231         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2232         return;
2233 errout:
2234         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2235 }
2236
2237 /**
2238  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2239  */
2240 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2241                      struct nlattr *tb[],
2242                      struct net_device *dev,
2243                      const unsigned char *addr,
2244                      u16 flags)
2245 {
2246         int err = -EINVAL;
2247
2248         /* If aging addresses are supported device will need to
2249          * implement its own handler for this.
2250          */
2251         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2252                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2253                 return err;
2254         }
2255
2256         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2257                 err = dev_uc_add_excl(dev, addr);
2258         else if (is_multicast_ether_addr(addr))
2259                 err = dev_mc_add_excl(dev, addr);
2260
2261         /* Only return duplicate errors if NLM_F_EXCL is set */
2262         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2263                 err = 0;
2264
2265         return err;
2266 }
2267 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2268
2269 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2270 {
2271         struct net *net = sock_net(skb->sk);
2272         struct ndmsg *ndm;
2273         struct nlattr *tb[NDA_MAX+1];
2274         struct net_device *dev;
2275         u8 *addr;
2276         int err;
2277
2278         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2279         if (err < 0)
2280                 return err;
2281
2282         ndm = nlmsg_data(nlh);
2283         if (ndm->ndm_ifindex == 0) {
2284                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2285                 return -EINVAL;
2286         }
2287
2288         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2289         if (dev == NULL) {
2290                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2291                 return -ENODEV;
2292         }
2293
2294         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2295                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2296                 return -EINVAL;
2297         }
2298
2299         addr = nla_data(tb[NDA_LLADDR]);
2300
2301         err = -EOPNOTSUPP;
2302
2303         /* Support fdb on master device the net/bridge default case */
2304         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2305             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2306                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2307                 const struct net_device_ops *ops = br_dev->netdev_ops;
2308
2309                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2310                 if (err)
2311                         goto out;
2312                 else
2313                         ndm->ndm_flags &= ~NTF_MASTER;
2314         }
2315
2316         /* Embedded bridge, macvlan, and any other device support */
2317         if ((ndm->ndm_flags & NTF_SELF)) {
2318                 if (dev->netdev_ops->ndo_fdb_add)
2319                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2320                                                            nlh->nlmsg_flags);
2321                 else
2322                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2323                                                nlh->nlmsg_flags);
2324
2325                 if (!err) {
2326                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2327                         ndm->ndm_flags &= ~NTF_SELF;
2328                 }
2329         }
2330 out:
2331         return err;
2332 }
2333
2334 /**
2335  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2336  */
2337 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2338                      struct nlattr *tb[],
2339                      struct net_device *dev,
2340                      const unsigned char *addr)
2341 {
2342         int err = -EOPNOTSUPP;
2343
2344         /* If aging addresses are supported device will need to
2345          * implement its own handler for this.
2346          */
2347         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2348                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2349                 return -EINVAL;
2350         }
2351
2352         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2353                 err = dev_uc_del(dev, addr);
2354         else if (is_multicast_ether_addr(addr))
2355                 err = dev_mc_del(dev, addr);
2356         else
2357                 err = -EINVAL;
2358
2359         return err;
2360 }
2361 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2362
2363 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2364 {
2365         struct net *net = sock_net(skb->sk);
2366         struct ndmsg *ndm;
2367         struct nlattr *tb[NDA_MAX+1];
2368         struct net_device *dev;
2369         int err = -EINVAL;
2370         __u8 *addr;
2371
2372         if (!netlink_capable(skb, CAP_NET_ADMIN))
2373                 return -EPERM;
2374
2375         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2376         if (err < 0)
2377                 return err;
2378
2379         ndm = nlmsg_data(nlh);
2380         if (ndm->ndm_ifindex == 0) {
2381                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2382                 return -EINVAL;
2383         }
2384
2385         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2386         if (dev == NULL) {
2387                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2388                 return -ENODEV;
2389         }
2390
2391         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2392                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2393                 return -EINVAL;
2394         }
2395
2396         addr = nla_data(tb[NDA_LLADDR]);
2397
2398         err = -EOPNOTSUPP;
2399
2400         /* Support fdb on master device the net/bridge default case */
2401         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2402             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2403                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2404                 const struct net_device_ops *ops = br_dev->netdev_ops;
2405
2406                 if (ops->ndo_fdb_del)
2407                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2408
2409                 if (err)
2410                         goto out;
2411                 else
2412                         ndm->ndm_flags &= ~NTF_MASTER;
2413         }
2414
2415         /* Embedded bridge, macvlan, and any other device support */
2416         if (ndm->ndm_flags & NTF_SELF) {
2417                 if (dev->netdev_ops->ndo_fdb_del)
2418                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2419                 else
2420                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2421
2422                 if (!err) {
2423                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2424                         ndm->ndm_flags &= ~NTF_SELF;
2425                 }
2426         }
2427 out:
2428         return err;
2429 }
2430
2431 static int nlmsg_populate_fdb(struct sk_buff *skb,
2432                               struct netlink_callback *cb,
2433                               struct net_device *dev,
2434                               int *idx,
2435                               struct netdev_hw_addr_list *list)
2436 {
2437         struct netdev_hw_addr *ha;
2438         int err;
2439         u32 portid, seq;
2440
2441         portid = NETLINK_CB(cb->skb).portid;
2442         seq = cb->nlh->nlmsg_seq;
2443
2444         list_for_each_entry(ha, &list->list, list) {
2445                 if (*idx < cb->args[0])
2446                         goto skip;
2447
2448                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2449                                               portid, seq,
2450                                               RTM_NEWNEIGH, NTF_SELF,
2451                                               NLM_F_MULTI);
2452                 if (err < 0)
2453                         return err;
2454 skip:
2455                 *idx += 1;
2456         }
2457         return 0;
2458 }
2459
2460 /**
2461  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2462  * @nlh: netlink message header
2463  * @dev: netdevice
2464  *
2465  * Default netdevice operation to dump the existing unicast address list.
2466  * Returns number of addresses from list put in skb.
2467  */
2468 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2469                       struct netlink_callback *cb,
2470                       struct net_device *dev,
2471                       int idx)
2472 {
2473         int err;
2474
2475         netif_addr_lock_bh(dev);
2476         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2477         if (err)
2478                 goto out;
2479         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2480 out:
2481         netif_addr_unlock_bh(dev);
2482         return idx;
2483 }
2484 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2485
2486 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2487 {
2488         int idx = 0;
2489         struct net *net = sock_net(skb->sk);
2490         struct net_device *dev;
2491
2492         rcu_read_lock();
2493         for_each_netdev_rcu(net, dev) {
2494                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2495                         struct net_device *br_dev;
2496                         const struct net_device_ops *ops;
2497
2498                         br_dev = netdev_master_upper_dev_get(dev);
2499                         ops = br_dev->netdev_ops;
2500                         if (ops->ndo_fdb_dump)
2501                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2502                 }
2503
2504                 if (dev->netdev_ops->ndo_fdb_dump)
2505                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2506                 else
2507                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2508         }
2509         rcu_read_unlock();
2510
2511         cb->args[0] = idx;
2512         return skb->len;
2513 }
2514
2515 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2516                             struct net_device *dev, u16 mode)
2517 {
2518         struct nlmsghdr *nlh;
2519         struct ifinfomsg *ifm;
2520         struct nlattr *br_afspec;
2521         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2522         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2523
2524         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2525         if (nlh == NULL)
2526                 return -EMSGSIZE;
2527
2528         ifm = nlmsg_data(nlh);
2529         ifm->ifi_family = AF_BRIDGE;
2530         ifm->__ifi_pad = 0;
2531         ifm->ifi_type = dev->type;
2532         ifm->ifi_index = dev->ifindex;
2533         ifm->ifi_flags = dev_get_flags(dev);
2534         ifm->ifi_change = 0;
2535
2536
2537         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2538             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2539             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2540             (br_dev &&
2541              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2542             (dev->addr_len &&
2543              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2544             (dev->ifindex != dev->iflink &&
2545              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2546                 goto nla_put_failure;
2547
2548         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2549         if (!br_afspec)
2550                 goto nla_put_failure;
2551
2552         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2553             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2554                 nla_nest_cancel(skb, br_afspec);
2555                 goto nla_put_failure;
2556         }
2557         nla_nest_end(skb, br_afspec);
2558
2559         return nlmsg_end(skb, nlh);
2560 nla_put_failure:
2561         nlmsg_cancel(skb, nlh);
2562         return -EMSGSIZE;
2563 }
2564 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2565
2566 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2567 {
2568         struct net *net = sock_net(skb->sk);
2569         struct net_device *dev;
2570         int idx = 0;
2571         u32 portid = NETLINK_CB(cb->skb).portid;
2572         u32 seq = cb->nlh->nlmsg_seq;
2573         struct nlattr *extfilt;
2574         u32 filter_mask = 0;
2575
2576         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2577                                   IFLA_EXT_MASK);
2578         if (extfilt)
2579                 filter_mask = nla_get_u32(extfilt);
2580
2581         rcu_read_lock();
2582         for_each_netdev_rcu(net, dev) {
2583                 const struct net_device_ops *ops = dev->netdev_ops;
2584                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2585
2586                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2587                         if (idx >= cb->args[0] &&
2588                             br_dev->netdev_ops->ndo_bridge_getlink(
2589                                     skb, portid, seq, dev, filter_mask) < 0)
2590                                 break;
2591                         idx++;
2592                 }
2593
2594                 if (ops->ndo_bridge_getlink) {
2595                         if (idx >= cb->args[0] &&
2596                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2597                                                     filter_mask) < 0)
2598                                 break;
2599                         idx++;
2600                 }
2601         }
2602         rcu_read_unlock();
2603         cb->args[0] = idx;
2604
2605         return skb->len;
2606 }
2607
2608 static inline size_t bridge_nlmsg_size(void)
2609 {
2610         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2611                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2612                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2613                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2614                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2615                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2616                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2617                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2618                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2619                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2620                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2621 }
2622
2623 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2624 {
2625         struct net *net = dev_net(dev);
2626         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2627         struct sk_buff *skb;
2628         int err = -EOPNOTSUPP;
2629
2630         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2631         if (!skb) {
2632                 err = -ENOMEM;
2633                 goto errout;
2634         }
2635
2636         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2637             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2638                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2639                 if (err < 0)
2640                         goto errout;
2641         }
2642
2643         if ((flags & BRIDGE_FLAGS_SELF) &&
2644             dev->netdev_ops->ndo_bridge_getlink) {
2645                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2646                 if (err < 0)
2647                         goto errout;
2648         }
2649
2650         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2651         return 0;
2652 errout:
2653         WARN_ON(err == -EMSGSIZE);
2654         kfree_skb(skb);
2655         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2656         return err;
2657 }
2658
2659 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2660 {
2661         struct net *net = sock_net(skb->sk);
2662         struct ifinfomsg *ifm;
2663         struct net_device *dev;
2664         struct nlattr *br_spec, *attr = NULL;
2665         int rem, err = -EOPNOTSUPP;
2666         u16 oflags, flags = 0;
2667         bool have_flags = false;
2668
2669         if (nlmsg_len(nlh) < sizeof(*ifm))
2670                 return -EINVAL;
2671
2672         ifm = nlmsg_data(nlh);
2673         if (ifm->ifi_family != AF_BRIDGE)
2674                 return -EPFNOSUPPORT;
2675
2676         dev = __dev_get_by_index(net, ifm->ifi_index);
2677         if (!dev) {
2678                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2679                 return -ENODEV;
2680         }
2681
2682         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2683         if (br_spec) {
2684                 nla_for_each_nested(attr, br_spec, rem) {
2685                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2686                                 have_flags = true;
2687                                 flags = nla_get_u16(attr);
2688                                 break;
2689                         }
2690                 }
2691         }
2692
2693         oflags = flags;
2694
2695         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2696                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2697
2698                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2699                         err = -EOPNOTSUPP;
2700                         goto out;
2701                 }
2702
2703                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2704                 if (err)
2705                         goto out;
2706
2707                 flags &= ~BRIDGE_FLAGS_MASTER;
2708         }
2709
2710         if ((flags & BRIDGE_FLAGS_SELF)) {
2711                 if (!dev->netdev_ops->ndo_bridge_setlink)
2712                         err = -EOPNOTSUPP;
2713                 else
2714                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2715
2716                 if (!err)
2717                         flags &= ~BRIDGE_FLAGS_SELF;
2718         }
2719
2720         if (have_flags)
2721                 memcpy(nla_data(attr), &flags, sizeof(flags));
2722         /* Generate event to notify upper layer of bridge change */
2723         if (!err)
2724                 err = rtnl_bridge_notify(dev, oflags);
2725 out:
2726         return err;
2727 }
2728
2729 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2730 {
2731         struct net *net = sock_net(skb->sk);
2732         struct ifinfomsg *ifm;
2733         struct net_device *dev;
2734         struct nlattr *br_spec, *attr = NULL;
2735         int rem, err = -EOPNOTSUPP;
2736         u16 oflags, flags = 0;
2737         bool have_flags = false;
2738
2739         if (nlmsg_len(nlh) < sizeof(*ifm))
2740                 return -EINVAL;
2741
2742         ifm = nlmsg_data(nlh);
2743         if (ifm->ifi_family != AF_BRIDGE)
2744                 return -EPFNOSUPPORT;
2745
2746         dev = __dev_get_by_index(net, ifm->ifi_index);
2747         if (!dev) {
2748                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2749                 return -ENODEV;
2750         }
2751
2752         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2753         if (br_spec) {
2754                 nla_for_each_nested(attr, br_spec, rem) {
2755                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2756                                 have_flags = true;
2757                                 flags = nla_get_u16(attr);
2758                                 break;
2759                         }
2760                 }
2761         }
2762
2763         oflags = flags;
2764
2765         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2766                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2767
2768                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2769                         err = -EOPNOTSUPP;
2770                         goto out;
2771                 }
2772
2773                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2774                 if (err)
2775                         goto out;
2776
2777                 flags &= ~BRIDGE_FLAGS_MASTER;
2778         }
2779
2780         if ((flags & BRIDGE_FLAGS_SELF)) {
2781                 if (!dev->netdev_ops->ndo_bridge_dellink)
2782                         err = -EOPNOTSUPP;
2783                 else
2784                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2785
2786                 if (!err)
2787                         flags &= ~BRIDGE_FLAGS_SELF;
2788         }
2789
2790         if (have_flags)
2791                 memcpy(nla_data(attr), &flags, sizeof(flags));
2792         /* Generate event to notify upper layer of bridge change */
2793         if (!err)
2794                 err = rtnl_bridge_notify(dev, oflags);
2795 out:
2796         return err;
2797 }
2798
2799 /* Process one rtnetlink message. */
2800
2801 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2802 {
2803         struct net *net = sock_net(skb->sk);
2804         rtnl_doit_func doit;
2805         int sz_idx, kind;
2806         int family;
2807         int type;
2808         int err;
2809
2810         type = nlh->nlmsg_type;
2811         if (type > RTM_MAX)
2812                 return -EOPNOTSUPP;
2813
2814         type -= RTM_BASE;
2815
2816         /* All the messages must have at least 1 byte length */
2817         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2818                 return 0;
2819
2820         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2821         sz_idx = type>>2;
2822         kind = type&3;
2823
2824         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2825                 return -EPERM;
2826
2827         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2828                 struct sock *rtnl;
2829                 rtnl_dumpit_func dumpit;
2830                 rtnl_calcit_func calcit;
2831                 u16 min_dump_alloc = 0;
2832
2833                 dumpit = rtnl_get_dumpit(family, type);
2834                 if (dumpit == NULL)
2835                         return -EOPNOTSUPP;
2836                 calcit = rtnl_get_calcit(family, type);
2837                 if (calcit)
2838                         min_dump_alloc = calcit(skb, nlh);
2839
2840                 __rtnl_unlock();
2841                 rtnl = net->rtnl;
2842                 {
2843                         struct netlink_dump_control c = {
2844                                 .dump           = dumpit,
2845                                 .min_dump_alloc = min_dump_alloc,
2846                         };
2847                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2848                 }
2849                 rtnl_lock();
2850                 return err;
2851         }
2852
2853         doit = rtnl_get_doit(family, type);
2854         if (doit == NULL)
2855                 return -EOPNOTSUPP;
2856
2857         return doit(skb, nlh);
2858 }
2859
2860 static void rtnetlink_rcv(struct sk_buff *skb)
2861 {
2862         rtnl_lock();
2863         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2864         rtnl_unlock();
2865 }
2866
2867 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2868 {
2869         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2870
2871         switch (event) {
2872         case NETDEV_UP:
2873         case NETDEV_DOWN:
2874         case NETDEV_PRE_UP:
2875         case NETDEV_POST_INIT:
2876         case NETDEV_REGISTER:
2877         case NETDEV_CHANGE:
2878         case NETDEV_PRE_TYPE_CHANGE:
2879         case NETDEV_GOING_DOWN:
2880         case NETDEV_UNREGISTER:
2881         case NETDEV_UNREGISTER_FINAL:
2882         case NETDEV_RELEASE:
2883         case NETDEV_JOIN:
2884                 break;
2885         default:
2886                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2887                 break;
2888         }
2889         return NOTIFY_DONE;
2890 }
2891
2892 static struct notifier_block rtnetlink_dev_notifier = {
2893         .notifier_call  = rtnetlink_event,
2894 };
2895
2896
2897 static int __net_init rtnetlink_net_init(struct net *net)
2898 {
2899         struct sock *sk;
2900         struct netlink_kernel_cfg cfg = {
2901                 .groups         = RTNLGRP_MAX,
2902                 .input          = rtnetlink_rcv,
2903                 .cb_mutex       = &rtnl_mutex,
2904                 .flags          = NL_CFG_F_NONROOT_RECV,
2905         };
2906
2907         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2908         if (!sk)
2909                 return -ENOMEM;
2910         net->rtnl = sk;
2911         return 0;
2912 }
2913
2914 static void __net_exit rtnetlink_net_exit(struct net *net)
2915 {
2916         netlink_kernel_release(net->rtnl);
2917         net->rtnl = NULL;
2918 }
2919
2920 static struct pernet_operations rtnetlink_net_ops = {
2921         .init = rtnetlink_net_init,
2922         .exit = rtnetlink_net_exit,
2923 };
2924
2925 void __init rtnetlink_init(void)
2926 {
2927         if (register_pernet_subsys(&rtnetlink_net_ops))
2928                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2929
2930         register_netdevice_notifier(&rtnetlink_dev_notifier);
2931
2932         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2933                       rtnl_dump_ifinfo, rtnl_calcit);
2934         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2935         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2936         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2937
2938         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2939         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2940
2941         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2942         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2943         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2944
2945         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2946         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2947         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2948 }
2949