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(1); /* IFLA_PROTO_DOWN */
 913
 914 }
 915
 916 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 917 {
 918         struct nlattr *vf_ports;
 919         struct nlattr *vf_port;
 920         int vf;
 921         int err;
 922
 923         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 924         if (!vf_ports)
 925                 return -EMSGSIZE;
 926
 927         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 928                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 929                 if (!vf_port)
 930                         goto nla_put_failure;
 931                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 932                         goto nla_put_failure;
 933                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 934                 if (err == -EMSGSIZE)
 935                         goto nla_put_failure;
 936                 if (err) {
 937                         nla_nest_cancel(skb, vf_port);
 938                         continue;
 939                 }
 940                 nla_nest_end(skb, vf_port);
 941         }
 942
 943         nla_nest_end(skb, vf_ports);
 944
 945         return 0;
 946
 947 nla_put_failure:
 948         nla_nest_cancel(skb, vf_ports);
 949         return -EMSGSIZE;
 950 }
 951
 952 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 953 {
 954         struct nlattr *port_self;
 955         int err;
 956
 957         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 958         if (!port_self)
 959                 return -EMSGSIZE;
 960
 961         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 962         if (err) {
 963                 nla_nest_cancel(skb, port_self);
 964                 return (err == -EMSGSIZE) ? err : 0;
 965         }
 966
 967         nla_nest_end(skb, port_self);
 968
 969         return 0;
 970 }
 971
 972 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 973                           u32 ext_filter_mask)
 974 {
 975         int err;
 976
 977         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 978             !(ext_filter_mask & RTEXT_FILTER_VF))
 979                 return 0;
 980
 981         err = rtnl_port_self_fill(skb, dev);
 982         if (err)
 983                 return err;
 984
 985         if (dev_num_vf(dev->dev.parent)) {
 986                 err = rtnl_vf_ports_fill(skb, dev);
 987                 if (err)
 988                         return err;
 989         }
 990
 991         return 0;
 992 }
 993
 994 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 995 {
 996         int err;
 997         struct netdev_phys_item_id ppid;
 998
 999         err = dev_get_phys_port_id(dev, &ppid);
1000         if (err) {
1001                 if (err == -EOPNOTSUPP)
1002                         return 0;
1003                 return err;
1004         }
1005
1006         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1007                 return -EMSGSIZE;
1008
1009         return 0;
1010 }
1011
1012 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1013 {
1014         char name[IFNAMSIZ];
1015         int err;
1016
1017         err = dev_get_phys_port_name(dev, name, sizeof(name));
1018         if (err) {
1019                 if (err == -EOPNOTSUPP)
1020                         return 0;
1021                 return err;
1022         }
1023
1024         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1025                 return -EMSGSIZE;
1026
1027         return 0;
1028 }
1029
1030 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1031 {
1032         int err;
1033         struct switchdev_attr attr = {
1034                 .orig_dev = dev,
1035                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1036                 .flags = SWITCHDEV_F_NO_RECURSE,
1037         };
1038
1039         err = switchdev_port_attr_get(dev, &attr);
1040         if (err) {
1041                 if (err == -EOPNOTSUPP)
1042                         return 0;
1043                 return err;
1044         }
1045
1046         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1047                     attr.u.ppid.id))
1048                 return -EMSGSIZE;
1049
1050         return 0;
1051 }
1052
1053 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1054                                               struct net_device *dev)
1055 {
1056         const struct rtnl_link_stats64 *stats;
1057         struct rtnl_link_stats64 temp;
1058         struct nlattr *attr;
1059
1060         stats = dev_get_stats(dev, &temp);
1061
1062         attr = nla_reserve(skb, IFLA_STATS,
1063                            sizeof(struct rtnl_link_stats));
1064         if (!attr)
1065                 return -EMSGSIZE;
1066
1067         copy_rtnl_link_stats(nla_data(attr), stats);
1068
1069         attr = nla_reserve(skb, IFLA_STATS64,
1070                            sizeof(struct rtnl_link_stats64));
1071         if (!attr)
1072                 return -EMSGSIZE;
1073
1074         copy_rtnl_link_stats64(nla_data(attr), stats);
1075
1076         return 0;
1077 }
1078
1079 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1080                                                struct net_device *dev,
1081                                                int vfs_num,
1082                                                struct nlattr *vfinfo)
1083 {
1084         struct ifla_vf_rss_query_en vf_rss_query_en;
1085         struct ifla_vf_link_state vf_linkstate;
1086         struct ifla_vf_spoofchk vf_spoofchk;
1087         struct ifla_vf_tx_rate vf_tx_rate;
1088         struct ifla_vf_stats vf_stats;
1089         struct ifla_vf_trust vf_trust;
1090         struct ifla_vf_vlan vf_vlan;
1091         struct ifla_vf_rate vf_rate;
1092         struct nlattr *vf, *vfstats;
1093         struct ifla_vf_mac vf_mac;
1094         struct ifla_vf_info ivi;
1095
1096         /* Not all SR-IOV capable drivers support the
1097          * spoofcheck and "RSS query enable" query.  Preset to
1098          * -1 so the user space tool can detect that the driver
1099          * didn't report anything.
1100          */
1101         ivi.spoofchk = -1;
1102         ivi.rss_query_en = -1;
1103         ivi.trusted = -1;
1104         memset(ivi.mac, 0, sizeof(ivi.mac));
1105         /* The default value for VF link state is "auto"
1106          * IFLA_VF_LINK_STATE_AUTO which equals zero
1107          */
1108         ivi.linkstate = 0;
1109         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1110                 return 0;
1111
1112         vf_mac.vf =
1113                 vf_vlan.vf =
1114                 vf_rate.vf =
1115                 vf_tx_rate.vf =
1116                 vf_spoofchk.vf =
1117                 vf_linkstate.vf =
1118                 vf_rss_query_en.vf =
1119                 vf_trust.vf = ivi.vf;
1120
1121         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1122         vf_vlan.vlan = ivi.vlan;
1123         vf_vlan.qos = ivi.qos;
1124         vf_tx_rate.rate = ivi.max_tx_rate;
1125         vf_rate.min_tx_rate = ivi.min_tx_rate;
1126         vf_rate.max_tx_rate = ivi.max_tx_rate;
1127         vf_spoofchk.setting = ivi.spoofchk;
1128         vf_linkstate.link_state = ivi.linkstate;
1129         vf_rss_query_en.setting = ivi.rss_query_en;
1130         vf_trust.setting = ivi.trusted;
1131         vf = nla_nest_start(skb, IFLA_VF_INFO);
1132         if (!vf) {
1133                 nla_nest_cancel(skb, vfinfo);
1134                 return -EMSGSIZE;
1135         }
1136         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1137             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1138             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1139                     &vf_rate) ||
1140             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1141                     &vf_tx_rate) ||
1142             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1143                     &vf_spoofchk) ||
1144             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1145                     &vf_linkstate) ||
1146             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1147                     sizeof(vf_rss_query_en),
1148                     &vf_rss_query_en) ||
1149             nla_put(skb, IFLA_VF_TRUST,
1150                     sizeof(vf_trust), &vf_trust))
1151                 return -EMSGSIZE;
1152         memset(&vf_stats, 0, sizeof(vf_stats));
1153         if (dev->netdev_ops->ndo_get_vf_stats)
1154                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1155                                                 &vf_stats);
1156         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1157         if (!vfstats) {
1158                 nla_nest_cancel(skb, vf);
1159                 nla_nest_cancel(skb, vfinfo);
1160                 return -EMSGSIZE;
1161         }
1162         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1163                         vf_stats.rx_packets) ||
1164             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1165                         vf_stats.tx_packets) ||
1166             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1167                         vf_stats.rx_bytes) ||
1168             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1169                         vf_stats.tx_bytes) ||
1170             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1171                         vf_stats.broadcast) ||
1172             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1173                         vf_stats.multicast))
1174                 return -EMSGSIZE;
1175         nla_nest_end(skb, vfstats);
1176         nla_nest_end(skb, vf);
1177         return 0;
1178 }
1179
1180 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1181 {
1182         struct rtnl_link_ifmap map = {
1183                 .mem_start   = dev->mem_start,
1184                 .mem_end     = dev->mem_end,
1185                 .base_addr   = dev->base_addr,
1186                 .irq         = dev->irq,
1187                 .dma         = dev->dma,
1188                 .port        = dev->if_port,
1189         };
1190         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1191                 return -EMSGSIZE;
1192
1193         return 0;
1194 }
1195
1196 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1197                             int type, u32 pid, u32 seq, u32 change,
1198                             unsigned int flags, u32 ext_filter_mask)
1199 {
1200         struct ifinfomsg *ifm;
1201         struct nlmsghdr *nlh;
1202         struct nlattr *af_spec;
1203         struct rtnl_af_ops *af_ops;
1204         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1205
1206         ASSERT_RTNL();
1207         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1208         if (nlh == NULL)
1209                 return -EMSGSIZE;
1210
1211         ifm = nlmsg_data(nlh);
1212         ifm->ifi_family = AF_UNSPEC;
1213         ifm->__ifi_pad = 0;
1214         ifm->ifi_type = dev->type;
1215         ifm->ifi_index = dev->ifindex;
1216         ifm->ifi_flags = dev_get_flags(dev);
1217         ifm->ifi_change = change;
1218
1219         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1220             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1221             nla_put_u8(skb, IFLA_OPERSTATE,
1222                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1223             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1224             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1225             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1226             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1227             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1228             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1229             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1230 #ifdef CONFIG_RPS
1231             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1232 #endif
1233             (dev->ifindex != dev_get_iflink(dev) &&
1234              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1235             (upper_dev &&
1236              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1237             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1238             (dev->qdisc &&
1239              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1240             (dev->ifalias &&
1241              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1242             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1243                         atomic_read(&dev->carrier_changes)) ||
1244             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1245                 goto nla_put_failure;
1246
1247         if (rtnl_fill_link_ifmap(skb, dev))
1248                 goto nla_put_failure;
1249
1250         if (dev->addr_len) {
1251                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1252                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1253                         goto nla_put_failure;
1254         }
1255
1256         if (rtnl_phys_port_id_fill(skb, dev))
1257                 goto nla_put_failure;
1258
1259         if (rtnl_phys_port_name_fill(skb, dev))
1260                 goto nla_put_failure;
1261
1262         if (rtnl_phys_switch_id_fill(skb, dev))
1263                 goto nla_put_failure;
1264
1265         if (rtnl_fill_stats(skb, dev))
1266                 goto nla_put_failure;
1267
1268         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1269             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1270                 goto nla_put_failure;
1271
1272         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1273             ext_filter_mask & RTEXT_FILTER_VF) {
1274                 int i;
1275                 struct nlattr *vfinfo;
1276                 int num_vfs = dev_num_vf(dev->dev.parent);
1277
1278                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1279                 if (!vfinfo)
1280                         goto nla_put_failure;
1281                 for (i = 0; i < num_vfs; i++) {
1282                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1283                                 goto nla_put_failure;
1284                 }
1285
1286                 nla_nest_end(skb, vfinfo);
1287         }
1288
1289         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1290                 goto nla_put_failure;
1291
1292         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1293                 if (rtnl_link_fill(skb, dev) < 0)
1294                         goto nla_put_failure;
1295         }
1296
1297         if (dev->rtnl_link_ops &&
1298             dev->rtnl_link_ops->get_link_net) {
1299                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1300
1301                 if (!net_eq(dev_net(dev), link_net)) {
1302                         int id = peernet2id_alloc(dev_net(dev), link_net);
1303
1304                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1305                                 goto nla_put_failure;
1306                 }
1307         }
1308
1309         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1310                 goto nla_put_failure;
1311
1312         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1313                 if (af_ops->fill_link_af) {
1314                         struct nlattr *af;
1315                         int err;
1316
1317                         if (!(af = nla_nest_start(skb, af_ops->family)))
1318                                 goto nla_put_failure;
1319
1320                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1321
1322                         /*
1323                          * Caller may return ENODATA to indicate that there
1324                          * was no data to be dumped. This is not an error, it
1325                          * means we should trim the attribute header and
1326                          * continue.
1327                          */
1328                         if (err == -ENODATA)
1329                                 nla_nest_cancel(skb, af);
1330                         else if (err < 0)
1331                                 goto nla_put_failure;
1332
1333                         nla_nest_end(skb, af);
1334                 }
1335         }
1336
1337         nla_nest_end(skb, af_spec);
1338
1339         nlmsg_end(skb, nlh);
1340         return 0;
1341
1342 nla_put_failure:
1343         nlmsg_cancel(skb, nlh);
1344         return -EMSGSIZE;
1345 }
1346
1347 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1348         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1349         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1350         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1351         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1352         [IFLA_MTU]              = { .type = NLA_U32 },
1353         [IFLA_LINK]             = { .type = NLA_U32 },
1354         [IFLA_MASTER]           = { .type = NLA_U32 },
1355         [IFLA_CARRIER]          = { .type = NLA_U8 },
1356         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1357         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1358         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1359         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1360         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1361         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1362         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1363         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1364         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1365         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1366         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1367         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1368         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1369         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1370         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1371         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1372         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1373         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1374         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1375         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1376         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1377 };
1378
1379 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1380         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1381         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1382         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1383         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1384 };
1385
1386 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1387         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1388         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1389         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1390         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1391         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1392         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1393         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1394         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1395         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1396         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1397         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1398 };
1399
1400 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1401         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1402         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1403                                     .len = PORT_PROFILE_MAX },
1404         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1405                                     .len = sizeof(struct ifla_port_vsi)},
1406         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1407                                       .len = PORT_UUID_MAX },
1408         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1409                                     .len = PORT_UUID_MAX },
1410         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1411         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1412 };
1413
1414 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1415 {
1416         const struct rtnl_link_ops *ops = NULL;
1417         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1418
1419         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1420                 return NULL;
1421
1422         if (linfo[IFLA_INFO_KIND]) {
1423                 char kind[MODULE_NAME_LEN];
1424
1425                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1426                 ops = rtnl_link_ops_get(kind);
1427         }
1428
1429         return ops;
1430 }
1431
1432 static bool link_master_filtered(struct net_device *dev, int master_idx)
1433 {
1434         struct net_device *master;
1435
1436         if (!master_idx)
1437                 return false;
1438
1439         master = netdev_master_upper_dev_get(dev);
1440         if (!master || master->ifindex != master_idx)
1441                 return true;
1442
1443         return false;
1444 }
1445
1446 static bool link_kind_filtered(const struct net_device *dev,
1447                                const struct rtnl_link_ops *kind_ops)
1448 {
1449         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1450                 return true;
1451
1452         return false;
1453 }
1454
1455 static bool link_dump_filtered(struct net_device *dev,
1456                                int master_idx,
1457                                const struct rtnl_link_ops *kind_ops)
1458 {
1459         if (link_master_filtered(dev, master_idx) ||
1460             link_kind_filtered(dev, kind_ops))
1461                 return true;
1462
1463         return false;
1464 }
1465
1466 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1467 {
1468         struct net *net = sock_net(skb->sk);
1469         int h, s_h;
1470         int idx = 0, s_idx;
1471         struct net_device *dev;
1472         struct hlist_head *head;
1473         struct nlattr *tb[IFLA_MAX+1];
1474         u32 ext_filter_mask = 0;
1475         const struct rtnl_link_ops *kind_ops = NULL;
1476         unsigned int flags = NLM_F_MULTI;
1477         int master_idx = 0;
1478         int err;
1479         int hdrlen;
1480
1481         s_h = cb->args[0];
1482         s_idx = cb->args[1];
1483
1484         cb->seq = net->dev_base_seq;
1485
1486         /* A hack to preserve kernel<->userspace interface.
1487          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1488          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1489          * what iproute2 < v3.9.0 used.
1490          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1491          * attribute, its netlink message is shorter than struct ifinfomsg.
1492          */
1493         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1494                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1495
1496         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1497
1498                 if (tb[IFLA_EXT_MASK])
1499                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1500
1501                 if (tb[IFLA_MASTER])
1502                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1503
1504                 if (tb[IFLA_LINKINFO])
1505                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1506
1507                 if (master_idx || kind_ops)
1508                         flags |= NLM_F_DUMP_FILTERED;
1509         }
1510
1511         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1512                 idx = 0;
1513                 head = &net->dev_index_head[h];
1514                 hlist_for_each_entry(dev, head, index_hlist) {
1515                         if (link_dump_filtered(dev, master_idx, kind_ops))
1516                                 continue;
1517                         if (idx < s_idx)
1518                                 goto cont;
1519                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1520                                                NETLINK_CB(cb->skb).portid,
1521                                                cb->nlh->nlmsg_seq, 0,
1522                                                flags,
1523                                                ext_filter_mask);
1524                         /* If we ran out of room on the first message,
1525                          * we're in trouble
1526                          */
1527                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1528
1529                         if (err < 0)
1530                                 goto out;
1531
1532                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1533 cont:
1534                         idx++;
1535                 }
1536         }
1537 out:
1538         cb->args[1] = idx;
1539         cb->args[0] = h;
1540
1541         return skb->len;
1542 }
1543
1544 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1545 {
1546         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1547 }
1548 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1549
1550 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1551 {
1552         struct net *net;
1553         /* Examine the link attributes and figure out which
1554          * network namespace we are talking about.
1555          */
1556         if (tb[IFLA_NET_NS_PID])
1557                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1558         else if (tb[IFLA_NET_NS_FD])
1559                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1560         else
1561                 net = get_net(src_net);
1562         return net;
1563 }
1564 EXPORT_SYMBOL(rtnl_link_get_net);
1565
1566 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1567 {
1568         if (dev) {
1569                 if (tb[IFLA_ADDRESS] &&
1570                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1571                         return -EINVAL;
1572
1573                 if (tb[IFLA_BROADCAST] &&
1574                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1575                         return -EINVAL;
1576         }
1577
1578         if (tb[IFLA_AF_SPEC]) {
1579                 struct nlattr *af;
1580                 int rem, err;
1581
1582                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1583                         const struct rtnl_af_ops *af_ops;
1584
1585                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1586                                 return -EAFNOSUPPORT;
1587
1588                         if (!af_ops->set_link_af)
1589                                 return -EOPNOTSUPP;
1590
1591                         if (af_ops->validate_link_af) {
1592                                 err = af_ops->validate_link_af(dev, af);
1593                                 if (err < 0)
1594                                         return err;
1595                         }
1596                 }
1597         }
1598
1599         return 0;
1600 }
1601
1602 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1603                                   int guid_type)
1604 {
1605         const struct net_device_ops *ops = dev->netdev_ops;
1606
1607         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1608 }
1609
1610 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1611 {
1612         if (dev->type != ARPHRD_INFINIBAND)
1613                 return -EOPNOTSUPP;
1614
1615         return handle_infiniband_guid(dev, ivt, guid_type);
1616 }
1617
1618 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1619 {
1620         const struct net_device_ops *ops = dev->netdev_ops;
1621         int err = -EINVAL;
1622
1623         if (tb[IFLA_VF_MAC]) {
1624                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1625
1626                 err = -EOPNOTSUPP;
1627                 if (ops->ndo_set_vf_mac)
1628                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1629                                                   ivm->mac);
1630                 if (err < 0)
1631                         return err;
1632         }
1633
1634         if (tb[IFLA_VF_VLAN]) {
1635                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1636
1637                 err = -EOPNOTSUPP;
1638                 if (ops->ndo_set_vf_vlan)
1639                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1640                                                    ivv->qos);
1641                 if (err < 0)
1642                         return err;
1643         }
1644
1645         if (tb[IFLA_VF_TX_RATE]) {
1646                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1647                 struct ifla_vf_info ivf;
1648
1649                 err = -EOPNOTSUPP;
1650                 if (ops->ndo_get_vf_config)
1651                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1652                 if (err < 0)
1653                         return err;
1654
1655                 err = -EOPNOTSUPP;
1656                 if (ops->ndo_set_vf_rate)
1657                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1658                                                    ivf.min_tx_rate,
1659                                                    ivt->rate);
1660                 if (err < 0)
1661                         return err;
1662         }
1663
1664         if (tb[IFLA_VF_RATE]) {
1665                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1666
1667                 err = -EOPNOTSUPP;
1668                 if (ops->ndo_set_vf_rate)
1669                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1670                                                    ivt->min_tx_rate,
1671                                                    ivt->max_tx_rate);
1672                 if (err < 0)
1673                         return err;
1674         }
1675
1676         if (tb[IFLA_VF_SPOOFCHK]) {
1677                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1678
1679                 err = -EOPNOTSUPP;
1680                 if (ops->ndo_set_vf_spoofchk)
1681                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1682                                                        ivs->setting);
1683                 if (err < 0)
1684                         return err;
1685         }
1686
1687         if (tb[IFLA_VF_LINK_STATE]) {
1688                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1689
1690                 err = -EOPNOTSUPP;
1691                 if (ops->ndo_set_vf_link_state)
1692                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1693                                                          ivl->link_state);
1694                 if (err < 0)
1695                         return err;
1696         }
1697
1698         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1699                 struct ifla_vf_rss_query_en *ivrssq_en;
1700
1701                 err = -EOPNOTSUPP;
1702                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1703                 if (ops->ndo_set_vf_rss_query_en)
1704                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1705                                                            ivrssq_en->setting);
1706                 if (err < 0)
1707                         return err;
1708         }
1709
1710         if (tb[IFLA_VF_TRUST]) {
1711                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1712
1713                 err = -EOPNOTSUPP;
1714                 if (ops->ndo_set_vf_trust)
1715                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1716                 if (err < 0)
1717                         return err;
1718         }
1719
1720         if (tb[IFLA_VF_IB_NODE_GUID]) {
1721                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1722
1723                 if (!ops->ndo_set_vf_guid)
1724                         return -EOPNOTSUPP;
1725
1726                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1727         }
1728
1729         if (tb[IFLA_VF_IB_PORT_GUID]) {
1730                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1731
1732                 if (!ops->ndo_set_vf_guid)
1733                         return -EOPNOTSUPP;
1734
1735                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1736         }
1737
1738         return err;
1739 }
1740
1741 static int do_set_master(struct net_device *dev, int ifindex)
1742 {
1743         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1744         const struct net_device_ops *ops;
1745         int err;
1746
1747         if (upper_dev) {
1748                 if (upper_dev->ifindex == ifindex)
1749                         return 0;
1750                 ops = upper_dev->netdev_ops;
1751                 if (ops->ndo_del_slave) {
1752                         err = ops->ndo_del_slave(upper_dev, dev);
1753                         if (err)
1754                                 return err;
1755                 } else {
1756                         return -EOPNOTSUPP;
1757                 }
1758         }
1759
1760         if (ifindex) {
1761                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1762                 if (!upper_dev)
1763                         return -EINVAL;
1764                 ops = upper_dev->netdev_ops;
1765                 if (ops->ndo_add_slave) {
1766                         err = ops->ndo_add_slave(upper_dev, dev);
1767                         if (err)
1768                                 return err;
1769                 } else {
1770                         return -EOPNOTSUPP;
1771                 }
1772         }
1773         return 0;
1774 }
1775
1776 #define DO_SETLINK_MODIFIED     0x01
1777 /* notify flag means notify + modified. */
1778 #define DO_SETLINK_NOTIFY       0x03
1779 static int do_setlink(const struct sk_buff *skb,
1780                       struct net_device *dev, struct ifinfomsg *ifm,
1781                       struct nlattr **tb, char *ifname, int status)
1782 {
1783         const struct net_device_ops *ops = dev->netdev_ops;
1784         int err;
1785
1786         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1787                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1788                 if (IS_ERR(net)) {
1789                         err = PTR_ERR(net);
1790                         goto errout;
1791                 }
1792                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1793                         put_net(net);
1794                         err = -EPERM;
1795                         goto errout;
1796                 }
1797                 err = dev_change_net_namespace(dev, net, ifname);
1798                 put_net(net);
1799                 if (err)
1800                         goto errout;
1801                 status |= DO_SETLINK_MODIFIED;
1802         }
1803
1804         if (tb[IFLA_MAP]) {
1805                 struct rtnl_link_ifmap *u_map;
1806                 struct ifmap k_map;
1807
1808                 if (!ops->ndo_set_config) {
1809                         err = -EOPNOTSUPP;
1810                         goto errout;
1811                 }
1812
1813                 if (!netif_device_present(dev)) {
1814                         err = -ENODEV;
1815                         goto errout;
1816                 }
1817
1818                 u_map = nla_data(tb[IFLA_MAP]);
1819                 k_map.mem_start = (unsigned long) u_map->mem_start;
1820                 k_map.mem_end = (unsigned long) u_map->mem_end;
1821                 k_map.base_addr = (unsigned short) u_map->base_addr;
1822                 k_map.irq = (unsigned char) u_map->irq;
1823                 k_map.dma = (unsigned char) u_map->dma;
1824                 k_map.port = (unsigned char) u_map->port;
1825
1826                 err = ops->ndo_set_config(dev, &k_map);
1827                 if (err < 0)
1828                         goto errout;
1829
1830                 status |= DO_SETLINK_NOTIFY;
1831         }
1832
1833         if (tb[IFLA_ADDRESS]) {
1834                 struct sockaddr *sa;
1835                 int len;
1836
1837                 len = sizeof(sa_family_t) + dev->addr_len;
1838                 sa = kmalloc(len, GFP_KERNEL);
1839                 if (!sa) {
1840                         err = -ENOMEM;
1841                         goto errout;
1842                 }
1843                 sa->sa_family = dev->type;
1844                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1845                        dev->addr_len);
1846                 err = dev_set_mac_address(dev, sa);
1847                 kfree(sa);
1848                 if (err)
1849                         goto errout;
1850                 status |= DO_SETLINK_MODIFIED;
1851         }
1852
1853         if (tb[IFLA_MTU]) {
1854                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1855                 if (err < 0)
1856                         goto errout;
1857                 status |= DO_SETLINK_MODIFIED;
1858         }
1859
1860         if (tb[IFLA_GROUP]) {
1861                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1862                 status |= DO_SETLINK_NOTIFY;
1863         }
1864
1865         /*
1866          * Interface selected by interface index but interface
1867          * name provided implies that a name change has been
1868          * requested.
1869          */
1870         if (ifm->ifi_index > 0 && ifname[0]) {
1871                 err = dev_change_name(dev, ifname);
1872                 if (err < 0)
1873                         goto errout;
1874                 status |= DO_SETLINK_MODIFIED;
1875         }
1876
1877         if (tb[IFLA_IFALIAS]) {
1878                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1879                                     nla_len(tb[IFLA_IFALIAS]));
1880                 if (err < 0)
1881                         goto errout;
1882                 status |= DO_SETLINK_NOTIFY;
1883         }
1884
1885         if (tb[IFLA_BROADCAST]) {
1886                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1887                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1888         }
1889
1890         if (ifm->ifi_flags || ifm->ifi_change) {
1891                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1892                 if (err < 0)
1893                         goto errout;
1894         }
1895
1896         if (tb[IFLA_MASTER]) {
1897                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1898                 if (err)
1899                         goto errout;
1900                 status |= DO_SETLINK_MODIFIED;
1901         }
1902
1903         if (tb[IFLA_CARRIER]) {
1904                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1905                 if (err)
1906                         goto errout;
1907                 status |= DO_SETLINK_MODIFIED;
1908         }
1909
1910         if (tb[IFLA_TXQLEN]) {
1911                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1912
1913                 if (dev->tx_queue_len ^ value)
1914                         status |= DO_SETLINK_NOTIFY;
1915
1916                 dev->tx_queue_len = value;
1917         }
1918
1919         if (tb[IFLA_OPERSTATE])
1920                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1921
1922         if (tb[IFLA_LINKMODE]) {
1923                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1924
1925                 write_lock_bh(&dev_base_lock);
1926                 if (dev->link_mode ^ value)
1927                         status |= DO_SETLINK_NOTIFY;
1928                 dev->link_mode = value;
1929                 write_unlock_bh(&dev_base_lock);
1930         }
1931
1932         if (tb[IFLA_VFINFO_LIST]) {
1933                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1934                 struct nlattr *attr;
1935                 int rem;
1936
1937                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1938                         if (nla_type(attr) != IFLA_VF_INFO ||
1939                             nla_len(attr) < NLA_HDRLEN) {
1940                                 err = -EINVAL;
1941                                 goto errout;
1942                         }
1943                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1944                                                ifla_vf_policy);
1945                         if (err < 0)
1946                                 goto errout;
1947                         err = do_setvfinfo(dev, vfinfo);
1948                         if (err < 0)
1949                                 goto errout;
1950                         status |= DO_SETLINK_NOTIFY;
1951                 }
1952         }
1953         err = 0;
1954
1955         if (tb[IFLA_VF_PORTS]) {
1956                 struct nlattr *port[IFLA_PORT_MAX+1];
1957                 struct nlattr *attr;
1958                 int vf;
1959                 int rem;
1960
1961                 err = -EOPNOTSUPP;
1962                 if (!ops->ndo_set_vf_port)
1963                         goto errout;
1964
1965                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1966                         if (nla_type(attr) != IFLA_VF_PORT ||
1967                             nla_len(attr) < NLA_HDRLEN) {
1968                                 err = -EINVAL;
1969                                 goto errout;
1970                         }
1971                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1972                                                ifla_port_policy);
1973                         if (err < 0)
1974                                 goto errout;
1975                         if (!port[IFLA_PORT_VF]) {
1976                                 err = -EOPNOTSUPP;
1977                                 goto errout;
1978                         }
1979                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1980                         err = ops->ndo_set_vf_port(dev, vf, port);
1981                         if (err < 0)
1982                                 goto errout;
1983                         status |= DO_SETLINK_NOTIFY;
1984                 }
1985         }
1986         err = 0;
1987
1988         if (tb[IFLA_PORT_SELF]) {
1989                 struct nlattr *port[IFLA_PORT_MAX+1];
1990
1991                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1992                         tb[IFLA_PORT_SELF], ifla_port_policy);
1993                 if (err < 0)
1994                         goto errout;
1995
1996                 err = -EOPNOTSUPP;
1997                 if (ops->ndo_set_vf_port)
1998                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1999                 if (err < 0)
2000                         goto errout;
2001                 status |= DO_SETLINK_NOTIFY;
2002         }
2003
2004         if (tb[IFLA_AF_SPEC]) {
2005                 struct nlattr *af;
2006                 int rem;
2007
2008                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2009                         const struct rtnl_af_ops *af_ops;
2010
2011                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2012                                 BUG();
2013
2014                         err = af_ops->set_link_af(dev, af);
2015                         if (err < 0)
2016                                 goto errout;
2017
2018                         status |= DO_SETLINK_NOTIFY;
2019                 }
2020         }
2021         err = 0;
2022
2023         if (tb[IFLA_PROTO_DOWN]) {
2024                 err = dev_change_proto_down(dev,
2025                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2026                 if (err)
2027                         goto errout;
2028                 status |= DO_SETLINK_NOTIFY;
2029         }
2030
2031 errout:
2032         if (status & DO_SETLINK_MODIFIED) {
2033                 if (status & DO_SETLINK_NOTIFY)
2034                         netdev_state_change(dev);
2035
2036                 if (err < 0)
2037                         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",
2038                                              dev->name);
2039         }
2040
2041         return err;
2042 }
2043
2044 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2045 {
2046         struct net *net = sock_net(skb->sk);
2047         struct ifinfomsg *ifm;
2048         struct net_device *dev;
2049         int err;
2050         struct nlattr *tb[IFLA_MAX+1];
2051         char ifname[IFNAMSIZ];
2052
2053         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2054         if (err < 0)
2055                 goto errout;
2056
2057         if (tb[IFLA_IFNAME])
2058                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2059         else
2060                 ifname[0] = '\0';
2061
2062         err = -EINVAL;
2063         ifm = nlmsg_data(nlh);
2064         if (ifm->ifi_index > 0)
2065                 dev = __dev_get_by_index(net, ifm->ifi_index);
2066         else if (tb[IFLA_IFNAME])
2067                 dev = __dev_get_by_name(net, ifname);
2068         else
2069                 goto errout;
2070
2071         if (dev == NULL) {
2072                 err = -ENODEV;
2073                 goto errout;
2074         }
2075
2076         err = validate_linkmsg(dev, tb);
2077         if (err < 0)
2078                 goto errout;
2079
2080         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2081 errout:
2082         return err;
2083 }
2084
2085 static int rtnl_group_dellink(const struct net *net, int group)
2086 {
2087         struct net_device *dev, *aux;
2088         LIST_HEAD(list_kill);
2089         bool found = false;
2090
2091         if (!group)
2092                 return -EPERM;
2093
2094         for_each_netdev(net, dev) {
2095                 if (dev->group == group) {
2096                         const struct rtnl_link_ops *ops;
2097
2098                         found = true;
2099                         ops = dev->rtnl_link_ops;
2100                         if (!ops || !ops->dellink)
2101                                 return -EOPNOTSUPP;
2102                 }
2103         }
2104
2105         if (!found)
2106                 return -ENODEV;
2107
2108         for_each_netdev_safe(net, dev, aux) {
2109                 if (dev->group == group) {
2110                         const struct rtnl_link_ops *ops;
2111
2112                         ops = dev->rtnl_link_ops;
2113                         ops->dellink(dev, &list_kill);
2114                 }
2115         }
2116         unregister_netdevice_many(&list_kill);
2117
2118         return 0;
2119 }
2120
2121 int rtnl_delete_link(struct net_device *dev)
2122 {
2123         const struct rtnl_link_ops *ops;
2124         LIST_HEAD(list_kill);
2125
2126         ops = dev->rtnl_link_ops;
2127         if (!ops || !ops->dellink)
2128                 return -EOPNOTSUPP;
2129
2130         ops->dellink(dev, &list_kill);
2131         unregister_netdevice_many(&list_kill);
2132
2133         return 0;
2134 }
2135 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2136
2137 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2138 {
2139         struct net *net = sock_net(skb->sk);
2140         struct net_device *dev;
2141         struct ifinfomsg *ifm;
2142         char ifname[IFNAMSIZ];
2143         struct nlattr *tb[IFLA_MAX+1];
2144         int err;
2145
2146         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2147         if (err < 0)
2148                 return err;
2149
2150         if (tb[IFLA_IFNAME])
2151                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2152
2153         ifm = nlmsg_data(nlh);
2154         if (ifm->ifi_index > 0)
2155                 dev = __dev_get_by_index(net, ifm->ifi_index);
2156         else if (tb[IFLA_IFNAME])
2157                 dev = __dev_get_by_name(net, ifname);
2158         else if (tb[IFLA_GROUP])
2159                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2160         else
2161                 return -EINVAL;
2162
2163         if (!dev)
2164                 return -ENODEV;
2165
2166         return rtnl_delete_link(dev);
2167 }
2168
2169 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2170 {
2171         unsigned int old_flags;
2172         int err;
2173
2174         old_flags = dev->flags;
2175         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2176                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2177                 if (err < 0)
2178                         return err;
2179         }
2180
2181         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2182
2183         __dev_notify_flags(dev, old_flags, ~0U);
2184         return 0;
2185 }
2186 EXPORT_SYMBOL(rtnl_configure_link);
2187
2188 struct net_device *rtnl_create_link(struct net *net,
2189         const char *ifname, unsigned char name_assign_type,
2190         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2191 {
2192         int err;
2193         struct net_device *dev;
2194         unsigned int num_tx_queues = 1;
2195         unsigned int num_rx_queues = 1;
2196
2197         if (tb[IFLA_NUM_TX_QUEUES])
2198                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2199         else if (ops->get_num_tx_queues)
2200                 num_tx_queues = ops->get_num_tx_queues();
2201
2202         if (tb[IFLA_NUM_RX_QUEUES])
2203                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2204         else if (ops->get_num_rx_queues)
2205                 num_rx_queues = ops->get_num_rx_queues();
2206
2207         err = -ENOMEM;
2208         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2209                                ops->setup, num_tx_queues, num_rx_queues);
2210         if (!dev)
2211                 goto err;
2212
2213         dev_net_set(dev, net);
2214         dev->rtnl_link_ops = ops;
2215         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2216
2217         if (tb[IFLA_MTU])
2218                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2219         if (tb[IFLA_ADDRESS]) {
2220                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2221                                 nla_len(tb[IFLA_ADDRESS]));
2222                 dev->addr_assign_type = NET_ADDR_SET;
2223         }
2224         if (tb[IFLA_BROADCAST])
2225                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2226                                 nla_len(tb[IFLA_BROADCAST]));
2227         if (tb[IFLA_TXQLEN])
2228                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2229         if (tb[IFLA_OPERSTATE])
2230                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2231         if (tb[IFLA_LINKMODE])
2232                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2233         if (tb[IFLA_GROUP])
2234                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2235
2236         return dev;
2237
2238 err:
2239         return ERR_PTR(err);
2240 }
2241 EXPORT_SYMBOL(rtnl_create_link);
2242
2243 static int rtnl_group_changelink(const struct sk_buff *skb,
2244                 struct net *net, int group,
2245                 struct ifinfomsg *ifm,
2246                 struct nlattr **tb)
2247 {
2248         struct net_device *dev, *aux;
2249         int err;
2250
2251         for_each_netdev_safe(net, dev, aux) {
2252                 if (dev->group == group) {
2253                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2254                         if (err < 0)
2255                                 return err;
2256                 }
2257         }
2258
2259         return 0;
2260 }
2261
2262 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2263 {
2264         struct net *net = sock_net(skb->sk);
2265         const struct rtnl_link_ops *ops;
2266         const struct rtnl_link_ops *m_ops = NULL;
2267         struct net_device *dev;
2268         struct net_device *master_dev = NULL;
2269         struct ifinfomsg *ifm;
2270         char kind[MODULE_NAME_LEN];
2271         char ifname[IFNAMSIZ];
2272         struct nlattr *tb[IFLA_MAX+1];
2273         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2274         unsigned char name_assign_type = NET_NAME_USER;
2275         int err;
2276
2277 #ifdef CONFIG_MODULES
2278 replay:
2279 #endif
2280         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2281         if (err < 0)
2282                 return err;
2283
2284         if (tb[IFLA_IFNAME])
2285                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2286         else
2287                 ifname[0] = '\0';
2288
2289         ifm = nlmsg_data(nlh);
2290         if (ifm->ifi_index > 0)
2291                 dev = __dev_get_by_index(net, ifm->ifi_index);
2292         else {
2293                 if (ifname[0])
2294                         dev = __dev_get_by_name(net, ifname);
2295                 else
2296                         dev = NULL;
2297         }
2298
2299         if (dev) {
2300                 master_dev = netdev_master_upper_dev_get(dev);
2301                 if (master_dev)
2302                         m_ops = master_dev->rtnl_link_ops;
2303         }
2304
2305         err = validate_linkmsg(dev, tb);
2306         if (err < 0)
2307                 return err;
2308
2309         if (tb[IFLA_LINKINFO]) {
2310                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2311                                        tb[IFLA_LINKINFO], ifla_info_policy);
2312                 if (err < 0)
2313                         return err;
2314         } else
2315                 memset(linkinfo, 0, sizeof(linkinfo));
2316
2317         if (linkinfo[IFLA_INFO_KIND]) {
2318                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2319                 ops = rtnl_link_ops_get(kind);
2320         } else {
2321                 kind[0] = '\0';
2322                 ops = NULL;
2323         }
2324
2325         if (1) {
2326                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2327                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2328                 struct nlattr **data = NULL;
2329                 struct nlattr **slave_data = NULL;
2330                 struct net *dest_net, *link_net = NULL;
2331
2332                 if (ops) {
2333                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2334                                 err = nla_parse_nested(attr, ops->maxtype,
2335                                                        linkinfo[IFLA_INFO_DATA],
2336                                                        ops->policy);
2337                                 if (err < 0)
2338                                         return err;
2339                                 data = attr;
2340                         }
2341                         if (ops->validate) {
2342                                 err = ops->validate(tb, data);
2343                                 if (err < 0)
2344                                         return err;
2345                         }
2346                 }
2347
2348                 if (m_ops) {
2349                         if (m_ops->slave_maxtype &&
2350                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2351                                 err = nla_parse_nested(slave_attr,
2352                                                        m_ops->slave_maxtype,
2353                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2354                                                        m_ops->slave_policy);
2355                                 if (err < 0)
2356                                         return err;
2357                                 slave_data = slave_attr;
2358                         }
2359                         if (m_ops->slave_validate) {
2360                                 err = m_ops->slave_validate(tb, slave_data);
2361                                 if (err < 0)
2362                                         return err;
2363                         }
2364                 }
2365
2366                 if (dev) {
2367                         int status = 0;
2368
2369                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2370                                 return -EEXIST;
2371                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2372                                 return -EOPNOTSUPP;
2373
2374                         if (linkinfo[IFLA_INFO_DATA]) {
2375                                 if (!ops || ops != dev->rtnl_link_ops ||
2376                                     !ops->changelink)
2377                                         return -EOPNOTSUPP;
2378
2379                                 err = ops->changelink(dev, tb, data);
2380                                 if (err < 0)
2381                                         return err;
2382                                 status |= DO_SETLINK_NOTIFY;
2383                         }
2384
2385                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2386                                 if (!m_ops || !m_ops->slave_changelink)
2387                                         return -EOPNOTSUPP;
2388
2389                                 err = m_ops->slave_changelink(master_dev, dev,
2390                                                               tb, slave_data);
2391                                 if (err < 0)
2392                                         return err;
2393                                 status |= DO_SETLINK_NOTIFY;
2394                         }
2395
2396                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2397                 }
2398
2399                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2400                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2401                                 return rtnl_group_changelink(skb, net,
2402                                                 nla_get_u32(tb[IFLA_GROUP]),
2403                                                 ifm, tb);
2404                         return -ENODEV;
2405                 }
2406
2407                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2408                         return -EOPNOTSUPP;
2409
2410                 if (!ops) {
2411 #ifdef CONFIG_MODULES
2412                         if (kind[0]) {
2413                                 __rtnl_unlock();
2414                                 request_module("rtnl-link-%s", kind);
2415                                 rtnl_lock();
2416                                 ops = rtnl_link_ops_get(kind);
2417                                 if (ops)
2418                                         goto replay;
2419                         }
2420 #endif
2421                         return -EOPNOTSUPP;
2422                 }
2423
2424                 if (!ops->setup)
2425                         return -EOPNOTSUPP;
2426
2427                 if (!ifname[0]) {
2428                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2429                         name_assign_type = NET_NAME_ENUM;
2430                 }
2431
2432                 dest_net = rtnl_link_get_net(net, tb);
2433                 if (IS_ERR(dest_net))
2434                         return PTR_ERR(dest_net);
2435
2436                 err = -EPERM;
2437                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2438                         goto out;
2439
2440                 if (tb[IFLA_LINK_NETNSID]) {
2441                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2442
2443                         link_net = get_net_ns_by_id(dest_net, id);
2444                         if (!link_net) {
2445                                 err =  -EINVAL;
2446                                 goto out;
2447                         }
2448                         err = -EPERM;
2449                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2450                                 goto out;
2451                 }
2452
2453                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2454                                        name_assign_type, ops, tb);
2455                 if (IS_ERR(dev)) {
2456                         err = PTR_ERR(dev);
2457                         goto out;
2458                 }
2459
2460                 dev->ifindex = ifm->ifi_index;
2461
2462                 if (ops->newlink) {
2463                         err = ops->newlink(link_net ? : net, dev, tb, data);
2464                         /* Drivers should call free_netdev() in ->destructor
2465                          * and unregister it on failure after registration
2466                          * so that device could be finally freed in rtnl_unlock.
2467                          */
2468                         if (err < 0) {
2469                                 /* If device is not registered at all, free it now */
2470                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2471                                         free_netdev(dev);
2472                                 goto out;
2473                         }
2474                 } else {
2475                         err = register_netdevice(dev);
2476                         if (err < 0) {
2477                                 free_netdev(dev);
2478                                 goto out;
2479                         }
2480                 }
2481                 err = rtnl_configure_link(dev, ifm);
2482                 if (err < 0)
2483                         goto out_unregister;
2484                 if (link_net) {
2485                         err = dev_change_net_namespace(dev, dest_net, ifname);
2486                         if (err < 0)
2487                                 goto out_unregister;
2488                 }
2489 out:
2490                 if (link_net)
2491                         put_net(link_net);
2492                 put_net(dest_net);
2493                 return err;
2494 out_unregister:
2495                 if (ops->newlink) {
2496                         LIST_HEAD(list_kill);
2497
2498                         ops->dellink(dev, &list_kill);
2499                         unregister_netdevice_many(&list_kill);
2500                 } else {
2501                         unregister_netdevice(dev);
2502                 }
2503                 goto out;
2504         }
2505 }
2506
2507 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2508 {
2509         struct net *net = sock_net(skb->sk);
2510         struct ifinfomsg *ifm;
2511         char ifname[IFNAMSIZ];
2512         struct nlattr *tb[IFLA_MAX+1];
2513         struct net_device *dev = NULL;
2514         struct sk_buff *nskb;
2515         int err;
2516         u32 ext_filter_mask = 0;
2517
2518         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2519         if (err < 0)
2520                 return err;
2521
2522         if (tb[IFLA_IFNAME])
2523                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2524
2525         if (tb[IFLA_EXT_MASK])
2526                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2527
2528         ifm = nlmsg_data(nlh);
2529         if (ifm->ifi_index > 0)
2530                 dev = __dev_get_by_index(net, ifm->ifi_index);
2531         else if (tb[IFLA_IFNAME])
2532                 dev = __dev_get_by_name(net, ifname);
2533         else
2534                 return -EINVAL;
2535
2536         if (dev == NULL)
2537                 return -ENODEV;
2538
2539         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2540         if (nskb == NULL)
2541                 return -ENOBUFS;
2542
2543         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2544                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2545         if (err < 0) {
2546                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2547                 WARN_ON(err == -EMSGSIZE);
2548                 kfree_skb(nskb);
2549         } else
2550                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2551
2552         return err;
2553 }
2554
2555 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2556 {
2557         struct net *net = sock_net(skb->sk);
2558         struct net_device *dev;
2559         struct nlattr *tb[IFLA_MAX+1];
2560         u32 ext_filter_mask = 0;
2561         u16 min_ifinfo_dump_size = 0;
2562         int hdrlen;
2563
2564         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2565         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2566                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2567
2568         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2569                 if (tb[IFLA_EXT_MASK])
2570                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2571         }
2572
2573         if (!ext_filter_mask)
2574                 return NLMSG_GOODSIZE;
2575         /*
2576          * traverse the list of net devices and compute the minimum
2577          * buffer size based upon the filter mask.
2578          */
2579         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2580                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2581                                              if_nlmsg_size(dev,
2582                                                            ext_filter_mask));
2583         }
2584
2585         return min_ifinfo_dump_size;
2586 }
2587
2588 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2589 {
2590         int idx;
2591         int s_idx = cb->family;
2592
2593         if (s_idx == 0)
2594                 s_idx = 1;
2595         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2596                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2597                 if (idx < s_idx || idx == PF_PACKET)
2598                         continue;
2599                 if (rtnl_msg_handlers[idx] == NULL ||
2600                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2601                         continue;
2602                 if (idx > s_idx) {
2603                         memset(&cb->args[0], 0, sizeof(cb->args));
2604                         cb->prev_seq = 0;
2605                         cb->seq = 0;
2606                 }
2607                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2608                         break;
2609         }
2610         cb->family = idx;
2611
2612         return skb->len;
2613 }
2614
2615 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2616                                        unsigned int change, gfp_t flags)
2617 {
2618         struct net *net = dev_net(dev);
2619         struct sk_buff *skb;
2620         int err = -ENOBUFS;
2621         size_t if_info_size;
2622
2623         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2624         if (skb == NULL)
2625                 goto errout;
2626
2627         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2628         if (err < 0) {
2629                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2630                 WARN_ON(err == -EMSGSIZE);
2631                 kfree_skb(skb);
2632                 goto errout;
2633         }
2634         return skb;
2635 errout:
2636         if (err < 0)
2637                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2638         return NULL;
2639 }
2640
2641 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2642 {
2643         struct net *net = dev_net(dev);
2644
2645         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2646 }
2647
2648 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2649                   gfp_t flags)
2650 {
2651         struct sk_buff *skb;
2652
2653         if (dev->reg_state != NETREG_REGISTERED)
2654                 return;
2655
2656         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2657         if (skb)
2658                 rtmsg_ifinfo_send(skb, dev, flags);
2659 }
2660 EXPORT_SYMBOL(rtmsg_ifinfo);
2661
2662 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2663                                    struct net_device *dev,
2664                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2665                                    int type, unsigned int flags,
2666                                    int nlflags, u16 ndm_state)
2667 {
2668         struct nlmsghdr *nlh;
2669         struct ndmsg *ndm;
2670
2671         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2672         if (!nlh)
2673                 return -EMSGSIZE;
2674
2675         ndm = nlmsg_data(nlh);
2676         ndm->ndm_family  = AF_BRIDGE;
2677         ndm->ndm_pad1    = 0;
2678         ndm->ndm_pad2    = 0;
2679         ndm->ndm_flags   = flags;
2680         ndm->ndm_type    = 0;
2681         ndm->ndm_ifindex = dev->ifindex;
2682         ndm->ndm_state   = ndm_state;
2683
2684         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2685                 goto nla_put_failure;
2686         if (vid)
2687                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2688                         goto nla_put_failure;
2689
2690         nlmsg_end(skb, nlh);
2691         return 0;
2692
2693 nla_put_failure:
2694         nlmsg_cancel(skb, nlh);
2695         return -EMSGSIZE;
2696 }
2697
2698 static inline size_t rtnl_fdb_nlmsg_size(void)
2699 {
2700         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2701 }
2702
2703 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2704                             u16 ndm_state)
2705 {
2706         struct net *net = dev_net(dev);
2707         struct sk_buff *skb;
2708         int err = -ENOBUFS;
2709
2710         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2711         if (!skb)
2712                 goto errout;
2713
2714         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2715                                       0, 0, type, NTF_SELF, 0, ndm_state);
2716         if (err < 0) {
2717                 kfree_skb(skb);
2718                 goto errout;
2719         }
2720
2721         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2722         return;
2723 errout:
2724         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2725 }
2726
2727 /**
2728  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2729  */
2730 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2731                      struct nlattr *tb[],
2732                      struct net_device *dev,
2733                      const unsigned char *addr, u16 vid,
2734                      u16 flags)
2735 {
2736         int err = -EINVAL;
2737
2738         /* If aging addresses are supported device will need to
2739          * implement its own handler for this.
2740          */
2741         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2742                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2743                 return err;
2744         }
2745
2746         if (vid) {
2747                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2748                 return err;
2749         }
2750
2751         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2752                 err = dev_uc_add_excl(dev, addr);
2753         else if (is_multicast_ether_addr(addr))
2754                 err = dev_mc_add_excl(dev, addr);
2755
2756         /* Only return duplicate errors if NLM_F_EXCL is set */
2757         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2758                 err = 0;
2759
2760         return err;
2761 }
2762 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2763
2764 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2765 {
2766         u16 vid = 0;
2767
2768         if (vlan_attr) {
2769                 if (nla_len(vlan_attr) != sizeof(u16)) {
2770                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2771                         return -EINVAL;
2772                 }
2773
2774                 vid = nla_get_u16(vlan_attr);
2775
2776                 if (!vid || vid >= VLAN_VID_MASK) {
2777                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2778                                 vid);
2779                         return -EINVAL;
2780                 }
2781         }
2782         *p_vid = vid;
2783         return 0;
2784 }
2785
2786 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2787 {
2788         struct net *net = sock_net(skb->sk);
2789         struct ndmsg *ndm;
2790         struct nlattr *tb[NDA_MAX+1];
2791         struct net_device *dev;
2792         u8 *addr;
2793         u16 vid;
2794         int err;
2795
2796         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2797         if (err < 0)
2798                 return err;
2799
2800         ndm = nlmsg_data(nlh);
2801         if (ndm->ndm_ifindex == 0) {
2802                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2803                 return -EINVAL;
2804         }
2805
2806         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2807         if (dev == NULL) {
2808                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2809                 return -ENODEV;
2810         }
2811
2812         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2813                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2814                 return -EINVAL;
2815         }
2816
2817         addr = nla_data(tb[NDA_LLADDR]);
2818
2819         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2820         if (err)
2821                 return err;
2822
2823         err = -EOPNOTSUPP;
2824
2825         /* Support fdb on master device the net/bridge default case */
2826         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2827             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2828                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2829                 const struct net_device_ops *ops = br_dev->netdev_ops;
2830
2831                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2832                                        nlh->nlmsg_flags);
2833                 if (err)
2834                         goto out;
2835                 else
2836                         ndm->ndm_flags &= ~NTF_MASTER;
2837         }
2838
2839         /* Embedded bridge, macvlan, and any other device support */
2840         if ((ndm->ndm_flags & NTF_SELF)) {
2841                 if (dev->netdev_ops->ndo_fdb_add)
2842                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2843                                                            vid,
2844                                                            nlh->nlmsg_flags);
2845                 else
2846                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2847                                                nlh->nlmsg_flags);
2848
2849                 if (!err) {
2850                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2851                                         ndm->ndm_state);
2852                         ndm->ndm_flags &= ~NTF_SELF;
2853                 }
2854         }
2855 out:
2856         return err;
2857 }
2858
2859 /**
2860  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2861  */
2862 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2863                      struct nlattr *tb[],
2864                      struct net_device *dev,
2865                      const unsigned char *addr, u16 vid)
2866 {
2867         int err = -EINVAL;
2868
2869         /* If aging addresses are supported device will need to
2870          * implement its own handler for this.
2871          */
2872         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2873                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2874                 return err;
2875         }
2876
2877         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2878                 err = dev_uc_del(dev, addr);
2879         else if (is_multicast_ether_addr(addr))
2880                 err = dev_mc_del(dev, addr);
2881
2882         return err;
2883 }
2884 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2885
2886 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2887 {
2888         struct net *net = sock_net(skb->sk);
2889         struct ndmsg *ndm;
2890         struct nlattr *tb[NDA_MAX+1];
2891         struct net_device *dev;
2892         int err = -EINVAL;
2893         __u8 *addr;
2894         u16 vid;
2895
2896         if (!netlink_capable(skb, CAP_NET_ADMIN))
2897                 return -EPERM;
2898
2899         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2900         if (err < 0)
2901                 return err;
2902
2903         ndm = nlmsg_data(nlh);
2904         if (ndm->ndm_ifindex == 0) {
2905                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2906                 return -EINVAL;
2907         }
2908
2909         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2910         if (dev == NULL) {
2911                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2912                 return -ENODEV;
2913         }
2914
2915         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2916                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2917                 return -EINVAL;
2918         }
2919
2920         addr = nla_data(tb[NDA_LLADDR]);
2921
2922         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2923         if (err)
2924                 return err;
2925
2926         err = -EOPNOTSUPP;
2927
2928         /* Support fdb on master device the net/bridge default case */
2929         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2930             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2931                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2932                 const struct net_device_ops *ops = br_dev->netdev_ops;
2933
2934                 if (ops->ndo_fdb_del)
2935                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2936
2937                 if (err)
2938                         goto out;
2939                 else
2940                         ndm->ndm_flags &= ~NTF_MASTER;
2941         }
2942
2943         /* Embedded bridge, macvlan, and any other device support */
2944         if (ndm->ndm_flags & NTF_SELF) {
2945                 if (dev->netdev_ops->ndo_fdb_del)
2946                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2947                                                            vid);
2948                 else
2949                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2950
2951                 if (!err) {
2952                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2953                                         ndm->ndm_state);
2954                         ndm->ndm_flags &= ~NTF_SELF;
2955                 }
2956         }
2957 out:
2958         return err;
2959 }
2960
2961 static int nlmsg_populate_fdb(struct sk_buff *skb,
2962                               struct netlink_callback *cb,
2963                               struct net_device *dev,
2964                               int *idx,
2965                               struct netdev_hw_addr_list *list)
2966 {
2967         struct netdev_hw_addr *ha;
2968         int err;
2969         u32 portid, seq;
2970
2971         portid = NETLINK_CB(cb->skb).portid;
2972         seq = cb->nlh->nlmsg_seq;
2973
2974         list_for_each_entry(ha, &list->list, list) {
2975                 if (*idx < cb->args[0])
2976                         goto skip;
2977
2978                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2979                                               portid, seq,
2980                                               RTM_NEWNEIGH, NTF_SELF,
2981                                               NLM_F_MULTI, NUD_PERMANENT);
2982                 if (err < 0)
2983                         return err;
2984 skip:
2985                 *idx += 1;
2986         }
2987         return 0;
2988 }
2989
2990 /**
2991  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2992  * @nlh: netlink message header
2993  * @dev: netdevice
2994  *
2995  * Default netdevice operation to dump the existing unicast address list.
2996  * Returns number of addresses from list put in skb.
2997  */
2998 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2999                       struct netlink_callback *cb,
3000                       struct net_device *dev,
3001                       struct net_device *filter_dev,
3002                       int idx)
3003 {
3004         int err;
3005
3006         netif_addr_lock_bh(dev);
3007         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
3008         if (err)
3009                 goto out;
3010         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
3011 out:
3012         netif_addr_unlock_bh(dev);
3013         cb->args[1] = err;
3014         return idx;
3015 }
3016 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3017
3018 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3019 {
3020         struct net_device *dev;
3021         struct nlattr *tb[IFLA_MAX+1];
3022         struct net_device *br_dev = NULL;
3023         const struct net_device_ops *ops = NULL;
3024         const struct net_device_ops *cops = NULL;
3025         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3026         struct net *net = sock_net(skb->sk);
3027         int brport_idx = 0;
3028         int br_idx = 0;
3029         int idx = 0;
3030
3031         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3032                         ifla_policy) == 0) {
3033                 if (tb[IFLA_MASTER])
3034                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3035         }
3036
3037         brport_idx = ifm->ifi_index;
3038
3039         if (br_idx) {
3040                 br_dev = __dev_get_by_index(net, br_idx);
3041                 if (!br_dev)
3042                         return -ENODEV;
3043
3044                 ops = br_dev->netdev_ops;
3045         }
3046
3047         cb->args[1] = 0;
3048         for_each_netdev(net, dev) {
3049                 if (brport_idx && (dev->ifindex != brport_idx))
3050                         continue;
3051
3052                 if (!br_idx) { /* user did not specify a specific bridge */
3053                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3054                                 br_dev = netdev_master_upper_dev_get(dev);
3055                                 cops = br_dev->netdev_ops;
3056                         }
3057
3058                 } else {
3059                         if (dev != br_dev &&
3060                             !(dev->priv_flags & IFF_BRIDGE_PORT))
3061                                 continue;
3062
3063                         if (br_dev != netdev_master_upper_dev_get(dev) &&
3064                             !(dev->priv_flags & IFF_EBRIDGE))
3065                                 continue;
3066
3067                         cops = ops;
3068                 }
3069
3070                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3071                         if (cops && cops->ndo_fdb_dump)
3072                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3073                                                          idx);
3074                 }
3075                 if (cb->args[1] == -EMSGSIZE)
3076                         break;
3077
3078                 if (dev->netdev_ops->ndo_fdb_dump)
3079                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3080                                                             idx);
3081                 else
3082                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3083                 if (cb->args[1] == -EMSGSIZE)
3084                         break;
3085
3086                 cops = NULL;
3087         }
3088
3089         cb->args[0] = idx;
3090         return skb->len;
3091 }
3092
3093 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3094                                unsigned int attrnum, unsigned int flag)
3095 {
3096         if (mask & flag)
3097                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3098         return 0;
3099 }
3100
3101 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3102                             struct net_device *dev, u16 mode,
3103                             u32 flags, u32 mask, int nlflags,
3104                             u32 filter_mask,
3105                             int (*vlan_fill)(struct sk_buff *skb,
3106                                              struct net_device *dev,
3107                                              u32 filter_mask))
3108 {
3109         struct nlmsghdr *nlh;
3110         struct ifinfomsg *ifm;
3111         struct nlattr *br_afspec;
3112         struct nlattr *protinfo;
3113         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3114         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3115         int err = 0;
3116
3117         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3118         if (nlh == NULL)
3119                 return -EMSGSIZE;
3120
3121         ifm = nlmsg_data(nlh);
3122         ifm->ifi_family = AF_BRIDGE;
3123         ifm->__ifi_pad = 0;
3124         ifm->ifi_type = dev->type;
3125         ifm->ifi_index = dev->ifindex;
3126         ifm->ifi_flags = dev_get_flags(dev);
3127         ifm->ifi_change = 0;
3128
3129
3130         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3131             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3132             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3133             (br_dev &&
3134              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3135             (dev->addr_len &&
3136              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3137             (dev->ifindex != dev_get_iflink(dev) &&
3138              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3139                 goto nla_put_failure;
3140
3141         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3142         if (!br_afspec)
3143                 goto nla_put_failure;
3144
3145         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3146                 nla_nest_cancel(skb, br_afspec);
3147                 goto nla_put_failure;
3148         }
3149
3150         if (mode != BRIDGE_MODE_UNDEF) {
3151                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3152                         nla_nest_cancel(skb, br_afspec);
3153                         goto nla_put_failure;
3154                 }
3155         }
3156         if (vlan_fill) {
3157                 err = vlan_fill(skb, dev, filter_mask);
3158                 if (err) {
3159                         nla_nest_cancel(skb, br_afspec);
3160                         goto nla_put_failure;
3161                 }
3162         }
3163         nla_nest_end(skb, br_afspec);
3164
3165         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3166         if (!protinfo)
3167                 goto nla_put_failure;
3168
3169         if (brport_nla_put_flag(skb, flags, mask,
3170                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3171             brport_nla_put_flag(skb, flags, mask,
3172                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3173             brport_nla_put_flag(skb, flags, mask,
3174                                 IFLA_BRPORT_FAST_LEAVE,
3175                                 BR_MULTICAST_FAST_LEAVE) ||
3176             brport_nla_put_flag(skb, flags, mask,
3177                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3178             brport_nla_put_flag(skb, flags, mask,
3179                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3180             brport_nla_put_flag(skb, flags, mask,
3181                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3182             brport_nla_put_flag(skb, flags, mask,
3183                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3184             brport_nla_put_flag(skb, flags, mask,
3185                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3186                 nla_nest_cancel(skb, protinfo);
3187                 goto nla_put_failure;
3188         }
3189
3190         nla_nest_end(skb, protinfo);
3191
3192         nlmsg_end(skb, nlh);
3193         return 0;
3194 nla_put_failure:
3195         nlmsg_cancel(skb, nlh);
3196         return err ? err : -EMSGSIZE;
3197 }
3198 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3199
3200 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3201 {
3202         struct net *net = sock_net(skb->sk);
3203         struct net_device *dev;
3204         int idx = 0;
3205         u32 portid = NETLINK_CB(cb->skb).portid;
3206         u32 seq = cb->nlh->nlmsg_seq;
3207         u32 filter_mask = 0;
3208         int err;
3209
3210         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3211                 struct nlattr *extfilt;
3212
3213                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3214                                           IFLA_EXT_MASK);
3215                 if (extfilt) {
3216                         if (nla_len(extfilt) < sizeof(filter_mask))
3217                                 return -EINVAL;
3218
3219                         filter_mask = nla_get_u32(extfilt);
3220                 }
3221         }
3222
3223         rcu_read_lock();
3224         for_each_netdev_rcu(net, dev) {
3225                 const struct net_device_ops *ops = dev->netdev_ops;
3226                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3227
3228                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3229                         if (idx >= cb->args[0]) {
3230                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3231                                                 skb, portid, seq, dev,
3232                                                 filter_mask, NLM_F_MULTI);
3233                                 if (err < 0 && err != -EOPNOTSUPP)
3234                                         break;
3235                         }
3236                         idx++;
3237                 }
3238
3239                 if (ops->ndo_bridge_getlink) {
3240                         if (idx >= cb->args[0]) {
3241                                 err = ops->ndo_bridge_getlink(skb, portid,
3242                                                               seq, dev,
3243                                                               filter_mask,
3244                                                               NLM_F_MULTI);
3245                                 if (err < 0 && err != -EOPNOTSUPP)
3246                                         break;
3247                         }
3248                         idx++;
3249                 }
3250         }
3251         rcu_read_unlock();
3252         cb->args[0] = idx;
3253
3254         return skb->len;
3255 }
3256
3257 static inline size_t bridge_nlmsg_size(void)
3258 {
3259         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3260                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3261                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3262                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3263                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3264                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3265                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3266                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3267                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3268                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3269                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3270 }
3271
3272 static int rtnl_bridge_notify(struct net_device *dev)
3273 {
3274         struct net *net = dev_net(dev);
3275         struct sk_buff *skb;
3276         int err = -EOPNOTSUPP;
3277
3278         if (!dev->netdev_ops->ndo_bridge_getlink)
3279                 return 0;
3280
3281         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3282         if (!skb) {
3283                 err = -ENOMEM;
3284                 goto errout;
3285         }
3286
3287         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3288         if (err < 0)
3289                 goto errout;
3290
3291         if (!skb->len)
3292                 goto errout;
3293
3294         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3295         return 0;
3296 errout:
3297         WARN_ON(err == -EMSGSIZE);
3298         kfree_skb(skb);
3299         if (err)
3300                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3301         return err;
3302 }
3303
3304 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3305 {
3306         struct net *net = sock_net(skb->sk);
3307         struct ifinfomsg *ifm;
3308         struct net_device *dev;
3309         struct nlattr *br_spec, *attr = NULL;
3310         int rem, err = -EOPNOTSUPP;
3311         u16 flags = 0;
3312         bool have_flags = false;
3313
3314         if (nlmsg_len(nlh) < sizeof(*ifm))
3315                 return -EINVAL;
3316
3317         ifm = nlmsg_data(nlh);
3318         if (ifm->ifi_family != AF_BRIDGE)
3319                 return -EPFNOSUPPORT;
3320
3321         dev = __dev_get_by_index(net, ifm->ifi_index);
3322         if (!dev) {
3323                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3324                 return -ENODEV;
3325         }
3326
3327         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3328         if (br_spec) {
3329                 nla_for_each_nested(attr, br_spec, rem) {
3330                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3331                                 if (nla_len(attr) < sizeof(flags))
3332                                         return -EINVAL;
3333
3334                                 have_flags = true;
3335                                 flags = nla_get_u16(attr);
3336                                 break;
3337                         }
3338                 }
3339         }
3340
3341         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3342                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3343
3344                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3345                         err = -EOPNOTSUPP;
3346                         goto out;
3347                 }
3348
3349                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3350                 if (err)
3351                         goto out;
3352
3353                 flags &= ~BRIDGE_FLAGS_MASTER;
3354         }
3355
3356         if ((flags & BRIDGE_FLAGS_SELF)) {
3357                 if (!dev->netdev_ops->ndo_bridge_setlink)
3358                         err = -EOPNOTSUPP;
3359                 else
3360                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3361                                                                   flags);
3362                 if (!err) {
3363                         flags &= ~BRIDGE_FLAGS_SELF;
3364
3365                         /* Generate event to notify upper layer of bridge
3366                          * change
3367                          */
3368                         err = rtnl_bridge_notify(dev);
3369                 }
3370         }
3371
3372         if (have_flags)
3373                 memcpy(nla_data(attr), &flags, sizeof(flags));
3374 out:
3375         return err;
3376 }
3377
3378 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3379 {
3380         struct net *net = sock_net(skb->sk);
3381         struct ifinfomsg *ifm;
3382         struct net_device *dev;
3383         struct nlattr *br_spec, *attr = NULL;
3384         int rem, err = -EOPNOTSUPP;
3385         u16 flags = 0;
3386         bool have_flags = false;
3387
3388         if (nlmsg_len(nlh) < sizeof(*ifm))
3389                 return -EINVAL;
3390
3391         ifm = nlmsg_data(nlh);
3392         if (ifm->ifi_family != AF_BRIDGE)
3393                 return -EPFNOSUPPORT;
3394
3395         dev = __dev_get_by_index(net, ifm->ifi_index);
3396         if (!dev) {
3397                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3398                 return -ENODEV;
3399         }
3400
3401         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3402         if (br_spec) {
3403                 nla_for_each_nested(attr, br_spec, rem) {
3404                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3405                                 if (nla_len(attr) < sizeof(flags))
3406                                         return -EINVAL;
3407
3408                                 have_flags = true;
3409                                 flags = nla_get_u16(attr);
3410                                 break;
3411                         }
3412                 }
3413         }
3414
3415         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3416                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3417
3418                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3419                         err = -EOPNOTSUPP;
3420                         goto out;
3421                 }
3422
3423                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3424                 if (err)
3425                         goto out;
3426
3427                 flags &= ~BRIDGE_FLAGS_MASTER;
3428         }
3429
3430         if ((flags & BRIDGE_FLAGS_SELF)) {
3431                 if (!dev->netdev_ops->ndo_bridge_dellink)
3432                         err = -EOPNOTSUPP;
3433                 else
3434                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3435                                                                   flags);
3436
3437                 if (!err) {
3438                         flags &= ~BRIDGE_FLAGS_SELF;
3439
3440                         /* Generate event to notify upper layer of bridge
3441                          * change
3442                          */
3443                         err = rtnl_bridge_notify(dev);
3444                 }
3445         }
3446
3447         if (have_flags)
3448                 memcpy(nla_data(attr), &flags, sizeof(flags));
3449 out:
3450         return err;
3451 }
3452
3453 /* Process one rtnetlink message. */
3454
3455 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3456 {
3457         struct net *net = sock_net(skb->sk);
3458         rtnl_doit_func doit;
3459         int kind;
3460         int family;
3461         int type;
3462         int err;
3463
3464         type = nlh->nlmsg_type;
3465         if (type > RTM_MAX)
3466                 return -EOPNOTSUPP;
3467
3468         type -= RTM_BASE;
3469
3470         /* All the messages must have at least 1 byte length */
3471         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3472                 return 0;
3473
3474         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3475         kind = type&3;
3476
3477         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3478                 return -EPERM;
3479
3480         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3481                 struct sock *rtnl;
3482                 rtnl_dumpit_func dumpit;
3483                 rtnl_calcit_func calcit;
3484                 u16 min_dump_alloc = 0;
3485
3486                 dumpit = rtnl_get_dumpit(family, type);
3487                 if (dumpit == NULL)
3488                         return -EOPNOTSUPP;
3489                 calcit = rtnl_get_calcit(family, type);
3490                 if (calcit)
3491                         min_dump_alloc = calcit(skb, nlh);
3492
3493                 __rtnl_unlock();
3494                 rtnl = net->rtnl;
3495                 {
3496                         struct netlink_dump_control c = {
3497                                 .dump           = dumpit,
3498                                 .min_dump_alloc = min_dump_alloc,
3499                         };
3500                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3501                 }
3502                 rtnl_lock();
3503                 return err;
3504         }
3505
3506         doit = rtnl_get_doit(family, type);
3507         if (doit == NULL)
3508                 return -EOPNOTSUPP;
3509
3510         return doit(skb, nlh);
3511 }
3512
3513 static void rtnetlink_rcv(struct sk_buff *skb)
3514 {
3515         rtnl_lock();
3516         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3517         rtnl_unlock();
3518 }
3519
3520 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3521 {
3522         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3523
3524         switch (event) {
3525         case NETDEV_UP:
3526         case NETDEV_DOWN:
3527         case NETDEV_PRE_UP:
3528         case NETDEV_POST_INIT:
3529         case NETDEV_REGISTER:
3530         case NETDEV_CHANGE:
3531         case NETDEV_PRE_TYPE_CHANGE:
3532         case NETDEV_GOING_DOWN:
3533         case NETDEV_UNREGISTER:
3534         case NETDEV_UNREGISTER_FINAL:
3535         case NETDEV_RELEASE:
3536         case NETDEV_JOIN:
3537         case NETDEV_BONDING_INFO:
3538                 break;
3539         default:
3540                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3541                 break;
3542         }
3543         return NOTIFY_DONE;
3544 }
3545
3546 static struct notifier_block rtnetlink_dev_notifier = {
3547         .notifier_call  = rtnetlink_event,
3548 };
3549
3550
3551 static int __net_init rtnetlink_net_init(struct net *net)
3552 {
3553         struct sock *sk;
3554         struct netlink_kernel_cfg cfg = {
3555                 .groups         = RTNLGRP_MAX,
3556                 .input          = rtnetlink_rcv,
3557                 .cb_mutex       = &rtnl_mutex,
3558                 .flags          = NL_CFG_F_NONROOT_RECV,
3559         };
3560
3561         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3562         if (!sk)
3563                 return -ENOMEM;
3564         net->rtnl = sk;
3565         return 0;
3566 }
3567
3568 static void __net_exit rtnetlink_net_exit(struct net *net)
3569 {
3570         netlink_kernel_release(net->rtnl);
3571         net->rtnl = NULL;
3572 }
3573
3574 static struct pernet_operations rtnetlink_net_ops = {
3575         .init = rtnetlink_net_init,
3576         .exit = rtnetlink_net_exit,
3577 };
3578
3579 void __init rtnetlink_init(void)
3580 {
3581         if (register_pernet_subsys(&rtnetlink_net_ops))
3582                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3583
3584         register_netdevice_notifier(&rtnetlink_dev_notifier);
3585
3586         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3587                       rtnl_dump_ifinfo, rtnl_calcit);
3588         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3589         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3590         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3591
3592         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3593         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3594
3595         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3596         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3597         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3598
3599         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3600         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3601         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3602 }