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(1) /* IFLA_OPERSTATE */
 899                + nla_total_size(1) /* IFLA_LINKMODE */
 900                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 901                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 902                + nla_total_size(ext_filter_mask
 903                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 904                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 905                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 906                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 907                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 908                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 909                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 910                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 911
 912 }
 913
 914 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 915 {
 916         struct nlattr *vf_ports;
 917         struct nlattr *vf_port;
 918         int vf;
 919         int err;
 920
 921         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 922         if (!vf_ports)
 923                 return -EMSGSIZE;
 924
 925         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 926                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 927                 if (!vf_port)
 928                         goto nla_put_failure;
 929                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 930                         goto nla_put_failure;
 931                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 932                 if (err == -EMSGSIZE)
 933                         goto nla_put_failure;
 934                 if (err) {
 935                         nla_nest_cancel(skb, vf_port);
 936                         continue;
 937                 }
 938                 nla_nest_end(skb, vf_port);
 939         }
 940
 941         nla_nest_end(skb, vf_ports);
 942
 943         return 0;
 944
 945 nla_put_failure:
 946         nla_nest_cancel(skb, vf_ports);
 947         return -EMSGSIZE;
 948 }
 949
 950 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 951 {
 952         struct nlattr *port_self;
 953         int err;
 954
 955         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 956         if (!port_self)
 957                 return -EMSGSIZE;
 958
 959         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 960         if (err) {
 961                 nla_nest_cancel(skb, port_self);
 962                 return (err == -EMSGSIZE) ? err : 0;
 963         }
 964
 965         nla_nest_end(skb, port_self);
 966
 967         return 0;
 968 }
 969
 970 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 971                           u32 ext_filter_mask)
 972 {
 973         int err;
 974
 975         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 976             !(ext_filter_mask & RTEXT_FILTER_VF))
 977                 return 0;
 978
 979         err = rtnl_port_self_fill(skb, dev);
 980         if (err)
 981                 return err;
 982
 983         if (dev_num_vf(dev->dev.parent)) {
 984                 err = rtnl_vf_ports_fill(skb, dev);
 985                 if (err)
 986                         return err;
 987         }
 988
 989         return 0;
 990 }
 991
 992 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 993 {
 994         int err;
 995         struct netdev_phys_item_id ppid;
 996
 997         err = dev_get_phys_port_id(dev, &ppid);
 998         if (err) {
 999                 if (err == -EOPNOTSUPP)
1000                         return 0;
1001                 return err;
1002         }
1003
1004         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1005                 return -EMSGSIZE;
1006
1007         return 0;
1008 }
1009
1010 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1011 {
1012         char name[IFNAMSIZ];
1013         int err;
1014
1015         err = dev_get_phys_port_name(dev, name, sizeof(name));
1016         if (err) {
1017                 if (err == -EOPNOTSUPP)
1018                         return 0;
1019                 return err;
1020         }
1021
1022         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1023                 return -EMSGSIZE;
1024
1025         return 0;
1026 }
1027
1028 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1029 {
1030         int err;
1031         struct switchdev_attr attr = {
1032                 .orig_dev = dev,
1033                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1034                 .flags = SWITCHDEV_F_NO_RECURSE,
1035         };
1036
1037         err = switchdev_port_attr_get(dev, &attr);
1038         if (err) {
1039                 if (err == -EOPNOTSUPP)
1040                         return 0;
1041                 return err;
1042         }
1043
1044         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1045                     attr.u.ppid.id))
1046                 return -EMSGSIZE;
1047
1048         return 0;
1049 }
1050
1051 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1052                                               struct net_device *dev)
1053 {
1054         const struct rtnl_link_stats64 *stats;
1055         struct rtnl_link_stats64 temp;
1056         struct nlattr *attr;
1057
1058         stats = dev_get_stats(dev, &temp);
1059
1060         attr = nla_reserve(skb, IFLA_STATS,
1061                            sizeof(struct rtnl_link_stats));
1062         if (!attr)
1063                 return -EMSGSIZE;
1064
1065         copy_rtnl_link_stats(nla_data(attr), stats);
1066
1067         attr = nla_reserve(skb, IFLA_STATS64,
1068                            sizeof(struct rtnl_link_stats64));
1069         if (!attr)
1070                 return -EMSGSIZE;
1071
1072         copy_rtnl_link_stats64(nla_data(attr), stats);
1073
1074         return 0;
1075 }
1076
1077 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1078                                                struct net_device *dev,
1079                                                int vfs_num,
1080                                                struct nlattr *vfinfo)
1081 {
1082         struct ifla_vf_rss_query_en vf_rss_query_en;
1083         struct ifla_vf_link_state vf_linkstate;
1084         struct ifla_vf_spoofchk vf_spoofchk;
1085         struct ifla_vf_tx_rate vf_tx_rate;
1086         struct ifla_vf_stats vf_stats;
1087         struct ifla_vf_trust vf_trust;
1088         struct ifla_vf_vlan vf_vlan;
1089         struct ifla_vf_rate vf_rate;
1090         struct nlattr *vf, *vfstats;
1091         struct ifla_vf_mac vf_mac;
1092         struct ifla_vf_info ivi;
1093
1094         /* Not all SR-IOV capable drivers support the
1095          * spoofcheck and "RSS query enable" query.  Preset to
1096          * -1 so the user space tool can detect that the driver
1097          * didn't report anything.
1098          */
1099         ivi.spoofchk = -1;
1100         ivi.rss_query_en = -1;
1101         ivi.trusted = -1;
1102         memset(ivi.mac, 0, sizeof(ivi.mac));
1103         /* The default value for VF link state is "auto"
1104          * IFLA_VF_LINK_STATE_AUTO which equals zero
1105          */
1106         ivi.linkstate = 0;
1107         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1108                 return 0;
1109
1110         vf_mac.vf =
1111                 vf_vlan.vf =
1112                 vf_rate.vf =
1113                 vf_tx_rate.vf =
1114                 vf_spoofchk.vf =
1115                 vf_linkstate.vf =
1116                 vf_rss_query_en.vf =
1117                 vf_trust.vf = ivi.vf;
1118
1119         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1120         vf_vlan.vlan = ivi.vlan;
1121         vf_vlan.qos = ivi.qos;
1122         vf_tx_rate.rate = ivi.max_tx_rate;
1123         vf_rate.min_tx_rate = ivi.min_tx_rate;
1124         vf_rate.max_tx_rate = ivi.max_tx_rate;
1125         vf_spoofchk.setting = ivi.spoofchk;
1126         vf_linkstate.link_state = ivi.linkstate;
1127         vf_rss_query_en.setting = ivi.rss_query_en;
1128         vf_trust.setting = ivi.trusted;
1129         vf = nla_nest_start(skb, IFLA_VF_INFO);
1130         if (!vf) {
1131                 nla_nest_cancel(skb, vfinfo);
1132                 return -EMSGSIZE;
1133         }
1134         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1135             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1136             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1137                     &vf_rate) ||
1138             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1139                     &vf_tx_rate) ||
1140             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1141                     &vf_spoofchk) ||
1142             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1143                     &vf_linkstate) ||
1144             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1145                     sizeof(vf_rss_query_en),
1146                     &vf_rss_query_en) ||
1147             nla_put(skb, IFLA_VF_TRUST,
1148                     sizeof(vf_trust), &vf_trust))
1149                 return -EMSGSIZE;
1150         memset(&vf_stats, 0, sizeof(vf_stats));
1151         if (dev->netdev_ops->ndo_get_vf_stats)
1152                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1153                                                 &vf_stats);
1154         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1155         if (!vfstats) {
1156                 nla_nest_cancel(skb, vf);
1157                 nla_nest_cancel(skb, vfinfo);
1158                 return -EMSGSIZE;
1159         }
1160         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1161                         vf_stats.rx_packets) ||
1162             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1163                         vf_stats.tx_packets) ||
1164             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1165                         vf_stats.rx_bytes) ||
1166             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1167                         vf_stats.tx_bytes) ||
1168             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1169                         vf_stats.broadcast) ||
1170             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1171                         vf_stats.multicast))
1172                 return -EMSGSIZE;
1173         nla_nest_end(skb, vfstats);
1174         nla_nest_end(skb, vf);
1175         return 0;
1176 }
1177
1178 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1179 {
1180         struct rtnl_link_ifmap map = {
1181                 .mem_start   = dev->mem_start,
1182                 .mem_end     = dev->mem_end,
1183                 .base_addr   = dev->base_addr,
1184                 .irq         = dev->irq,
1185                 .dma         = dev->dma,
1186                 .port        = dev->if_port,
1187         };
1188         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1189                 return -EMSGSIZE;
1190
1191         return 0;
1192 }
1193
1194 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1195                             int type, u32 pid, u32 seq, u32 change,
1196                             unsigned int flags, u32 ext_filter_mask)
1197 {
1198         struct ifinfomsg *ifm;
1199         struct nlmsghdr *nlh;
1200         struct nlattr *af_spec;
1201         struct rtnl_af_ops *af_ops;
1202         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1203
1204         ASSERT_RTNL();
1205         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1206         if (nlh == NULL)
1207                 return -EMSGSIZE;
1208
1209         ifm = nlmsg_data(nlh);
1210         ifm->ifi_family = AF_UNSPEC;
1211         ifm->__ifi_pad = 0;
1212         ifm->ifi_type = dev->type;
1213         ifm->ifi_index = dev->ifindex;
1214         ifm->ifi_flags = dev_get_flags(dev);
1215         ifm->ifi_change = change;
1216
1217         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1218             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1219             nla_put_u8(skb, IFLA_OPERSTATE,
1220                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1221             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1222             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1223             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1224             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1225             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1226 #ifdef CONFIG_RPS
1227             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1228 #endif
1229             (dev->ifindex != dev_get_iflink(dev) &&
1230              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1231             (upper_dev &&
1232              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1233             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1234             (dev->qdisc &&
1235              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1236             (dev->ifalias &&
1237              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1238             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1239                         atomic_read(&dev->carrier_changes)) ||
1240             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1241                 goto nla_put_failure;
1242
1243         if (rtnl_fill_link_ifmap(skb, dev))
1244                 goto nla_put_failure;
1245
1246         if (dev->addr_len) {
1247                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1248                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1249                         goto nla_put_failure;
1250         }
1251
1252         if (rtnl_phys_port_id_fill(skb, dev))
1253                 goto nla_put_failure;
1254
1255         if (rtnl_phys_port_name_fill(skb, dev))
1256                 goto nla_put_failure;
1257
1258         if (rtnl_phys_switch_id_fill(skb, dev))
1259                 goto nla_put_failure;
1260
1261         if (rtnl_fill_stats(skb, dev))
1262                 goto nla_put_failure;
1263
1264         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1265             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1266                 goto nla_put_failure;
1267
1268         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1269             ext_filter_mask & RTEXT_FILTER_VF) {
1270                 int i;
1271                 struct nlattr *vfinfo;
1272                 int num_vfs = dev_num_vf(dev->dev.parent);
1273
1274                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1275                 if (!vfinfo)
1276                         goto nla_put_failure;
1277                 for (i = 0; i < num_vfs; i++) {
1278                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1279                                 goto nla_put_failure;
1280                 }
1281
1282                 nla_nest_end(skb, vfinfo);
1283         }
1284
1285         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1286                 goto nla_put_failure;
1287
1288         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1289                 if (rtnl_link_fill(skb, dev) < 0)
1290                         goto nla_put_failure;
1291         }
1292
1293         if (dev->rtnl_link_ops &&
1294             dev->rtnl_link_ops->get_link_net) {
1295                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1296
1297                 if (!net_eq(dev_net(dev), link_net)) {
1298                         int id = peernet2id_alloc(dev_net(dev), link_net);
1299
1300                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1301                                 goto nla_put_failure;
1302                 }
1303         }
1304
1305         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1306                 goto nla_put_failure;
1307
1308         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1309                 if (af_ops->fill_link_af) {
1310                         struct nlattr *af;
1311                         int err;
1312
1313                         if (!(af = nla_nest_start(skb, af_ops->family)))
1314                                 goto nla_put_failure;
1315
1316                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1317
1318                         /*
1319                          * Caller may return ENODATA to indicate that there
1320                          * was no data to be dumped. This is not an error, it
1321                          * means we should trim the attribute header and
1322                          * continue.
1323                          */
1324                         if (err == -ENODATA)
1325                                 nla_nest_cancel(skb, af);
1326                         else if (err < 0)
1327                                 goto nla_put_failure;
1328
1329                         nla_nest_end(skb, af);
1330                 }
1331         }
1332
1333         nla_nest_end(skb, af_spec);
1334
1335         nlmsg_end(skb, nlh);
1336         return 0;
1337
1338 nla_put_failure:
1339         nlmsg_cancel(skb, nlh);
1340         return -EMSGSIZE;
1341 }
1342
1343 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1344         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1345         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1346         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1347         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1348         [IFLA_MTU]              = { .type = NLA_U32 },
1349         [IFLA_LINK]             = { .type = NLA_U32 },
1350         [IFLA_MASTER]           = { .type = NLA_U32 },
1351         [IFLA_CARRIER]          = { .type = NLA_U8 },
1352         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1353         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1354         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1355         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1356         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1357         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1358         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1359         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1360         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1361         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1362         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1363         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1364         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1365         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1366         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1367         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1368         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1369         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1370         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1371         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1372         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1373 };
1374
1375 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1376         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1377         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1378         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1379         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1380 };
1381
1382 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1383         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1384         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1385         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1386         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1387         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1388         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1389         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1390         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1391         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1392 };
1393
1394 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1395         [IFLA_VF_STATS_RX_PACKETS]      = { .type = NLA_U64 },
1396         [IFLA_VF_STATS_TX_PACKETS]      = { .type = NLA_U64 },
1397         [IFLA_VF_STATS_RX_BYTES]        = { .type = NLA_U64 },
1398         [IFLA_VF_STATS_TX_BYTES]        = { .type = NLA_U64 },
1399         [IFLA_VF_STATS_BROADCAST]       = { .type = NLA_U64 },
1400         [IFLA_VF_STATS_MULTICAST]       = { .type = NLA_U64 },
1401 };
1402
1403 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1404         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1405         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1406                                     .len = PORT_PROFILE_MAX },
1407         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1408                                     .len = sizeof(struct ifla_port_vsi)},
1409         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1410                                       .len = PORT_UUID_MAX },
1411         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1412                                     .len = PORT_UUID_MAX },
1413         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1414         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1415 };
1416
1417 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1418 {
1419         const struct rtnl_link_ops *ops = NULL;
1420         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1421
1422         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1423                 return NULL;
1424
1425         if (linfo[IFLA_INFO_KIND]) {
1426                 char kind[MODULE_NAME_LEN];
1427
1428                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1429                 ops = rtnl_link_ops_get(kind);
1430         }
1431
1432         return ops;
1433 }
1434
1435 static bool link_master_filtered(struct net_device *dev, int master_idx)
1436 {
1437         struct net_device *master;
1438
1439         if (!master_idx)
1440                 return false;
1441
1442         master = netdev_master_upper_dev_get(dev);
1443         if (!master || master->ifindex != master_idx)
1444                 return true;
1445
1446         return false;
1447 }
1448
1449 static bool link_kind_filtered(const struct net_device *dev,
1450                                const struct rtnl_link_ops *kind_ops)
1451 {
1452         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1453                 return true;
1454
1455         return false;
1456 }
1457
1458 static bool link_dump_filtered(struct net_device *dev,
1459                                int master_idx,
1460                                const struct rtnl_link_ops *kind_ops)
1461 {
1462         if (link_master_filtered(dev, master_idx) ||
1463             link_kind_filtered(dev, kind_ops))
1464                 return true;
1465
1466         return false;
1467 }
1468
1469 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1470 {
1471         struct net *net = sock_net(skb->sk);
1472         int h, s_h;
1473         int idx = 0, s_idx;
1474         struct net_device *dev;
1475         struct hlist_head *head;
1476         struct nlattr *tb[IFLA_MAX+1];
1477         u32 ext_filter_mask = 0;
1478         const struct rtnl_link_ops *kind_ops = NULL;
1479         unsigned int flags = NLM_F_MULTI;
1480         int master_idx = 0;
1481         int err;
1482         int hdrlen;
1483
1484         s_h = cb->args[0];
1485         s_idx = cb->args[1];
1486
1487         cb->seq = net->dev_base_seq;
1488
1489         /* A hack to preserve kernel<->userspace interface.
1490          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1491          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1492          * what iproute2 < v3.9.0 used.
1493          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1494          * attribute, its netlink message is shorter than struct ifinfomsg.
1495          */
1496         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1497                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1498
1499         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1500
1501                 if (tb[IFLA_EXT_MASK])
1502                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1503
1504                 if (tb[IFLA_MASTER])
1505                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1506
1507                 if (tb[IFLA_LINKINFO])
1508                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1509
1510                 if (master_idx || kind_ops)
1511                         flags |= NLM_F_DUMP_FILTERED;
1512         }
1513
1514         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1515                 idx = 0;
1516                 head = &net->dev_index_head[h];
1517                 hlist_for_each_entry(dev, head, index_hlist) {
1518                         if (link_dump_filtered(dev, master_idx, kind_ops))
1519                                 continue;
1520                         if (idx < s_idx)
1521                                 goto cont;
1522                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1523                                                NETLINK_CB(cb->skb).portid,
1524                                                cb->nlh->nlmsg_seq, 0,
1525                                                flags,
1526                                                ext_filter_mask);
1527                         /* If we ran out of room on the first message,
1528                          * we're in trouble
1529                          */
1530                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1531
1532                         if (err < 0)
1533                                 goto out;
1534
1535                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1536 cont:
1537                         idx++;
1538                 }
1539         }
1540 out:
1541         cb->args[1] = idx;
1542         cb->args[0] = h;
1543
1544         return skb->len;
1545 }
1546
1547 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1548 {
1549         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1550 }
1551 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1552
1553 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1554 {
1555         struct net *net;
1556         /* Examine the link attributes and figure out which
1557          * network namespace we are talking about.
1558          */
1559         if (tb[IFLA_NET_NS_PID])
1560                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1561         else if (tb[IFLA_NET_NS_FD])
1562                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1563         else
1564                 net = get_net(src_net);
1565         return net;
1566 }
1567 EXPORT_SYMBOL(rtnl_link_get_net);
1568
1569 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1570 {
1571         if (dev) {
1572                 if (tb[IFLA_ADDRESS] &&
1573                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1574                         return -EINVAL;
1575
1576                 if (tb[IFLA_BROADCAST] &&
1577                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1578                         return -EINVAL;
1579         }
1580
1581         if (tb[IFLA_AF_SPEC]) {
1582                 struct nlattr *af;
1583                 int rem, err;
1584
1585                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1586                         const struct rtnl_af_ops *af_ops;
1587
1588                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1589                                 return -EAFNOSUPPORT;
1590
1591                         if (!af_ops->set_link_af)
1592                                 return -EOPNOTSUPP;
1593
1594                         if (af_ops->validate_link_af) {
1595                                 err = af_ops->validate_link_af(dev, af);
1596                                 if (err < 0)
1597                                         return err;
1598                         }
1599                 }
1600         }
1601
1602         return 0;
1603 }
1604
1605 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1606 {
1607         const struct net_device_ops *ops = dev->netdev_ops;
1608         int err = -EINVAL;
1609
1610         if (tb[IFLA_VF_MAC]) {
1611                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1612
1613                 err = -EOPNOTSUPP;
1614                 if (ops->ndo_set_vf_mac)
1615                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1616                                                   ivm->mac);
1617                 if (err < 0)
1618                         return err;
1619         }
1620
1621         if (tb[IFLA_VF_VLAN]) {
1622                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1623
1624                 err = -EOPNOTSUPP;
1625                 if (ops->ndo_set_vf_vlan)
1626                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1627                                                    ivv->qos);
1628                 if (err < 0)
1629                         return err;
1630         }
1631
1632         if (tb[IFLA_VF_TX_RATE]) {
1633                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1634                 struct ifla_vf_info ivf;
1635
1636                 err = -EOPNOTSUPP;
1637                 if (ops->ndo_get_vf_config)
1638                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1639                 if (err < 0)
1640                         return err;
1641
1642                 err = -EOPNOTSUPP;
1643                 if (ops->ndo_set_vf_rate)
1644                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1645                                                    ivf.min_tx_rate,
1646                                                    ivt->rate);
1647                 if (err < 0)
1648                         return err;
1649         }
1650
1651         if (tb[IFLA_VF_RATE]) {
1652                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1653
1654                 err = -EOPNOTSUPP;
1655                 if (ops->ndo_set_vf_rate)
1656                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1657                                                    ivt->min_tx_rate,
1658                                                    ivt->max_tx_rate);
1659                 if (err < 0)
1660                         return err;
1661         }
1662
1663         if (tb[IFLA_VF_SPOOFCHK]) {
1664                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1665
1666                 err = -EOPNOTSUPP;
1667                 if (ops->ndo_set_vf_spoofchk)
1668                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1669                                                        ivs->setting);
1670                 if (err < 0)
1671                         return err;
1672         }
1673
1674         if (tb[IFLA_VF_LINK_STATE]) {
1675                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1676
1677                 err = -EOPNOTSUPP;
1678                 if (ops->ndo_set_vf_link_state)
1679                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1680                                                          ivl->link_state);
1681                 if (err < 0)
1682                         return err;
1683         }
1684
1685         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1686                 struct ifla_vf_rss_query_en *ivrssq_en;
1687
1688                 err = -EOPNOTSUPP;
1689                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1690                 if (ops->ndo_set_vf_rss_query_en)
1691                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1692                                                            ivrssq_en->setting);
1693                 if (err < 0)
1694                         return err;
1695         }
1696
1697         if (tb[IFLA_VF_TRUST]) {
1698                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1699
1700                 err = -EOPNOTSUPP;
1701                 if (ops->ndo_set_vf_trust)
1702                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1703                 if (err < 0)
1704                         return err;
1705         }
1706
1707         return err;
1708 }
1709
1710 static int do_set_master(struct net_device *dev, int ifindex)
1711 {
1712         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1713         const struct net_device_ops *ops;
1714         int err;
1715
1716         if (upper_dev) {
1717                 if (upper_dev->ifindex == ifindex)
1718                         return 0;
1719                 ops = upper_dev->netdev_ops;
1720                 if (ops->ndo_del_slave) {
1721                         err = ops->ndo_del_slave(upper_dev, dev);
1722                         if (err)
1723                                 return err;
1724                 } else {
1725                         return -EOPNOTSUPP;
1726                 }
1727         }
1728
1729         if (ifindex) {
1730                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1731                 if (!upper_dev)
1732                         return -EINVAL;
1733                 ops = upper_dev->netdev_ops;
1734                 if (ops->ndo_add_slave) {
1735                         err = ops->ndo_add_slave(upper_dev, dev);
1736                         if (err)
1737                                 return err;
1738                 } else {
1739                         return -EOPNOTSUPP;
1740                 }
1741         }
1742         return 0;
1743 }
1744
1745 #define DO_SETLINK_MODIFIED     0x01
1746 /* notify flag means notify + modified. */
1747 #define DO_SETLINK_NOTIFY       0x03
1748 static int do_setlink(const struct sk_buff *skb,
1749                       struct net_device *dev, struct ifinfomsg *ifm,
1750                       struct nlattr **tb, char *ifname, int status)
1751 {
1752         const struct net_device_ops *ops = dev->netdev_ops;
1753         int err;
1754
1755         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1756                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1757                 if (IS_ERR(net)) {
1758                         err = PTR_ERR(net);
1759                         goto errout;
1760                 }
1761                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1762                         put_net(net);
1763                         err = -EPERM;
1764                         goto errout;
1765                 }
1766                 err = dev_change_net_namespace(dev, net, ifname);
1767                 put_net(net);
1768                 if (err)
1769                         goto errout;
1770                 status |= DO_SETLINK_MODIFIED;
1771         }
1772
1773         if (tb[IFLA_MAP]) {
1774                 struct rtnl_link_ifmap *u_map;
1775                 struct ifmap k_map;
1776
1777                 if (!ops->ndo_set_config) {
1778                         err = -EOPNOTSUPP;
1779                         goto errout;
1780                 }
1781
1782                 if (!netif_device_present(dev)) {
1783                         err = -ENODEV;
1784                         goto errout;
1785                 }
1786
1787                 u_map = nla_data(tb[IFLA_MAP]);
1788                 k_map.mem_start = (unsigned long) u_map->mem_start;
1789                 k_map.mem_end = (unsigned long) u_map->mem_end;
1790                 k_map.base_addr = (unsigned short) u_map->base_addr;
1791                 k_map.irq = (unsigned char) u_map->irq;
1792                 k_map.dma = (unsigned char) u_map->dma;
1793                 k_map.port = (unsigned char) u_map->port;
1794
1795                 err = ops->ndo_set_config(dev, &k_map);
1796                 if (err < 0)
1797                         goto errout;
1798
1799                 status |= DO_SETLINK_NOTIFY;
1800         }
1801
1802         if (tb[IFLA_ADDRESS]) {
1803                 struct sockaddr *sa;
1804                 int len;
1805
1806                 len = sizeof(sa_family_t) + dev->addr_len;
1807                 sa = kmalloc(len, GFP_KERNEL);
1808                 if (!sa) {
1809                         err = -ENOMEM;
1810                         goto errout;
1811                 }
1812                 sa->sa_family = dev->type;
1813                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1814                        dev->addr_len);
1815                 err = dev_set_mac_address(dev, sa);
1816                 kfree(sa);
1817                 if (err)
1818                         goto errout;
1819                 status |= DO_SETLINK_MODIFIED;
1820         }
1821
1822         if (tb[IFLA_MTU]) {
1823                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1824                 if (err < 0)
1825                         goto errout;
1826                 status |= DO_SETLINK_MODIFIED;
1827         }
1828
1829         if (tb[IFLA_GROUP]) {
1830                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1831                 status |= DO_SETLINK_NOTIFY;
1832         }
1833
1834         /*
1835          * Interface selected by interface index but interface
1836          * name provided implies that a name change has been
1837          * requested.
1838          */
1839         if (ifm->ifi_index > 0 && ifname[0]) {
1840                 err = dev_change_name(dev, ifname);
1841                 if (err < 0)
1842                         goto errout;
1843                 status |= DO_SETLINK_MODIFIED;
1844         }
1845
1846         if (tb[IFLA_IFALIAS]) {
1847                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1848                                     nla_len(tb[IFLA_IFALIAS]));
1849                 if (err < 0)
1850                         goto errout;
1851                 status |= DO_SETLINK_NOTIFY;
1852         }
1853
1854         if (tb[IFLA_BROADCAST]) {
1855                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1856                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1857         }
1858
1859         if (ifm->ifi_flags || ifm->ifi_change) {
1860                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1861                 if (err < 0)
1862                         goto errout;
1863         }
1864
1865         if (tb[IFLA_MASTER]) {
1866                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1867                 if (err)
1868                         goto errout;
1869                 status |= DO_SETLINK_MODIFIED;
1870         }
1871
1872         if (tb[IFLA_CARRIER]) {
1873                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1874                 if (err)
1875                         goto errout;
1876                 status |= DO_SETLINK_MODIFIED;
1877         }
1878
1879         if (tb[IFLA_TXQLEN]) {
1880                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1881
1882                 if (dev->tx_queue_len ^ value)
1883                         status |= DO_SETLINK_NOTIFY;
1884
1885                 dev->tx_queue_len = value;
1886         }
1887
1888         if (tb[IFLA_OPERSTATE])
1889                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1890
1891         if (tb[IFLA_LINKMODE]) {
1892                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1893
1894                 write_lock_bh(&dev_base_lock);
1895                 if (dev->link_mode ^ value)
1896                         status |= DO_SETLINK_NOTIFY;
1897                 dev->link_mode = value;
1898                 write_unlock_bh(&dev_base_lock);
1899         }
1900
1901         if (tb[IFLA_VFINFO_LIST]) {
1902                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1903                 struct nlattr *attr;
1904                 int rem;
1905
1906                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1907                         if (nla_type(attr) != IFLA_VF_INFO ||
1908                             nla_len(attr) < NLA_HDRLEN) {
1909                                 err = -EINVAL;
1910                                 goto errout;
1911                         }
1912                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1913                                                ifla_vf_policy);
1914                         if (err < 0)
1915                                 goto errout;
1916                         err = do_setvfinfo(dev, vfinfo);
1917                         if (err < 0)
1918                                 goto errout;
1919                         status |= DO_SETLINK_NOTIFY;
1920                 }
1921         }
1922         err = 0;
1923
1924         if (tb[IFLA_VF_PORTS]) {
1925                 struct nlattr *port[IFLA_PORT_MAX+1];
1926                 struct nlattr *attr;
1927                 int vf;
1928                 int rem;
1929
1930                 err = -EOPNOTSUPP;
1931                 if (!ops->ndo_set_vf_port)
1932                         goto errout;
1933
1934                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1935                         if (nla_type(attr) != IFLA_VF_PORT ||
1936                             nla_len(attr) < NLA_HDRLEN) {
1937                                 err = -EINVAL;
1938                                 goto errout;
1939                         }
1940                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1941                                                ifla_port_policy);
1942                         if (err < 0)
1943                                 goto errout;
1944                         if (!port[IFLA_PORT_VF]) {
1945                                 err = -EOPNOTSUPP;
1946                                 goto errout;
1947                         }
1948                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1949                         err = ops->ndo_set_vf_port(dev, vf, port);
1950                         if (err < 0)
1951                                 goto errout;
1952                         status |= DO_SETLINK_NOTIFY;
1953                 }
1954         }
1955         err = 0;
1956
1957         if (tb[IFLA_PORT_SELF]) {
1958                 struct nlattr *port[IFLA_PORT_MAX+1];
1959
1960                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1961                         tb[IFLA_PORT_SELF], ifla_port_policy);
1962                 if (err < 0)
1963                         goto errout;
1964
1965                 err = -EOPNOTSUPP;
1966                 if (ops->ndo_set_vf_port)
1967                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1968                 if (err < 0)
1969                         goto errout;
1970                 status |= DO_SETLINK_NOTIFY;
1971         }
1972
1973         if (tb[IFLA_AF_SPEC]) {
1974                 struct nlattr *af;
1975                 int rem;
1976
1977                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1978                         const struct rtnl_af_ops *af_ops;
1979
1980                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1981                                 BUG();
1982
1983                         err = af_ops->set_link_af(dev, af);
1984                         if (err < 0)
1985                                 goto errout;
1986
1987                         status |= DO_SETLINK_NOTIFY;
1988                 }
1989         }
1990         err = 0;
1991
1992         if (tb[IFLA_PROTO_DOWN]) {
1993                 err = dev_change_proto_down(dev,
1994                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
1995                 if (err)
1996                         goto errout;
1997                 status |= DO_SETLINK_NOTIFY;
1998         }
1999
2000 errout:
2001         if (status & DO_SETLINK_MODIFIED) {
2002                 if (status & DO_SETLINK_NOTIFY)
2003                         netdev_state_change(dev);
2004
2005                 if (err < 0)
2006                         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",
2007                                              dev->name);
2008         }
2009
2010         return err;
2011 }
2012
2013 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2014 {
2015         struct net *net = sock_net(skb->sk);
2016         struct ifinfomsg *ifm;
2017         struct net_device *dev;
2018         int err;
2019         struct nlattr *tb[IFLA_MAX+1];
2020         char ifname[IFNAMSIZ];
2021
2022         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2023         if (err < 0)
2024                 goto errout;
2025
2026         if (tb[IFLA_IFNAME])
2027                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2028         else
2029                 ifname[0] = '\0';
2030
2031         err = -EINVAL;
2032         ifm = nlmsg_data(nlh);
2033         if (ifm->ifi_index > 0)
2034                 dev = __dev_get_by_index(net, ifm->ifi_index);
2035         else if (tb[IFLA_IFNAME])
2036                 dev = __dev_get_by_name(net, ifname);
2037         else
2038                 goto errout;
2039
2040         if (dev == NULL) {
2041                 err = -ENODEV;
2042                 goto errout;
2043         }
2044
2045         err = validate_linkmsg(dev, tb);
2046         if (err < 0)
2047                 goto errout;
2048
2049         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2050 errout:
2051         return err;
2052 }
2053
2054 static int rtnl_group_dellink(const struct net *net, int group)
2055 {
2056         struct net_device *dev, *aux;
2057         LIST_HEAD(list_kill);
2058         bool found = false;
2059
2060         if (!group)
2061                 return -EPERM;
2062
2063         for_each_netdev(net, dev) {
2064                 if (dev->group == group) {
2065                         const struct rtnl_link_ops *ops;
2066
2067                         found = true;
2068                         ops = dev->rtnl_link_ops;
2069                         if (!ops || !ops->dellink)
2070                                 return -EOPNOTSUPP;
2071                 }
2072         }
2073
2074         if (!found)
2075                 return -ENODEV;
2076
2077         for_each_netdev_safe(net, dev, aux) {
2078                 if (dev->group == group) {
2079                         const struct rtnl_link_ops *ops;
2080
2081                         ops = dev->rtnl_link_ops;
2082                         ops->dellink(dev, &list_kill);
2083                 }
2084         }
2085         unregister_netdevice_many(&list_kill);
2086
2087         return 0;
2088 }
2089
2090 int rtnl_delete_link(struct net_device *dev)
2091 {
2092         const struct rtnl_link_ops *ops;
2093         LIST_HEAD(list_kill);
2094
2095         ops = dev->rtnl_link_ops;
2096         if (!ops || !ops->dellink)
2097                 return -EOPNOTSUPP;
2098
2099         ops->dellink(dev, &list_kill);
2100         unregister_netdevice_many(&list_kill);
2101
2102         return 0;
2103 }
2104 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2105
2106 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2107 {
2108         struct net *net = sock_net(skb->sk);
2109         struct net_device *dev;
2110         struct ifinfomsg *ifm;
2111         char ifname[IFNAMSIZ];
2112         struct nlattr *tb[IFLA_MAX+1];
2113         int err;
2114
2115         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2116         if (err < 0)
2117                 return err;
2118
2119         if (tb[IFLA_IFNAME])
2120                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2121
2122         ifm = nlmsg_data(nlh);
2123         if (ifm->ifi_index > 0)
2124                 dev = __dev_get_by_index(net, ifm->ifi_index);
2125         else if (tb[IFLA_IFNAME])
2126                 dev = __dev_get_by_name(net, ifname);
2127         else if (tb[IFLA_GROUP])
2128                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2129         else
2130                 return -EINVAL;
2131
2132         if (!dev)
2133                 return -ENODEV;
2134
2135         return rtnl_delete_link(dev);
2136 }
2137
2138 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2139 {
2140         unsigned int old_flags;
2141         int err;
2142
2143         old_flags = dev->flags;
2144         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2145                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2146                 if (err < 0)
2147                         return err;
2148         }
2149
2150         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2151
2152         __dev_notify_flags(dev, old_flags, ~0U);
2153         return 0;
2154 }
2155 EXPORT_SYMBOL(rtnl_configure_link);
2156
2157 struct net_device *rtnl_create_link(struct net *net,
2158         const char *ifname, unsigned char name_assign_type,
2159         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2160 {
2161         int err;
2162         struct net_device *dev;
2163         unsigned int num_tx_queues = 1;
2164         unsigned int num_rx_queues = 1;
2165
2166         if (tb[IFLA_NUM_TX_QUEUES])
2167                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2168         else if (ops->get_num_tx_queues)
2169                 num_tx_queues = ops->get_num_tx_queues();
2170
2171         if (tb[IFLA_NUM_RX_QUEUES])
2172                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2173         else if (ops->get_num_rx_queues)
2174                 num_rx_queues = ops->get_num_rx_queues();
2175
2176         err = -ENOMEM;
2177         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2178                                ops->setup, num_tx_queues, num_rx_queues);
2179         if (!dev)
2180                 goto err;
2181
2182         dev_net_set(dev, net);
2183         dev->rtnl_link_ops = ops;
2184         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2185
2186         if (tb[IFLA_MTU])
2187                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2188         if (tb[IFLA_ADDRESS]) {
2189                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2190                                 nla_len(tb[IFLA_ADDRESS]));
2191                 dev->addr_assign_type = NET_ADDR_SET;
2192         }
2193         if (tb[IFLA_BROADCAST])
2194                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2195                                 nla_len(tb[IFLA_BROADCAST]));
2196         if (tb[IFLA_TXQLEN])
2197                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2198         if (tb[IFLA_OPERSTATE])
2199                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2200         if (tb[IFLA_LINKMODE])
2201                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2202         if (tb[IFLA_GROUP])
2203                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2204
2205         return dev;
2206
2207 err:
2208         return ERR_PTR(err);
2209 }
2210 EXPORT_SYMBOL(rtnl_create_link);
2211
2212 static int rtnl_group_changelink(const struct sk_buff *skb,
2213                 struct net *net, int group,
2214                 struct ifinfomsg *ifm,
2215                 struct nlattr **tb)
2216 {
2217         struct net_device *dev, *aux;
2218         int err;
2219
2220         for_each_netdev_safe(net, dev, aux) {
2221                 if (dev->group == group) {
2222                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2223                         if (err < 0)
2224                                 return err;
2225                 }
2226         }
2227
2228         return 0;
2229 }
2230
2231 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2232 {
2233         struct net *net = sock_net(skb->sk);
2234         const struct rtnl_link_ops *ops;
2235         const struct rtnl_link_ops *m_ops = NULL;
2236         struct net_device *dev;
2237         struct net_device *master_dev = NULL;
2238         struct ifinfomsg *ifm;
2239         char kind[MODULE_NAME_LEN];
2240         char ifname[IFNAMSIZ];
2241         struct nlattr *tb[IFLA_MAX+1];
2242         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2243         unsigned char name_assign_type = NET_NAME_USER;
2244         int err;
2245
2246 #ifdef CONFIG_MODULES
2247 replay:
2248 #endif
2249         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2250         if (err < 0)
2251                 return err;
2252
2253         if (tb[IFLA_IFNAME])
2254                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2255         else
2256                 ifname[0] = '\0';
2257
2258         ifm = nlmsg_data(nlh);
2259         if (ifm->ifi_index > 0)
2260                 dev = __dev_get_by_index(net, ifm->ifi_index);
2261         else {
2262                 if (ifname[0])
2263                         dev = __dev_get_by_name(net, ifname);
2264                 else
2265                         dev = NULL;
2266         }
2267
2268         if (dev) {
2269                 master_dev = netdev_master_upper_dev_get(dev);
2270                 if (master_dev)
2271                         m_ops = master_dev->rtnl_link_ops;
2272         }
2273
2274         err = validate_linkmsg(dev, tb);
2275         if (err < 0)
2276                 return err;
2277
2278         if (tb[IFLA_LINKINFO]) {
2279                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2280                                        tb[IFLA_LINKINFO], ifla_info_policy);
2281                 if (err < 0)
2282                         return err;
2283         } else
2284                 memset(linkinfo, 0, sizeof(linkinfo));
2285
2286         if (linkinfo[IFLA_INFO_KIND]) {
2287                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2288                 ops = rtnl_link_ops_get(kind);
2289         } else {
2290                 kind[0] = '\0';
2291                 ops = NULL;
2292         }
2293
2294         if (1) {
2295                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2296                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2297                 struct nlattr **data = NULL;
2298                 struct nlattr **slave_data = NULL;
2299                 struct net *dest_net, *link_net = NULL;
2300
2301                 if (ops) {
2302                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2303                                 err = nla_parse_nested(attr, ops->maxtype,
2304                                                        linkinfo[IFLA_INFO_DATA],
2305                                                        ops->policy);
2306                                 if (err < 0)
2307                                         return err;
2308                                 data = attr;
2309                         }
2310                         if (ops->validate) {
2311                                 err = ops->validate(tb, data);
2312                                 if (err < 0)
2313                                         return err;
2314                         }
2315                 }
2316
2317                 if (m_ops) {
2318                         if (m_ops->slave_maxtype &&
2319                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2320                                 err = nla_parse_nested(slave_attr,
2321                                                        m_ops->slave_maxtype,
2322                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2323                                                        m_ops->slave_policy);
2324                                 if (err < 0)
2325                                         return err;
2326                                 slave_data = slave_attr;
2327                         }
2328                         if (m_ops->slave_validate) {
2329                                 err = m_ops->slave_validate(tb, slave_data);
2330                                 if (err < 0)
2331                                         return err;
2332                         }
2333                 }
2334
2335                 if (dev) {
2336                         int status = 0;
2337
2338                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2339                                 return -EEXIST;
2340                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2341                                 return -EOPNOTSUPP;
2342
2343                         if (linkinfo[IFLA_INFO_DATA]) {
2344                                 if (!ops || ops != dev->rtnl_link_ops ||
2345                                     !ops->changelink)
2346                                         return -EOPNOTSUPP;
2347
2348                                 err = ops->changelink(dev, tb, data);
2349                                 if (err < 0)
2350                                         return err;
2351                                 status |= DO_SETLINK_NOTIFY;
2352                         }
2353
2354                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2355                                 if (!m_ops || !m_ops->slave_changelink)
2356                                         return -EOPNOTSUPP;
2357
2358                                 err = m_ops->slave_changelink(master_dev, dev,
2359                                                               tb, slave_data);
2360                                 if (err < 0)
2361                                         return err;
2362                                 status |= DO_SETLINK_NOTIFY;
2363                         }
2364
2365                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2366                 }
2367
2368                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2369                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2370                                 return rtnl_group_changelink(skb, net,
2371                                                 nla_get_u32(tb[IFLA_GROUP]),
2372                                                 ifm, tb);
2373                         return -ENODEV;
2374                 }
2375
2376                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2377                         return -EOPNOTSUPP;
2378
2379                 if (!ops) {
2380 #ifdef CONFIG_MODULES
2381                         if (kind[0]) {
2382                                 __rtnl_unlock();
2383                                 request_module("rtnl-link-%s", kind);
2384                                 rtnl_lock();
2385                                 ops = rtnl_link_ops_get(kind);
2386                                 if (ops)
2387                                         goto replay;
2388                         }
2389 #endif
2390                         return -EOPNOTSUPP;
2391                 }
2392
2393                 if (!ops->setup)
2394                         return -EOPNOTSUPP;
2395
2396                 if (!ifname[0]) {
2397                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2398                         name_assign_type = NET_NAME_ENUM;
2399                 }
2400
2401                 dest_net = rtnl_link_get_net(net, tb);
2402                 if (IS_ERR(dest_net))
2403                         return PTR_ERR(dest_net);
2404
2405                 err = -EPERM;
2406                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2407                         goto out;
2408
2409                 if (tb[IFLA_LINK_NETNSID]) {
2410                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2411
2412                         link_net = get_net_ns_by_id(dest_net, id);
2413                         if (!link_net) {
2414                                 err =  -EINVAL;
2415                                 goto out;
2416                         }
2417                         err = -EPERM;
2418                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2419                                 goto out;
2420                 }
2421
2422                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2423                                        name_assign_type, ops, tb);
2424                 if (IS_ERR(dev)) {
2425                         err = PTR_ERR(dev);
2426                         goto out;
2427                 }
2428
2429                 dev->ifindex = ifm->ifi_index;
2430
2431                 if (ops->newlink) {
2432                         err = ops->newlink(link_net ? : net, dev, tb, data);
2433                         /* Drivers should call free_netdev() in ->destructor
2434                          * and unregister it on failure after registration
2435                          * so that device could be finally freed in rtnl_unlock.
2436                          */
2437                         if (err < 0) {
2438                                 /* If device is not registered at all, free it now */
2439                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2440                                         free_netdev(dev);
2441                                 goto out;
2442                         }
2443                 } else {
2444                         err = register_netdevice(dev);
2445                         if (err < 0) {
2446                                 free_netdev(dev);
2447                                 goto out;
2448                         }
2449                 }
2450                 err = rtnl_configure_link(dev, ifm);
2451                 if (err < 0)
2452                         goto out_unregister;
2453                 if (link_net) {
2454                         err = dev_change_net_namespace(dev, dest_net, ifname);
2455                         if (err < 0)
2456                                 goto out_unregister;
2457                 }
2458 out:
2459                 if (link_net)
2460                         put_net(link_net);
2461                 put_net(dest_net);
2462                 return err;
2463 out_unregister:
2464                 if (ops->newlink) {
2465                         LIST_HEAD(list_kill);
2466
2467                         ops->dellink(dev, &list_kill);
2468                         unregister_netdevice_many(&list_kill);
2469                 } else {
2470                         unregister_netdevice(dev);
2471                 }
2472                 goto out;
2473         }
2474 }
2475
2476 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2477 {
2478         struct net *net = sock_net(skb->sk);
2479         struct ifinfomsg *ifm;
2480         char ifname[IFNAMSIZ];
2481         struct nlattr *tb[IFLA_MAX+1];
2482         struct net_device *dev = NULL;
2483         struct sk_buff *nskb;
2484         int err;
2485         u32 ext_filter_mask = 0;
2486
2487         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2488         if (err < 0)
2489                 return err;
2490
2491         if (tb[IFLA_IFNAME])
2492                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2493
2494         if (tb[IFLA_EXT_MASK])
2495                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2496
2497         ifm = nlmsg_data(nlh);
2498         if (ifm->ifi_index > 0)
2499                 dev = __dev_get_by_index(net, ifm->ifi_index);
2500         else if (tb[IFLA_IFNAME])
2501                 dev = __dev_get_by_name(net, ifname);
2502         else
2503                 return -EINVAL;
2504
2505         if (dev == NULL)
2506                 return -ENODEV;
2507
2508         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2509         if (nskb == NULL)
2510                 return -ENOBUFS;
2511
2512         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2513                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2514         if (err < 0) {
2515                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2516                 WARN_ON(err == -EMSGSIZE);
2517                 kfree_skb(nskb);
2518         } else
2519                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2520
2521         return err;
2522 }
2523
2524 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2525 {
2526         struct net *net = sock_net(skb->sk);
2527         struct net_device *dev;
2528         struct nlattr *tb[IFLA_MAX+1];
2529         u32 ext_filter_mask = 0;
2530         u16 min_ifinfo_dump_size = 0;
2531         int hdrlen;
2532
2533         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2534         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2535                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2536
2537         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2538                 if (tb[IFLA_EXT_MASK])
2539                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2540         }
2541
2542         if (!ext_filter_mask)
2543                 return NLMSG_GOODSIZE;
2544         /*
2545          * traverse the list of net devices and compute the minimum
2546          * buffer size based upon the filter mask.
2547          */
2548         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2549                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2550                                              if_nlmsg_size(dev,
2551                                                            ext_filter_mask));
2552         }
2553
2554         return min_ifinfo_dump_size;
2555 }
2556
2557 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2558 {
2559         int idx;
2560         int s_idx = cb->family;
2561
2562         if (s_idx == 0)
2563                 s_idx = 1;
2564         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2565                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2566                 if (idx < s_idx || idx == PF_PACKET)
2567                         continue;
2568                 if (rtnl_msg_handlers[idx] == NULL ||
2569                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2570                         continue;
2571                 if (idx > s_idx) {
2572                         memset(&cb->args[0], 0, sizeof(cb->args));
2573                         cb->prev_seq = 0;
2574                         cb->seq = 0;
2575                 }
2576                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2577                         break;
2578         }
2579         cb->family = idx;
2580
2581         return skb->len;
2582 }
2583
2584 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2585                                        unsigned int change, gfp_t flags)
2586 {
2587         struct net *net = dev_net(dev);
2588         struct sk_buff *skb;
2589         int err = -ENOBUFS;
2590         size_t if_info_size;
2591
2592         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2593         if (skb == NULL)
2594                 goto errout;
2595
2596         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2597         if (err < 0) {
2598                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2599                 WARN_ON(err == -EMSGSIZE);
2600                 kfree_skb(skb);
2601                 goto errout;
2602         }
2603         return skb;
2604 errout:
2605         if (err < 0)
2606                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2607         return NULL;
2608 }
2609
2610 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2611 {
2612         struct net *net = dev_net(dev);
2613
2614         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2615 }
2616
2617 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2618                   gfp_t flags)
2619 {
2620         struct sk_buff *skb;
2621
2622         if (dev->reg_state != NETREG_REGISTERED)
2623                 return;
2624
2625         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2626         if (skb)
2627                 rtmsg_ifinfo_send(skb, dev, flags);
2628 }
2629 EXPORT_SYMBOL(rtmsg_ifinfo);
2630
2631 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2632                                    struct net_device *dev,
2633                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2634                                    int type, unsigned int flags,
2635                                    int nlflags, u16 ndm_state)
2636 {
2637         struct nlmsghdr *nlh;
2638         struct ndmsg *ndm;
2639
2640         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2641         if (!nlh)
2642                 return -EMSGSIZE;
2643
2644         ndm = nlmsg_data(nlh);
2645         ndm->ndm_family  = AF_BRIDGE;
2646         ndm->ndm_pad1    = 0;
2647         ndm->ndm_pad2    = 0;
2648         ndm->ndm_flags   = flags;
2649         ndm->ndm_type    = 0;
2650         ndm->ndm_ifindex = dev->ifindex;
2651         ndm->ndm_state   = ndm_state;
2652
2653         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2654                 goto nla_put_failure;
2655         if (vid)
2656                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2657                         goto nla_put_failure;
2658
2659         nlmsg_end(skb, nlh);
2660         return 0;
2661
2662 nla_put_failure:
2663         nlmsg_cancel(skb, nlh);
2664         return -EMSGSIZE;
2665 }
2666
2667 static inline size_t rtnl_fdb_nlmsg_size(void)
2668 {
2669         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2670 }
2671
2672 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2673                             u16 ndm_state)
2674 {
2675         struct net *net = dev_net(dev);
2676         struct sk_buff *skb;
2677         int err = -ENOBUFS;
2678
2679         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2680         if (!skb)
2681                 goto errout;
2682
2683         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2684                                       0, 0, type, NTF_SELF, 0, ndm_state);
2685         if (err < 0) {
2686                 kfree_skb(skb);
2687                 goto errout;
2688         }
2689
2690         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2691         return;
2692 errout:
2693         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2694 }
2695
2696 /**
2697  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2698  */
2699 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2700                      struct nlattr *tb[],
2701                      struct net_device *dev,
2702                      const unsigned char *addr, u16 vid,
2703                      u16 flags)
2704 {
2705         int err = -EINVAL;
2706
2707         /* If aging addresses are supported device will need to
2708          * implement its own handler for this.
2709          */
2710         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2711                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2712                 return err;
2713         }
2714
2715         if (vid) {
2716                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2717                 return err;
2718         }
2719
2720         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2721                 err = dev_uc_add_excl(dev, addr);
2722         else if (is_multicast_ether_addr(addr))
2723                 err = dev_mc_add_excl(dev, addr);
2724
2725         /* Only return duplicate errors if NLM_F_EXCL is set */
2726         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2727                 err = 0;
2728
2729         return err;
2730 }
2731 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2732
2733 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2734 {
2735         u16 vid = 0;
2736
2737         if (vlan_attr) {
2738                 if (nla_len(vlan_attr) != sizeof(u16)) {
2739                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2740                         return -EINVAL;
2741                 }
2742
2743                 vid = nla_get_u16(vlan_attr);
2744
2745                 if (!vid || vid >= VLAN_VID_MASK) {
2746                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2747                                 vid);
2748                         return -EINVAL;
2749                 }
2750         }
2751         *p_vid = vid;
2752         return 0;
2753 }
2754
2755 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2756 {
2757         struct net *net = sock_net(skb->sk);
2758         struct ndmsg *ndm;
2759         struct nlattr *tb[NDA_MAX+1];
2760         struct net_device *dev;
2761         u8 *addr;
2762         u16 vid;
2763         int err;
2764
2765         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2766         if (err < 0)
2767                 return err;
2768
2769         ndm = nlmsg_data(nlh);
2770         if (ndm->ndm_ifindex == 0) {
2771                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2772                 return -EINVAL;
2773         }
2774
2775         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2776         if (dev == NULL) {
2777                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2778                 return -ENODEV;
2779         }
2780
2781         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2782                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2783                 return -EINVAL;
2784         }
2785
2786         addr = nla_data(tb[NDA_LLADDR]);
2787
2788         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2789         if (err)
2790                 return err;
2791
2792         err = -EOPNOTSUPP;
2793
2794         /* Support fdb on master device the net/bridge default case */
2795         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2796             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2797                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2798                 const struct net_device_ops *ops = br_dev->netdev_ops;
2799
2800                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2801                                        nlh->nlmsg_flags);
2802                 if (err)
2803                         goto out;
2804                 else
2805                         ndm->ndm_flags &= ~NTF_MASTER;
2806         }
2807
2808         /* Embedded bridge, macvlan, and any other device support */
2809         if ((ndm->ndm_flags & NTF_SELF)) {
2810                 if (dev->netdev_ops->ndo_fdb_add)
2811                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2812                                                            vid,
2813                                                            nlh->nlmsg_flags);
2814                 else
2815                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2816                                                nlh->nlmsg_flags);
2817
2818                 if (!err) {
2819                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2820                                         ndm->ndm_state);
2821                         ndm->ndm_flags &= ~NTF_SELF;
2822                 }
2823         }
2824 out:
2825         return err;
2826 }
2827
2828 /**
2829  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2830  */
2831 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2832                      struct nlattr *tb[],
2833                      struct net_device *dev,
2834                      const unsigned char *addr, u16 vid)
2835 {
2836         int err = -EINVAL;
2837
2838         /* If aging addresses are supported device will need to
2839          * implement its own handler for this.
2840          */
2841         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2842                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2843                 return err;
2844         }
2845
2846         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2847                 err = dev_uc_del(dev, addr);
2848         else if (is_multicast_ether_addr(addr))
2849                 err = dev_mc_del(dev, addr);
2850
2851         return err;
2852 }
2853 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2854
2855 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2856 {
2857         struct net *net = sock_net(skb->sk);
2858         struct ndmsg *ndm;
2859         struct nlattr *tb[NDA_MAX+1];
2860         struct net_device *dev;
2861         int err = -EINVAL;
2862         __u8 *addr;
2863         u16 vid;
2864
2865         if (!netlink_capable(skb, CAP_NET_ADMIN))
2866                 return -EPERM;
2867
2868         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2869         if (err < 0)
2870                 return err;
2871
2872         ndm = nlmsg_data(nlh);
2873         if (ndm->ndm_ifindex == 0) {
2874                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2875                 return -EINVAL;
2876         }
2877
2878         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2879         if (dev == NULL) {
2880                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2881                 return -ENODEV;
2882         }
2883
2884         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2885                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2886                 return -EINVAL;
2887         }
2888
2889         addr = nla_data(tb[NDA_LLADDR]);
2890
2891         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2892         if (err)
2893                 return err;
2894
2895         err = -EOPNOTSUPP;
2896
2897         /* Support fdb on master device the net/bridge default case */
2898         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2899             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2900                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2901                 const struct net_device_ops *ops = br_dev->netdev_ops;
2902
2903                 if (ops->ndo_fdb_del)
2904                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2905
2906                 if (err)
2907                         goto out;
2908                 else
2909                         ndm->ndm_flags &= ~NTF_MASTER;
2910         }
2911
2912         /* Embedded bridge, macvlan, and any other device support */
2913         if (ndm->ndm_flags & NTF_SELF) {
2914                 if (dev->netdev_ops->ndo_fdb_del)
2915                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2916                                                            vid);
2917                 else
2918                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2919
2920                 if (!err) {
2921                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2922                                         ndm->ndm_state);
2923                         ndm->ndm_flags &= ~NTF_SELF;
2924                 }
2925         }
2926 out:
2927         return err;
2928 }
2929
2930 static int nlmsg_populate_fdb(struct sk_buff *skb,
2931                               struct netlink_callback *cb,
2932                               struct net_device *dev,
2933                               int *idx,
2934                               struct netdev_hw_addr_list *list)
2935 {
2936         struct netdev_hw_addr *ha;
2937         int err;
2938         u32 portid, seq;
2939
2940         portid = NETLINK_CB(cb->skb).portid;
2941         seq = cb->nlh->nlmsg_seq;
2942
2943         list_for_each_entry(ha, &list->list, list) {
2944                 if (*idx < cb->args[0])
2945                         goto skip;
2946
2947                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2948                                               portid, seq,
2949                                               RTM_NEWNEIGH, NTF_SELF,
2950                                               NLM_F_MULTI, NUD_PERMANENT);
2951                 if (err < 0)
2952                         return err;
2953 skip:
2954                 *idx += 1;
2955         }
2956         return 0;
2957 }
2958
2959 /**
2960  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2961  * @nlh: netlink message header
2962  * @dev: netdevice
2963  *
2964  * Default netdevice operation to dump the existing unicast address list.
2965  * Returns number of addresses from list put in skb.
2966  */
2967 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2968                       struct netlink_callback *cb,
2969                       struct net_device *dev,
2970                       struct net_device *filter_dev,
2971                       int idx)
2972 {
2973         int err;
2974
2975         netif_addr_lock_bh(dev);
2976         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2977         if (err)
2978                 goto out;
2979         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2980 out:
2981         netif_addr_unlock_bh(dev);
2982         return idx;
2983 }
2984 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2985
2986 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2987 {
2988         struct net_device *dev;
2989         struct nlattr *tb[IFLA_MAX+1];
2990         struct net_device *br_dev = NULL;
2991         const struct net_device_ops *ops = NULL;
2992         const struct net_device_ops *cops = NULL;
2993         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2994         struct net *net = sock_net(skb->sk);
2995         int brport_idx = 0;
2996         int br_idx = 0;
2997         int idx = 0;
2998
2999         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3000                         ifla_policy) == 0) {
3001                 if (tb[IFLA_MASTER])
3002                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3003         }
3004
3005         brport_idx = ifm->ifi_index;
3006
3007         if (br_idx) {
3008                 br_dev = __dev_get_by_index(net, br_idx);
3009                 if (!br_dev)
3010                         return -ENODEV;
3011
3012                 ops = br_dev->netdev_ops;
3013         }
3014
3015         for_each_netdev(net, dev) {
3016                 if (brport_idx && (dev->ifindex != brport_idx))
3017                         continue;
3018
3019                 if (!br_idx) { /* user did not specify a specific bridge */
3020                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3021                                 br_dev = netdev_master_upper_dev_get(dev);
3022                                 cops = br_dev->netdev_ops;
3023                         }
3024
3025                 } else {
3026                         if (dev != br_dev &&
3027                             !(dev->priv_flags & IFF_BRIDGE_PORT))
3028                                 continue;
3029
3030                         if (br_dev != netdev_master_upper_dev_get(dev) &&
3031                             !(dev->priv_flags & IFF_EBRIDGE))
3032                                 continue;
3033
3034                         cops = ops;
3035                 }
3036
3037                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3038                         if (cops && cops->ndo_fdb_dump)
3039                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3040                                                          idx);
3041                 }
3042
3043                 if (dev->netdev_ops->ndo_fdb_dump)
3044                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3045                                                             idx);
3046                 else
3047                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3048
3049                 cops = NULL;
3050         }
3051
3052         cb->args[0] = idx;
3053         return skb->len;
3054 }
3055
3056 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3057                                unsigned int attrnum, unsigned int flag)
3058 {
3059         if (mask & flag)
3060                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3061         return 0;
3062 }
3063
3064 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3065                             struct net_device *dev, u16 mode,
3066                             u32 flags, u32 mask, int nlflags,
3067                             u32 filter_mask,
3068                             int (*vlan_fill)(struct sk_buff *skb,
3069                                              struct net_device *dev,
3070                                              u32 filter_mask))
3071 {
3072         struct nlmsghdr *nlh;
3073         struct ifinfomsg *ifm;
3074         struct nlattr *br_afspec;
3075         struct nlattr *protinfo;
3076         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3077         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3078         int err = 0;
3079
3080         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3081         if (nlh == NULL)
3082                 return -EMSGSIZE;
3083
3084         ifm = nlmsg_data(nlh);
3085         ifm->ifi_family = AF_BRIDGE;
3086         ifm->__ifi_pad = 0;
3087         ifm->ifi_type = dev->type;
3088         ifm->ifi_index = dev->ifindex;
3089         ifm->ifi_flags = dev_get_flags(dev);
3090         ifm->ifi_change = 0;
3091
3092
3093         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3094             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3095             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3096             (br_dev &&
3097              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3098             (dev->addr_len &&
3099              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3100             (dev->ifindex != dev_get_iflink(dev) &&
3101              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3102                 goto nla_put_failure;
3103
3104         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3105         if (!br_afspec)
3106                 goto nla_put_failure;
3107
3108         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3109                 nla_nest_cancel(skb, br_afspec);
3110                 goto nla_put_failure;
3111         }
3112
3113         if (mode != BRIDGE_MODE_UNDEF) {
3114                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3115                         nla_nest_cancel(skb, br_afspec);
3116                         goto nla_put_failure;
3117                 }
3118         }
3119         if (vlan_fill) {
3120                 err = vlan_fill(skb, dev, filter_mask);
3121                 if (err) {
3122                         nla_nest_cancel(skb, br_afspec);
3123                         goto nla_put_failure;
3124                 }
3125         }
3126         nla_nest_end(skb, br_afspec);
3127
3128         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3129         if (!protinfo)
3130                 goto nla_put_failure;
3131
3132         if (brport_nla_put_flag(skb, flags, mask,
3133                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3134             brport_nla_put_flag(skb, flags, mask,
3135                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3136             brport_nla_put_flag(skb, flags, mask,
3137                                 IFLA_BRPORT_FAST_LEAVE,
3138                                 BR_MULTICAST_FAST_LEAVE) ||
3139             brport_nla_put_flag(skb, flags, mask,
3140                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3141             brport_nla_put_flag(skb, flags, mask,
3142                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3143             brport_nla_put_flag(skb, flags, mask,
3144                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3145             brport_nla_put_flag(skb, flags, mask,
3146                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3147             brport_nla_put_flag(skb, flags, mask,
3148                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3149                 nla_nest_cancel(skb, protinfo);
3150                 goto nla_put_failure;
3151         }
3152
3153         nla_nest_end(skb, protinfo);
3154
3155         nlmsg_end(skb, nlh);
3156         return 0;
3157 nla_put_failure:
3158         nlmsg_cancel(skb, nlh);
3159         return err ? err : -EMSGSIZE;
3160 }
3161 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3162
3163 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3164 {
3165         struct net *net = sock_net(skb->sk);
3166         struct net_device *dev;
3167         int idx = 0;
3168         u32 portid = NETLINK_CB(cb->skb).portid;
3169         u32 seq = cb->nlh->nlmsg_seq;
3170         u32 filter_mask = 0;
3171         int err;
3172
3173         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3174                 struct nlattr *extfilt;
3175
3176                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3177                                           IFLA_EXT_MASK);
3178                 if (extfilt) {
3179                         if (nla_len(extfilt) < sizeof(filter_mask))
3180                                 return -EINVAL;
3181
3182                         filter_mask = nla_get_u32(extfilt);
3183                 }
3184         }
3185
3186         rcu_read_lock();
3187         for_each_netdev_rcu(net, dev) {
3188                 const struct net_device_ops *ops = dev->netdev_ops;
3189                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3190
3191                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3192                         if (idx >= cb->args[0]) {
3193                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3194                                                 skb, portid, seq, dev,
3195                                                 filter_mask, NLM_F_MULTI);
3196                                 if (err < 0 && err != -EOPNOTSUPP)
3197                                         break;
3198                         }
3199                         idx++;
3200                 }
3201
3202                 if (ops->ndo_bridge_getlink) {
3203                         if (idx >= cb->args[0]) {
3204                                 err = ops->ndo_bridge_getlink(skb, portid,
3205                                                               seq, dev,
3206                                                               filter_mask,
3207                                                               NLM_F_MULTI);
3208                                 if (err < 0 && err != -EOPNOTSUPP)
3209                                         break;
3210                         }
3211                         idx++;
3212                 }
3213         }
3214         rcu_read_unlock();
3215         cb->args[0] = idx;
3216
3217         return skb->len;
3218 }
3219
3220 static inline size_t bridge_nlmsg_size(void)
3221 {
3222         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3223                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3224                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3225                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3226                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3227                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3228                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3229                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3230                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3231                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3232                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3233 }
3234
3235 static int rtnl_bridge_notify(struct net_device *dev)
3236 {
3237         struct net *net = dev_net(dev);
3238         struct sk_buff *skb;
3239         int err = -EOPNOTSUPP;
3240
3241         if (!dev->netdev_ops->ndo_bridge_getlink)
3242                 return 0;
3243
3244         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3245         if (!skb) {
3246                 err = -ENOMEM;
3247                 goto errout;
3248         }
3249
3250         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3251         if (err < 0)
3252                 goto errout;
3253
3254         if (!skb->len)
3255                 goto errout;
3256
3257         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3258         return 0;
3259 errout:
3260         WARN_ON(err == -EMSGSIZE);
3261         kfree_skb(skb);
3262         if (err)
3263                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3264         return err;
3265 }
3266
3267 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3268 {
3269         struct net *net = sock_net(skb->sk);
3270         struct ifinfomsg *ifm;
3271         struct net_device *dev;
3272         struct nlattr *br_spec, *attr = NULL;
3273         int rem, err = -EOPNOTSUPP;
3274         u16 flags = 0;
3275         bool have_flags = false;
3276
3277         if (nlmsg_len(nlh) < sizeof(*ifm))
3278                 return -EINVAL;
3279
3280         ifm = nlmsg_data(nlh);
3281         if (ifm->ifi_family != AF_BRIDGE)
3282                 return -EPFNOSUPPORT;
3283
3284         dev = __dev_get_by_index(net, ifm->ifi_index);
3285         if (!dev) {
3286                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3287                 return -ENODEV;
3288         }
3289
3290         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3291         if (br_spec) {
3292                 nla_for_each_nested(attr, br_spec, rem) {
3293                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3294                                 if (nla_len(attr) < sizeof(flags))
3295                                         return -EINVAL;
3296
3297                                 have_flags = true;
3298                                 flags = nla_get_u16(attr);
3299                                 break;
3300                         }
3301                 }
3302         }
3303
3304         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3305                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3306
3307                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3308                         err = -EOPNOTSUPP;
3309                         goto out;
3310                 }
3311
3312                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3313                 if (err)
3314                         goto out;
3315
3316                 flags &= ~BRIDGE_FLAGS_MASTER;
3317         }
3318
3319         if ((flags & BRIDGE_FLAGS_SELF)) {
3320                 if (!dev->netdev_ops->ndo_bridge_setlink)
3321                         err = -EOPNOTSUPP;
3322                 else
3323                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3324                                                                   flags);
3325                 if (!err) {
3326                         flags &= ~BRIDGE_FLAGS_SELF;
3327
3328                         /* Generate event to notify upper layer of bridge
3329                          * change
3330                          */
3331                         err = rtnl_bridge_notify(dev);
3332                 }
3333         }
3334
3335         if (have_flags)
3336                 memcpy(nla_data(attr), &flags, sizeof(flags));
3337 out:
3338         return err;
3339 }
3340
3341 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3342 {
3343         struct net *net = sock_net(skb->sk);
3344         struct ifinfomsg *ifm;
3345         struct net_device *dev;
3346         struct nlattr *br_spec, *attr = NULL;
3347         int rem, err = -EOPNOTSUPP;
3348         u16 flags = 0;
3349         bool have_flags = false;
3350
3351         if (nlmsg_len(nlh) < sizeof(*ifm))
3352                 return -EINVAL;
3353
3354         ifm = nlmsg_data(nlh);
3355         if (ifm->ifi_family != AF_BRIDGE)
3356                 return -EPFNOSUPPORT;
3357
3358         dev = __dev_get_by_index(net, ifm->ifi_index);
3359         if (!dev) {
3360                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3361                 return -ENODEV;
3362         }
3363
3364         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3365         if (br_spec) {
3366                 nla_for_each_nested(attr, br_spec, rem) {
3367                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3368                                 if (nla_len(attr) < sizeof(flags))
3369                                         return -EINVAL;
3370
3371                                 have_flags = true;
3372                                 flags = nla_get_u16(attr);
3373                                 break;
3374                         }
3375                 }
3376         }
3377
3378         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3379                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3380
3381                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3382                         err = -EOPNOTSUPP;
3383                         goto out;
3384                 }
3385
3386                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3387                 if (err)
3388                         goto out;
3389
3390                 flags &= ~BRIDGE_FLAGS_MASTER;
3391         }
3392
3393         if ((flags & BRIDGE_FLAGS_SELF)) {
3394                 if (!dev->netdev_ops->ndo_bridge_dellink)
3395                         err = -EOPNOTSUPP;
3396                 else
3397                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3398                                                                   flags);
3399
3400                 if (!err) {
3401                         flags &= ~BRIDGE_FLAGS_SELF;
3402
3403                         /* Generate event to notify upper layer of bridge
3404                          * change
3405                          */
3406                         err = rtnl_bridge_notify(dev);
3407                 }
3408         }
3409
3410         if (have_flags)
3411                 memcpy(nla_data(attr), &flags, sizeof(flags));
3412 out:
3413         return err;
3414 }
3415
3416 /* Process one rtnetlink message. */
3417
3418 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3419 {
3420         struct net *net = sock_net(skb->sk);
3421         rtnl_doit_func doit;
3422         int kind;
3423         int family;
3424         int type;
3425         int err;
3426
3427         type = nlh->nlmsg_type;
3428         if (type > RTM_MAX)
3429                 return -EOPNOTSUPP;
3430
3431         type -= RTM_BASE;
3432
3433         /* All the messages must have at least 1 byte length */
3434         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3435                 return 0;
3436
3437         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3438         kind = type&3;
3439
3440         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3441                 return -EPERM;
3442
3443         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3444                 struct sock *rtnl;
3445                 rtnl_dumpit_func dumpit;
3446                 rtnl_calcit_func calcit;
3447                 u16 min_dump_alloc = 0;
3448
3449                 dumpit = rtnl_get_dumpit(family, type);
3450                 if (dumpit == NULL)
3451                         return -EOPNOTSUPP;
3452                 calcit = rtnl_get_calcit(family, type);
3453                 if (calcit)
3454                         min_dump_alloc = calcit(skb, nlh);
3455
3456                 __rtnl_unlock();
3457                 rtnl = net->rtnl;
3458                 {
3459                         struct netlink_dump_control c = {
3460                                 .dump           = dumpit,
3461                                 .min_dump_alloc = min_dump_alloc,
3462                         };
3463                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3464                 }
3465                 rtnl_lock();
3466                 return err;
3467         }
3468
3469         doit = rtnl_get_doit(family, type);
3470         if (doit == NULL)
3471                 return -EOPNOTSUPP;
3472
3473         return doit(skb, nlh);
3474 }
3475
3476 static void rtnetlink_rcv(struct sk_buff *skb)
3477 {
3478         rtnl_lock();
3479         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3480         rtnl_unlock();
3481 }
3482
3483 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3484 {
3485         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3486
3487         switch (event) {
3488         case NETDEV_UP:
3489         case NETDEV_DOWN:
3490         case NETDEV_PRE_UP:
3491         case NETDEV_POST_INIT:
3492         case NETDEV_REGISTER:
3493         case NETDEV_CHANGE:
3494         case NETDEV_PRE_TYPE_CHANGE:
3495         case NETDEV_GOING_DOWN:
3496         case NETDEV_UNREGISTER:
3497         case NETDEV_UNREGISTER_FINAL:
3498         case NETDEV_RELEASE:
3499         case NETDEV_JOIN:
3500         case NETDEV_BONDING_INFO:
3501                 break;
3502         default:
3503                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3504                 break;
3505         }
3506         return NOTIFY_DONE;
3507 }
3508
3509 static struct notifier_block rtnetlink_dev_notifier = {
3510         .notifier_call  = rtnetlink_event,
3511 };
3512
3513
3514 static int __net_init rtnetlink_net_init(struct net *net)
3515 {
3516         struct sock *sk;
3517         struct netlink_kernel_cfg cfg = {
3518                 .groups         = RTNLGRP_MAX,
3519                 .input          = rtnetlink_rcv,
3520                 .cb_mutex       = &rtnl_mutex,
3521                 .flags          = NL_CFG_F_NONROOT_RECV,
3522         };
3523
3524         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3525         if (!sk)
3526                 return -ENOMEM;
3527         net->rtnl = sk;
3528         return 0;
3529 }
3530
3531 static void __net_exit rtnetlink_net_exit(struct net *net)
3532 {
3533         netlink_kernel_release(net->rtnl);
3534         net->rtnl = NULL;
3535 }
3536
3537 static struct pernet_operations rtnetlink_net_ops = {
3538         .init = rtnetlink_net_init,
3539         .exit = rtnetlink_net_exit,
3540 };
3541
3542 void __init rtnetlink_init(void)
3543 {
3544         if (register_pernet_subsys(&rtnetlink_net_ops))
3545                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3546
3547         register_netdevice_notifier(&rtnetlink_dev_notifier);
3548
3549         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3550                       rtnl_dump_ifinfo, rtnl_calcit);
3551         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3552         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3553         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3554
3555         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3556         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3557
3558         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3559         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3560         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3561
3562         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3563         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3564         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3565 }