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