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/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 int lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256
 257         return 0;
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_unregister);
 260
 261 /**
 262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 263  * @protocol : Protocol family or PF_UNSPEC
 264  *
 265  * Identical to calling rtnl_unregster() for all registered message types
 266  * of a certain protocol family.
 267  */
 268 void rtnl_unregister_all(int protocol)
 269 {
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271
 272         kfree(rtnl_msg_handlers[protocol]);
 273         rtnl_msg_handlers[protocol] = NULL;
 274 }
 275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 276
 277 static LIST_HEAD(link_ops);
 278
 279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 280 {
 281         const struct rtnl_link_ops *ops;
 282
 283         list_for_each_entry(ops, &link_ops, list) {
 284                 if (!strcmp(ops->kind, kind))
 285                         return ops;
 286         }
 287         return NULL;
 288 }
 289
 290 /**
 291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 292  * @ops: struct rtnl_link_ops * to register
 293  *
 294  * The caller must hold the rtnl_mutex. This function should be used
 295  * by drivers that create devices during module initialization. It
 296  * must be called before registering the devices.
 297  *
 298  * Returns 0 on success or a negative error code.
 299  */
 300 int __rtnl_link_register(struct rtnl_link_ops *ops)
 301 {
 302         if (rtnl_link_ops_get(ops->kind))
 303                 return -EEXIST;
 304
 305         /* The check for setup is here because if ops
 306          * does not have that filled up, it is not possible
 307          * to use the ops for creating device. So do not
 308          * fill up dellink as well. That disables rtnl_dellink.
 309          */
 310         if (ops->setup && !ops->dellink)
 311                 ops->dellink = unregister_netdevice_queue;
 312
 313         list_add_tail(&ops->list, &link_ops);
 314         return 0;
 315 }
 316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 317
 318 /**
 319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 320  * @ops: struct rtnl_link_ops * to register
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_link_register(struct rtnl_link_ops *ops)
 325 {
 326         int err;
 327
 328         rtnl_lock();
 329         err = __rtnl_link_register(ops);
 330         rtnl_unlock();
 331         return err;
 332 }
 333 EXPORT_SYMBOL_GPL(rtnl_link_register);
 334
 335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 336 {
 337         struct net_device *dev;
 338         LIST_HEAD(list_kill);
 339
 340         for_each_netdev(net, dev) {
 341                 if (dev->rtnl_link_ops == ops)
 342                         ops->dellink(dev, &list_kill);
 343         }
 344         unregister_netdevice_many(&list_kill);
 345 }
 346
 347 /**
 348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 349  * @ops: struct rtnl_link_ops * to unregister
 350  *
 351  * The caller must hold the rtnl_mutex.
 352  */
 353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 354 {
 355         struct net *net;
 356
 357         for_each_net(net) {
 358                 __rtnl_kill_links(net, ops);
 359         }
 360         list_del(&ops->list);
 361 }
 362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 363
 364 /* Return with the rtnl_lock held when there are no network
 365  * devices unregistering in any network namespace.
 366  */
 367 static void rtnl_lock_unregistering_all(void)
 368 {
 369         struct net *net;
 370         bool unregistering;
 371         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 372
 373         add_wait_queue(&netdev_unregistering_wq, &wait);
 374         for (;;) {
 375                 unregistering = false;
 376                 rtnl_lock();
 377                 for_each_net(net) {
 378                         if (net->dev_unreg_count > 0) {
 379                                 unregistering = true;
 380                                 break;
 381                         }
 382                 }
 383                 if (!unregistering)
 384                         break;
 385                 __rtnl_unlock();
 386
 387                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 388         }
 389         remove_wait_queue(&netdev_unregistering_wq, &wait);
 390 }
 391
 392 /**
 393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 394  * @ops: struct rtnl_link_ops * to unregister
 395  */
 396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 397 {
 398         /* Close the race with cleanup_net() */
 399         mutex_lock(&net_mutex);
 400         rtnl_lock_unregistering_all();
 401         __rtnl_link_unregister(ops);
 402         rtnl_unlock();
 403         mutex_unlock(&net_mutex);
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 406
 407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 408 {
 409         struct net_device *master_dev;
 410         const struct rtnl_link_ops *ops;
 411
 412         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 413         if (!master_dev)
 414                 return 0;
 415         ops = master_dev->rtnl_link_ops;
 416         if (!ops || !ops->get_slave_size)
 417                 return 0;
 418         /* IFLA_INFO_SLAVE_DATA + nested data */
 419         return nla_total_size(sizeof(struct nlattr)) +
 420                ops->get_slave_size(master_dev, dev);
 421 }
 422
 423 static size_t rtnl_link_get_size(const struct net_device *dev)
 424 {
 425         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 426         size_t size;
 427
 428         if (!ops)
 429                 return 0;
 430
 431         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 432                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 433
 434         if (ops->get_size)
 435                 /* IFLA_INFO_DATA + nested data */
 436                 size += nla_total_size(sizeof(struct nlattr)) +
 437                         ops->get_size(dev);
 438
 439         if (ops->get_xstats_size)
 440                 /* IFLA_INFO_XSTATS */
 441                 size += nla_total_size(ops->get_xstats_size(dev));
 442
 443         size += rtnl_link_get_slave_info_data_size(dev);
 444
 445         return size;
 446 }
 447
 448 static LIST_HEAD(rtnl_af_ops);
 449
 450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 451 {
 452         const struct rtnl_af_ops *ops;
 453
 454         list_for_each_entry(ops, &rtnl_af_ops, list) {
 455                 if (ops->family == family)
 456                         return ops;
 457         }
 458
 459         return NULL;
 460 }
 461
 462 /**
 463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 464  * @ops: struct rtnl_af_ops * to register
 465  *
 466  * Returns 0 on success or a negative error code.
 467  */
 468 void rtnl_af_register(struct rtnl_af_ops *ops)
 469 {
 470         rtnl_lock();
 471         list_add_tail(&ops->list, &rtnl_af_ops);
 472         rtnl_unlock();
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_af_register);
 475
 476 /**
 477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 478  * @ops: struct rtnl_af_ops * to unregister
 479  *
 480  * The caller must hold the rtnl_mutex.
 481  */
 482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         list_del(&ops->list);
 485 }
 486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 487
 488 /**
 489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 490  * @ops: struct rtnl_af_ops * to unregister
 491  */
 492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 493 {
 494         rtnl_lock();
 495         __rtnl_af_unregister(ops);
 496         rtnl_unlock();
 497 }
 498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 499
 500 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 501 {
 502         struct rtnl_af_ops *af_ops;
 503         size_t size;
 504
 505         /* IFLA_AF_SPEC */
 506         size = nla_total_size(sizeof(struct nlattr));
 507
 508         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 509                 if (af_ops->get_link_af_size) {
 510                         /* AF_* + nested data */
 511                         size += nla_total_size(sizeof(struct nlattr)) +
 512                                 af_ops->get_link_af_size(dev);
 513                 }
 514         }
 515
 516         return size;
 517 }
 518
 519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 520 {
 521         struct net_device *master_dev;
 522
 523         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 524         if (master_dev && master_dev->rtnl_link_ops)
 525                 return true;
 526         return false;
 527 }
 528
 529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 530                                      const struct net_device *dev)
 531 {
 532         struct net_device *master_dev;
 533         const struct rtnl_link_ops *ops;
 534         struct nlattr *slave_data;
 535         int err;
 536
 537         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 538         if (!master_dev)
 539                 return 0;
 540         ops = master_dev->rtnl_link_ops;
 541         if (!ops)
 542                 return 0;
 543         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 544                 return -EMSGSIZE;
 545         if (ops->fill_slave_info) {
 546                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 547                 if (!slave_data)
 548                         return -EMSGSIZE;
 549                 err = ops->fill_slave_info(skb, master_dev, dev);
 550                 if (err < 0)
 551                         goto err_cancel_slave_data;
 552                 nla_nest_end(skb, slave_data);
 553         }
 554         return 0;
 555
 556 err_cancel_slave_data:
 557         nla_nest_cancel(skb, slave_data);
 558         return err;
 559 }
 560
 561 static int rtnl_link_info_fill(struct sk_buff *skb,
 562                                const struct net_device *dev)
 563 {
 564         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 565         struct nlattr *data;
 566         int err;
 567
 568         if (!ops)
 569                 return 0;
 570         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 571                 return -EMSGSIZE;
 572         if (ops->fill_xstats) {
 573                 err = ops->fill_xstats(skb, dev);
 574                 if (err < 0)
 575                         return err;
 576         }
 577         if (ops->fill_info) {
 578                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 579                 if (data == NULL)
 580                         return -EMSGSIZE;
 581                 err = ops->fill_info(skb, dev);
 582                 if (err < 0)
 583                         goto err_cancel_data;
 584                 nla_nest_end(skb, data);
 585         }
 586         return 0;
 587
 588 err_cancel_data:
 589         nla_nest_cancel(skb, data);
 590         return err;
 591 }
 592
 593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 594 {
 595         struct nlattr *linkinfo;
 596         int err = -EMSGSIZE;
 597
 598         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 599         if (linkinfo == NULL)
 600                 goto out;
 601
 602         err = rtnl_link_info_fill(skb, dev);
 603         if (err < 0)
 604                 goto err_cancel_link;
 605
 606         err = rtnl_link_slave_info_fill(skb, dev);
 607         if (err < 0)
 608                 goto err_cancel_link;
 609
 610         nla_nest_end(skb, linkinfo);
 611         return 0;
 612
 613 err_cancel_link:
 614         nla_nest_cancel(skb, linkinfo);
 615 out:
 616         return err;
 617 }
 618
 619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 620 {
 621         struct sock *rtnl = net->rtnl;
 622         int err = 0;
 623
 624         NETLINK_CB(skb).dst_group = group;
 625         if (echo)
 626                 atomic_inc(&skb->users);
 627         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 628         if (echo)
 629                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 630         return err;
 631 }
 632
 633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 634 {
 635         struct sock *rtnl = net->rtnl;
 636
 637         return nlmsg_unicast(rtnl, skb, pid);
 638 }
 639 EXPORT_SYMBOL(rtnl_unicast);
 640
 641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 642                  struct nlmsghdr *nlh, gfp_t flags)
 643 {
 644         struct sock *rtnl = net->rtnl;
 645         int report = 0;
 646
 647         if (nlh)
 648                 report = nlmsg_report(nlh);
 649
 650         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 651 }
 652 EXPORT_SYMBOL(rtnl_notify);
 653
 654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 655 {
 656         struct sock *rtnl = net->rtnl;
 657
 658         netlink_set_err(rtnl, 0, group, error);
 659 }
 660 EXPORT_SYMBOL(rtnl_set_sk_err);
 661
 662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 663 {
 664         struct nlattr *mx;
 665         int i, valid = 0;
 666
 667         mx = nla_nest_start(skb, RTA_METRICS);
 668         if (mx == NULL)
 669                 return -ENOBUFS;
 670
 671         for (i = 0; i < RTAX_MAX; i++) {
 672                 if (metrics[i]) {
 673                         if (i == RTAX_CC_ALGO - 1) {
 674                                 char tmp[TCP_CA_NAME_MAX], *name;
 675
 676                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 677                                 if (!name)
 678                                         continue;
 679                                 if (nla_put_string(skb, i + 1, name))
 680                                         goto nla_put_failure;
 681                         } else {
 682                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 683                                         goto nla_put_failure;
 684                         }
 685                         valid++;
 686                 }
 687         }
 688
 689         if (!valid) {
 690                 nla_nest_cancel(skb, mx);
 691                 return 0;
 692         }
 693
 694         return nla_nest_end(skb, mx);
 695
 696 nla_put_failure:
 697         nla_nest_cancel(skb, mx);
 698         return -EMSGSIZE;
 699 }
 700 EXPORT_SYMBOL(rtnetlink_put_metrics);
 701
 702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 703                        long expires, u32 error)
 704 {
 705         struct rta_cacheinfo ci = {
 706                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 707                 .rta_used = dst->__use,
 708                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 709                 .rta_error = error,
 710                 .rta_id =  id,
 711         };
 712
 713         if (expires) {
 714                 unsigned long clock;
 715
 716                 clock = jiffies_to_clock_t(abs(expires));
 717                 clock = min_t(unsigned long, clock, INT_MAX);
 718                 ci.rta_expires = (expires > 0) ? clock : -clock;
 719         }
 720         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 721 }
 722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 723
 724 static void set_operstate(struct net_device *dev, unsigned char transition)
 725 {
 726         unsigned char operstate = dev->operstate;
 727
 728         switch (transition) {
 729         case IF_OPER_UP:
 730                 if ((operstate == IF_OPER_DORMANT ||
 731                      operstate == IF_OPER_UNKNOWN) &&
 732                     !netif_dormant(dev))
 733                         operstate = IF_OPER_UP;
 734                 break;
 735
 736         case IF_OPER_DORMANT:
 737                 if (operstate == IF_OPER_UP ||
 738                     operstate == IF_OPER_UNKNOWN)
 739                         operstate = IF_OPER_DORMANT;
 740                 break;
 741         }
 742
 743         if (dev->operstate != operstate) {
 744                 write_lock_bh(&dev_base_lock);
 745                 dev->operstate = operstate;
 746                 write_unlock_bh(&dev_base_lock);
 747                 netdev_state_change(dev);
 748         }
 749 }
 750
 751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 752 {
 753         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 754                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 755 }
 756
 757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 758                                            const struct ifinfomsg *ifm)
 759 {
 760         unsigned int flags = ifm->ifi_flags;
 761
 762         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 763         if (ifm->ifi_change)
 764                 flags = (flags & ifm->ifi_change) |
 765                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 766
 767         return flags;
 768 }
 769
 770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 771                                  const struct rtnl_link_stats64 *b)
 772 {
 773         a->rx_packets = b->rx_packets;
 774         a->tx_packets = b->tx_packets;
 775         a->rx_bytes = b->rx_bytes;
 776         a->tx_bytes = b->tx_bytes;
 777         a->rx_errors = b->rx_errors;
 778         a->tx_errors = b->tx_errors;
 779         a->rx_dropped = b->rx_dropped;
 780         a->tx_dropped = b->tx_dropped;
 781
 782         a->multicast = b->multicast;
 783         a->collisions = b->collisions;
 784
 785         a->rx_length_errors = b->rx_length_errors;
 786         a->rx_over_errors = b->rx_over_errors;
 787         a->rx_crc_errors = b->rx_crc_errors;
 788         a->rx_frame_errors = b->rx_frame_errors;
 789         a->rx_fifo_errors = b->rx_fifo_errors;
 790         a->rx_missed_errors = b->rx_missed_errors;
 791
 792         a->tx_aborted_errors = b->tx_aborted_errors;
 793         a->tx_carrier_errors = b->tx_carrier_errors;
 794         a->tx_fifo_errors = b->tx_fifo_errors;
 795         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 796         a->tx_window_errors = b->tx_window_errors;
 797
 798         a->rx_compressed = b->rx_compressed;
 799         a->tx_compressed = b->tx_compressed;
 800 }
 801
 802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 803 {
 804         memcpy(v, b, sizeof(*b));
 805 }
 806
 807 /* All VF info */
 808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 809                                    u32 ext_filter_mask)
 810 {
 811         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 812             (ext_filter_mask & RTEXT_FILTER_VF)) {
 813                 int num_vfs = dev_num_vf(dev->dev.parent);
 814                 size_t size = nla_total_size(sizeof(struct nlattr));
 815                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 816                 size += num_vfs *
 817                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 818                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 819                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 820                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 821                          nla_total_size(sizeof(struct ifla_vf_link_state)));
 822                 return size;
 823         } else
 824                 return 0;
 825 }
 826
 827 static size_t rtnl_port_size(const struct net_device *dev,
 828                              u32 ext_filter_mask)
 829 {
 830         size_t port_size = nla_total_size(4)            /* PORT_VF */
 831                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 832                 + nla_total_size(sizeof(struct ifla_port_vsi))
 833                                                         /* PORT_VSI_TYPE */
 834                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 835                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 836                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 837                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 838         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 839         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 840                 + port_size;
 841         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 842                 + port_size;
 843
 844         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 845             !(ext_filter_mask & RTEXT_FILTER_VF))
 846                 return 0;
 847         if (dev_num_vf(dev->dev.parent))
 848                 return port_self_size + vf_ports_size +
 849                         vf_port_size * dev_num_vf(dev->dev.parent);
 850         else
 851                 return port_self_size;
 852 }
 853
 854 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 855                                      u32 ext_filter_mask)
 856 {
 857         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 858                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 859                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 860                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 861                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 862                + nla_total_size(sizeof(struct rtnl_link_stats))
 863                + nla_total_size(sizeof(struct rtnl_link_stats64))
 864                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 865                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 866                + nla_total_size(4) /* IFLA_TXQLEN */
 867                + nla_total_size(4) /* IFLA_WEIGHT */
 868                + nla_total_size(4) /* IFLA_MTU */
 869                + nla_total_size(4) /* IFLA_LINK */
 870                + nla_total_size(4) /* IFLA_MASTER */
 871                + nla_total_size(1) /* IFLA_CARRIER */
 872                + nla_total_size(4) /* IFLA_PROMISCUITY */
 873                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 874                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 875                + nla_total_size(1) /* IFLA_OPERSTATE */
 876                + nla_total_size(1) /* IFLA_LINKMODE */
 877                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 878                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 879                + nla_total_size(ext_filter_mask
 880                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 881                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 882                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 883                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 884                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 885                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 886                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 887 }
 888
 889 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 890 {
 891         struct nlattr *vf_ports;
 892         struct nlattr *vf_port;
 893         int vf;
 894         int err;
 895
 896         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 897         if (!vf_ports)
 898                 return -EMSGSIZE;
 899
 900         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 901                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 902                 if (!vf_port)
 903                         goto nla_put_failure;
 904                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 905                         goto nla_put_failure;
 906                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 907                 if (err == -EMSGSIZE)
 908                         goto nla_put_failure;
 909                 if (err) {
 910                         nla_nest_cancel(skb, vf_port);
 911                         continue;
 912                 }
 913                 nla_nest_end(skb, vf_port);
 914         }
 915
 916         nla_nest_end(skb, vf_ports);
 917
 918         return 0;
 919
 920 nla_put_failure:
 921         nla_nest_cancel(skb, vf_ports);
 922         return -EMSGSIZE;
 923 }
 924
 925 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 926 {
 927         struct nlattr *port_self;
 928         int err;
 929
 930         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 931         if (!port_self)
 932                 return -EMSGSIZE;
 933
 934         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 935         if (err) {
 936                 nla_nest_cancel(skb, port_self);
 937                 return (err == -EMSGSIZE) ? err : 0;
 938         }
 939
 940         nla_nest_end(skb, port_self);
 941
 942         return 0;
 943 }
 944
 945 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 946                           u32 ext_filter_mask)
 947 {
 948         int err;
 949
 950         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 951             !(ext_filter_mask & RTEXT_FILTER_VF))
 952                 return 0;
 953
 954         err = rtnl_port_self_fill(skb, dev);
 955         if (err)
 956                 return err;
 957
 958         if (dev_num_vf(dev->dev.parent)) {
 959                 err = rtnl_vf_ports_fill(skb, dev);
 960                 if (err)
 961                         return err;
 962         }
 963
 964         return 0;
 965 }
 966
 967 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 968 {
 969         int err;
 970         struct netdev_phys_item_id ppid;
 971
 972         err = dev_get_phys_port_id(dev, &ppid);
 973         if (err) {
 974                 if (err == -EOPNOTSUPP)
 975                         return 0;
 976                 return err;
 977         }
 978
 979         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 980                 return -EMSGSIZE;
 981
 982         return 0;
 983 }
 984
 985 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
 986 {
 987         char name[IFNAMSIZ];
 988         int err;
 989
 990         err = dev_get_phys_port_name(dev, name, sizeof(name));
 991         if (err) {
 992                 if (err == -EOPNOTSUPP)
 993                         return 0;
 994                 return err;
 995         }
 996
 997         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
 998                 return -EMSGSIZE;
 999
1000         return 0;
1001 }
1002
1003 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1004 {
1005         int err;
1006         struct netdev_phys_item_id psid;
1007
1008         err = netdev_switch_parent_id_get(dev, &psid);
1009         if (err) {
1010                 if (err == -EOPNOTSUPP)
1011                         return 0;
1012                 return err;
1013         }
1014
1015         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
1016                 return -EMSGSIZE;
1017
1018         return 0;
1019 }
1020
1021 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1022                             int type, u32 pid, u32 seq, u32 change,
1023                             unsigned int flags, u32 ext_filter_mask)
1024 {
1025         struct ifinfomsg *ifm;
1026         struct nlmsghdr *nlh;
1027         struct rtnl_link_stats64 temp;
1028         const struct rtnl_link_stats64 *stats;
1029         struct nlattr *attr, *af_spec;
1030         struct rtnl_af_ops *af_ops;
1031         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1032
1033         ASSERT_RTNL();
1034         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1035         if (nlh == NULL)
1036                 return -EMSGSIZE;
1037
1038         ifm = nlmsg_data(nlh);
1039         ifm->ifi_family = AF_UNSPEC;
1040         ifm->__ifi_pad = 0;
1041         ifm->ifi_type = dev->type;
1042         ifm->ifi_index = dev->ifindex;
1043         ifm->ifi_flags = dev_get_flags(dev);
1044         ifm->ifi_change = change;
1045
1046         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1047             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1048             nla_put_u8(skb, IFLA_OPERSTATE,
1049                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1050             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1051             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1052             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1053             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1054             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1055 #ifdef CONFIG_RPS
1056             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1057 #endif
1058             (dev->ifindex != dev->iflink &&
1059              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1060             (upper_dev &&
1061              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1062             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1063             (dev->qdisc &&
1064              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1065             (dev->ifalias &&
1066              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1067             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1068                         atomic_read(&dev->carrier_changes)))
1069                 goto nla_put_failure;
1070
1071         if (1) {
1072                 struct rtnl_link_ifmap map = {
1073                         .mem_start   = dev->mem_start,
1074                         .mem_end     = dev->mem_end,
1075                         .base_addr   = dev->base_addr,
1076                         .irq         = dev->irq,
1077                         .dma         = dev->dma,
1078                         .port        = dev->if_port,
1079                 };
1080                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1081                         goto nla_put_failure;
1082         }
1083
1084         if (dev->addr_len) {
1085                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1086                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1087                         goto nla_put_failure;
1088         }
1089
1090         if (rtnl_phys_port_id_fill(skb, dev))
1091                 goto nla_put_failure;
1092
1093         if (rtnl_phys_port_name_fill(skb, dev))
1094                 goto nla_put_failure;
1095
1096         if (rtnl_phys_switch_id_fill(skb, dev))
1097                 goto nla_put_failure;
1098
1099         attr = nla_reserve(skb, IFLA_STATS,
1100                         sizeof(struct rtnl_link_stats));
1101         if (attr == NULL)
1102                 goto nla_put_failure;
1103
1104         stats = dev_get_stats(dev, &temp);
1105         copy_rtnl_link_stats(nla_data(attr), stats);
1106
1107         attr = nla_reserve(skb, IFLA_STATS64,
1108                         sizeof(struct rtnl_link_stats64));
1109         if (attr == NULL)
1110                 goto nla_put_failure;
1111         copy_rtnl_link_stats64(nla_data(attr), stats);
1112
1113         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1114             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1115                 goto nla_put_failure;
1116
1117         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1118             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1119                 int i;
1120
1121                 struct nlattr *vfinfo, *vf;
1122                 int num_vfs = dev_num_vf(dev->dev.parent);
1123
1124                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1125                 if (!vfinfo)
1126                         goto nla_put_failure;
1127                 for (i = 0; i < num_vfs; i++) {
1128                         struct ifla_vf_info ivi;
1129                         struct ifla_vf_mac vf_mac;
1130                         struct ifla_vf_vlan vf_vlan;
1131                         struct ifla_vf_rate vf_rate;
1132                         struct ifla_vf_tx_rate vf_tx_rate;
1133                         struct ifla_vf_spoofchk vf_spoofchk;
1134                         struct ifla_vf_link_state vf_linkstate;
1135
1136                         /*
1137                          * Not all SR-IOV capable drivers support the
1138                          * spoofcheck query.  Preset to -1 so the user
1139                          * space tool can detect that the driver didn't
1140                          * report anything.
1141                          */
1142                         ivi.spoofchk = -1;
1143                         memset(ivi.mac, 0, sizeof(ivi.mac));
1144                         /* The default value for VF link state is "auto"
1145                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1146                          */
1147                         ivi.linkstate = 0;
1148                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1149                                 break;
1150                         vf_mac.vf =
1151                                 vf_vlan.vf =
1152                                 vf_rate.vf =
1153                                 vf_tx_rate.vf =
1154                                 vf_spoofchk.vf =
1155                                 vf_linkstate.vf = ivi.vf;
1156
1157                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1158                         vf_vlan.vlan = ivi.vlan;
1159                         vf_vlan.qos = ivi.qos;
1160                         vf_tx_rate.rate = ivi.max_tx_rate;
1161                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1162                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1163                         vf_spoofchk.setting = ivi.spoofchk;
1164                         vf_linkstate.link_state = ivi.linkstate;
1165                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1166                         if (!vf) {
1167                                 nla_nest_cancel(skb, vfinfo);
1168                                 goto nla_put_failure;
1169                         }
1170                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1171                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1172                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1173                                     &vf_rate) ||
1174                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1175                                     &vf_tx_rate) ||
1176                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1177                                     &vf_spoofchk) ||
1178                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1179                                     &vf_linkstate))
1180                                 goto nla_put_failure;
1181                         nla_nest_end(skb, vf);
1182                 }
1183                 nla_nest_end(skb, vfinfo);
1184         }
1185
1186         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1187                 goto nla_put_failure;
1188
1189         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1190                 if (rtnl_link_fill(skb, dev) < 0)
1191                         goto nla_put_failure;
1192         }
1193
1194         if (dev->rtnl_link_ops &&
1195             dev->rtnl_link_ops->get_link_net) {
1196                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1197
1198                 if (!net_eq(dev_net(dev), link_net)) {
1199                         int id = peernet2id(dev_net(dev), link_net);
1200
1201                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1202                                 goto nla_put_failure;
1203                 }
1204         }
1205
1206         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1207                 goto nla_put_failure;
1208
1209         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1210                 if (af_ops->fill_link_af) {
1211                         struct nlattr *af;
1212                         int err;
1213
1214                         if (!(af = nla_nest_start(skb, af_ops->family)))
1215                                 goto nla_put_failure;
1216
1217                         err = af_ops->fill_link_af(skb, dev);
1218
1219                         /*
1220                          * Caller may return ENODATA to indicate that there
1221                          * was no data to be dumped. This is not an error, it
1222                          * means we should trim the attribute header and
1223                          * continue.
1224                          */
1225                         if (err == -ENODATA)
1226                                 nla_nest_cancel(skb, af);
1227                         else if (err < 0)
1228                                 goto nla_put_failure;
1229
1230                         nla_nest_end(skb, af);
1231                 }
1232         }
1233
1234         nla_nest_end(skb, af_spec);
1235
1236         nlmsg_end(skb, nlh);
1237         return 0;
1238
1239 nla_put_failure:
1240         nlmsg_cancel(skb, nlh);
1241         return -EMSGSIZE;
1242 }
1243
1244 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1245         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1246         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1247         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1248         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1249         [IFLA_MTU]              = { .type = NLA_U32 },
1250         [IFLA_LINK]             = { .type = NLA_U32 },
1251         [IFLA_MASTER]           = { .type = NLA_U32 },
1252         [IFLA_CARRIER]          = { .type = NLA_U8 },
1253         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1254         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1255         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1256         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1257         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1258         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1259         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1260         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1261         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1262         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1263         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1264         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1265         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1266         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1267         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1268         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1269         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1270         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1271         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1272         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1273 };
1274
1275 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1276         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1277         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1278         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1279         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1280 };
1281
1282 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1283         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1284 };
1285
1286 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1287         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1288         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1289         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1290         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1291         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1292         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1293 };
1294
1295 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1296         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1297         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1298                                     .len = PORT_PROFILE_MAX },
1299         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1300                                     .len = sizeof(struct ifla_port_vsi)},
1301         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1302                                       .len = PORT_UUID_MAX },
1303         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1304                                     .len = PORT_UUID_MAX },
1305         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1306         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1307 };
1308
1309 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1310 {
1311         struct net *net = sock_net(skb->sk);
1312         int h, s_h;
1313         int idx = 0, s_idx;
1314         struct net_device *dev;
1315         struct hlist_head *head;
1316         struct nlattr *tb[IFLA_MAX+1];
1317         u32 ext_filter_mask = 0;
1318         int err;
1319         int hdrlen;
1320
1321         s_h = cb->args[0];
1322         s_idx = cb->args[1];
1323
1324         cb->seq = net->dev_base_seq;
1325
1326         /* A hack to preserve kernel<->userspace interface.
1327          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1328          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1329          * what iproute2 < v3.9.0 used.
1330          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1331          * attribute, its netlink message is shorter than struct ifinfomsg.
1332          */
1333         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1334                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1335
1336         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1337
1338                 if (tb[IFLA_EXT_MASK])
1339                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1340         }
1341
1342         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1343                 idx = 0;
1344                 head = &net->dev_index_head[h];
1345                 hlist_for_each_entry(dev, head, index_hlist) {
1346                         if (idx < s_idx)
1347                                 goto cont;
1348                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1349                                                NETLINK_CB(cb->skb).portid,
1350                                                cb->nlh->nlmsg_seq, 0,
1351                                                NLM_F_MULTI,
1352                                                ext_filter_mask);
1353                         /* If we ran out of room on the first message,
1354                          * we're in trouble
1355                          */
1356                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1357
1358                         if (err < 0)
1359                                 goto out;
1360
1361                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1362 cont:
1363                         idx++;
1364                 }
1365         }
1366 out:
1367         cb->args[1] = idx;
1368         cb->args[0] = h;
1369
1370         return skb->len;
1371 }
1372
1373 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1374 {
1375         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1376 }
1377 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1378
1379 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1380 {
1381         struct net *net;
1382         /* Examine the link attributes and figure out which
1383          * network namespace we are talking about.
1384          */
1385         if (tb[IFLA_NET_NS_PID])
1386                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1387         else if (tb[IFLA_NET_NS_FD])
1388                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1389         else
1390                 net = get_net(src_net);
1391         return net;
1392 }
1393 EXPORT_SYMBOL(rtnl_link_get_net);
1394
1395 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1396 {
1397         if (dev) {
1398                 if (tb[IFLA_ADDRESS] &&
1399                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1400                         return -EINVAL;
1401
1402                 if (tb[IFLA_BROADCAST] &&
1403                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1404                         return -EINVAL;
1405         }
1406
1407         if (tb[IFLA_AF_SPEC]) {
1408                 struct nlattr *af;
1409                 int rem, err;
1410
1411                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1412                         const struct rtnl_af_ops *af_ops;
1413
1414                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1415                                 return -EAFNOSUPPORT;
1416
1417                         if (!af_ops->set_link_af)
1418                                 return -EOPNOTSUPP;
1419
1420                         if (af_ops->validate_link_af) {
1421                                 err = af_ops->validate_link_af(dev, af);
1422                                 if (err < 0)
1423                                         return err;
1424                         }
1425                 }
1426         }
1427
1428         return 0;
1429 }
1430
1431 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1432 {
1433         int rem, err = -EINVAL;
1434         struct nlattr *vf;
1435         const struct net_device_ops *ops = dev->netdev_ops;
1436
1437         nla_for_each_nested(vf, attr, rem) {
1438                 switch (nla_type(vf)) {
1439                 case IFLA_VF_MAC: {
1440                         struct ifla_vf_mac *ivm;
1441                         ivm = nla_data(vf);
1442                         err = -EOPNOTSUPP;
1443                         if (ops->ndo_set_vf_mac)
1444                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1445                                                           ivm->mac);
1446                         break;
1447                 }
1448                 case IFLA_VF_VLAN: {
1449                         struct ifla_vf_vlan *ivv;
1450                         ivv = nla_data(vf);
1451                         err = -EOPNOTSUPP;
1452                         if (ops->ndo_set_vf_vlan)
1453                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1454                                                            ivv->vlan,
1455                                                            ivv->qos);
1456                         break;
1457                 }
1458                 case IFLA_VF_TX_RATE: {
1459                         struct ifla_vf_tx_rate *ivt;
1460                         struct ifla_vf_info ivf;
1461                         ivt = nla_data(vf);
1462                         err = -EOPNOTSUPP;
1463                         if (ops->ndo_get_vf_config)
1464                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1465                                                              &ivf);
1466                         if (err)
1467                                 break;
1468                         err = -EOPNOTSUPP;
1469                         if (ops->ndo_set_vf_rate)
1470                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1471                                                            ivf.min_tx_rate,
1472                                                            ivt->rate);
1473                         break;
1474                 }
1475                 case IFLA_VF_RATE: {
1476                         struct ifla_vf_rate *ivt;
1477                         ivt = nla_data(vf);
1478                         err = -EOPNOTSUPP;
1479                         if (ops->ndo_set_vf_rate)
1480                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1481                                                            ivt->min_tx_rate,
1482                                                            ivt->max_tx_rate);
1483                         break;
1484                 }
1485                 case IFLA_VF_SPOOFCHK: {
1486                         struct ifla_vf_spoofchk *ivs;
1487                         ivs = nla_data(vf);
1488                         err = -EOPNOTSUPP;
1489                         if (ops->ndo_set_vf_spoofchk)
1490                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1491                                                                ivs->setting);
1492                         break;
1493                 }
1494                 case IFLA_VF_LINK_STATE: {
1495                         struct ifla_vf_link_state *ivl;
1496                         ivl = nla_data(vf);
1497                         err = -EOPNOTSUPP;
1498                         if (ops->ndo_set_vf_link_state)
1499                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1500                                                                  ivl->link_state);
1501                         break;
1502                 }
1503                 default:
1504                         err = -EINVAL;
1505                         break;
1506                 }
1507                 if (err)
1508                         break;
1509         }
1510         return err;
1511 }
1512
1513 static int do_set_master(struct net_device *dev, int ifindex)
1514 {
1515         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1516         const struct net_device_ops *ops;
1517         int err;
1518
1519         if (upper_dev) {
1520                 if (upper_dev->ifindex == ifindex)
1521                         return 0;
1522                 ops = upper_dev->netdev_ops;
1523                 if (ops->ndo_del_slave) {
1524                         err = ops->ndo_del_slave(upper_dev, dev);
1525                         if (err)
1526                                 return err;
1527                 } else {
1528                         return -EOPNOTSUPP;
1529                 }
1530         }
1531
1532         if (ifindex) {
1533                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1534                 if (!upper_dev)
1535                         return -EINVAL;
1536                 ops = upper_dev->netdev_ops;
1537                 if (ops->ndo_add_slave) {
1538                         err = ops->ndo_add_slave(upper_dev, dev);
1539                         if (err)
1540                                 return err;
1541                 } else {
1542                         return -EOPNOTSUPP;
1543                 }
1544         }
1545         return 0;
1546 }
1547
1548 #define DO_SETLINK_MODIFIED     0x01
1549 /* notify flag means notify + modified. */
1550 #define DO_SETLINK_NOTIFY       0x03
1551 static int do_setlink(const struct sk_buff *skb,
1552                       struct net_device *dev, struct ifinfomsg *ifm,
1553                       struct nlattr **tb, char *ifname, int status)
1554 {
1555         const struct net_device_ops *ops = dev->netdev_ops;
1556         int err;
1557
1558         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1559                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1560                 if (IS_ERR(net)) {
1561                         err = PTR_ERR(net);
1562                         goto errout;
1563                 }
1564                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1565                         put_net(net);
1566                         err = -EPERM;
1567                         goto errout;
1568                 }
1569                 err = dev_change_net_namespace(dev, net, ifname);
1570                 put_net(net);
1571                 if (err)
1572                         goto errout;
1573                 status |= DO_SETLINK_MODIFIED;
1574         }
1575
1576         if (tb[IFLA_MAP]) {
1577                 struct rtnl_link_ifmap *u_map;
1578                 struct ifmap k_map;
1579
1580                 if (!ops->ndo_set_config) {
1581                         err = -EOPNOTSUPP;
1582                         goto errout;
1583                 }
1584
1585                 if (!netif_device_present(dev)) {
1586                         err = -ENODEV;
1587                         goto errout;
1588                 }
1589
1590                 u_map = nla_data(tb[IFLA_MAP]);
1591                 k_map.mem_start = (unsigned long) u_map->mem_start;
1592                 k_map.mem_end = (unsigned long) u_map->mem_end;
1593                 k_map.base_addr = (unsigned short) u_map->base_addr;
1594                 k_map.irq = (unsigned char) u_map->irq;
1595                 k_map.dma = (unsigned char) u_map->dma;
1596                 k_map.port = (unsigned char) u_map->port;
1597
1598                 err = ops->ndo_set_config(dev, &k_map);
1599                 if (err < 0)
1600                         goto errout;
1601
1602                 status |= DO_SETLINK_NOTIFY;
1603         }
1604
1605         if (tb[IFLA_ADDRESS]) {
1606                 struct sockaddr *sa;
1607                 int len;
1608
1609                 len = sizeof(sa_family_t) + dev->addr_len;
1610                 sa = kmalloc(len, GFP_KERNEL);
1611                 if (!sa) {
1612                         err = -ENOMEM;
1613                         goto errout;
1614                 }
1615                 sa->sa_family = dev->type;
1616                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1617                        dev->addr_len);
1618                 err = dev_set_mac_address(dev, sa);
1619                 kfree(sa);
1620                 if (err)
1621                         goto errout;
1622                 status |= DO_SETLINK_MODIFIED;
1623         }
1624
1625         if (tb[IFLA_MTU]) {
1626                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1627                 if (err < 0)
1628                         goto errout;
1629                 status |= DO_SETLINK_MODIFIED;
1630         }
1631
1632         if (tb[IFLA_GROUP]) {
1633                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1634                 status |= DO_SETLINK_NOTIFY;
1635         }
1636
1637         /*
1638          * Interface selected by interface index but interface
1639          * name provided implies that a name change has been
1640          * requested.
1641          */
1642         if (ifm->ifi_index > 0 && ifname[0]) {
1643                 err = dev_change_name(dev, ifname);
1644                 if (err < 0)
1645                         goto errout;
1646                 status |= DO_SETLINK_MODIFIED;
1647         }
1648
1649         if (tb[IFLA_IFALIAS]) {
1650                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1651                                     nla_len(tb[IFLA_IFALIAS]));
1652                 if (err < 0)
1653                         goto errout;
1654                 status |= DO_SETLINK_NOTIFY;
1655         }
1656
1657         if (tb[IFLA_BROADCAST]) {
1658                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1659                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1660         }
1661
1662         if (ifm->ifi_flags || ifm->ifi_change) {
1663                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1664                 if (err < 0)
1665                         goto errout;
1666         }
1667
1668         if (tb[IFLA_MASTER]) {
1669                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1670                 if (err)
1671                         goto errout;
1672                 status |= DO_SETLINK_MODIFIED;
1673         }
1674
1675         if (tb[IFLA_CARRIER]) {
1676                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1677                 if (err)
1678                         goto errout;
1679                 status |= DO_SETLINK_MODIFIED;
1680         }
1681
1682         if (tb[IFLA_TXQLEN]) {
1683                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1684
1685                 if (dev->tx_queue_len ^ value)
1686                         status |= DO_SETLINK_NOTIFY;
1687
1688                 dev->tx_queue_len = value;
1689         }
1690
1691         if (tb[IFLA_OPERSTATE])
1692                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1693
1694         if (tb[IFLA_LINKMODE]) {
1695                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1696
1697                 write_lock_bh(&dev_base_lock);
1698                 if (dev->link_mode ^ value)
1699                         status |= DO_SETLINK_NOTIFY;
1700                 dev->link_mode = value;
1701                 write_unlock_bh(&dev_base_lock);
1702         }
1703
1704         if (tb[IFLA_VFINFO_LIST]) {
1705                 struct nlattr *attr;
1706                 int rem;
1707                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1708                         if (nla_type(attr) != IFLA_VF_INFO) {
1709                                 err = -EINVAL;
1710                                 goto errout;
1711                         }
1712                         err = do_setvfinfo(dev, attr);
1713                         if (err < 0)
1714                                 goto errout;
1715                         status |= DO_SETLINK_NOTIFY;
1716                 }
1717         }
1718         err = 0;
1719
1720         if (tb[IFLA_VF_PORTS]) {
1721                 struct nlattr *port[IFLA_PORT_MAX+1];
1722                 struct nlattr *attr;
1723                 int vf;
1724                 int rem;
1725
1726                 err = -EOPNOTSUPP;
1727                 if (!ops->ndo_set_vf_port)
1728                         goto errout;
1729
1730                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1731                         if (nla_type(attr) != IFLA_VF_PORT)
1732                                 continue;
1733                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1734                                 attr, ifla_port_policy);
1735                         if (err < 0)
1736                                 goto errout;
1737                         if (!port[IFLA_PORT_VF]) {
1738                                 err = -EOPNOTSUPP;
1739                                 goto errout;
1740                         }
1741                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1742                         err = ops->ndo_set_vf_port(dev, vf, port);
1743                         if (err < 0)
1744                                 goto errout;
1745                         status |= DO_SETLINK_NOTIFY;
1746                 }
1747         }
1748         err = 0;
1749
1750         if (tb[IFLA_PORT_SELF]) {
1751                 struct nlattr *port[IFLA_PORT_MAX+1];
1752
1753                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1754                         tb[IFLA_PORT_SELF], ifla_port_policy);
1755                 if (err < 0)
1756                         goto errout;
1757
1758                 err = -EOPNOTSUPP;
1759                 if (ops->ndo_set_vf_port)
1760                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1761                 if (err < 0)
1762                         goto errout;
1763                 status |= DO_SETLINK_NOTIFY;
1764         }
1765
1766         if (tb[IFLA_AF_SPEC]) {
1767                 struct nlattr *af;
1768                 int rem;
1769
1770                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1771                         const struct rtnl_af_ops *af_ops;
1772
1773                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1774                                 BUG();
1775
1776                         err = af_ops->set_link_af(dev, af);
1777                         if (err < 0)
1778                                 goto errout;
1779
1780                         status |= DO_SETLINK_NOTIFY;
1781                 }
1782         }
1783         err = 0;
1784
1785 errout:
1786         if (status & DO_SETLINK_MODIFIED) {
1787                 if (status & DO_SETLINK_NOTIFY)
1788                         netdev_state_change(dev);
1789
1790                 if (err < 0)
1791                         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",
1792                                              dev->name);
1793         }
1794
1795         return err;
1796 }
1797
1798 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1799 {
1800         struct net *net = sock_net(skb->sk);
1801         struct ifinfomsg *ifm;
1802         struct net_device *dev;
1803         int err;
1804         struct nlattr *tb[IFLA_MAX+1];
1805         char ifname[IFNAMSIZ];
1806
1807         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1808         if (err < 0)
1809                 goto errout;
1810
1811         if (tb[IFLA_IFNAME])
1812                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1813         else
1814                 ifname[0] = '\0';
1815
1816         err = -EINVAL;
1817         ifm = nlmsg_data(nlh);
1818         if (ifm->ifi_index > 0)
1819                 dev = __dev_get_by_index(net, ifm->ifi_index);
1820         else if (tb[IFLA_IFNAME])
1821                 dev = __dev_get_by_name(net, ifname);
1822         else
1823                 goto errout;
1824
1825         if (dev == NULL) {
1826                 err = -ENODEV;
1827                 goto errout;
1828         }
1829
1830         err = validate_linkmsg(dev, tb);
1831         if (err < 0)
1832                 goto errout;
1833
1834         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1835 errout:
1836         return err;
1837 }
1838
1839 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1840 {
1841         struct net *net = sock_net(skb->sk);
1842         const struct rtnl_link_ops *ops;
1843         struct net_device *dev;
1844         struct ifinfomsg *ifm;
1845         char ifname[IFNAMSIZ];
1846         struct nlattr *tb[IFLA_MAX+1];
1847         int err;
1848         LIST_HEAD(list_kill);
1849
1850         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1851         if (err < 0)
1852                 return err;
1853
1854         if (tb[IFLA_IFNAME])
1855                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1856
1857         ifm = nlmsg_data(nlh);
1858         if (ifm->ifi_index > 0)
1859                 dev = __dev_get_by_index(net, ifm->ifi_index);
1860         else if (tb[IFLA_IFNAME])
1861                 dev = __dev_get_by_name(net, ifname);
1862         else
1863                 return -EINVAL;
1864
1865         if (!dev)
1866                 return -ENODEV;
1867
1868         ops = dev->rtnl_link_ops;
1869         if (!ops || !ops->dellink)
1870                 return -EOPNOTSUPP;
1871
1872         ops->dellink(dev, &list_kill);
1873         unregister_netdevice_many(&list_kill);
1874         return 0;
1875 }
1876
1877 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1878 {
1879         unsigned int old_flags;
1880         int err;
1881
1882         old_flags = dev->flags;
1883         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1884                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1885                 if (err < 0)
1886                         return err;
1887         }
1888
1889         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1890
1891         __dev_notify_flags(dev, old_flags, ~0U);
1892         return 0;
1893 }
1894 EXPORT_SYMBOL(rtnl_configure_link);
1895
1896 struct net_device *rtnl_create_link(struct net *net,
1897         char *ifname, unsigned char name_assign_type,
1898         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1899 {
1900         int err;
1901         struct net_device *dev;
1902         unsigned int num_tx_queues = 1;
1903         unsigned int num_rx_queues = 1;
1904
1905         if (tb[IFLA_NUM_TX_QUEUES])
1906                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1907         else if (ops->get_num_tx_queues)
1908                 num_tx_queues = ops->get_num_tx_queues();
1909
1910         if (tb[IFLA_NUM_RX_QUEUES])
1911                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1912         else if (ops->get_num_rx_queues)
1913                 num_rx_queues = ops->get_num_rx_queues();
1914
1915         err = -ENOMEM;
1916         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1917                                ops->setup, num_tx_queues, num_rx_queues);
1918         if (!dev)
1919                 goto err;
1920
1921         dev_net_set(dev, net);
1922         dev->rtnl_link_ops = ops;
1923         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1924
1925         if (tb[IFLA_MTU])
1926                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1927         if (tb[IFLA_ADDRESS]) {
1928                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1929                                 nla_len(tb[IFLA_ADDRESS]));
1930                 dev->addr_assign_type = NET_ADDR_SET;
1931         }
1932         if (tb[IFLA_BROADCAST])
1933                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1934                                 nla_len(tb[IFLA_BROADCAST]));
1935         if (tb[IFLA_TXQLEN])
1936                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1937         if (tb[IFLA_OPERSTATE])
1938                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1939         if (tb[IFLA_LINKMODE])
1940                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1941         if (tb[IFLA_GROUP])
1942                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1943
1944         return dev;
1945
1946 err:
1947         return ERR_PTR(err);
1948 }
1949 EXPORT_SYMBOL(rtnl_create_link);
1950
1951 static int rtnl_group_changelink(const struct sk_buff *skb,
1952                 struct net *net, int group,
1953                 struct ifinfomsg *ifm,
1954                 struct nlattr **tb)
1955 {
1956         struct net_device *dev;
1957         int err;
1958
1959         for_each_netdev(net, dev) {
1960                 if (dev->group == group) {
1961                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1962                         if (err < 0)
1963                                 return err;
1964                 }
1965         }
1966
1967         return 0;
1968 }
1969
1970 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1971 {
1972         struct net *net = sock_net(skb->sk);
1973         const struct rtnl_link_ops *ops;
1974         const struct rtnl_link_ops *m_ops = NULL;
1975         struct net_device *dev;
1976         struct net_device *master_dev = NULL;
1977         struct ifinfomsg *ifm;
1978         char kind[MODULE_NAME_LEN];
1979         char ifname[IFNAMSIZ];
1980         struct nlattr *tb[IFLA_MAX+1];
1981         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1982         unsigned char name_assign_type = NET_NAME_USER;
1983         int err;
1984
1985 #ifdef CONFIG_MODULES
1986 replay:
1987 #endif
1988         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1989         if (err < 0)
1990                 return err;
1991
1992         if (tb[IFLA_IFNAME])
1993                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1994         else
1995                 ifname[0] = '\0';
1996
1997         ifm = nlmsg_data(nlh);
1998         if (ifm->ifi_index > 0)
1999                 dev = __dev_get_by_index(net, ifm->ifi_index);
2000         else {
2001                 if (ifname[0])
2002                         dev = __dev_get_by_name(net, ifname);
2003                 else
2004                         dev = NULL;
2005         }
2006
2007         if (dev) {
2008                 master_dev = netdev_master_upper_dev_get(dev);
2009                 if (master_dev)
2010                         m_ops = master_dev->rtnl_link_ops;
2011         }
2012
2013         err = validate_linkmsg(dev, tb);
2014         if (err < 0)
2015                 return err;
2016
2017         if (tb[IFLA_LINKINFO]) {
2018                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2019                                        tb[IFLA_LINKINFO], ifla_info_policy);
2020                 if (err < 0)
2021                         return err;
2022         } else
2023                 memset(linkinfo, 0, sizeof(linkinfo));
2024
2025         if (linkinfo[IFLA_INFO_KIND]) {
2026                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2027                 ops = rtnl_link_ops_get(kind);
2028         } else {
2029                 kind[0] = '\0';
2030                 ops = NULL;
2031         }
2032
2033         if (1) {
2034                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2035                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2036                 struct nlattr **data = NULL;
2037                 struct nlattr **slave_data = NULL;
2038                 struct net *dest_net, *link_net = NULL;
2039
2040                 if (ops) {
2041                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2042                                 err = nla_parse_nested(attr, ops->maxtype,
2043                                                        linkinfo[IFLA_INFO_DATA],
2044                                                        ops->policy);
2045                                 if (err < 0)
2046                                         return err;
2047                                 data = attr;
2048                         }
2049                         if (ops->validate) {
2050                                 err = ops->validate(tb, data);
2051                                 if (err < 0)
2052                                         return err;
2053                         }
2054                 }
2055
2056                 if (m_ops) {
2057                         if (m_ops->slave_maxtype &&
2058                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2059                                 err = nla_parse_nested(slave_attr,
2060                                                        m_ops->slave_maxtype,
2061                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2062                                                        m_ops->slave_policy);
2063                                 if (err < 0)
2064                                         return err;
2065                                 slave_data = slave_attr;
2066                         }
2067                         if (m_ops->slave_validate) {
2068                                 err = m_ops->slave_validate(tb, slave_data);
2069                                 if (err < 0)
2070                                         return err;
2071                         }
2072                 }
2073
2074                 if (dev) {
2075                         int status = 0;
2076
2077                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2078                                 return -EEXIST;
2079                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2080                                 return -EOPNOTSUPP;
2081
2082                         if (linkinfo[IFLA_INFO_DATA]) {
2083                                 if (!ops || ops != dev->rtnl_link_ops ||
2084                                     !ops->changelink)
2085                                         return -EOPNOTSUPP;
2086
2087                                 err = ops->changelink(dev, tb, data);
2088                                 if (err < 0)
2089                                         return err;
2090                                 status |= DO_SETLINK_NOTIFY;
2091                         }
2092
2093                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2094                                 if (!m_ops || !m_ops->slave_changelink)
2095                                         return -EOPNOTSUPP;
2096
2097                                 err = m_ops->slave_changelink(master_dev, dev,
2098                                                               tb, slave_data);
2099                                 if (err < 0)
2100                                         return err;
2101                                 status |= DO_SETLINK_NOTIFY;
2102                         }
2103
2104                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2105                 }
2106
2107                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2108                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2109                                 return rtnl_group_changelink(skb, net,
2110                                                 nla_get_u32(tb[IFLA_GROUP]),
2111                                                 ifm, tb);
2112                         return -ENODEV;
2113                 }
2114
2115                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2116                         return -EOPNOTSUPP;
2117
2118                 if (!ops) {
2119 #ifdef CONFIG_MODULES
2120                         if (kind[0]) {
2121                                 __rtnl_unlock();
2122                                 request_module("rtnl-link-%s", kind);
2123                                 rtnl_lock();
2124                                 ops = rtnl_link_ops_get(kind);
2125                                 if (ops)
2126                                         goto replay;
2127                         }
2128 #endif
2129                         return -EOPNOTSUPP;
2130                 }
2131
2132                 if (!ops->setup)
2133                         return -EOPNOTSUPP;
2134
2135                 if (!ifname[0]) {
2136                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2137                         name_assign_type = NET_NAME_ENUM;
2138                 }
2139
2140                 dest_net = rtnl_link_get_net(net, tb);
2141                 if (IS_ERR(dest_net))
2142                         return PTR_ERR(dest_net);
2143
2144                 err = -EPERM;
2145                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2146                         goto out;
2147
2148                 if (tb[IFLA_LINK_NETNSID]) {
2149                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2150
2151                         link_net = get_net_ns_by_id(dest_net, id);
2152                         if (!link_net) {
2153                                 err =  -EINVAL;
2154                                 goto out;
2155                         }
2156                         err = -EPERM;
2157                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2158                                 goto out;
2159                 }
2160
2161                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2162                                        name_assign_type, ops, tb);
2163                 if (IS_ERR(dev)) {
2164                         err = PTR_ERR(dev);
2165                         goto out;
2166                 }
2167
2168                 dev->ifindex = ifm->ifi_index;
2169
2170                 if (ops->newlink) {
2171                         err = ops->newlink(link_net ? : net, dev, tb, data);
2172                         /* Drivers should call free_netdev() in ->destructor
2173                          * and unregister it on failure after registration
2174                          * so that device could be finally freed in rtnl_unlock.
2175                          */
2176                         if (err < 0) {
2177                                 /* If device is not registered at all, free it now */
2178                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2179                                         free_netdev(dev);
2180                                 goto out;
2181                         }
2182                 } else {
2183                         err = register_netdevice(dev);
2184                         if (err < 0) {
2185                                 free_netdev(dev);
2186                                 goto out;
2187                         }
2188                 }
2189                 err = rtnl_configure_link(dev, ifm);
2190                 if (err < 0)
2191                         goto out_unregister;
2192                 if (link_net) {
2193                         err = dev_change_net_namespace(dev, dest_net, ifname);
2194                         if (err < 0)
2195                                 goto out_unregister;
2196                 }
2197 out:
2198                 if (link_net)
2199                         put_net(link_net);
2200                 put_net(dest_net);
2201                 return err;
2202 out_unregister:
2203                 if (ops->newlink) {
2204                         LIST_HEAD(list_kill);
2205
2206                         ops->dellink(dev, &list_kill);
2207                         unregister_netdevice_many(&list_kill);
2208                 } else {
2209                         unregister_netdevice(dev);
2210                 }
2211                 goto out;
2212         }
2213 }
2214
2215 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2216 {
2217         struct net *net = sock_net(skb->sk);
2218         struct ifinfomsg *ifm;
2219         char ifname[IFNAMSIZ];
2220         struct nlattr *tb[IFLA_MAX+1];
2221         struct net_device *dev = NULL;
2222         struct sk_buff *nskb;
2223         int err;
2224         u32 ext_filter_mask = 0;
2225
2226         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2227         if (err < 0)
2228                 return err;
2229
2230         if (tb[IFLA_IFNAME])
2231                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2232
2233         if (tb[IFLA_EXT_MASK])
2234                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2235
2236         ifm = nlmsg_data(nlh);
2237         if (ifm->ifi_index > 0)
2238                 dev = __dev_get_by_index(net, ifm->ifi_index);
2239         else if (tb[IFLA_IFNAME])
2240                 dev = __dev_get_by_name(net, ifname);
2241         else
2242                 return -EINVAL;
2243
2244         if (dev == NULL)
2245                 return -ENODEV;
2246
2247         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2248         if (nskb == NULL)
2249                 return -ENOBUFS;
2250
2251         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2252                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2253         if (err < 0) {
2254                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2255                 WARN_ON(err == -EMSGSIZE);
2256                 kfree_skb(nskb);
2257         } else
2258                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2259
2260         return err;
2261 }
2262
2263 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2264 {
2265         struct net *net = sock_net(skb->sk);
2266         struct net_device *dev;
2267         struct nlattr *tb[IFLA_MAX+1];
2268         u32 ext_filter_mask = 0;
2269         u16 min_ifinfo_dump_size = 0;
2270         int hdrlen;
2271
2272         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2273         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2274                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2275
2276         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2277                 if (tb[IFLA_EXT_MASK])
2278                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2279         }
2280
2281         if (!ext_filter_mask)
2282                 return NLMSG_GOODSIZE;
2283         /*
2284          * traverse the list of net devices and compute the minimum
2285          * buffer size based upon the filter mask.
2286          */
2287         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2288                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2289                                              if_nlmsg_size(dev,
2290                                                            ext_filter_mask));
2291         }
2292
2293         return min_ifinfo_dump_size;
2294 }
2295
2296 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2297 {
2298         int idx;
2299         int s_idx = cb->family;
2300
2301         if (s_idx == 0)
2302                 s_idx = 1;
2303         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2304                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2305                 if (idx < s_idx || idx == PF_PACKET)
2306                         continue;
2307                 if (rtnl_msg_handlers[idx] == NULL ||
2308                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2309                         continue;
2310                 if (idx > s_idx) {
2311                         memset(&cb->args[0], 0, sizeof(cb->args));
2312                         cb->prev_seq = 0;
2313                         cb->seq = 0;
2314                 }
2315                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2316                         break;
2317         }
2318         cb->family = idx;
2319
2320         return skb->len;
2321 }
2322
2323 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2324                                        unsigned int change, gfp_t flags)
2325 {
2326         struct net *net = dev_net(dev);
2327         struct sk_buff *skb;
2328         int err = -ENOBUFS;
2329         size_t if_info_size;
2330
2331         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2332         if (skb == NULL)
2333                 goto errout;
2334
2335         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2336         if (err < 0) {
2337                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2338                 WARN_ON(err == -EMSGSIZE);
2339                 kfree_skb(skb);
2340                 goto errout;
2341         }
2342         return skb;
2343 errout:
2344         if (err < 0)
2345                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2346         return NULL;
2347 }
2348
2349 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2350 {
2351         struct net *net = dev_net(dev);
2352
2353         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2354 }
2355
2356 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2357                   gfp_t flags)
2358 {
2359         struct sk_buff *skb;
2360
2361         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2362         if (skb)
2363                 rtmsg_ifinfo_send(skb, dev, flags);
2364 }
2365 EXPORT_SYMBOL(rtmsg_ifinfo);
2366
2367 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2368                                    struct net_device *dev,
2369                                    u8 *addr, u32 pid, u32 seq,
2370                                    int type, unsigned int flags,
2371                                    int nlflags)
2372 {
2373         struct nlmsghdr *nlh;
2374         struct ndmsg *ndm;
2375
2376         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2377         if (!nlh)
2378                 return -EMSGSIZE;
2379
2380         ndm = nlmsg_data(nlh);
2381         ndm->ndm_family  = AF_BRIDGE;
2382         ndm->ndm_pad1    = 0;
2383         ndm->ndm_pad2    = 0;
2384         ndm->ndm_flags   = flags;
2385         ndm->ndm_type    = 0;
2386         ndm->ndm_ifindex = dev->ifindex;
2387         ndm->ndm_state   = NUD_PERMANENT;
2388
2389         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2390                 goto nla_put_failure;
2391
2392         nlmsg_end(skb, nlh);
2393         return 0;
2394
2395 nla_put_failure:
2396         nlmsg_cancel(skb, nlh);
2397         return -EMSGSIZE;
2398 }
2399
2400 static inline size_t rtnl_fdb_nlmsg_size(void)
2401 {
2402         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2403 }
2404
2405 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2406 {
2407         struct net *net = dev_net(dev);
2408         struct sk_buff *skb;
2409         int err = -ENOBUFS;
2410
2411         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2412         if (!skb)
2413                 goto errout;
2414
2415         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2416         if (err < 0) {
2417                 kfree_skb(skb);
2418                 goto errout;
2419         }
2420
2421         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2422         return;
2423 errout:
2424         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2425 }
2426
2427 /**
2428  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2429  */
2430 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2431                      struct nlattr *tb[],
2432                      struct net_device *dev,
2433                      const unsigned char *addr, u16 vid,
2434                      u16 flags)
2435 {
2436         int err = -EINVAL;
2437
2438         /* If aging addresses are supported device will need to
2439          * implement its own handler for this.
2440          */
2441         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2442                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2443                 return err;
2444         }
2445
2446         if (vid) {
2447                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2448                 return err;
2449         }
2450
2451         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2452                 err = dev_uc_add_excl(dev, addr);
2453         else if (is_multicast_ether_addr(addr))
2454                 err = dev_mc_add_excl(dev, addr);
2455
2456         /* Only return duplicate errors if NLM_F_EXCL is set */
2457         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2458                 err = 0;
2459
2460         return err;
2461 }
2462 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2463
2464 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2465 {
2466         u16 vid = 0;
2467
2468         if (vlan_attr) {
2469                 if (nla_len(vlan_attr) != sizeof(u16)) {
2470                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2471                         return -EINVAL;
2472                 }
2473
2474                 vid = nla_get_u16(vlan_attr);
2475
2476                 if (!vid || vid >= VLAN_VID_MASK) {
2477                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2478                                 vid);
2479                         return -EINVAL;
2480                 }
2481         }
2482         *p_vid = vid;
2483         return 0;
2484 }
2485
2486 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2487 {
2488         struct net *net = sock_net(skb->sk);
2489         struct ndmsg *ndm;
2490         struct nlattr *tb[NDA_MAX+1];
2491         struct net_device *dev;
2492         u8 *addr;
2493         u16 vid;
2494         int err;
2495
2496         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2497         if (err < 0)
2498                 return err;
2499
2500         ndm = nlmsg_data(nlh);
2501         if (ndm->ndm_ifindex == 0) {
2502                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2503                 return -EINVAL;
2504         }
2505
2506         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2507         if (dev == NULL) {
2508                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2509                 return -ENODEV;
2510         }
2511
2512         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2513                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2514                 return -EINVAL;
2515         }
2516
2517         addr = nla_data(tb[NDA_LLADDR]);
2518
2519         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2520         if (err)
2521                 return err;
2522
2523         err = -EOPNOTSUPP;
2524
2525         /* Support fdb on master device the net/bridge default case */
2526         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2527             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2528                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2529                 const struct net_device_ops *ops = br_dev->netdev_ops;
2530
2531                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2532                                        nlh->nlmsg_flags);
2533                 if (err)
2534                         goto out;
2535                 else
2536                         ndm->ndm_flags &= ~NTF_MASTER;
2537         }
2538
2539         /* Embedded bridge, macvlan, and any other device support */
2540         if ((ndm->ndm_flags & NTF_SELF)) {
2541                 if (dev->netdev_ops->ndo_fdb_add)
2542                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2543                                                            vid,
2544                                                            nlh->nlmsg_flags);
2545                 else
2546                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2547                                                nlh->nlmsg_flags);
2548
2549                 if (!err) {
2550                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2551                         ndm->ndm_flags &= ~NTF_SELF;
2552                 }
2553         }
2554 out:
2555         return err;
2556 }
2557
2558 /**
2559  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2560  */
2561 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2562                      struct nlattr *tb[],
2563                      struct net_device *dev,
2564                      const unsigned char *addr, u16 vid)
2565 {
2566         int err = -EINVAL;
2567
2568         /* If aging addresses are supported device will need to
2569          * implement its own handler for this.
2570          */
2571         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2572                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2573                 return err;
2574         }
2575
2576         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2577                 err = dev_uc_del(dev, addr);
2578         else if (is_multicast_ether_addr(addr))
2579                 err = dev_mc_del(dev, addr);
2580
2581         return err;
2582 }
2583 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2584
2585 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2586 {
2587         struct net *net = sock_net(skb->sk);
2588         struct ndmsg *ndm;
2589         struct nlattr *tb[NDA_MAX+1];
2590         struct net_device *dev;
2591         int err = -EINVAL;
2592         __u8 *addr;
2593         u16 vid;
2594
2595         if (!netlink_capable(skb, CAP_NET_ADMIN))
2596                 return -EPERM;
2597
2598         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2599         if (err < 0)
2600                 return err;
2601
2602         ndm = nlmsg_data(nlh);
2603         if (ndm->ndm_ifindex == 0) {
2604                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2605                 return -EINVAL;
2606         }
2607
2608         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2609         if (dev == NULL) {
2610                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2611                 return -ENODEV;
2612         }
2613
2614         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2615                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2616                 return -EINVAL;
2617         }
2618
2619         addr = nla_data(tb[NDA_LLADDR]);
2620
2621         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2622         if (err)
2623                 return err;
2624
2625         err = -EOPNOTSUPP;
2626
2627         /* Support fdb on master device the net/bridge default case */
2628         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2629             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2630                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2631                 const struct net_device_ops *ops = br_dev->netdev_ops;
2632
2633                 if (ops->ndo_fdb_del)
2634                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2635
2636                 if (err)
2637                         goto out;
2638                 else
2639                         ndm->ndm_flags &= ~NTF_MASTER;
2640         }
2641
2642         /* Embedded bridge, macvlan, and any other device support */
2643         if (ndm->ndm_flags & NTF_SELF) {
2644                 if (dev->netdev_ops->ndo_fdb_del)
2645                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2646                                                            vid);
2647                 else
2648                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2649
2650                 if (!err) {
2651                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2652                         ndm->ndm_flags &= ~NTF_SELF;
2653                 }
2654         }
2655 out:
2656         return err;
2657 }
2658
2659 static int nlmsg_populate_fdb(struct sk_buff *skb,
2660                               struct netlink_callback *cb,
2661                               struct net_device *dev,
2662                               int *idx,
2663                               struct netdev_hw_addr_list *list)
2664 {
2665         struct netdev_hw_addr *ha;
2666         int err;
2667         u32 portid, seq;
2668
2669         portid = NETLINK_CB(cb->skb).portid;
2670         seq = cb->nlh->nlmsg_seq;
2671
2672         list_for_each_entry(ha, &list->list, list) {
2673                 if (*idx < cb->args[0])
2674                         goto skip;
2675
2676                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2677                                               portid, seq,
2678                                               RTM_NEWNEIGH, NTF_SELF,
2679                                               NLM_F_MULTI);
2680                 if (err < 0)
2681                         return err;
2682 skip:
2683                 *idx += 1;
2684         }
2685         return 0;
2686 }
2687
2688 /**
2689  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2690  * @nlh: netlink message header
2691  * @dev: netdevice
2692  *
2693  * Default netdevice operation to dump the existing unicast address list.
2694  * Returns number of addresses from list put in skb.
2695  */
2696 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2697                       struct netlink_callback *cb,
2698                       struct net_device *dev,
2699                       struct net_device *filter_dev,
2700                       int idx)
2701 {
2702         int err;
2703
2704         netif_addr_lock_bh(dev);
2705         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2706         if (err)
2707                 goto out;
2708         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2709 out:
2710         netif_addr_unlock_bh(dev);
2711         return idx;
2712 }
2713 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2714
2715 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2716 {
2717         struct net_device *dev;
2718         struct nlattr *tb[IFLA_MAX+1];
2719         struct net_device *bdev = NULL;
2720         struct net_device *br_dev = NULL;
2721         const struct net_device_ops *ops = NULL;
2722         const struct net_device_ops *cops = NULL;
2723         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2724         struct net *net = sock_net(skb->sk);
2725         int brport_idx = 0;
2726         int br_idx = 0;
2727         int idx = 0;
2728
2729         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2730                         ifla_policy) == 0) {
2731                 if (tb[IFLA_MASTER])
2732                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2733         }
2734
2735         brport_idx = ifm->ifi_index;
2736
2737         if (br_idx) {
2738                 br_dev = __dev_get_by_index(net, br_idx);
2739                 if (!br_dev)
2740                         return -ENODEV;
2741
2742                 ops = br_dev->netdev_ops;
2743                 bdev = br_dev;
2744         }
2745
2746         for_each_netdev(net, dev) {
2747                 if (brport_idx && (dev->ifindex != brport_idx))
2748                         continue;
2749
2750                 if (!br_idx) { /* user did not specify a specific bridge */
2751                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2752                                 br_dev = netdev_master_upper_dev_get(dev);
2753                                 cops = br_dev->netdev_ops;
2754                         }
2755
2756                         bdev = dev;
2757                 } else {
2758                         if (dev != br_dev &&
2759                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2760                                 continue;
2761
2762                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2763                             !(dev->priv_flags & IFF_EBRIDGE))
2764                                 continue;
2765
2766                         bdev = br_dev;
2767                         cops = ops;
2768                 }
2769
2770                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2771                         if (cops && cops->ndo_fdb_dump)
2772                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2773                                                          idx);
2774                 }
2775
2776                 if (dev->netdev_ops->ndo_fdb_dump)
2777                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2778                                                             idx);
2779                 else
2780                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2781
2782                 cops = NULL;
2783         }
2784
2785         cb->args[0] = idx;
2786         return skb->len;
2787 }
2788
2789 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2790                                unsigned int attrnum, unsigned int flag)
2791 {
2792         if (mask & flag)
2793                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2794         return 0;
2795 }
2796
2797 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2798                             struct net_device *dev, u16 mode,
2799                             u32 flags, u32 mask)
2800 {
2801         struct nlmsghdr *nlh;
2802         struct ifinfomsg *ifm;
2803         struct nlattr *br_afspec;
2804         struct nlattr *protinfo;
2805         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2806         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2807
2808         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2809         if (nlh == NULL)
2810                 return -EMSGSIZE;
2811
2812         ifm = nlmsg_data(nlh);
2813         ifm->ifi_family = AF_BRIDGE;
2814         ifm->__ifi_pad = 0;
2815         ifm->ifi_type = dev->type;
2816         ifm->ifi_index = dev->ifindex;
2817         ifm->ifi_flags = dev_get_flags(dev);
2818         ifm->ifi_change = 0;
2819
2820
2821         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2822             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2823             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2824             (br_dev &&
2825              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2826             (dev->addr_len &&
2827              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2828             (dev->ifindex != dev->iflink &&
2829              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2830                 goto nla_put_failure;
2831
2832         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2833         if (!br_afspec)
2834                 goto nla_put_failure;
2835
2836         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2837                 nla_nest_cancel(skb, br_afspec);
2838                 goto nla_put_failure;
2839         }
2840
2841         if (mode != BRIDGE_MODE_UNDEF) {
2842                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2843                         nla_nest_cancel(skb, br_afspec);
2844                         goto nla_put_failure;
2845                 }
2846         }
2847         nla_nest_end(skb, br_afspec);
2848
2849         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2850         if (!protinfo)
2851                 goto nla_put_failure;
2852
2853         if (brport_nla_put_flag(skb, flags, mask,
2854                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2855             brport_nla_put_flag(skb, flags, mask,
2856                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2857             brport_nla_put_flag(skb, flags, mask,
2858                                 IFLA_BRPORT_FAST_LEAVE,
2859                                 BR_MULTICAST_FAST_LEAVE) ||
2860             brport_nla_put_flag(skb, flags, mask,
2861                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2862             brport_nla_put_flag(skb, flags, mask,
2863                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2864             brport_nla_put_flag(skb, flags, mask,
2865                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2866             brport_nla_put_flag(skb, flags, mask,
2867                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2868             brport_nla_put_flag(skb, flags, mask,
2869                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2870                 nla_nest_cancel(skb, protinfo);
2871                 goto nla_put_failure;
2872         }
2873
2874         nla_nest_end(skb, protinfo);
2875
2876         nlmsg_end(skb, nlh);
2877         return 0;
2878 nla_put_failure:
2879         nlmsg_cancel(skb, nlh);
2880         return -EMSGSIZE;
2881 }
2882 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2883
2884 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2885 {
2886         struct net *net = sock_net(skb->sk);
2887         struct net_device *dev;
2888         int idx = 0;
2889         u32 portid = NETLINK_CB(cb->skb).portid;
2890         u32 seq = cb->nlh->nlmsg_seq;
2891         u32 filter_mask = 0;
2892
2893         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2894                 struct nlattr *extfilt;
2895
2896                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2897                                           IFLA_EXT_MASK);
2898                 if (extfilt) {
2899                         if (nla_len(extfilt) < sizeof(filter_mask))
2900                                 return -EINVAL;
2901
2902                         filter_mask = nla_get_u32(extfilt);
2903                 }
2904         }
2905
2906         rcu_read_lock();
2907         for_each_netdev_rcu(net, dev) {
2908                 const struct net_device_ops *ops = dev->netdev_ops;
2909                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2910
2911                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2912                         if (idx >= cb->args[0] &&
2913                             br_dev->netdev_ops->ndo_bridge_getlink(
2914                                     skb, portid, seq, dev, filter_mask) < 0)
2915                                 break;
2916                         idx++;
2917                 }
2918
2919                 if (ops->ndo_bridge_getlink) {
2920                         if (idx >= cb->args[0] &&
2921                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2922                                                     filter_mask) < 0)
2923                                 break;
2924                         idx++;
2925                 }
2926         }
2927         rcu_read_unlock();
2928         cb->args[0] = idx;
2929
2930         return skb->len;
2931 }
2932
2933 static inline size_t bridge_nlmsg_size(void)
2934 {
2935         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2936                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2937                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2938                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2939                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2940                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2941                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2942                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2943                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2944                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2945                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2946 }
2947
2948 static int rtnl_bridge_notify(struct net_device *dev)
2949 {
2950         struct net *net = dev_net(dev);
2951         struct sk_buff *skb;
2952         int err = -EOPNOTSUPP;
2953
2954         if (!dev->netdev_ops->ndo_bridge_getlink)
2955                 return 0;
2956
2957         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2958         if (!skb) {
2959                 err = -ENOMEM;
2960                 goto errout;
2961         }
2962
2963         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2964         if (err < 0)
2965                 goto errout;
2966
2967         if (!skb->len)
2968                 goto errout;
2969
2970         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2971         return 0;
2972 errout:
2973         WARN_ON(err == -EMSGSIZE);
2974         kfree_skb(skb);
2975         if (err)
2976                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2977         return err;
2978 }
2979
2980 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2981 {
2982         struct net *net = sock_net(skb->sk);
2983         struct ifinfomsg *ifm;
2984         struct net_device *dev;
2985         struct nlattr *br_spec, *attr = NULL;
2986         int rem, err = -EOPNOTSUPP;
2987         u16 flags = 0;
2988         bool have_flags = false;
2989
2990         if (nlmsg_len(nlh) < sizeof(*ifm))
2991                 return -EINVAL;
2992
2993         ifm = nlmsg_data(nlh);
2994         if (ifm->ifi_family != AF_BRIDGE)
2995                 return -EPFNOSUPPORT;
2996
2997         dev = __dev_get_by_index(net, ifm->ifi_index);
2998         if (!dev) {
2999                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3000                 return -ENODEV;
3001         }
3002
3003         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3004         if (br_spec) {
3005                 nla_for_each_nested(attr, br_spec, rem) {
3006                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3007                                 if (nla_len(attr) < sizeof(flags))
3008                                         return -EINVAL;
3009
3010                                 have_flags = true;
3011                                 flags = nla_get_u16(attr);
3012                                 break;
3013                         }
3014                 }
3015         }
3016
3017         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3018                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3019
3020                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3021                         err = -EOPNOTSUPP;
3022                         goto out;
3023                 }
3024
3025                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3026                 if (err)
3027                         goto out;
3028
3029                 flags &= ~BRIDGE_FLAGS_MASTER;
3030         }
3031
3032         if ((flags & BRIDGE_FLAGS_SELF)) {
3033                 if (!dev->netdev_ops->ndo_bridge_setlink)
3034                         err = -EOPNOTSUPP;
3035                 else
3036                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3037                                                                   flags);
3038                 if (!err) {
3039                         flags &= ~BRIDGE_FLAGS_SELF;
3040
3041                         /* Generate event to notify upper layer of bridge
3042                          * change
3043                          */
3044                         err = rtnl_bridge_notify(dev);
3045                 }
3046         }
3047
3048         if (have_flags)
3049                 memcpy(nla_data(attr), &flags, sizeof(flags));
3050 out:
3051         return err;
3052 }
3053
3054 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3055 {
3056         struct net *net = sock_net(skb->sk);
3057         struct ifinfomsg *ifm;
3058         struct net_device *dev;
3059         struct nlattr *br_spec, *attr = NULL;
3060         int rem, err = -EOPNOTSUPP;
3061         u16 flags = 0;
3062         bool have_flags = false;
3063
3064         if (nlmsg_len(nlh) < sizeof(*ifm))
3065                 return -EINVAL;
3066
3067         ifm = nlmsg_data(nlh);
3068         if (ifm->ifi_family != AF_BRIDGE)
3069                 return -EPFNOSUPPORT;
3070
3071         dev = __dev_get_by_index(net, ifm->ifi_index);
3072         if (!dev) {
3073                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3074                 return -ENODEV;
3075         }
3076
3077         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3078         if (br_spec) {
3079                 nla_for_each_nested(attr, br_spec, rem) {
3080                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3081                                 if (nla_len(attr) < sizeof(flags))
3082                                         return -EINVAL;
3083
3084                                 have_flags = true;
3085                                 flags = nla_get_u16(attr);
3086                                 break;
3087                         }
3088                 }
3089         }
3090
3091         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3092                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3093
3094                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3095                         err = -EOPNOTSUPP;
3096                         goto out;
3097                 }
3098
3099                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3100                 if (err)
3101                         goto out;
3102
3103                 flags &= ~BRIDGE_FLAGS_MASTER;
3104         }
3105
3106         if ((flags & BRIDGE_FLAGS_SELF)) {
3107                 if (!dev->netdev_ops->ndo_bridge_dellink)
3108                         err = -EOPNOTSUPP;
3109                 else
3110                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3111                                                                   flags);
3112
3113                 if (!err) {
3114                         flags &= ~BRIDGE_FLAGS_SELF;
3115
3116                         /* Generate event to notify upper layer of bridge
3117                          * change
3118                          */
3119                         err = rtnl_bridge_notify(dev);
3120                 }
3121         }
3122
3123         if (have_flags)
3124                 memcpy(nla_data(attr), &flags, sizeof(flags));
3125 out:
3126         return err;
3127 }
3128
3129 /* Process one rtnetlink message. */
3130
3131 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3132 {
3133         struct net *net = sock_net(skb->sk);
3134         rtnl_doit_func doit;
3135         int sz_idx, kind;
3136         int family;
3137         int type;
3138         int err;
3139
3140         type = nlh->nlmsg_type;
3141         if (type > RTM_MAX)
3142                 return -EOPNOTSUPP;
3143
3144         type -= RTM_BASE;
3145
3146         /* All the messages must have at least 1 byte length */
3147         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3148                 return 0;
3149
3150         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3151         sz_idx = type>>2;
3152         kind = type&3;
3153
3154         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3155                 return -EPERM;
3156
3157         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3158                 struct sock *rtnl;
3159                 rtnl_dumpit_func dumpit;
3160                 rtnl_calcit_func calcit;
3161                 u16 min_dump_alloc = 0;
3162
3163                 dumpit = rtnl_get_dumpit(family, type);
3164                 if (dumpit == NULL)
3165                         return -EOPNOTSUPP;
3166                 calcit = rtnl_get_calcit(family, type);
3167                 if (calcit)
3168                         min_dump_alloc = calcit(skb, nlh);
3169
3170                 __rtnl_unlock();
3171                 rtnl = net->rtnl;
3172                 {
3173                         struct netlink_dump_control c = {
3174                                 .dump           = dumpit,
3175                                 .min_dump_alloc = min_dump_alloc,
3176                         };
3177                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3178                 }
3179                 rtnl_lock();
3180                 return err;
3181         }
3182
3183         doit = rtnl_get_doit(family, type);
3184         if (doit == NULL)
3185                 return -EOPNOTSUPP;
3186
3187         return doit(skb, nlh);
3188 }
3189
3190 static void rtnetlink_rcv(struct sk_buff *skb)
3191 {
3192         rtnl_lock();
3193         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3194         rtnl_unlock();
3195 }
3196
3197 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3198 {
3199         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3200
3201         switch (event) {
3202         case NETDEV_UP:
3203         case NETDEV_DOWN:
3204         case NETDEV_PRE_UP:
3205         case NETDEV_POST_INIT:
3206         case NETDEV_REGISTER:
3207         case NETDEV_CHANGE:
3208         case NETDEV_PRE_TYPE_CHANGE:
3209         case NETDEV_GOING_DOWN:
3210         case NETDEV_UNREGISTER:
3211         case NETDEV_UNREGISTER_FINAL:
3212         case NETDEV_RELEASE:
3213         case NETDEV_JOIN:
3214         case NETDEV_BONDING_INFO:
3215                 break;
3216         default:
3217                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3218                 break;
3219         }
3220         return NOTIFY_DONE;
3221 }
3222
3223 static struct notifier_block rtnetlink_dev_notifier = {
3224         .notifier_call  = rtnetlink_event,
3225 };
3226
3227
3228 static int __net_init rtnetlink_net_init(struct net *net)
3229 {
3230         struct sock *sk;
3231         struct netlink_kernel_cfg cfg = {
3232                 .groups         = RTNLGRP_MAX,
3233                 .input          = rtnetlink_rcv,
3234                 .cb_mutex       = &rtnl_mutex,
3235                 .flags          = NL_CFG_F_NONROOT_RECV,
3236         };
3237
3238         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3239         if (!sk)
3240                 return -ENOMEM;
3241         net->rtnl = sk;
3242         return 0;
3243 }
3244
3245 static void __net_exit rtnetlink_net_exit(struct net *net)
3246 {
3247         netlink_kernel_release(net->rtnl);
3248         net->rtnl = NULL;
3249 }
3250
3251 static struct pernet_operations rtnetlink_net_ops = {
3252         .init = rtnetlink_net_init,
3253         .exit = rtnetlink_net_exit,
3254 };
3255
3256 void __init rtnetlink_init(void)
3257 {
3258         if (register_pernet_subsys(&rtnetlink_net_ops))
3259                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3260
3261         register_netdevice_notifier(&rtnetlink_dev_notifier);
3262
3263         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3264                       rtnl_dump_ifinfo, rtnl_calcit);
3265         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3266         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3267         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3268
3269         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3270         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3271
3272         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3273         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3274         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3275
3276         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3277         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3278         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3279 }
3280