net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 bool lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256
 257         return 0;
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_unregister);
 260
 261 /**
 262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 263  * @protocol : Protocol family or PF_UNSPEC
 264  *
 265  * Identical to calling rtnl_unregster() for all registered message types
 266  * of a certain protocol family.
 267  */
 268 void rtnl_unregister_all(int protocol)
 269 {
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271
 272         kfree(rtnl_msg_handlers[protocol]);
 273         rtnl_msg_handlers[protocol] = NULL;
 274 }
 275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 276
 277 static LIST_HEAD(link_ops);
 278
 279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 280 {
 281         const struct rtnl_link_ops *ops;
 282
 283         list_for_each_entry(ops, &link_ops, list) {
 284                 if (!strcmp(ops->kind, kind))
 285                         return ops;
 286         }
 287         return NULL;
 288 }
 289
 290 /**
 291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 292  * @ops: struct rtnl_link_ops * to register
 293  *
 294  * The caller must hold the rtnl_mutex. This function should be used
 295  * by drivers that create devices during module initialization. It
 296  * must be called before registering the devices.
 297  *
 298  * Returns 0 on success or a negative error code.
 299  */
 300 int __rtnl_link_register(struct rtnl_link_ops *ops)
 301 {
 302         if (rtnl_link_ops_get(ops->kind))
 303                 return -EEXIST;
 304
 305         /* The check for setup is here because if ops
 306          * does not have that filled up, it is not possible
 307          * to use the ops for creating device. So do not
 308          * fill up dellink as well. That disables rtnl_dellink.
 309          */
 310         if (ops->setup && !ops->dellink)
 311                 ops->dellink = unregister_netdevice_queue;
 312
 313         list_add_tail(&ops->list, &link_ops);
 314         return 0;
 315 }
 316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 317
 318 /**
 319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 320  * @ops: struct rtnl_link_ops * to register
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_link_register(struct rtnl_link_ops *ops)
 325 {
 326         int err;
 327
 328         rtnl_lock();
 329         err = __rtnl_link_register(ops);
 330         rtnl_unlock();
 331         return err;
 332 }
 333 EXPORT_SYMBOL_GPL(rtnl_link_register);
 334
 335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 336 {
 337         struct net_device *dev;
 338         LIST_HEAD(list_kill);
 339
 340         for_each_netdev(net, dev) {
 341                 if (dev->rtnl_link_ops == ops)
 342                         ops->dellink(dev, &list_kill);
 343         }
 344         unregister_netdevice_many(&list_kill);
 345 }
 346
 347 /**
 348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 349  * @ops: struct rtnl_link_ops * to unregister
 350  *
 351  * The caller must hold the rtnl_mutex.
 352  */
 353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 354 {
 355         struct net *net;
 356
 357         for_each_net(net) {
 358                 __rtnl_kill_links(net, ops);
 359         }
 360         list_del(&ops->list);
 361 }
 362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 363
 364 /* Return with the rtnl_lock held when there are no network
 365  * devices unregistering in any network namespace.
 366  */
 367 static void rtnl_lock_unregistering_all(void)
 368 {
 369         struct net *net;
 370         bool unregistering;
 371         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 372
 373         add_wait_queue(&netdev_unregistering_wq, &wait);
 374         for (;;) {
 375                 unregistering = false;
 376                 rtnl_lock();
 377                 for_each_net(net) {
 378                         if (net->dev_unreg_count > 0) {
 379                                 unregistering = true;
 380                                 break;
 381                         }
 382                 }
 383                 if (!unregistering)
 384                         break;
 385                 __rtnl_unlock();
 386
 387                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 388         }
 389         remove_wait_queue(&netdev_unregistering_wq, &wait);
 390 }
 391
 392 /**
 393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 394  * @ops: struct rtnl_link_ops * to unregister
 395  */
 396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 397 {
 398         /* Close the race with cleanup_net() */
 399         mutex_lock(&net_mutex);
 400         rtnl_lock_unregistering_all();
 401         __rtnl_link_unregister(ops);
 402         rtnl_unlock();
 403         mutex_unlock(&net_mutex);
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 406
 407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 408 {
 409         struct net_device *master_dev;
 410         const struct rtnl_link_ops *ops;
 411
 412         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 413         if (!master_dev)
 414                 return 0;
 415         ops = master_dev->rtnl_link_ops;
 416         if (!ops || !ops->get_slave_size)
 417                 return 0;
 418         /* IFLA_INFO_SLAVE_DATA + nested data */
 419         return nla_total_size(sizeof(struct nlattr)) +
 420                ops->get_slave_size(master_dev, dev);
 421 }
 422
 423 static size_t rtnl_link_get_size(const struct net_device *dev)
 424 {
 425         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 426         size_t size;
 427
 428         if (!ops)
 429                 return 0;
 430
 431         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 432                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 433
 434         if (ops->get_size)
 435                 /* IFLA_INFO_DATA + nested data */
 436                 size += nla_total_size(sizeof(struct nlattr)) +
 437                         ops->get_size(dev);
 438
 439         if (ops->get_xstats_size)
 440                 /* IFLA_INFO_XSTATS */
 441                 size += nla_total_size(ops->get_xstats_size(dev));
 442
 443         size += rtnl_link_get_slave_info_data_size(dev);
 444
 445         return size;
 446 }
 447
 448 static LIST_HEAD(rtnl_af_ops);
 449
 450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 451 {
 452         const struct rtnl_af_ops *ops;
 453
 454         list_for_each_entry(ops, &rtnl_af_ops, list) {
 455                 if (ops->family == family)
 456                         return ops;
 457         }
 458
 459         return NULL;
 460 }
 461
 462 /**
 463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 464  * @ops: struct rtnl_af_ops * to register
 465  *
 466  * Returns 0 on success or a negative error code.
 467  */
 468 void rtnl_af_register(struct rtnl_af_ops *ops)
 469 {
 470         rtnl_lock();
 471         list_add_tail(&ops->list, &rtnl_af_ops);
 472         rtnl_unlock();
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_af_register);
 475
 476 /**
 477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 478  * @ops: struct rtnl_af_ops * to unregister
 479  *
 480  * The caller must hold the rtnl_mutex.
 481  */
 482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         list_del(&ops->list);
 485 }
 486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 487
 488 /**
 489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 490  * @ops: struct rtnl_af_ops * to unregister
 491  */
 492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 493 {
 494         rtnl_lock();
 495         __rtnl_af_unregister(ops);
 496         rtnl_unlock();
 497 }
 498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 499
 500 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 501                                     u32 ext_filter_mask)
 502 {
 503         struct rtnl_af_ops *af_ops;
 504         size_t size;
 505
 506         /* IFLA_AF_SPEC */
 507         size = nla_total_size(sizeof(struct nlattr));
 508
 509         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 510                 if (af_ops->get_link_af_size) {
 511                         /* AF_* + nested data */
 512                         size += nla_total_size(sizeof(struct nlattr)) +
 513                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 514                 }
 515         }
 516
 517         return size;
 518 }
 519
 520 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 521 {
 522         struct net_device *master_dev;
 523
 524         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 525         if (master_dev && master_dev->rtnl_link_ops)
 526                 return true;
 527         return false;
 528 }
 529
 530 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 531                                      const struct net_device *dev)
 532 {
 533         struct net_device *master_dev;
 534         const struct rtnl_link_ops *ops;
 535         struct nlattr *slave_data;
 536         int err;
 537
 538         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 539         if (!master_dev)
 540                 return 0;
 541         ops = master_dev->rtnl_link_ops;
 542         if (!ops)
 543                 return 0;
 544         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 545                 return -EMSGSIZE;
 546         if (ops->fill_slave_info) {
 547                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 548                 if (!slave_data)
 549                         return -EMSGSIZE;
 550                 err = ops->fill_slave_info(skb, master_dev, dev);
 551                 if (err < 0)
 552                         goto err_cancel_slave_data;
 553                 nla_nest_end(skb, slave_data);
 554         }
 555         return 0;
 556
 557 err_cancel_slave_data:
 558         nla_nest_cancel(skb, slave_data);
 559         return err;
 560 }
 561
 562 static int rtnl_link_info_fill(struct sk_buff *skb,
 563                                const struct net_device *dev)
 564 {
 565         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 566         struct nlattr *data;
 567         int err;
 568
 569         if (!ops)
 570                 return 0;
 571         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 572                 return -EMSGSIZE;
 573         if (ops->fill_xstats) {
 574                 err = ops->fill_xstats(skb, dev);
 575                 if (err < 0)
 576                         return err;
 577         }
 578         if (ops->fill_info) {
 579                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 580                 if (data == NULL)
 581                         return -EMSGSIZE;
 582                 err = ops->fill_info(skb, dev);
 583                 if (err < 0)
 584                         goto err_cancel_data;
 585                 nla_nest_end(skb, data);
 586         }
 587         return 0;
 588
 589 err_cancel_data:
 590         nla_nest_cancel(skb, data);
 591         return err;
 592 }
 593
 594 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 595 {
 596         struct nlattr *linkinfo;
 597         int err = -EMSGSIZE;
 598
 599         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 600         if (linkinfo == NULL)
 601                 goto out;
 602
 603         err = rtnl_link_info_fill(skb, dev);
 604         if (err < 0)
 605                 goto err_cancel_link;
 606
 607         err = rtnl_link_slave_info_fill(skb, dev);
 608         if (err < 0)
 609                 goto err_cancel_link;
 610
 611         nla_nest_end(skb, linkinfo);
 612         return 0;
 613
 614 err_cancel_link:
 615         nla_nest_cancel(skb, linkinfo);
 616 out:
 617         return err;
 618 }
 619
 620 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 621 {
 622         struct sock *rtnl = net->rtnl;
 623         int err = 0;
 624
 625         NETLINK_CB(skb).dst_group = group;
 626         if (echo)
 627                 atomic_inc(&skb->users);
 628         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 629         if (echo)
 630                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 631         return err;
 632 }
 633
 634 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 635 {
 636         struct sock *rtnl = net->rtnl;
 637
 638         return nlmsg_unicast(rtnl, skb, pid);
 639 }
 640 EXPORT_SYMBOL(rtnl_unicast);
 641
 642 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 643                  struct nlmsghdr *nlh, gfp_t flags)
 644 {
 645         struct sock *rtnl = net->rtnl;
 646         int report = 0;
 647
 648         if (nlh)
 649                 report = nlmsg_report(nlh);
 650
 651         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 652 }
 653 EXPORT_SYMBOL(rtnl_notify);
 654
 655 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 656 {
 657         struct sock *rtnl = net->rtnl;
 658
 659         netlink_set_err(rtnl, 0, group, error);
 660 }
 661 EXPORT_SYMBOL(rtnl_set_sk_err);
 662
 663 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 664 {
 665         struct nlattr *mx;
 666         int i, valid = 0;
 667
 668         mx = nla_nest_start(skb, RTA_METRICS);
 669         if (mx == NULL)
 670                 return -ENOBUFS;
 671
 672         for (i = 0; i < RTAX_MAX; i++) {
 673                 if (metrics[i]) {
 674                         if (i == RTAX_CC_ALGO - 1) {
 675                                 char tmp[TCP_CA_NAME_MAX], *name;
 676
 677                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 678                                 if (!name)
 679                                         continue;
 680                                 if (nla_put_string(skb, i + 1, name))
 681                                         goto nla_put_failure;
 682                         } else if (i == RTAX_FEATURES - 1) {
 683                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 684
 685                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 686                                 if (nla_put_u32(skb, i + 1, user_features))
 687                                         goto nla_put_failure;
 688                         } else {
 689                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 690                                         goto nla_put_failure;
 691                         }
 692                         valid++;
 693                 }
 694         }
 695
 696         if (!valid) {
 697                 nla_nest_cancel(skb, mx);
 698                 return 0;
 699         }
 700
 701         return nla_nest_end(skb, mx);
 702
 703 nla_put_failure:
 704         nla_nest_cancel(skb, mx);
 705         return -EMSGSIZE;
 706 }
 707 EXPORT_SYMBOL(rtnetlink_put_metrics);
 708
 709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 710                        long expires, u32 error)
 711 {
 712         struct rta_cacheinfo ci = {
 713                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 714                 .rta_used = dst->__use,
 715                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 716                 .rta_error = error,
 717                 .rta_id =  id,
 718         };
 719
 720         if (expires) {
 721                 unsigned long clock;
 722
 723                 clock = jiffies_to_clock_t(abs(expires));
 724                 clock = min_t(unsigned long, clock, INT_MAX);
 725                 ci.rta_expires = (expires > 0) ? clock : -clock;
 726         }
 727         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 728 }
 729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 730
 731 static void set_operstate(struct net_device *dev, unsigned char transition)
 732 {
 733         unsigned char operstate = dev->operstate;
 734
 735         switch (transition) {
 736         case IF_OPER_UP:
 737                 if ((operstate == IF_OPER_DORMANT ||
 738                      operstate == IF_OPER_UNKNOWN) &&
 739                     !netif_dormant(dev))
 740                         operstate = IF_OPER_UP;
 741                 break;
 742
 743         case IF_OPER_DORMANT:
 744                 if (operstate == IF_OPER_UP ||
 745                     operstate == IF_OPER_UNKNOWN)
 746                         operstate = IF_OPER_DORMANT;
 747                 break;
 748         }
 749
 750         if (dev->operstate != operstate) {
 751                 write_lock_bh(&dev_base_lock);
 752                 dev->operstate = operstate;
 753                 write_unlock_bh(&dev_base_lock);
 754                 netdev_state_change(dev);
 755         }
 756 }
 757
 758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 759 {
 760         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 761                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 762 }
 763
 764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 765                                            const struct ifinfomsg *ifm)
 766 {
 767         unsigned int flags = ifm->ifi_flags;
 768
 769         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 770         if (ifm->ifi_change)
 771                 flags = (flags & ifm->ifi_change) |
 772                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 773
 774         return flags;
 775 }
 776
 777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 778                                  const struct rtnl_link_stats64 *b)
 779 {
 780         a->rx_packets = b->rx_packets;
 781         a->tx_packets = b->tx_packets;
 782         a->rx_bytes = b->rx_bytes;
 783         a->tx_bytes = b->tx_bytes;
 784         a->rx_errors = b->rx_errors;
 785         a->tx_errors = b->tx_errors;
 786         a->rx_dropped = b->rx_dropped;
 787         a->tx_dropped = b->tx_dropped;
 788
 789         a->multicast = b->multicast;
 790         a->collisions = b->collisions;
 791
 792         a->rx_length_errors = b->rx_length_errors;
 793         a->rx_over_errors = b->rx_over_errors;
 794         a->rx_crc_errors = b->rx_crc_errors;
 795         a->rx_frame_errors = b->rx_frame_errors;
 796         a->rx_fifo_errors = b->rx_fifo_errors;
 797         a->rx_missed_errors = b->rx_missed_errors;
 798
 799         a->tx_aborted_errors = b->tx_aborted_errors;
 800         a->tx_carrier_errors = b->tx_carrier_errors;
 801         a->tx_fifo_errors = b->tx_fifo_errors;
 802         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 803         a->tx_window_errors = b->tx_window_errors;
 804
 805         a->rx_compressed = b->rx_compressed;
 806         a->tx_compressed = b->tx_compressed;
 807
 808         a->rx_nohandler = b->rx_nohandler;
 809 }
 810
 811 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 812 {
 813         memcpy(v, b, sizeof(*b));
 814 }
 815
 816 /* All VF info */
 817 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 818                                    u32 ext_filter_mask)
 819 {
 820         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 821             (ext_filter_mask & RTEXT_FILTER_VF)) {
 822                 int num_vfs = dev_num_vf(dev->dev.parent);
 823                 size_t size = nla_total_size(sizeof(struct nlattr));
 824                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 825                 size += num_vfs *
 826                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 827                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 828                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 829                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 830                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 831                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 832                          /* IFLA_VF_STATS_RX_PACKETS */
 833                          nla_total_size(sizeof(__u64)) +
 834                          /* IFLA_VF_STATS_TX_PACKETS */
 835                          nla_total_size(sizeof(__u64)) +
 836                          /* IFLA_VF_STATS_RX_BYTES */
 837                          nla_total_size(sizeof(__u64)) +
 838                          /* IFLA_VF_STATS_TX_BYTES */
 839                          nla_total_size(sizeof(__u64)) +
 840                          /* IFLA_VF_STATS_BROADCAST */
 841                          nla_total_size(sizeof(__u64)) +
 842                          /* IFLA_VF_STATS_MULTICAST */
 843                          nla_total_size(sizeof(__u64)) +
 844                          nla_total_size(sizeof(struct ifla_vf_trust)));
 845                 return size;
 846         } else
 847                 return 0;
 848 }
 849
 850 static size_t rtnl_port_size(const struct net_device *dev,
 851                              u32 ext_filter_mask)
 852 {
 853         size_t port_size = nla_total_size(4)            /* PORT_VF */
 854                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 855                 + nla_total_size(sizeof(struct ifla_port_vsi))
 856                                                         /* PORT_VSI_TYPE */
 857                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 858                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 859                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 860                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 861         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 862         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 863                 + port_size;
 864         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 865                 + port_size;
 866
 867         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 868             !(ext_filter_mask & RTEXT_FILTER_VF))
 869                 return 0;
 870         if (dev_num_vf(dev->dev.parent))
 871                 return port_self_size + vf_ports_size +
 872                         vf_port_size * dev_num_vf(dev->dev.parent);
 873         else
 874                 return port_self_size;
 875 }
 876
 877 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 878                                      u32 ext_filter_mask)
 879 {
 880         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 881                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 882                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 883                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 884                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 885                + nla_total_size(sizeof(struct rtnl_link_stats))
 886                + nla_total_size(sizeof(struct rtnl_link_stats64))
 887                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 888                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 889                + nla_total_size(4) /* IFLA_TXQLEN */
 890                + nla_total_size(4) /* IFLA_WEIGHT */
 891                + nla_total_size(4) /* IFLA_MTU */
 892                + nla_total_size(4) /* IFLA_LINK */
 893                + nla_total_size(4) /* IFLA_MASTER */
 894                + nla_total_size(1) /* IFLA_CARRIER */
 895                + nla_total_size(4) /* IFLA_PROMISCUITY */
 896                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 897                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 898                + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
 899                + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
 900                + nla_total_size(1) /* IFLA_OPERSTATE */
 901                + nla_total_size(1) /* IFLA_LINKMODE */
 902                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 903                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 904                + nla_total_size(ext_filter_mask
 905                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 906                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 907                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 908                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 909                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 910                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 911                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 912                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 913                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 914
 915 }
 916
 917 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 918 {
 919         struct nlattr *vf_ports;
 920         struct nlattr *vf_port;
 921         int vf;
 922         int err;
 923
 924         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 925         if (!vf_ports)
 926                 return -EMSGSIZE;
 927
 928         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 929                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 930                 if (!vf_port)
 931                         goto nla_put_failure;
 932                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 933                         goto nla_put_failure;
 934                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 935                 if (err == -EMSGSIZE)
 936                         goto nla_put_failure;
 937                 if (err) {
 938                         nla_nest_cancel(skb, vf_port);
 939                         continue;
 940                 }
 941                 nla_nest_end(skb, vf_port);
 942         }
 943
 944         nla_nest_end(skb, vf_ports);
 945
 946         return 0;
 947
 948 nla_put_failure:
 949         nla_nest_cancel(skb, vf_ports);
 950         return -EMSGSIZE;
 951 }
 952
 953 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 954 {
 955         struct nlattr *port_self;
 956         int err;
 957
 958         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 959         if (!port_self)
 960                 return -EMSGSIZE;
 961
 962         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 963         if (err) {
 964                 nla_nest_cancel(skb, port_self);
 965                 return (err == -EMSGSIZE) ? err : 0;
 966         }
 967
 968         nla_nest_end(skb, port_self);
 969
 970         return 0;
 971 }
 972
 973 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 974                           u32 ext_filter_mask)
 975 {
 976         int err;
 977
 978         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 979             !(ext_filter_mask & RTEXT_FILTER_VF))
 980                 return 0;
 981
 982         err = rtnl_port_self_fill(skb, dev);
 983         if (err)
 984                 return err;
 985
 986         if (dev_num_vf(dev->dev.parent)) {
 987                 err = rtnl_vf_ports_fill(skb, dev);
 988                 if (err)
 989                         return err;
 990         }
 991
 992         return 0;
 993 }
 994
 995 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 996 {
 997         int err;
 998         struct netdev_phys_item_id ppid;
 999
1000         err = dev_get_phys_port_id(dev, &ppid);
1001         if (err) {
1002                 if (err == -EOPNOTSUPP)
1003                         return 0;
1004                 return err;
1005         }
1006
1007         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1008                 return -EMSGSIZE;
1009
1010         return 0;
1011 }
1012
1013 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1014 {
1015         char name[IFNAMSIZ];
1016         int err;
1017
1018         err = dev_get_phys_port_name(dev, name, sizeof(name));
1019         if (err) {
1020                 if (err == -EOPNOTSUPP)
1021                         return 0;
1022                 return err;
1023         }
1024
1025         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1026                 return -EMSGSIZE;
1027
1028         return 0;
1029 }
1030
1031 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 {
1033         int err;
1034         struct switchdev_attr attr = {
1035                 .orig_dev = dev,
1036                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1037                 .flags = SWITCHDEV_F_NO_RECURSE,
1038         };
1039
1040         err = switchdev_port_attr_get(dev, &attr);
1041         if (err) {
1042                 if (err == -EOPNOTSUPP)
1043                         return 0;
1044                 return err;
1045         }
1046
1047         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1048                     attr.u.ppid.id))
1049                 return -EMSGSIZE;
1050
1051         return 0;
1052 }
1053
1054 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1055                                               struct net_device *dev)
1056 {
1057         const struct rtnl_link_stats64 *stats;
1058         struct rtnl_link_stats64 temp;
1059         struct nlattr *attr;
1060
1061         stats = dev_get_stats(dev, &temp);
1062
1063         attr = nla_reserve(skb, IFLA_STATS,
1064                            sizeof(struct rtnl_link_stats));
1065         if (!attr)
1066                 return -EMSGSIZE;
1067
1068         copy_rtnl_link_stats(nla_data(attr), stats);
1069
1070         attr = nla_reserve(skb, IFLA_STATS64,
1071                            sizeof(struct rtnl_link_stats64));
1072         if (!attr)
1073                 return -EMSGSIZE;
1074
1075         copy_rtnl_link_stats64(nla_data(attr), stats);
1076
1077         return 0;
1078 }
1079
1080 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1081                                                struct net_device *dev,
1082                                                int vfs_num,
1083                                                struct nlattr *vfinfo)
1084 {
1085         struct ifla_vf_rss_query_en vf_rss_query_en;
1086         struct ifla_vf_link_state vf_linkstate;
1087         struct ifla_vf_spoofchk vf_spoofchk;
1088         struct ifla_vf_tx_rate vf_tx_rate;
1089         struct ifla_vf_stats vf_stats;
1090         struct ifla_vf_trust vf_trust;
1091         struct ifla_vf_vlan vf_vlan;
1092         struct ifla_vf_rate vf_rate;
1093         struct nlattr *vf, *vfstats;
1094         struct ifla_vf_mac vf_mac;
1095         struct ifla_vf_info ivi;
1096
1097         /* Not all SR-IOV capable drivers support the
1098          * spoofcheck and "RSS query enable" query.  Preset to
1099          * -1 so the user space tool can detect that the driver
1100          * didn't report anything.
1101          */
1102         ivi.spoofchk = -1;
1103         ivi.rss_query_en = -1;
1104         ivi.trusted = -1;
1105         memset(ivi.mac, 0, sizeof(ivi.mac));
1106         /* The default value for VF link state is "auto"
1107          * IFLA_VF_LINK_STATE_AUTO which equals zero
1108          */
1109         ivi.linkstate = 0;
1110         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1111                 return 0;
1112
1113         vf_mac.vf =
1114                 vf_vlan.vf =
1115                 vf_rate.vf =
1116                 vf_tx_rate.vf =
1117                 vf_spoofchk.vf =
1118                 vf_linkstate.vf =
1119                 vf_rss_query_en.vf =
1120                 vf_trust.vf = ivi.vf;
1121
1122         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1123         vf_vlan.vlan = ivi.vlan;
1124         vf_vlan.qos = ivi.qos;
1125         vf_tx_rate.rate = ivi.max_tx_rate;
1126         vf_rate.min_tx_rate = ivi.min_tx_rate;
1127         vf_rate.max_tx_rate = ivi.max_tx_rate;
1128         vf_spoofchk.setting = ivi.spoofchk;
1129         vf_linkstate.link_state = ivi.linkstate;
1130         vf_rss_query_en.setting = ivi.rss_query_en;
1131         vf_trust.setting = ivi.trusted;
1132         vf = nla_nest_start(skb, IFLA_VF_INFO);
1133         if (!vf) {
1134                 nla_nest_cancel(skb, vfinfo);
1135                 return -EMSGSIZE;
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             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1148                     sizeof(vf_rss_query_en),
1149                     &vf_rss_query_en) ||
1150             nla_put(skb, IFLA_VF_TRUST,
1151                     sizeof(vf_trust), &vf_trust))
1152                 return -EMSGSIZE;
1153         memset(&vf_stats, 0, sizeof(vf_stats));
1154         if (dev->netdev_ops->ndo_get_vf_stats)
1155                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1156                                                 &vf_stats);
1157         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1158         if (!vfstats) {
1159                 nla_nest_cancel(skb, vf);
1160                 nla_nest_cancel(skb, vfinfo);
1161                 return -EMSGSIZE;
1162         }
1163         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1164                         vf_stats.rx_packets) ||
1165             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1166                         vf_stats.tx_packets) ||
1167             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1168                         vf_stats.rx_bytes) ||
1169             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1170                         vf_stats.tx_bytes) ||
1171             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1172                         vf_stats.broadcast) ||
1173             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1174                         vf_stats.multicast))
1175                 return -EMSGSIZE;
1176         nla_nest_end(skb, vfstats);
1177         nla_nest_end(skb, vf);
1178         return 0;
1179 }
1180
1181 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1182 {
1183         struct rtnl_link_ifmap map;
1184
1185         memset(&map, 0, sizeof(map));
1186         map.mem_start   = dev->mem_start;
1187         map.mem_end     = dev->mem_end;
1188         map.base_addr   = dev->base_addr;
1189         map.irq         = dev->irq;
1190         map.dma         = dev->dma;
1191         map.port        = dev->if_port;
1192
1193         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1194                 return -EMSGSIZE;
1195
1196         return 0;
1197 }
1198
1199 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1200                             int type, u32 pid, u32 seq, u32 change,
1201                             unsigned int flags, u32 ext_filter_mask)
1202 {
1203         struct ifinfomsg *ifm;
1204         struct nlmsghdr *nlh;
1205         struct nlattr *af_spec;
1206         struct rtnl_af_ops *af_ops;
1207         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1208
1209         ASSERT_RTNL();
1210         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1211         if (nlh == NULL)
1212                 return -EMSGSIZE;
1213
1214         ifm = nlmsg_data(nlh);
1215         ifm->ifi_family = AF_UNSPEC;
1216         ifm->__ifi_pad = 0;
1217         ifm->ifi_type = dev->type;
1218         ifm->ifi_index = dev->ifindex;
1219         ifm->ifi_flags = dev_get_flags(dev);
1220         ifm->ifi_change = change;
1221
1222         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1223             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1224             nla_put_u8(skb, IFLA_OPERSTATE,
1225                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1226             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1227             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1228             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1229             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1230             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1231             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1232             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1233 #ifdef CONFIG_RPS
1234             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1235 #endif
1236             (dev->ifindex != dev_get_iflink(dev) &&
1237              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1238             (upper_dev &&
1239              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1240             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1241             (dev->qdisc &&
1242              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1243             (dev->ifalias &&
1244              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1245             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1246                         atomic_read(&dev->carrier_changes)) ||
1247             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1248                 goto nla_put_failure;
1249
1250         if (rtnl_fill_link_ifmap(skb, dev))
1251                 goto nla_put_failure;
1252
1253         if (dev->addr_len) {
1254                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1255                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1256                         goto nla_put_failure;
1257         }
1258
1259         if (rtnl_phys_port_id_fill(skb, dev))
1260                 goto nla_put_failure;
1261
1262         if (rtnl_phys_port_name_fill(skb, dev))
1263                 goto nla_put_failure;
1264
1265         if (rtnl_phys_switch_id_fill(skb, dev))
1266                 goto nla_put_failure;
1267
1268         if (rtnl_fill_stats(skb, dev))
1269                 goto nla_put_failure;
1270
1271         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1272             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1273                 goto nla_put_failure;
1274
1275         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1276             ext_filter_mask & RTEXT_FILTER_VF) {
1277                 int i;
1278                 struct nlattr *vfinfo;
1279                 int num_vfs = dev_num_vf(dev->dev.parent);
1280
1281                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1282                 if (!vfinfo)
1283                         goto nla_put_failure;
1284                 for (i = 0; i < num_vfs; i++) {
1285                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1286                                 goto nla_put_failure;
1287                 }
1288
1289                 nla_nest_end(skb, vfinfo);
1290         }
1291
1292         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1293                 goto nla_put_failure;
1294
1295         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1296                 if (rtnl_link_fill(skb, dev) < 0)
1297                         goto nla_put_failure;
1298         }
1299
1300         if (dev->rtnl_link_ops &&
1301             dev->rtnl_link_ops->get_link_net) {
1302                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1303
1304                 if (!net_eq(dev_net(dev), link_net)) {
1305                         int id = peernet2id_alloc(dev_net(dev), link_net);
1306
1307                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1308                                 goto nla_put_failure;
1309                 }
1310         }
1311
1312         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1313                 goto nla_put_failure;
1314
1315         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1316                 if (af_ops->fill_link_af) {
1317                         struct nlattr *af;
1318                         int err;
1319
1320                         if (!(af = nla_nest_start(skb, af_ops->family)))
1321                                 goto nla_put_failure;
1322
1323                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1324
1325                         /*
1326                          * Caller may return ENODATA to indicate that there
1327                          * was no data to be dumped. This is not an error, it
1328                          * means we should trim the attribute header and
1329                          * continue.
1330                          */
1331                         if (err == -ENODATA)
1332                                 nla_nest_cancel(skb, af);
1333                         else if (err < 0)
1334                                 goto nla_put_failure;
1335
1336                         nla_nest_end(skb, af);
1337                 }
1338         }
1339
1340         nla_nest_end(skb, af_spec);
1341
1342         nlmsg_end(skb, nlh);
1343         return 0;
1344
1345 nla_put_failure:
1346         nlmsg_cancel(skb, nlh);
1347         return -EMSGSIZE;
1348 }
1349
1350 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1351         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1352         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1353         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1354         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1355         [IFLA_MTU]              = { .type = NLA_U32 },
1356         [IFLA_LINK]             = { .type = NLA_U32 },
1357         [IFLA_MASTER]           = { .type = NLA_U32 },
1358         [IFLA_CARRIER]          = { .type = NLA_U8 },
1359         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1360         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1361         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1362         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1363         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1364         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1365         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1366         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1367         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1368         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1369         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1370         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1371         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1372         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1373         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1374         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1375         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1376         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1377         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1378         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1379         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1380 };
1381
1382 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1383         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1384         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1385         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1386         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1387 };
1388
1389 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1390         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1391         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1392         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1393         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1394         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1395         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1396         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1397         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1398         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1399         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1400         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1401 };
1402
1403 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1404         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1405         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1406                                     .len = PORT_PROFILE_MAX },
1407         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1408                                     .len = sizeof(struct ifla_port_vsi)},
1409         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1410                                       .len = PORT_UUID_MAX },
1411         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1412                                     .len = PORT_UUID_MAX },
1413         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1414         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1415 };
1416
1417 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1418 {
1419         const struct rtnl_link_ops *ops = NULL;
1420         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1421
1422         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1423                 return NULL;
1424
1425         if (linfo[IFLA_INFO_KIND]) {
1426                 char kind[MODULE_NAME_LEN];
1427
1428                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1429                 ops = rtnl_link_ops_get(kind);
1430         }
1431
1432         return ops;
1433 }
1434
1435 static bool link_master_filtered(struct net_device *dev, int master_idx)
1436 {
1437         struct net_device *master;
1438
1439         if (!master_idx)
1440                 return false;
1441
1442         master = netdev_master_upper_dev_get(dev);
1443         if (!master || master->ifindex != master_idx)
1444                 return true;
1445
1446         return false;
1447 }
1448
1449 static bool link_kind_filtered(const struct net_device *dev,
1450                                const struct rtnl_link_ops *kind_ops)
1451 {
1452         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1453                 return true;
1454
1455         return false;
1456 }
1457
1458 static bool link_dump_filtered(struct net_device *dev,
1459                                int master_idx,
1460                                const struct rtnl_link_ops *kind_ops)
1461 {
1462         if (link_master_filtered(dev, master_idx) ||
1463             link_kind_filtered(dev, kind_ops))
1464                 return true;
1465
1466         return false;
1467 }
1468
1469 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1470 {
1471         struct net *net = sock_net(skb->sk);
1472         int h, s_h;
1473         int idx = 0, s_idx;
1474         struct net_device *dev;
1475         struct hlist_head *head;
1476         struct nlattr *tb[IFLA_MAX+1];
1477         u32 ext_filter_mask = 0;
1478         const struct rtnl_link_ops *kind_ops = NULL;
1479         unsigned int flags = NLM_F_MULTI;
1480         int master_idx = 0;
1481         int err;
1482         int hdrlen;
1483
1484         s_h = cb->args[0];
1485         s_idx = cb->args[1];
1486
1487         cb->seq = net->dev_base_seq;
1488
1489         /* A hack to preserve kernel<->userspace interface.
1490          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1491          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1492          * what iproute2 < v3.9.0 used.
1493          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1494          * attribute, its netlink message is shorter than struct ifinfomsg.
1495          */
1496         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1497                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1498
1499         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1500
1501                 if (tb[IFLA_EXT_MASK])
1502                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1503
1504                 if (tb[IFLA_MASTER])
1505                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1506
1507                 if (tb[IFLA_LINKINFO])
1508                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1509
1510                 if (master_idx || kind_ops)
1511                         flags |= NLM_F_DUMP_FILTERED;
1512         }
1513
1514         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1515                 idx = 0;
1516                 head = &net->dev_index_head[h];
1517                 hlist_for_each_entry(dev, head, index_hlist) {
1518                         if (link_dump_filtered(dev, master_idx, kind_ops))
1519                                 continue;
1520                         if (idx < s_idx)
1521                                 goto cont;
1522                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1523                                                NETLINK_CB(cb->skb).portid,
1524                                                cb->nlh->nlmsg_seq, 0,
1525                                                flags,
1526                                                ext_filter_mask);
1527                         /* If we ran out of room on the first message,
1528                          * we're in trouble
1529                          */
1530                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1531
1532                         if (err < 0)
1533                                 goto out;
1534
1535                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1536 cont:
1537                         idx++;
1538                 }
1539         }
1540 out:
1541         cb->args[1] = idx;
1542         cb->args[0] = h;
1543
1544         return skb->len;
1545 }
1546
1547 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1548 {
1549         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1550 }
1551 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1552
1553 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1554 {
1555         struct net *net;
1556         /* Examine the link attributes and figure out which
1557          * network namespace we are talking about.
1558          */
1559         if (tb[IFLA_NET_NS_PID])
1560                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1561         else if (tb[IFLA_NET_NS_FD])
1562                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1563         else
1564                 net = get_net(src_net);
1565         return net;
1566 }
1567 EXPORT_SYMBOL(rtnl_link_get_net);
1568
1569 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1570 {
1571         if (dev) {
1572                 if (tb[IFLA_ADDRESS] &&
1573                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1574                         return -EINVAL;
1575
1576                 if (tb[IFLA_BROADCAST] &&
1577                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1578                         return -EINVAL;
1579         }
1580
1581         if (tb[IFLA_AF_SPEC]) {
1582                 struct nlattr *af;
1583                 int rem, err;
1584
1585                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1586                         const struct rtnl_af_ops *af_ops;
1587
1588                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1589                                 return -EAFNOSUPPORT;
1590
1591                         if (!af_ops->set_link_af)
1592                                 return -EOPNOTSUPP;
1593
1594                         if (af_ops->validate_link_af) {
1595                                 err = af_ops->validate_link_af(dev, af);
1596                                 if (err < 0)
1597                                         return err;
1598                         }
1599                 }
1600         }
1601
1602         return 0;
1603 }
1604
1605 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1606                                   int guid_type)
1607 {
1608         const struct net_device_ops *ops = dev->netdev_ops;
1609
1610         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1611 }
1612
1613 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1614 {
1615         if (dev->type != ARPHRD_INFINIBAND)
1616                 return -EOPNOTSUPP;
1617
1618         return handle_infiniband_guid(dev, ivt, guid_type);
1619 }
1620
1621 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1622 {
1623         const struct net_device_ops *ops = dev->netdev_ops;
1624         int err = -EINVAL;
1625
1626         if (tb[IFLA_VF_MAC]) {
1627                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1628
1629                 err = -EOPNOTSUPP;
1630                 if (ops->ndo_set_vf_mac)
1631                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1632                                                   ivm->mac);
1633                 if (err < 0)
1634                         return err;
1635         }
1636
1637         if (tb[IFLA_VF_VLAN]) {
1638                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1639
1640                 err = -EOPNOTSUPP;
1641                 if (ops->ndo_set_vf_vlan)
1642                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1643                                                    ivv->qos);
1644                 if (err < 0)
1645                         return err;
1646         }
1647
1648         if (tb[IFLA_VF_TX_RATE]) {
1649                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1650                 struct ifla_vf_info ivf;
1651
1652                 err = -EOPNOTSUPP;
1653                 if (ops->ndo_get_vf_config)
1654                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1655                 if (err < 0)
1656                         return err;
1657
1658                 err = -EOPNOTSUPP;
1659                 if (ops->ndo_set_vf_rate)
1660                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1661                                                    ivf.min_tx_rate,
1662                                                    ivt->rate);
1663                 if (err < 0)
1664                         return err;
1665         }
1666
1667         if (tb[IFLA_VF_RATE]) {
1668                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1669
1670                 err = -EOPNOTSUPP;
1671                 if (ops->ndo_set_vf_rate)
1672                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1673                                                    ivt->min_tx_rate,
1674                                                    ivt->max_tx_rate);
1675                 if (err < 0)
1676                         return err;
1677         }
1678
1679         if (tb[IFLA_VF_SPOOFCHK]) {
1680                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1681
1682                 err = -EOPNOTSUPP;
1683                 if (ops->ndo_set_vf_spoofchk)
1684                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1685                                                        ivs->setting);
1686                 if (err < 0)
1687                         return err;
1688         }
1689
1690         if (tb[IFLA_VF_LINK_STATE]) {
1691                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1692
1693                 err = -EOPNOTSUPP;
1694                 if (ops->ndo_set_vf_link_state)
1695                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1696                                                          ivl->link_state);
1697                 if (err < 0)
1698                         return err;
1699         }
1700
1701         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1702                 struct ifla_vf_rss_query_en *ivrssq_en;
1703
1704                 err = -EOPNOTSUPP;
1705                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1706                 if (ops->ndo_set_vf_rss_query_en)
1707                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1708                                                            ivrssq_en->setting);
1709                 if (err < 0)
1710                         return err;
1711         }
1712
1713         if (tb[IFLA_VF_TRUST]) {
1714                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1715
1716                 err = -EOPNOTSUPP;
1717                 if (ops->ndo_set_vf_trust)
1718                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1719                 if (err < 0)
1720                         return err;
1721         }
1722
1723         if (tb[IFLA_VF_IB_NODE_GUID]) {
1724                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1725
1726                 if (!ops->ndo_set_vf_guid)
1727                         return -EOPNOTSUPP;
1728
1729                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1730         }
1731
1732         if (tb[IFLA_VF_IB_PORT_GUID]) {
1733                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1734
1735                 if (!ops->ndo_set_vf_guid)
1736                         return -EOPNOTSUPP;
1737
1738                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1739         }
1740
1741         return err;
1742 }
1743
1744 static int do_set_master(struct net_device *dev, int ifindex)
1745 {
1746         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1747         const struct net_device_ops *ops;
1748         int err;
1749
1750         if (upper_dev) {
1751                 if (upper_dev->ifindex == ifindex)
1752                         return 0;
1753                 ops = upper_dev->netdev_ops;
1754                 if (ops->ndo_del_slave) {
1755                         err = ops->ndo_del_slave(upper_dev, dev);
1756                         if (err)
1757                                 return err;
1758                 } else {
1759                         return -EOPNOTSUPP;
1760                 }
1761         }
1762
1763         if (ifindex) {
1764                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1765                 if (!upper_dev)
1766                         return -EINVAL;
1767                 ops = upper_dev->netdev_ops;
1768                 if (ops->ndo_add_slave) {
1769                         err = ops->ndo_add_slave(upper_dev, dev);
1770                         if (err)
1771                                 return err;
1772                 } else {
1773                         return -EOPNOTSUPP;
1774                 }
1775         }
1776         return 0;
1777 }
1778
1779 #define DO_SETLINK_MODIFIED     0x01
1780 /* notify flag means notify + modified. */
1781 #define DO_SETLINK_NOTIFY       0x03
1782 static int do_setlink(const struct sk_buff *skb,
1783                       struct net_device *dev, struct ifinfomsg *ifm,
1784                       struct nlattr **tb, char *ifname, int status)
1785 {
1786         const struct net_device_ops *ops = dev->netdev_ops;
1787         int err;
1788
1789         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1790                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1791                 if (IS_ERR(net)) {
1792                         err = PTR_ERR(net);
1793                         goto errout;
1794                 }
1795                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1796                         put_net(net);
1797                         err = -EPERM;
1798                         goto errout;
1799                 }
1800                 err = dev_change_net_namespace(dev, net, ifname);
1801                 put_net(net);
1802                 if (err)
1803                         goto errout;
1804                 status |= DO_SETLINK_MODIFIED;
1805         }
1806
1807         if (tb[IFLA_MAP]) {
1808                 struct rtnl_link_ifmap *u_map;
1809                 struct ifmap k_map;
1810
1811                 if (!ops->ndo_set_config) {
1812                         err = -EOPNOTSUPP;
1813                         goto errout;
1814                 }
1815
1816                 if (!netif_device_present(dev)) {
1817                         err = -ENODEV;
1818                         goto errout;
1819                 }
1820
1821                 u_map = nla_data(tb[IFLA_MAP]);
1822                 k_map.mem_start = (unsigned long) u_map->mem_start;
1823                 k_map.mem_end = (unsigned long) u_map->mem_end;
1824                 k_map.base_addr = (unsigned short) u_map->base_addr;
1825                 k_map.irq = (unsigned char) u_map->irq;
1826                 k_map.dma = (unsigned char) u_map->dma;
1827                 k_map.port = (unsigned char) u_map->port;
1828
1829                 err = ops->ndo_set_config(dev, &k_map);
1830                 if (err < 0)
1831                         goto errout;
1832
1833                 status |= DO_SETLINK_NOTIFY;
1834         }
1835
1836         if (tb[IFLA_ADDRESS]) {
1837                 struct sockaddr *sa;
1838                 int len;
1839
1840                 len = sizeof(sa_family_t) + dev->addr_len;
1841                 sa = kmalloc(len, GFP_KERNEL);
1842                 if (!sa) {
1843                         err = -ENOMEM;
1844                         goto errout;
1845                 }
1846                 sa->sa_family = dev->type;
1847                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1848                        dev->addr_len);
1849                 err = dev_set_mac_address(dev, sa);
1850                 kfree(sa);
1851                 if (err)
1852                         goto errout;
1853                 status |= DO_SETLINK_MODIFIED;
1854         }
1855
1856         if (tb[IFLA_MTU]) {
1857                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1858                 if (err < 0)
1859                         goto errout;
1860                 status |= DO_SETLINK_MODIFIED;
1861         }
1862
1863         if (tb[IFLA_GROUP]) {
1864                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1865                 status |= DO_SETLINK_NOTIFY;
1866         }
1867
1868         /*
1869          * Interface selected by interface index but interface
1870          * name provided implies that a name change has been
1871          * requested.
1872          */
1873         if (ifm->ifi_index > 0 && ifname[0]) {
1874                 err = dev_change_name(dev, ifname);
1875                 if (err < 0)
1876                         goto errout;
1877                 status |= DO_SETLINK_MODIFIED;
1878         }
1879
1880         if (tb[IFLA_IFALIAS]) {
1881                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1882                                     nla_len(tb[IFLA_IFALIAS]));
1883                 if (err < 0)
1884                         goto errout;
1885                 status |= DO_SETLINK_NOTIFY;
1886         }
1887
1888         if (tb[IFLA_BROADCAST]) {
1889                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1890                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1891         }
1892
1893         if (ifm->ifi_flags || ifm->ifi_change) {
1894                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1895                 if (err < 0)
1896                         goto errout;
1897         }
1898
1899         if (tb[IFLA_MASTER]) {
1900                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1901                 if (err)
1902                         goto errout;
1903                 status |= DO_SETLINK_MODIFIED;
1904         }
1905
1906         if (tb[IFLA_CARRIER]) {
1907                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1908                 if (err)
1909                         goto errout;
1910                 status |= DO_SETLINK_MODIFIED;
1911         }
1912
1913         if (tb[IFLA_TXQLEN]) {
1914                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1915
1916                 if (dev->tx_queue_len ^ value)
1917                         status |= DO_SETLINK_NOTIFY;
1918
1919                 dev->tx_queue_len = value;
1920         }
1921
1922         if (tb[IFLA_OPERSTATE])
1923                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1924
1925         if (tb[IFLA_LINKMODE]) {
1926                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1927
1928                 write_lock_bh(&dev_base_lock);
1929                 if (dev->link_mode ^ value)
1930                         status |= DO_SETLINK_NOTIFY;
1931                 dev->link_mode = value;
1932                 write_unlock_bh(&dev_base_lock);
1933         }
1934
1935         if (tb[IFLA_VFINFO_LIST]) {
1936                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1937                 struct nlattr *attr;
1938                 int rem;
1939
1940                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1941                         if (nla_type(attr) != IFLA_VF_INFO ||
1942                             nla_len(attr) < NLA_HDRLEN) {
1943                                 err = -EINVAL;
1944                                 goto errout;
1945                         }
1946                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1947                                                ifla_vf_policy);
1948                         if (err < 0)
1949                                 goto errout;
1950                         err = do_setvfinfo(dev, vfinfo);
1951                         if (err < 0)
1952                                 goto errout;
1953                         status |= DO_SETLINK_NOTIFY;
1954                 }
1955         }
1956         err = 0;
1957
1958         if (tb[IFLA_VF_PORTS]) {
1959                 struct nlattr *port[IFLA_PORT_MAX+1];
1960                 struct nlattr *attr;
1961                 int vf;
1962                 int rem;
1963
1964                 err = -EOPNOTSUPP;
1965                 if (!ops->ndo_set_vf_port)
1966                         goto errout;
1967
1968                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1969                         if (nla_type(attr) != IFLA_VF_PORT ||
1970                             nla_len(attr) < NLA_HDRLEN) {
1971                                 err = -EINVAL;
1972                                 goto errout;
1973                         }
1974                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1975                                                ifla_port_policy);
1976                         if (err < 0)
1977                                 goto errout;
1978                         if (!port[IFLA_PORT_VF]) {
1979                                 err = -EOPNOTSUPP;
1980                                 goto errout;
1981                         }
1982                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1983                         err = ops->ndo_set_vf_port(dev, vf, port);
1984                         if (err < 0)
1985                                 goto errout;
1986                         status |= DO_SETLINK_NOTIFY;
1987                 }
1988         }
1989         err = 0;
1990
1991         if (tb[IFLA_PORT_SELF]) {
1992                 struct nlattr *port[IFLA_PORT_MAX+1];
1993
1994                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1995                         tb[IFLA_PORT_SELF], ifla_port_policy);
1996                 if (err < 0)
1997                         goto errout;
1998
1999                 err = -EOPNOTSUPP;
2000                 if (ops->ndo_set_vf_port)
2001                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2002                 if (err < 0)
2003                         goto errout;
2004                 status |= DO_SETLINK_NOTIFY;
2005         }
2006
2007         if (tb[IFLA_AF_SPEC]) {
2008                 struct nlattr *af;
2009                 int rem;
2010
2011                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2012                         const struct rtnl_af_ops *af_ops;
2013
2014                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2015                                 BUG();
2016
2017                         err = af_ops->set_link_af(dev, af);
2018                         if (err < 0)
2019                                 goto errout;
2020
2021                         status |= DO_SETLINK_NOTIFY;
2022                 }
2023         }
2024         err = 0;
2025
2026         if (tb[IFLA_PROTO_DOWN]) {
2027                 err = dev_change_proto_down(dev,
2028                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2029                 if (err)
2030                         goto errout;
2031                 status |= DO_SETLINK_NOTIFY;
2032         }
2033
2034 errout:
2035         if (status & DO_SETLINK_MODIFIED) {
2036                 if (status & DO_SETLINK_NOTIFY)
2037                         netdev_state_change(dev);
2038
2039                 if (err < 0)
2040                         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",
2041                                              dev->name);
2042         }
2043
2044         return err;
2045 }
2046
2047 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2048 {
2049         struct net *net = sock_net(skb->sk);
2050         struct ifinfomsg *ifm;
2051         struct net_device *dev;
2052         int err;
2053         struct nlattr *tb[IFLA_MAX+1];
2054         char ifname[IFNAMSIZ];
2055
2056         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2057         if (err < 0)
2058                 goto errout;
2059
2060         if (tb[IFLA_IFNAME])
2061                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2062         else
2063                 ifname[0] = '\0';
2064
2065         err = -EINVAL;
2066         ifm = nlmsg_data(nlh);
2067         if (ifm->ifi_index > 0)
2068                 dev = __dev_get_by_index(net, ifm->ifi_index);
2069         else if (tb[IFLA_IFNAME])
2070                 dev = __dev_get_by_name(net, ifname);
2071         else
2072                 goto errout;
2073
2074         if (dev == NULL) {
2075                 err = -ENODEV;
2076                 goto errout;
2077         }
2078
2079         err = validate_linkmsg(dev, tb);
2080         if (err < 0)
2081                 goto errout;
2082
2083         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2084 errout:
2085         return err;
2086 }
2087
2088 static int rtnl_group_dellink(const struct net *net, int group)
2089 {
2090         struct net_device *dev, *aux;
2091         LIST_HEAD(list_kill);
2092         bool found = false;
2093
2094         if (!group)
2095                 return -EPERM;
2096
2097         for_each_netdev(net, dev) {
2098                 if (dev->group == group) {
2099                         const struct rtnl_link_ops *ops;
2100
2101                         found = true;
2102                         ops = dev->rtnl_link_ops;
2103                         if (!ops || !ops->dellink)
2104                                 return -EOPNOTSUPP;
2105                 }
2106         }
2107
2108         if (!found)
2109                 return -ENODEV;
2110
2111         for_each_netdev_safe(net, dev, aux) {
2112                 if (dev->group == group) {
2113                         const struct rtnl_link_ops *ops;
2114
2115                         ops = dev->rtnl_link_ops;
2116                         ops->dellink(dev, &list_kill);
2117                 }
2118         }
2119         unregister_netdevice_many(&list_kill);
2120
2121         return 0;
2122 }
2123
2124 int rtnl_delete_link(struct net_device *dev)
2125 {
2126         const struct rtnl_link_ops *ops;
2127         LIST_HEAD(list_kill);
2128
2129         ops = dev->rtnl_link_ops;
2130         if (!ops || !ops->dellink)
2131                 return -EOPNOTSUPP;
2132
2133         ops->dellink(dev, &list_kill);
2134         unregister_netdevice_many(&list_kill);
2135
2136         return 0;
2137 }
2138 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2139
2140 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2141 {
2142         struct net *net = sock_net(skb->sk);
2143         struct net_device *dev;
2144         struct ifinfomsg *ifm;
2145         char ifname[IFNAMSIZ];
2146         struct nlattr *tb[IFLA_MAX+1];
2147         int err;
2148
2149         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2150         if (err < 0)
2151                 return err;
2152
2153         if (tb[IFLA_IFNAME])
2154                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2155
2156         ifm = nlmsg_data(nlh);
2157         if (ifm->ifi_index > 0)
2158                 dev = __dev_get_by_index(net, ifm->ifi_index);
2159         else if (tb[IFLA_IFNAME])
2160                 dev = __dev_get_by_name(net, ifname);
2161         else if (tb[IFLA_GROUP])
2162                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2163         else
2164                 return -EINVAL;
2165
2166         if (!dev)
2167                 return -ENODEV;
2168
2169         return rtnl_delete_link(dev);
2170 }
2171
2172 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2173 {
2174         unsigned int old_flags;
2175         int err;
2176
2177         old_flags = dev->flags;
2178         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2179                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2180                 if (err < 0)
2181                         return err;
2182         }
2183
2184         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2185
2186         __dev_notify_flags(dev, old_flags, ~0U);
2187         return 0;
2188 }
2189 EXPORT_SYMBOL(rtnl_configure_link);
2190
2191 struct net_device *rtnl_create_link(struct net *net,
2192         const char *ifname, unsigned char name_assign_type,
2193         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2194 {
2195         int err;
2196         struct net_device *dev;
2197         unsigned int num_tx_queues = 1;
2198         unsigned int num_rx_queues = 1;
2199
2200         if (tb[IFLA_NUM_TX_QUEUES])
2201                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2202         else if (ops->get_num_tx_queues)
2203                 num_tx_queues = ops->get_num_tx_queues();
2204
2205         if (tb[IFLA_NUM_RX_QUEUES])
2206                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2207         else if (ops->get_num_rx_queues)
2208                 num_rx_queues = ops->get_num_rx_queues();
2209
2210         err = -ENOMEM;
2211         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2212                                ops->setup, num_tx_queues, num_rx_queues);
2213         if (!dev)
2214                 goto err;
2215
2216         dev_net_set(dev, net);
2217         dev->rtnl_link_ops = ops;
2218         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2219
2220         if (tb[IFLA_MTU])
2221                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2222         if (tb[IFLA_ADDRESS]) {
2223                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2224                                 nla_len(tb[IFLA_ADDRESS]));
2225                 dev->addr_assign_type = NET_ADDR_SET;
2226         }
2227         if (tb[IFLA_BROADCAST])
2228                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2229                                 nla_len(tb[IFLA_BROADCAST]));
2230         if (tb[IFLA_TXQLEN])
2231                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2232         if (tb[IFLA_OPERSTATE])
2233                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2234         if (tb[IFLA_LINKMODE])
2235                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2236         if (tb[IFLA_GROUP])
2237                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2238
2239         return dev;
2240
2241 err:
2242         return ERR_PTR(err);
2243 }
2244 EXPORT_SYMBOL(rtnl_create_link);
2245
2246 static int rtnl_group_changelink(const struct sk_buff *skb,
2247                 struct net *net, int group,
2248                 struct ifinfomsg *ifm,
2249                 struct nlattr **tb)
2250 {
2251         struct net_device *dev, *aux;
2252         int err;
2253
2254         for_each_netdev_safe(net, dev, aux) {
2255                 if (dev->group == group) {
2256                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2257                         if (err < 0)
2258                                 return err;
2259                 }
2260         }
2261
2262         return 0;
2263 }
2264
2265 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2266 {
2267         struct net *net = sock_net(skb->sk);
2268         const struct rtnl_link_ops *ops;
2269         const struct rtnl_link_ops *m_ops = NULL;
2270         struct net_device *dev;
2271         struct net_device *master_dev = NULL;
2272         struct ifinfomsg *ifm;
2273         char kind[MODULE_NAME_LEN];
2274         char ifname[IFNAMSIZ];
2275         struct nlattr *tb[IFLA_MAX+1];
2276         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2277         unsigned char name_assign_type = NET_NAME_USER;
2278         int err;
2279
2280 #ifdef CONFIG_MODULES
2281 replay:
2282 #endif
2283         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2284         if (err < 0)
2285                 return err;
2286
2287         if (tb[IFLA_IFNAME])
2288                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2289         else
2290                 ifname[0] = '\0';
2291
2292         ifm = nlmsg_data(nlh);
2293         if (ifm->ifi_index > 0)
2294                 dev = __dev_get_by_index(net, ifm->ifi_index);
2295         else {
2296                 if (ifname[0])
2297                         dev = __dev_get_by_name(net, ifname);
2298                 else
2299                         dev = NULL;
2300         }
2301
2302         if (dev) {
2303                 master_dev = netdev_master_upper_dev_get(dev);
2304                 if (master_dev)
2305                         m_ops = master_dev->rtnl_link_ops;
2306         }
2307
2308         err = validate_linkmsg(dev, tb);
2309         if (err < 0)
2310                 return err;
2311
2312         if (tb[IFLA_LINKINFO]) {
2313                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2314                                        tb[IFLA_LINKINFO], ifla_info_policy);
2315                 if (err < 0)
2316                         return err;
2317         } else
2318                 memset(linkinfo, 0, sizeof(linkinfo));
2319
2320         if (linkinfo[IFLA_INFO_KIND]) {
2321                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2322                 ops = rtnl_link_ops_get(kind);
2323         } else {
2324                 kind[0] = '\0';
2325                 ops = NULL;
2326         }
2327
2328         if (1) {
2329                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2330                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2331                 struct nlattr **data = NULL;
2332                 struct nlattr **slave_data = NULL;
2333                 struct net *dest_net, *link_net = NULL;
2334
2335                 if (ops) {
2336                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2337                                 err = nla_parse_nested(attr, ops->maxtype,
2338                                                        linkinfo[IFLA_INFO_DATA],
2339                                                        ops->policy);
2340                                 if (err < 0)
2341                                         return err;
2342                                 data = attr;
2343                         }
2344                         if (ops->validate) {
2345                                 err = ops->validate(tb, data);
2346                                 if (err < 0)
2347                                         return err;
2348                         }
2349                 }
2350
2351                 if (m_ops) {
2352                         if (m_ops->slave_maxtype &&
2353                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2354                                 err = nla_parse_nested(slave_attr,
2355                                                        m_ops->slave_maxtype,
2356                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2357                                                        m_ops->slave_policy);
2358                                 if (err < 0)
2359                                         return err;
2360                                 slave_data = slave_attr;
2361                         }
2362                         if (m_ops->slave_validate) {
2363                                 err = m_ops->slave_validate(tb, slave_data);
2364                                 if (err < 0)
2365                                         return err;
2366                         }
2367                 }
2368
2369                 if (dev) {
2370                         int status = 0;
2371
2372                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2373                                 return -EEXIST;
2374                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2375                                 return -EOPNOTSUPP;
2376
2377                         if (linkinfo[IFLA_INFO_DATA]) {
2378                                 if (!ops || ops != dev->rtnl_link_ops ||
2379                                     !ops->changelink)
2380                                         return -EOPNOTSUPP;
2381
2382                                 err = ops->changelink(dev, tb, data);
2383                                 if (err < 0)
2384                                         return err;
2385                                 status |= DO_SETLINK_NOTIFY;
2386                         }
2387
2388                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2389                                 if (!m_ops || !m_ops->slave_changelink)
2390                                         return -EOPNOTSUPP;
2391
2392                                 err = m_ops->slave_changelink(master_dev, dev,
2393                                                               tb, slave_data);
2394                                 if (err < 0)
2395                                         return err;
2396                                 status |= DO_SETLINK_NOTIFY;
2397                         }
2398
2399                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2400                 }
2401
2402                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2403                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2404                                 return rtnl_group_changelink(skb, net,
2405                                                 nla_get_u32(tb[IFLA_GROUP]),
2406                                                 ifm, tb);
2407                         return -ENODEV;
2408                 }
2409
2410                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2411                         return -EOPNOTSUPP;
2412
2413                 if (!ops) {
2414 #ifdef CONFIG_MODULES
2415                         if (kind[0]) {
2416                                 __rtnl_unlock();
2417                                 request_module("rtnl-link-%s", kind);
2418                                 rtnl_lock();
2419                                 ops = rtnl_link_ops_get(kind);
2420                                 if (ops)
2421                                         goto replay;
2422                         }
2423 #endif
2424                         return -EOPNOTSUPP;
2425                 }
2426
2427                 if (!ops->setup)
2428                         return -EOPNOTSUPP;
2429
2430                 if (!ifname[0]) {
2431                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2432                         name_assign_type = NET_NAME_ENUM;
2433                 }
2434
2435                 dest_net = rtnl_link_get_net(net, tb);
2436                 if (IS_ERR(dest_net))
2437                         return PTR_ERR(dest_net);
2438
2439                 err = -EPERM;
2440                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2441                         goto out;
2442
2443                 if (tb[IFLA_LINK_NETNSID]) {
2444                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2445
2446                         link_net = get_net_ns_by_id(dest_net, id);
2447                         if (!link_net) {
2448                                 err =  -EINVAL;
2449                                 goto out;
2450                         }
2451                         err = -EPERM;
2452                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2453                                 goto out;
2454                 }
2455
2456                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2457                                        name_assign_type, ops, tb);
2458                 if (IS_ERR(dev)) {
2459                         err = PTR_ERR(dev);
2460                         goto out;
2461                 }
2462
2463                 dev->ifindex = ifm->ifi_index;
2464
2465                 if (ops->newlink) {
2466                         err = ops->newlink(link_net ? : net, dev, tb, data);
2467                         /* Drivers should call free_netdev() in ->destructor
2468                          * and unregister it on failure after registration
2469                          * so that device could be finally freed in rtnl_unlock.
2470                          */
2471                         if (err < 0) {
2472                                 /* If device is not registered at all, free it now */
2473                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2474                                         free_netdev(dev);
2475                                 goto out;
2476                         }
2477                 } else {
2478                         err = register_netdevice(dev);
2479                         if (err < 0) {
2480                                 free_netdev(dev);
2481                                 goto out;
2482                         }
2483                 }
2484                 err = rtnl_configure_link(dev, ifm);
2485                 if (err < 0)
2486                         goto out_unregister;
2487                 if (link_net) {
2488                         err = dev_change_net_namespace(dev, dest_net, ifname);
2489                         if (err < 0)
2490                                 goto out_unregister;
2491                 }
2492 out:
2493                 if (link_net)
2494                         put_net(link_net);
2495                 put_net(dest_net);
2496                 return err;
2497 out_unregister:
2498                 if (ops->newlink) {
2499                         LIST_HEAD(list_kill);
2500
2501                         ops->dellink(dev, &list_kill);
2502                         unregister_netdevice_many(&list_kill);
2503                 } else {
2504                         unregister_netdevice(dev);
2505                 }
2506                 goto out;
2507         }
2508 }
2509
2510 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2511 {
2512         struct net *net = sock_net(skb->sk);
2513         struct ifinfomsg *ifm;
2514         char ifname[IFNAMSIZ];
2515         struct nlattr *tb[IFLA_MAX+1];
2516         struct net_device *dev = NULL;
2517         struct sk_buff *nskb;
2518         int err;
2519         u32 ext_filter_mask = 0;
2520
2521         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2522         if (err < 0)
2523                 return err;
2524
2525         if (tb[IFLA_IFNAME])
2526                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2527
2528         if (tb[IFLA_EXT_MASK])
2529                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2530
2531         ifm = nlmsg_data(nlh);
2532         if (ifm->ifi_index > 0)
2533                 dev = __dev_get_by_index(net, ifm->ifi_index);
2534         else if (tb[IFLA_IFNAME])
2535                 dev = __dev_get_by_name(net, ifname);
2536         else
2537                 return -EINVAL;
2538
2539         if (dev == NULL)
2540                 return -ENODEV;
2541
2542         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2543         if (nskb == NULL)
2544                 return -ENOBUFS;
2545
2546         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2547                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2548         if (err < 0) {
2549                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2550                 WARN_ON(err == -EMSGSIZE);
2551                 kfree_skb(nskb);
2552         } else
2553                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2554
2555         return err;
2556 }
2557
2558 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2559 {
2560         struct net *net = sock_net(skb->sk);
2561         struct net_device *dev;
2562         struct nlattr *tb[IFLA_MAX+1];
2563         u32 ext_filter_mask = 0;
2564         u16 min_ifinfo_dump_size = 0;
2565         int hdrlen;
2566
2567         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2568         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2569                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2570
2571         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2572                 if (tb[IFLA_EXT_MASK])
2573                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2574         }
2575
2576         if (!ext_filter_mask)
2577                 return NLMSG_GOODSIZE;
2578         /*
2579          * traverse the list of net devices and compute the minimum
2580          * buffer size based upon the filter mask.
2581          */
2582         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2583                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2584                                              if_nlmsg_size(dev,
2585                                                            ext_filter_mask));
2586         }
2587
2588         return min_ifinfo_dump_size;
2589 }
2590
2591 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2592 {
2593         int idx;
2594         int s_idx = cb->family;
2595
2596         if (s_idx == 0)
2597                 s_idx = 1;
2598         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2599                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2600                 if (idx < s_idx || idx == PF_PACKET)
2601                         continue;
2602                 if (rtnl_msg_handlers[idx] == NULL ||
2603                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2604                         continue;
2605                 if (idx > s_idx) {
2606                         memset(&cb->args[0], 0, sizeof(cb->args));
2607                         cb->prev_seq = 0;
2608                         cb->seq = 0;
2609                 }
2610                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2611                         break;
2612         }
2613         cb->family = idx;
2614
2615         return skb->len;
2616 }
2617
2618 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2619                                        unsigned int change, gfp_t flags)
2620 {
2621         struct net *net = dev_net(dev);
2622         struct sk_buff *skb;
2623         int err = -ENOBUFS;
2624         size_t if_info_size;
2625
2626         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2627         if (skb == NULL)
2628                 goto errout;
2629
2630         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2631         if (err < 0) {
2632                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2633                 WARN_ON(err == -EMSGSIZE);
2634                 kfree_skb(skb);
2635                 goto errout;
2636         }
2637         return skb;
2638 errout:
2639         if (err < 0)
2640                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2641         return NULL;
2642 }
2643
2644 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2645 {
2646         struct net *net = dev_net(dev);
2647
2648         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2649 }
2650
2651 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2652                   gfp_t flags)
2653 {
2654         struct sk_buff *skb;
2655
2656         if (dev->reg_state != NETREG_REGISTERED)
2657                 return;
2658
2659         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2660         if (skb)
2661                 rtmsg_ifinfo_send(skb, dev, flags);
2662 }
2663 EXPORT_SYMBOL(rtmsg_ifinfo);
2664
2665 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2666                                    struct net_device *dev,
2667                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2668                                    int type, unsigned int flags,
2669                                    int nlflags, u16 ndm_state)
2670 {
2671         struct nlmsghdr *nlh;
2672         struct ndmsg *ndm;
2673
2674         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2675         if (!nlh)
2676                 return -EMSGSIZE;
2677
2678         ndm = nlmsg_data(nlh);
2679         ndm->ndm_family  = AF_BRIDGE;
2680         ndm->ndm_pad1    = 0;
2681         ndm->ndm_pad2    = 0;
2682         ndm->ndm_flags   = flags;
2683         ndm->ndm_type    = 0;
2684         ndm->ndm_ifindex = dev->ifindex;
2685         ndm->ndm_state   = ndm_state;
2686
2687         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2688                 goto nla_put_failure;
2689         if (vid)
2690                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2691                         goto nla_put_failure;
2692
2693         nlmsg_end(skb, nlh);
2694         return 0;
2695
2696 nla_put_failure:
2697         nlmsg_cancel(skb, nlh);
2698         return -EMSGSIZE;
2699 }
2700
2701 static inline size_t rtnl_fdb_nlmsg_size(void)
2702 {
2703         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2704 }
2705
2706 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2707                             u16 ndm_state)
2708 {
2709         struct net *net = dev_net(dev);
2710         struct sk_buff *skb;
2711         int err = -ENOBUFS;
2712
2713         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2714         if (!skb)
2715                 goto errout;
2716
2717         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2718                                       0, 0, type, NTF_SELF, 0, ndm_state);
2719         if (err < 0) {
2720                 kfree_skb(skb);
2721                 goto errout;
2722         }
2723
2724         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2725         return;
2726 errout:
2727         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2728 }
2729
2730 /**
2731  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2732  */
2733 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2734                      struct nlattr *tb[],
2735                      struct net_device *dev,
2736                      const unsigned char *addr, u16 vid,
2737                      u16 flags)
2738 {
2739         int err = -EINVAL;
2740
2741         /* If aging addresses are supported device will need to
2742          * implement its own handler for this.
2743          */
2744         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2745                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2746                 return err;
2747         }
2748
2749         if (vid) {
2750                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2751                 return err;
2752         }
2753
2754         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2755                 err = dev_uc_add_excl(dev, addr);
2756         else if (is_multicast_ether_addr(addr))
2757                 err = dev_mc_add_excl(dev, addr);
2758
2759         /* Only return duplicate errors if NLM_F_EXCL is set */
2760         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2761                 err = 0;
2762
2763         return err;
2764 }
2765 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2766
2767 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2768 {
2769         u16 vid = 0;
2770
2771         if (vlan_attr) {
2772                 if (nla_len(vlan_attr) != sizeof(u16)) {
2773                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2774                         return -EINVAL;
2775                 }
2776
2777                 vid = nla_get_u16(vlan_attr);
2778
2779                 if (!vid || vid >= VLAN_VID_MASK) {
2780                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2781                                 vid);
2782                         return -EINVAL;
2783                 }
2784         }
2785         *p_vid = vid;
2786         return 0;
2787 }
2788
2789 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2790 {
2791         struct net *net = sock_net(skb->sk);
2792         struct ndmsg *ndm;
2793         struct nlattr *tb[NDA_MAX+1];
2794         struct net_device *dev;
2795         u8 *addr;
2796         u16 vid;
2797         int err;
2798
2799         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2800         if (err < 0)
2801                 return err;
2802
2803         ndm = nlmsg_data(nlh);
2804         if (ndm->ndm_ifindex == 0) {
2805                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2806                 return -EINVAL;
2807         }
2808
2809         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2810         if (dev == NULL) {
2811                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2812                 return -ENODEV;
2813         }
2814
2815         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2816                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2817                 return -EINVAL;
2818         }
2819
2820         addr = nla_data(tb[NDA_LLADDR]);
2821
2822         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2823         if (err)
2824                 return err;
2825
2826         err = -EOPNOTSUPP;
2827
2828         /* Support fdb on master device the net/bridge default case */
2829         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2830             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2831                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2832                 const struct net_device_ops *ops = br_dev->netdev_ops;
2833
2834                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2835                                        nlh->nlmsg_flags);
2836                 if (err)
2837                         goto out;
2838                 else
2839                         ndm->ndm_flags &= ~NTF_MASTER;
2840         }
2841
2842         /* Embedded bridge, macvlan, and any other device support */
2843         if ((ndm->ndm_flags & NTF_SELF)) {
2844                 if (dev->netdev_ops->ndo_fdb_add)
2845                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2846                                                            vid,
2847                                                            nlh->nlmsg_flags);
2848                 else
2849                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2850                                                nlh->nlmsg_flags);
2851
2852                 if (!err) {
2853                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2854                                         ndm->ndm_state);
2855                         ndm->ndm_flags &= ~NTF_SELF;
2856                 }
2857         }
2858 out:
2859         return err;
2860 }
2861
2862 /**
2863  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2864  */
2865 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2866                      struct nlattr *tb[],
2867                      struct net_device *dev,
2868                      const unsigned char *addr, u16 vid)
2869 {
2870         int err = -EINVAL;
2871
2872         /* If aging addresses are supported device will need to
2873          * implement its own handler for this.
2874          */
2875         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2876                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2877                 return err;
2878         }
2879
2880         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2881                 err = dev_uc_del(dev, addr);
2882         else if (is_multicast_ether_addr(addr))
2883                 err = dev_mc_del(dev, addr);
2884
2885         return err;
2886 }
2887 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2888
2889 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2890 {
2891         struct net *net = sock_net(skb->sk);
2892         struct ndmsg *ndm;
2893         struct nlattr *tb[NDA_MAX+1];
2894         struct net_device *dev;
2895         int err = -EINVAL;
2896         __u8 *addr;
2897         u16 vid;
2898
2899         if (!netlink_capable(skb, CAP_NET_ADMIN))
2900                 return -EPERM;
2901
2902         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2903         if (err < 0)
2904                 return err;
2905
2906         ndm = nlmsg_data(nlh);
2907         if (ndm->ndm_ifindex == 0) {
2908                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2909                 return -EINVAL;
2910         }
2911
2912         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2913         if (dev == NULL) {
2914                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2915                 return -ENODEV;
2916         }
2917
2918         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2919                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2920                 return -EINVAL;
2921         }
2922
2923         addr = nla_data(tb[NDA_LLADDR]);
2924
2925         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2926         if (err)
2927                 return err;
2928
2929         err = -EOPNOTSUPP;
2930
2931         /* Support fdb on master device the net/bridge default case */
2932         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2933             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2934                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2935                 const struct net_device_ops *ops = br_dev->netdev_ops;
2936
2937                 if (ops->ndo_fdb_del)
2938                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2939
2940                 if (err)
2941                         goto out;
2942                 else
2943                         ndm->ndm_flags &= ~NTF_MASTER;
2944         }
2945
2946         /* Embedded bridge, macvlan, and any other device support */
2947         if (ndm->ndm_flags & NTF_SELF) {
2948                 if (dev->netdev_ops->ndo_fdb_del)
2949                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2950                                                            vid);
2951                 else
2952                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2953
2954                 if (!err) {
2955                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2956                                         ndm->ndm_state);
2957                         ndm->ndm_flags &= ~NTF_SELF;
2958                 }
2959         }
2960 out:
2961         return err;
2962 }
2963
2964 static int nlmsg_populate_fdb(struct sk_buff *skb,
2965                               struct netlink_callback *cb,
2966                               struct net_device *dev,
2967                               int *idx,
2968                               struct netdev_hw_addr_list *list)
2969 {
2970         struct netdev_hw_addr *ha;
2971         int err;
2972         u32 portid, seq;
2973
2974         portid = NETLINK_CB(cb->skb).portid;
2975         seq = cb->nlh->nlmsg_seq;
2976
2977         list_for_each_entry(ha, &list->list, list) {
2978                 if (*idx < cb->args[0])
2979                         goto skip;
2980
2981                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2982                                               portid, seq,
2983                                               RTM_NEWNEIGH, NTF_SELF,
2984                                               NLM_F_MULTI, NUD_PERMANENT);
2985                 if (err < 0)
2986                         return err;
2987 skip:
2988                 *idx += 1;
2989         }
2990         return 0;
2991 }
2992
2993 /**
2994  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2995  * @nlh: netlink message header
2996  * @dev: netdevice
2997  *
2998  * Default netdevice operation to dump the existing unicast address list.
2999  * Returns number of addresses from list put in skb.
3000  */
3001 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3002                       struct netlink_callback *cb,
3003                       struct net_device *dev,
3004                       struct net_device *filter_dev,
3005                       int idx)
3006 {
3007         int err;
3008
3009         netif_addr_lock_bh(dev);
3010         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
3011         if (err)
3012                 goto out;
3013         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
3014 out:
3015         netif_addr_unlock_bh(dev);
3016         cb->args[1] = err;
3017         return idx;
3018 }
3019 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3020
3021 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3022 {
3023         struct net_device *dev;
3024         struct nlattr *tb[IFLA_MAX+1];
3025         struct net_device *br_dev = NULL;
3026         const struct net_device_ops *ops = NULL;
3027         const struct net_device_ops *cops = NULL;
3028         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3029         struct net *net = sock_net(skb->sk);
3030         int brport_idx = 0;
3031         int br_idx = 0;
3032         int idx = 0;
3033
3034         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3035                         ifla_policy) == 0) {
3036                 if (tb[IFLA_MASTER])
3037                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3038         }
3039
3040         brport_idx = ifm->ifi_index;
3041
3042         if (br_idx) {
3043                 br_dev = __dev_get_by_index(net, br_idx);
3044                 if (!br_dev)
3045                         return -ENODEV;
3046
3047                 ops = br_dev->netdev_ops;
3048         }
3049
3050         cb->args[1] = 0;
3051         for_each_netdev(net, dev) {
3052                 if (brport_idx && (dev->ifindex != brport_idx))
3053                         continue;
3054
3055                 if (!br_idx) { /* user did not specify a specific bridge */
3056                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3057                                 br_dev = netdev_master_upper_dev_get(dev);
3058                                 cops = br_dev->netdev_ops;
3059                         }
3060
3061                 } else {
3062                         if (dev != br_dev &&
3063                             !(dev->priv_flags & IFF_BRIDGE_PORT))
3064                                 continue;
3065
3066                         if (br_dev != netdev_master_upper_dev_get(dev) &&
3067                             !(dev->priv_flags & IFF_EBRIDGE))
3068                                 continue;
3069
3070                         cops = ops;
3071                 }
3072
3073                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3074                         if (cops && cops->ndo_fdb_dump)
3075                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3076                                                          idx);
3077                 }
3078                 if (cb->args[1] == -EMSGSIZE)
3079                         break;
3080
3081                 if (dev->netdev_ops->ndo_fdb_dump)
3082                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3083                                                             idx);
3084                 else
3085                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3086                 if (cb->args[1] == -EMSGSIZE)
3087                         break;
3088
3089                 cops = NULL;
3090         }
3091
3092         cb->args[0] = idx;
3093         return skb->len;
3094 }
3095
3096 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3097                                unsigned int attrnum, unsigned int flag)
3098 {
3099         if (mask & flag)
3100                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3101         return 0;
3102 }
3103
3104 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3105                             struct net_device *dev, u16 mode,
3106                             u32 flags, u32 mask, int nlflags,
3107                             u32 filter_mask,
3108                             int (*vlan_fill)(struct sk_buff *skb,
3109                                              struct net_device *dev,
3110                                              u32 filter_mask))
3111 {
3112         struct nlmsghdr *nlh;
3113         struct ifinfomsg *ifm;
3114         struct nlattr *br_afspec;
3115         struct nlattr *protinfo;
3116         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3117         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3118         int err = 0;
3119
3120         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3121         if (nlh == NULL)
3122                 return -EMSGSIZE;
3123
3124         ifm = nlmsg_data(nlh);
3125         ifm->ifi_family = AF_BRIDGE;
3126         ifm->__ifi_pad = 0;
3127         ifm->ifi_type = dev->type;
3128         ifm->ifi_index = dev->ifindex;
3129         ifm->ifi_flags = dev_get_flags(dev);
3130         ifm->ifi_change = 0;
3131
3132
3133         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3134             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3135             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3136             (br_dev &&
3137              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3138             (dev->addr_len &&
3139              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3140             (dev->ifindex != dev_get_iflink(dev) &&
3141              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3142                 goto nla_put_failure;
3143
3144         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3145         if (!br_afspec)
3146                 goto nla_put_failure;
3147
3148         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3149                 nla_nest_cancel(skb, br_afspec);
3150                 goto nla_put_failure;
3151         }
3152
3153         if (mode != BRIDGE_MODE_UNDEF) {
3154                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3155                         nla_nest_cancel(skb, br_afspec);
3156                         goto nla_put_failure;
3157                 }
3158         }
3159         if (vlan_fill) {
3160                 err = vlan_fill(skb, dev, filter_mask);
3161                 if (err) {
3162                         nla_nest_cancel(skb, br_afspec);
3163                         goto nla_put_failure;
3164                 }
3165         }
3166         nla_nest_end(skb, br_afspec);
3167
3168         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3169         if (!protinfo)
3170                 goto nla_put_failure;
3171
3172         if (brport_nla_put_flag(skb, flags, mask,
3173                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3174             brport_nla_put_flag(skb, flags, mask,
3175                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3176             brport_nla_put_flag(skb, flags, mask,
3177                                 IFLA_BRPORT_FAST_LEAVE,
3178                                 BR_MULTICAST_FAST_LEAVE) ||
3179             brport_nla_put_flag(skb, flags, mask,
3180                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3181             brport_nla_put_flag(skb, flags, mask,
3182                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3183             brport_nla_put_flag(skb, flags, mask,
3184                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3185             brport_nla_put_flag(skb, flags, mask,
3186                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3187             brport_nla_put_flag(skb, flags, mask,
3188                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3189                 nla_nest_cancel(skb, protinfo);
3190                 goto nla_put_failure;
3191         }
3192
3193         nla_nest_end(skb, protinfo);
3194
3195         nlmsg_end(skb, nlh);
3196         return 0;
3197 nla_put_failure:
3198         nlmsg_cancel(skb, nlh);
3199         return err ? err : -EMSGSIZE;
3200 }
3201 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3202
3203 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3204 {
3205         struct net *net = sock_net(skb->sk);
3206         struct net_device *dev;
3207         int idx = 0;
3208         u32 portid = NETLINK_CB(cb->skb).portid;
3209         u32 seq = cb->nlh->nlmsg_seq;
3210         u32 filter_mask = 0;
3211         int err;
3212
3213         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3214                 struct nlattr *extfilt;
3215
3216                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3217                                           IFLA_EXT_MASK);
3218                 if (extfilt) {
3219                         if (nla_len(extfilt) < sizeof(filter_mask))
3220                                 return -EINVAL;
3221
3222                         filter_mask = nla_get_u32(extfilt);
3223                 }
3224         }
3225
3226         rcu_read_lock();
3227         for_each_netdev_rcu(net, dev) {
3228                 const struct net_device_ops *ops = dev->netdev_ops;
3229                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3230
3231                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3232                         if (idx >= cb->args[0]) {
3233                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3234                                                 skb, portid, seq, dev,
3235                                                 filter_mask, NLM_F_MULTI);
3236                                 if (err < 0 && err != -EOPNOTSUPP)
3237                                         break;
3238                         }
3239                         idx++;
3240                 }
3241
3242                 if (ops->ndo_bridge_getlink) {
3243                         if (idx >= cb->args[0]) {
3244                                 err = ops->ndo_bridge_getlink(skb, portid,
3245                                                               seq, dev,
3246                                                               filter_mask,
3247                                                               NLM_F_MULTI);
3248                                 if (err < 0 && err != -EOPNOTSUPP)
3249                                         break;
3250                         }
3251                         idx++;
3252                 }
3253         }
3254         rcu_read_unlock();
3255         cb->args[0] = idx;
3256
3257         return skb->len;
3258 }
3259
3260 static inline size_t bridge_nlmsg_size(void)
3261 {
3262         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3263                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3264                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3265                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3266                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3267                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3268                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3269                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3270                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3271                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3272                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3273 }
3274
3275 static int rtnl_bridge_notify(struct net_device *dev)
3276 {
3277         struct net *net = dev_net(dev);
3278         struct sk_buff *skb;
3279         int err = -EOPNOTSUPP;
3280
3281         if (!dev->netdev_ops->ndo_bridge_getlink)
3282                 return 0;
3283
3284         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3285         if (!skb) {
3286                 err = -ENOMEM;
3287                 goto errout;
3288         }
3289
3290         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3291         if (err < 0)
3292                 goto errout;
3293
3294         if (!skb->len)
3295                 goto errout;
3296
3297         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3298         return 0;
3299 errout:
3300         WARN_ON(err == -EMSGSIZE);
3301         kfree_skb(skb);
3302         if (err)
3303                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3304         return err;
3305 }
3306
3307 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3308 {
3309         struct net *net = sock_net(skb->sk);
3310         struct ifinfomsg *ifm;
3311         struct net_device *dev;
3312         struct nlattr *br_spec, *attr = NULL;
3313         int rem, err = -EOPNOTSUPP;
3314         u16 flags = 0;
3315         bool have_flags = false;
3316
3317         if (nlmsg_len(nlh) < sizeof(*ifm))
3318                 return -EINVAL;
3319
3320         ifm = nlmsg_data(nlh);
3321         if (ifm->ifi_family != AF_BRIDGE)
3322                 return -EPFNOSUPPORT;
3323
3324         dev = __dev_get_by_index(net, ifm->ifi_index);
3325         if (!dev) {
3326                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3327                 return -ENODEV;
3328         }
3329
3330         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3331         if (br_spec) {
3332                 nla_for_each_nested(attr, br_spec, rem) {
3333                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3334                                 if (nla_len(attr) < sizeof(flags))
3335                                         return -EINVAL;
3336
3337                                 have_flags = true;
3338                                 flags = nla_get_u16(attr);
3339                                 break;
3340                         }
3341                 }
3342         }
3343
3344         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3345                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3346
3347                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3348                         err = -EOPNOTSUPP;
3349                         goto out;
3350                 }
3351
3352                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3353                 if (err)
3354                         goto out;
3355
3356                 flags &= ~BRIDGE_FLAGS_MASTER;
3357         }
3358
3359         if ((flags & BRIDGE_FLAGS_SELF)) {
3360                 if (!dev->netdev_ops->ndo_bridge_setlink)
3361                         err = -EOPNOTSUPP;
3362                 else
3363                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3364                                                                   flags);
3365                 if (!err) {
3366                         flags &= ~BRIDGE_FLAGS_SELF;
3367
3368                         /* Generate event to notify upper layer of bridge
3369                          * change
3370                          */
3371                         err = rtnl_bridge_notify(dev);
3372                 }
3373         }
3374
3375         if (have_flags)
3376                 memcpy(nla_data(attr), &flags, sizeof(flags));
3377 out:
3378         return err;
3379 }
3380
3381 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3382 {
3383         struct net *net = sock_net(skb->sk);
3384         struct ifinfomsg *ifm;
3385         struct net_device *dev;
3386         struct nlattr *br_spec, *attr = NULL;
3387         int rem, err = -EOPNOTSUPP;
3388         u16 flags = 0;
3389         bool have_flags = false;
3390
3391         if (nlmsg_len(nlh) < sizeof(*ifm))
3392                 return -EINVAL;
3393
3394         ifm = nlmsg_data(nlh);
3395         if (ifm->ifi_family != AF_BRIDGE)
3396                 return -EPFNOSUPPORT;
3397
3398         dev = __dev_get_by_index(net, ifm->ifi_index);
3399         if (!dev) {
3400                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3401                 return -ENODEV;
3402         }
3403
3404         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3405         if (br_spec) {
3406                 nla_for_each_nested(attr, br_spec, rem) {
3407                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3408                                 if (nla_len(attr) < sizeof(flags))
3409                                         return -EINVAL;
3410
3411                                 have_flags = true;
3412                                 flags = nla_get_u16(attr);
3413                                 break;
3414                         }
3415                 }
3416         }
3417
3418         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3419                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3420
3421                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3422                         err = -EOPNOTSUPP;
3423                         goto out;
3424                 }
3425
3426                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3427                 if (err)
3428                         goto out;
3429
3430                 flags &= ~BRIDGE_FLAGS_MASTER;
3431         }
3432
3433         if ((flags & BRIDGE_FLAGS_SELF)) {
3434                 if (!dev->netdev_ops->ndo_bridge_dellink)
3435                         err = -EOPNOTSUPP;
3436                 else
3437                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3438                                                                   flags);
3439
3440                 if (!err) {
3441                         flags &= ~BRIDGE_FLAGS_SELF;
3442
3443                         /* Generate event to notify upper layer of bridge
3444                          * change
3445                          */
3446                         err = rtnl_bridge_notify(dev);
3447                 }
3448         }
3449
3450         if (have_flags)
3451                 memcpy(nla_data(attr), &flags, sizeof(flags));
3452 out:
3453         return err;
3454 }
3455
3456 /* Process one rtnetlink message. */
3457
3458 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3459 {
3460         struct net *net = sock_net(skb->sk);
3461         rtnl_doit_func doit;
3462         int kind;
3463         int family;
3464         int type;
3465         int err;
3466
3467         type = nlh->nlmsg_type;
3468         if (type > RTM_MAX)
3469                 return -EOPNOTSUPP;
3470
3471         type -= RTM_BASE;
3472
3473         /* All the messages must have at least 1 byte length */
3474         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3475                 return 0;
3476
3477         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3478         kind = type&3;
3479
3480         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3481                 return -EPERM;
3482
3483         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3484                 struct sock *rtnl;
3485                 rtnl_dumpit_func dumpit;
3486                 rtnl_calcit_func calcit;
3487                 u16 min_dump_alloc = 0;
3488
3489                 dumpit = rtnl_get_dumpit(family, type);
3490                 if (dumpit == NULL)
3491                         return -EOPNOTSUPP;
3492                 calcit = rtnl_get_calcit(family, type);
3493                 if (calcit)
3494                         min_dump_alloc = calcit(skb, nlh);
3495
3496                 __rtnl_unlock();
3497                 rtnl = net->rtnl;
3498                 {
3499                         struct netlink_dump_control c = {
3500                                 .dump           = dumpit,
3501                                 .min_dump_alloc = min_dump_alloc,
3502                         };
3503                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3504                 }
3505                 rtnl_lock();
3506                 return err;
3507         }
3508
3509         doit = rtnl_get_doit(family, type);
3510         if (doit == NULL)
3511                 return -EOPNOTSUPP;
3512
3513         return doit(skb, nlh);
3514 }
3515
3516 static void rtnetlink_rcv(struct sk_buff *skb)
3517 {
3518         rtnl_lock();
3519         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3520         rtnl_unlock();
3521 }
3522
3523 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3524 {
3525         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3526
3527         switch (event) {
3528         case NETDEV_UP:
3529         case NETDEV_DOWN:
3530         case NETDEV_PRE_UP:
3531         case NETDEV_POST_INIT:
3532         case NETDEV_REGISTER:
3533         case NETDEV_CHANGE:
3534         case NETDEV_PRE_TYPE_CHANGE:
3535         case NETDEV_GOING_DOWN:
3536         case NETDEV_UNREGISTER:
3537         case NETDEV_UNREGISTER_FINAL:
3538         case NETDEV_RELEASE:
3539         case NETDEV_JOIN:
3540         case NETDEV_BONDING_INFO:
3541                 break;
3542         default:
3543                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3544                 break;
3545         }
3546         return NOTIFY_DONE;
3547 }
3548
3549 static struct notifier_block rtnetlink_dev_notifier = {
3550         .notifier_call  = rtnetlink_event,
3551 };
3552
3553
3554 static int __net_init rtnetlink_net_init(struct net *net)
3555 {
3556         struct sock *sk;
3557         struct netlink_kernel_cfg cfg = {
3558                 .groups         = RTNLGRP_MAX,
3559                 .input          = rtnetlink_rcv,
3560                 .cb_mutex       = &rtnl_mutex,
3561                 .flags          = NL_CFG_F_NONROOT_RECV,
3562         };
3563
3564         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3565         if (!sk)
3566                 return -ENOMEM;
3567         net->rtnl = sk;
3568         return 0;
3569 }
3570
3571 static void __net_exit rtnetlink_net_exit(struct net *net)
3572 {
3573         netlink_kernel_release(net->rtnl);
3574         net->rtnl = NULL;
3575 }
3576
3577 static struct pernet_operations rtnetlink_net_ops = {
3578         .init = rtnetlink_net_init,
3579         .exit = rtnetlink_net_exit,
3580 };
3581
3582 void __init rtnetlink_init(void)
3583 {
3584         if (register_pernet_subsys(&rtnetlink_net_ops))
3585                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3586
3587         register_netdevice_notifier(&rtnetlink_dev_notifier);
3588
3589         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3590                       rtnl_dump_ifinfo, rtnl_calcit);
3591         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3592         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3593         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3594
3595         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3596         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3597
3598         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3599         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3600         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3601
3602         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3603         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3604         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3605 }