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