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
 809 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 810 {
 811         memcpy(v, b, sizeof(*b));
 812 }
 813
 814 /* All VF info */
 815 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 816                                    u32 ext_filter_mask)
 817 {
 818         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 819             (ext_filter_mask & RTEXT_FILTER_VF)) {
 820                 int num_vfs = dev_num_vf(dev->dev.parent);
 821                 size_t size = nla_total_size(sizeof(struct nlattr));
 822                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 823                 size += num_vfs *
 824                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 825                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 826                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 827                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 828                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 829                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 830                          /* IFLA_VF_STATS_RX_PACKETS */
 831                          nla_total_size(sizeof(__u64)) +
 832                          /* IFLA_VF_STATS_TX_PACKETS */
 833                          nla_total_size(sizeof(__u64)) +
 834                          /* IFLA_VF_STATS_RX_BYTES */
 835                          nla_total_size(sizeof(__u64)) +
 836                          /* IFLA_VF_STATS_TX_BYTES */
 837                          nla_total_size(sizeof(__u64)) +
 838                          /* IFLA_VF_STATS_BROADCAST */
 839                          nla_total_size(sizeof(__u64)) +
 840                          /* IFLA_VF_STATS_MULTICAST */
 841                          nla_total_size(sizeof(__u64)) +
 842                          nla_total_size(sizeof(struct ifla_vf_trust)));
 843                 return size;
 844         } else
 845                 return 0;
 846 }
 847
 848 static size_t rtnl_port_size(const struct net_device *dev,
 849                              u32 ext_filter_mask)
 850 {
 851         size_t port_size = nla_total_size(4)            /* PORT_VF */
 852                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 853                 + nla_total_size(sizeof(struct ifla_port_vsi))
 854                                                         /* PORT_VSI_TYPE */
 855                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 856                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 857                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 858                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 859         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 860         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 861                 + port_size;
 862         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 863                 + port_size;
 864
 865         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 866             !(ext_filter_mask & RTEXT_FILTER_VF))
 867                 return 0;
 868         if (dev_num_vf(dev->dev.parent))
 869                 return port_self_size + vf_ports_size +
 870                         vf_port_size * dev_num_vf(dev->dev.parent);
 871         else
 872                 return port_self_size;
 873 }
 874
 875 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 876                                      u32 ext_filter_mask)
 877 {
 878         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 879                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 880                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 881                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 882                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 883                + nla_total_size(sizeof(struct rtnl_link_stats))
 884                + nla_total_size(sizeof(struct rtnl_link_stats64))
 885                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 886                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 887                + nla_total_size(4) /* IFLA_TXQLEN */
 888                + nla_total_size(4) /* IFLA_WEIGHT */
 889                + nla_total_size(4) /* IFLA_MTU */
 890                + nla_total_size(4) /* IFLA_LINK */
 891                + nla_total_size(4) /* IFLA_MASTER */
 892                + nla_total_size(1) /* IFLA_CARRIER */
 893                + nla_total_size(4) /* IFLA_PROMISCUITY */
 894                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 895                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 896                + nla_total_size(1) /* IFLA_OPERSTATE */
 897                + nla_total_size(1) /* IFLA_LINKMODE */
 898                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 899                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 900                + nla_total_size(ext_filter_mask
 901                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 902                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 903                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 904                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 905                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 906                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 907                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 908                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 909
 910 }
 911
 912 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 913 {
 914         struct nlattr *vf_ports;
 915         struct nlattr *vf_port;
 916         int vf;
 917         int err;
 918
 919         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 920         if (!vf_ports)
 921                 return -EMSGSIZE;
 922
 923         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 924                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 925                 if (!vf_port)
 926                         goto nla_put_failure;
 927                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 928                         goto nla_put_failure;
 929                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 930                 if (err == -EMSGSIZE)
 931                         goto nla_put_failure;
 932                 if (err) {
 933                         nla_nest_cancel(skb, vf_port);
 934                         continue;
 935                 }
 936                 nla_nest_end(skb, vf_port);
 937         }
 938
 939         nla_nest_end(skb, vf_ports);
 940
 941         return 0;
 942
 943 nla_put_failure:
 944         nla_nest_cancel(skb, vf_ports);
 945         return -EMSGSIZE;
 946 }
 947
 948 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 949 {
 950         struct nlattr *port_self;
 951         int err;
 952
 953         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 954         if (!port_self)
 955                 return -EMSGSIZE;
 956
 957         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 958         if (err) {
 959                 nla_nest_cancel(skb, port_self);
 960                 return (err == -EMSGSIZE) ? err : 0;
 961         }
 962
 963         nla_nest_end(skb, port_self);
 964
 965         return 0;
 966 }
 967
 968 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 969                           u32 ext_filter_mask)
 970 {
 971         int err;
 972
 973         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 974             !(ext_filter_mask & RTEXT_FILTER_VF))
 975                 return 0;
 976
 977         err = rtnl_port_self_fill(skb, dev);
 978         if (err)
 979                 return err;
 980
 981         if (dev_num_vf(dev->dev.parent)) {
 982                 err = rtnl_vf_ports_fill(skb, dev);
 983                 if (err)
 984                         return err;
 985         }
 986
 987         return 0;
 988 }
 989
 990 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 991 {
 992         int err;
 993         struct netdev_phys_item_id ppid;
 994
 995         err = dev_get_phys_port_id(dev, &ppid);
 996         if (err) {
 997                 if (err == -EOPNOTSUPP)
 998                         return 0;
 999                 return err;
1000         }
1001
1002         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1003                 return -EMSGSIZE;
1004
1005         return 0;
1006 }
1007
1008 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1009 {
1010         char name[IFNAMSIZ];
1011         int err;
1012
1013         err = dev_get_phys_port_name(dev, name, sizeof(name));
1014         if (err) {
1015                 if (err == -EOPNOTSUPP)
1016                         return 0;
1017                 return err;
1018         }
1019
1020         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1021                 return -EMSGSIZE;
1022
1023         return 0;
1024 }
1025
1026 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028         int err;
1029         struct switchdev_attr attr = {
1030                 .orig_dev = dev,
1031                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1032                 .flags = SWITCHDEV_F_NO_RECURSE,
1033         };
1034
1035         err = switchdev_port_attr_get(dev, &attr);
1036         if (err) {
1037                 if (err == -EOPNOTSUPP)
1038                         return 0;
1039                 return err;
1040         }
1041
1042         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1043                     attr.u.ppid.id))
1044                 return -EMSGSIZE;
1045
1046         return 0;
1047 }
1048
1049 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1050                                               struct net_device *dev)
1051 {
1052         const struct rtnl_link_stats64 *stats;
1053         struct rtnl_link_stats64 temp;
1054         struct nlattr *attr;
1055
1056         stats = dev_get_stats(dev, &temp);
1057
1058         attr = nla_reserve(skb, IFLA_STATS,
1059                            sizeof(struct rtnl_link_stats));
1060         if (!attr)
1061                 return -EMSGSIZE;
1062
1063         copy_rtnl_link_stats(nla_data(attr), stats);
1064
1065         attr = nla_reserve(skb, IFLA_STATS64,
1066                            sizeof(struct rtnl_link_stats64));
1067         if (!attr)
1068                 return -EMSGSIZE;
1069
1070         copy_rtnl_link_stats64(nla_data(attr), stats);
1071
1072         return 0;
1073 }
1074
1075 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1076                                                struct net_device *dev,
1077                                                int vfs_num,
1078                                                struct nlattr *vfinfo)
1079 {
1080         struct ifla_vf_rss_query_en vf_rss_query_en;
1081         struct ifla_vf_link_state vf_linkstate;
1082         struct ifla_vf_spoofchk vf_spoofchk;
1083         struct ifla_vf_tx_rate vf_tx_rate;
1084         struct ifla_vf_stats vf_stats;
1085         struct ifla_vf_trust vf_trust;
1086         struct ifla_vf_vlan vf_vlan;
1087         struct ifla_vf_rate vf_rate;
1088         struct nlattr *vf, *vfstats;
1089         struct ifla_vf_mac vf_mac;
1090         struct ifla_vf_info ivi;
1091
1092         /* Not all SR-IOV capable drivers support the
1093          * spoofcheck and "RSS query enable" query.  Preset to
1094          * -1 so the user space tool can detect that the driver
1095          * didn't report anything.
1096          */
1097         ivi.spoofchk = -1;
1098         ivi.rss_query_en = -1;
1099         ivi.trusted = -1;
1100         memset(ivi.mac, 0, sizeof(ivi.mac));
1101         /* The default value for VF link state is "auto"
1102          * IFLA_VF_LINK_STATE_AUTO which equals zero
1103          */
1104         ivi.linkstate = 0;
1105         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1106                 return 0;
1107
1108         vf_mac.vf =
1109                 vf_vlan.vf =
1110                 vf_rate.vf =
1111                 vf_tx_rate.vf =
1112                 vf_spoofchk.vf =
1113                 vf_linkstate.vf =
1114                 vf_rss_query_en.vf =
1115                 vf_trust.vf = ivi.vf;
1116
1117         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1118         vf_vlan.vlan = ivi.vlan;
1119         vf_vlan.qos = ivi.qos;
1120         vf_tx_rate.rate = ivi.max_tx_rate;
1121         vf_rate.min_tx_rate = ivi.min_tx_rate;
1122         vf_rate.max_tx_rate = ivi.max_tx_rate;
1123         vf_spoofchk.setting = ivi.spoofchk;
1124         vf_linkstate.link_state = ivi.linkstate;
1125         vf_rss_query_en.setting = ivi.rss_query_en;
1126         vf_trust.setting = ivi.trusted;
1127         vf = nla_nest_start(skb, IFLA_VF_INFO);
1128         if (!vf) {
1129                 nla_nest_cancel(skb, vfinfo);
1130                 return -EMSGSIZE;
1131         }
1132         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1133             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1134             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1135                     &vf_rate) ||
1136             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1137                     &vf_tx_rate) ||
1138             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1139                     &vf_spoofchk) ||
1140             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1141                     &vf_linkstate) ||
1142             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1143                     sizeof(vf_rss_query_en),
1144                     &vf_rss_query_en) ||
1145             nla_put(skb, IFLA_VF_TRUST,
1146                     sizeof(vf_trust), &vf_trust))
1147                 return -EMSGSIZE;
1148         memset(&vf_stats, 0, sizeof(vf_stats));
1149         if (dev->netdev_ops->ndo_get_vf_stats)
1150                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1151                                                 &vf_stats);
1152         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1153         if (!vfstats) {
1154                 nla_nest_cancel(skb, vf);
1155                 nla_nest_cancel(skb, vfinfo);
1156                 return -EMSGSIZE;
1157         }
1158         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1159                         vf_stats.rx_packets) ||
1160             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1161                         vf_stats.tx_packets) ||
1162             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1163                         vf_stats.rx_bytes) ||
1164             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1165                         vf_stats.tx_bytes) ||
1166             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1167                         vf_stats.broadcast) ||
1168             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1169                         vf_stats.multicast))
1170                 return -EMSGSIZE;
1171         nla_nest_end(skb, vfstats);
1172         nla_nest_end(skb, vf);
1173         return 0;
1174 }
1175
1176 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1177 {
1178         struct rtnl_link_ifmap map = {
1179                 .mem_start   = dev->mem_start,
1180                 .mem_end     = dev->mem_end,
1181                 .base_addr   = dev->base_addr,
1182                 .irq         = dev->irq,
1183                 .dma         = dev->dma,
1184                 .port        = dev->if_port,
1185         };
1186         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1187                 return -EMSGSIZE;
1188
1189         return 0;
1190 }
1191
1192 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1193                             int type, u32 pid, u32 seq, u32 change,
1194                             unsigned int flags, u32 ext_filter_mask)
1195 {
1196         struct ifinfomsg *ifm;
1197         struct nlmsghdr *nlh;
1198         struct nlattr *af_spec;
1199         struct rtnl_af_ops *af_ops;
1200         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1201
1202         ASSERT_RTNL();
1203         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1204         if (nlh == NULL)
1205                 return -EMSGSIZE;
1206
1207         ifm = nlmsg_data(nlh);
1208         ifm->ifi_family = AF_UNSPEC;
1209         ifm->__ifi_pad = 0;
1210         ifm->ifi_type = dev->type;
1211         ifm->ifi_index = dev->ifindex;
1212         ifm->ifi_flags = dev_get_flags(dev);
1213         ifm->ifi_change = change;
1214
1215         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1216             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1217             nla_put_u8(skb, IFLA_OPERSTATE,
1218                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1219             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1220             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1221             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1222             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1223             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1224 #ifdef CONFIG_RPS
1225             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1226 #endif
1227             (dev->ifindex != dev_get_iflink(dev) &&
1228              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1229             (upper_dev &&
1230              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1231             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1232             (dev->qdisc &&
1233              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1234             (dev->ifalias &&
1235              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1236             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1237                         atomic_read(&dev->carrier_changes)) ||
1238             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1239                 goto nla_put_failure;
1240
1241         if (rtnl_fill_link_ifmap(skb, dev))
1242                 goto nla_put_failure;
1243
1244         if (dev->addr_len) {
1245                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1246                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1247                         goto nla_put_failure;
1248         }
1249
1250         if (rtnl_phys_port_id_fill(skb, dev))
1251                 goto nla_put_failure;
1252
1253         if (rtnl_phys_port_name_fill(skb, dev))
1254                 goto nla_put_failure;
1255
1256         if (rtnl_phys_switch_id_fill(skb, dev))
1257                 goto nla_put_failure;
1258
1259         if (rtnl_fill_stats(skb, dev))
1260                 goto nla_put_failure;
1261
1262         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1263             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1264                 goto nla_put_failure;
1265
1266         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1267             ext_filter_mask & RTEXT_FILTER_VF) {
1268                 int i;
1269                 struct nlattr *vfinfo;
1270                 int num_vfs = dev_num_vf(dev->dev.parent);
1271
1272                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1273                 if (!vfinfo)
1274                         goto nla_put_failure;
1275                 for (i = 0; i < num_vfs; i++) {
1276                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1277                                 goto nla_put_failure;
1278                 }
1279
1280                 nla_nest_end(skb, vfinfo);
1281         }
1282
1283         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1284                 goto nla_put_failure;
1285
1286         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1287                 if (rtnl_link_fill(skb, dev) < 0)
1288                         goto nla_put_failure;
1289         }
1290
1291         if (dev->rtnl_link_ops &&
1292             dev->rtnl_link_ops->get_link_net) {
1293                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1294
1295                 if (!net_eq(dev_net(dev), link_net)) {
1296                         int id = peernet2id_alloc(dev_net(dev), link_net);
1297
1298                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1299                                 goto nla_put_failure;
1300                 }
1301         }
1302
1303         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1304                 goto nla_put_failure;
1305
1306         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1307                 if (af_ops->fill_link_af) {
1308                         struct nlattr *af;
1309                         int err;
1310
1311                         if (!(af = nla_nest_start(skb, af_ops->family)))
1312                                 goto nla_put_failure;
1313
1314                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1315
1316                         /*
1317                          * Caller may return ENODATA to indicate that there
1318                          * was no data to be dumped. This is not an error, it
1319                          * means we should trim the attribute header and
1320                          * continue.
1321                          */
1322                         if (err == -ENODATA)
1323                                 nla_nest_cancel(skb, af);
1324                         else if (err < 0)
1325                                 goto nla_put_failure;
1326
1327                         nla_nest_end(skb, af);
1328                 }
1329         }
1330
1331         nla_nest_end(skb, af_spec);
1332
1333         nlmsg_end(skb, nlh);
1334         return 0;
1335
1336 nla_put_failure:
1337         nlmsg_cancel(skb, nlh);
1338         return -EMSGSIZE;
1339 }
1340
1341 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1342         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1343         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1344         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1345         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1346         [IFLA_MTU]              = { .type = NLA_U32 },
1347         [IFLA_LINK]             = { .type = NLA_U32 },
1348         [IFLA_MASTER]           = { .type = NLA_U32 },
1349         [IFLA_CARRIER]          = { .type = NLA_U8 },
1350         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1351         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1352         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1353         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1354         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1355         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1356         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1357         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1358         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1359         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1360         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1361         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1362         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1363         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1364         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1365         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1366         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1367         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1368         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1369         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1370         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1371 };
1372
1373 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1374         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1375         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1376         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1377         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1378 };
1379
1380 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1381         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1382         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1383         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1384         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1385         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1386         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1387         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1388         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1389         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1390 };
1391
1392 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1393         [IFLA_VF_STATS_RX_PACKETS]      = { .type = NLA_U64 },
1394         [IFLA_VF_STATS_TX_PACKETS]      = { .type = NLA_U64 },
1395         [IFLA_VF_STATS_RX_BYTES]        = { .type = NLA_U64 },
1396         [IFLA_VF_STATS_TX_BYTES]        = { .type = NLA_U64 },
1397         [IFLA_VF_STATS_BROADCAST]       = { .type = NLA_U64 },
1398         [IFLA_VF_STATS_MULTICAST]       = { .type = NLA_U64 },
1399 };
1400
1401 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1402         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1403         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1404                                     .len = PORT_PROFILE_MAX },
1405         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1406                                     .len = sizeof(struct ifla_port_vsi)},
1407         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1408                                       .len = PORT_UUID_MAX },
1409         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1410                                     .len = PORT_UUID_MAX },
1411         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1412         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1413 };
1414
1415 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1416 {
1417         struct net *net = sock_net(skb->sk);
1418         int h, s_h;
1419         int idx = 0, s_idx;
1420         struct net_device *dev;
1421         struct hlist_head *head;
1422         struct nlattr *tb[IFLA_MAX+1];
1423         u32 ext_filter_mask = 0;
1424         int err;
1425         int hdrlen;
1426
1427         s_h = cb->args[0];
1428         s_idx = cb->args[1];
1429
1430         cb->seq = net->dev_base_seq;
1431
1432         /* A hack to preserve kernel<->userspace interface.
1433          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1434          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1435          * what iproute2 < v3.9.0 used.
1436          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1437          * attribute, its netlink message is shorter than struct ifinfomsg.
1438          */
1439         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1440                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1441
1442         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1443
1444                 if (tb[IFLA_EXT_MASK])
1445                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1446         }
1447
1448         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1449                 idx = 0;
1450                 head = &net->dev_index_head[h];
1451                 hlist_for_each_entry(dev, head, index_hlist) {
1452                         if (idx < s_idx)
1453                                 goto cont;
1454                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1455                                                NETLINK_CB(cb->skb).portid,
1456                                                cb->nlh->nlmsg_seq, 0,
1457                                                NLM_F_MULTI,
1458                                                ext_filter_mask);
1459                         /* If we ran out of room on the first message,
1460                          * we're in trouble
1461                          */
1462                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1463
1464                         if (err < 0)
1465                                 goto out;
1466
1467                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1468 cont:
1469                         idx++;
1470                 }
1471         }
1472 out:
1473         cb->args[1] = idx;
1474         cb->args[0] = h;
1475
1476         return skb->len;
1477 }
1478
1479 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1480 {
1481         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1482 }
1483 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1484
1485 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1486 {
1487         struct net *net;
1488         /* Examine the link attributes and figure out which
1489          * network namespace we are talking about.
1490          */
1491         if (tb[IFLA_NET_NS_PID])
1492                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1493         else if (tb[IFLA_NET_NS_FD])
1494                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1495         else
1496                 net = get_net(src_net);
1497         return net;
1498 }
1499 EXPORT_SYMBOL(rtnl_link_get_net);
1500
1501 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1502 {
1503         if (dev) {
1504                 if (tb[IFLA_ADDRESS] &&
1505                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1506                         return -EINVAL;
1507
1508                 if (tb[IFLA_BROADCAST] &&
1509                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1510                         return -EINVAL;
1511         }
1512
1513         if (tb[IFLA_AF_SPEC]) {
1514                 struct nlattr *af;
1515                 int rem, err;
1516
1517                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1518                         const struct rtnl_af_ops *af_ops;
1519
1520                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1521                                 return -EAFNOSUPPORT;
1522
1523                         if (!af_ops->set_link_af)
1524                                 return -EOPNOTSUPP;
1525
1526                         if (af_ops->validate_link_af) {
1527                                 err = af_ops->validate_link_af(dev, af);
1528                                 if (err < 0)
1529                                         return err;
1530                         }
1531                 }
1532         }
1533
1534         return 0;
1535 }
1536
1537 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1538 {
1539         const struct net_device_ops *ops = dev->netdev_ops;
1540         int err = -EINVAL;
1541
1542         if (tb[IFLA_VF_MAC]) {
1543                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1544
1545                 err = -EOPNOTSUPP;
1546                 if (ops->ndo_set_vf_mac)
1547                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1548                                                   ivm->mac);
1549                 if (err < 0)
1550                         return err;
1551         }
1552
1553         if (tb[IFLA_VF_VLAN]) {
1554                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1555
1556                 err = -EOPNOTSUPP;
1557                 if (ops->ndo_set_vf_vlan)
1558                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1559                                                    ivv->qos);
1560                 if (err < 0)
1561                         return err;
1562         }
1563
1564         if (tb[IFLA_VF_TX_RATE]) {
1565                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1566                 struct ifla_vf_info ivf;
1567
1568                 err = -EOPNOTSUPP;
1569                 if (ops->ndo_get_vf_config)
1570                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1571                 if (err < 0)
1572                         return err;
1573
1574                 err = -EOPNOTSUPP;
1575                 if (ops->ndo_set_vf_rate)
1576                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1577                                                    ivf.min_tx_rate,
1578                                                    ivt->rate);
1579                 if (err < 0)
1580                         return err;
1581         }
1582
1583         if (tb[IFLA_VF_RATE]) {
1584                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1585
1586                 err = -EOPNOTSUPP;
1587                 if (ops->ndo_set_vf_rate)
1588                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1589                                                    ivt->min_tx_rate,
1590                                                    ivt->max_tx_rate);
1591                 if (err < 0)
1592                         return err;
1593         }
1594
1595         if (tb[IFLA_VF_SPOOFCHK]) {
1596                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1597
1598                 err = -EOPNOTSUPP;
1599                 if (ops->ndo_set_vf_spoofchk)
1600                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1601                                                        ivs->setting);
1602                 if (err < 0)
1603                         return err;
1604         }
1605
1606         if (tb[IFLA_VF_LINK_STATE]) {
1607                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1608
1609                 err = -EOPNOTSUPP;
1610                 if (ops->ndo_set_vf_link_state)
1611                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1612                                                          ivl->link_state);
1613                 if (err < 0)
1614                         return err;
1615         }
1616
1617         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1618                 struct ifla_vf_rss_query_en *ivrssq_en;
1619
1620                 err = -EOPNOTSUPP;
1621                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1622                 if (ops->ndo_set_vf_rss_query_en)
1623                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1624                                                            ivrssq_en->setting);
1625                 if (err < 0)
1626                         return err;
1627         }
1628
1629         if (tb[IFLA_VF_TRUST]) {
1630                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1631
1632                 err = -EOPNOTSUPP;
1633                 if (ops->ndo_set_vf_trust)
1634                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1635                 if (err < 0)
1636                         return err;
1637         }
1638
1639         return err;
1640 }
1641
1642 static int do_set_master(struct net_device *dev, int ifindex)
1643 {
1644         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1645         const struct net_device_ops *ops;
1646         int err;
1647
1648         if (upper_dev) {
1649                 if (upper_dev->ifindex == ifindex)
1650                         return 0;
1651                 ops = upper_dev->netdev_ops;
1652                 if (ops->ndo_del_slave) {
1653                         err = ops->ndo_del_slave(upper_dev, dev);
1654                         if (err)
1655                                 return err;
1656                 } else {
1657                         return -EOPNOTSUPP;
1658                 }
1659         }
1660
1661         if (ifindex) {
1662                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1663                 if (!upper_dev)
1664                         return -EINVAL;
1665                 ops = upper_dev->netdev_ops;
1666                 if (ops->ndo_add_slave) {
1667                         err = ops->ndo_add_slave(upper_dev, dev);
1668                         if (err)
1669                                 return err;
1670                 } else {
1671                         return -EOPNOTSUPP;
1672                 }
1673         }
1674         return 0;
1675 }
1676
1677 #define DO_SETLINK_MODIFIED     0x01
1678 /* notify flag means notify + modified. */
1679 #define DO_SETLINK_NOTIFY       0x03
1680 static int do_setlink(const struct sk_buff *skb,
1681                       struct net_device *dev, struct ifinfomsg *ifm,
1682                       struct nlattr **tb, char *ifname, int status)
1683 {
1684         const struct net_device_ops *ops = dev->netdev_ops;
1685         int err;
1686
1687         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1688                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1689                 if (IS_ERR(net)) {
1690                         err = PTR_ERR(net);
1691                         goto errout;
1692                 }
1693                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1694                         put_net(net);
1695                         err = -EPERM;
1696                         goto errout;
1697                 }
1698                 err = dev_change_net_namespace(dev, net, ifname);
1699                 put_net(net);
1700                 if (err)
1701                         goto errout;
1702                 status |= DO_SETLINK_MODIFIED;
1703         }
1704
1705         if (tb[IFLA_MAP]) {
1706                 struct rtnl_link_ifmap *u_map;
1707                 struct ifmap k_map;
1708
1709                 if (!ops->ndo_set_config) {
1710                         err = -EOPNOTSUPP;
1711                         goto errout;
1712                 }
1713
1714                 if (!netif_device_present(dev)) {
1715                         err = -ENODEV;
1716                         goto errout;
1717                 }
1718
1719                 u_map = nla_data(tb[IFLA_MAP]);
1720                 k_map.mem_start = (unsigned long) u_map->mem_start;
1721                 k_map.mem_end = (unsigned long) u_map->mem_end;
1722                 k_map.base_addr = (unsigned short) u_map->base_addr;
1723                 k_map.irq = (unsigned char) u_map->irq;
1724                 k_map.dma = (unsigned char) u_map->dma;
1725                 k_map.port = (unsigned char) u_map->port;
1726
1727                 err = ops->ndo_set_config(dev, &k_map);
1728                 if (err < 0)
1729                         goto errout;
1730
1731                 status |= DO_SETLINK_NOTIFY;
1732         }
1733
1734         if (tb[IFLA_ADDRESS]) {
1735                 struct sockaddr *sa;
1736                 int len;
1737
1738                 len = sizeof(sa_family_t) + dev->addr_len;
1739                 sa = kmalloc(len, GFP_KERNEL);
1740                 if (!sa) {
1741                         err = -ENOMEM;
1742                         goto errout;
1743                 }
1744                 sa->sa_family = dev->type;
1745                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1746                        dev->addr_len);
1747                 err = dev_set_mac_address(dev, sa);
1748                 kfree(sa);
1749                 if (err)
1750                         goto errout;
1751                 status |= DO_SETLINK_MODIFIED;
1752         }
1753
1754         if (tb[IFLA_MTU]) {
1755                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1756                 if (err < 0)
1757                         goto errout;
1758                 status |= DO_SETLINK_MODIFIED;
1759         }
1760
1761         if (tb[IFLA_GROUP]) {
1762                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1763                 status |= DO_SETLINK_NOTIFY;
1764         }
1765
1766         /*
1767          * Interface selected by interface index but interface
1768          * name provided implies that a name change has been
1769          * requested.
1770          */
1771         if (ifm->ifi_index > 0 && ifname[0]) {
1772                 err = dev_change_name(dev, ifname);
1773                 if (err < 0)
1774                         goto errout;
1775                 status |= DO_SETLINK_MODIFIED;
1776         }
1777
1778         if (tb[IFLA_IFALIAS]) {
1779                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1780                                     nla_len(tb[IFLA_IFALIAS]));
1781                 if (err < 0)
1782                         goto errout;
1783                 status |= DO_SETLINK_NOTIFY;
1784         }
1785
1786         if (tb[IFLA_BROADCAST]) {
1787                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1788                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1789         }
1790
1791         if (ifm->ifi_flags || ifm->ifi_change) {
1792                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1793                 if (err < 0)
1794                         goto errout;
1795         }
1796
1797         if (tb[IFLA_MASTER]) {
1798                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1799                 if (err)
1800                         goto errout;
1801                 status |= DO_SETLINK_MODIFIED;
1802         }
1803
1804         if (tb[IFLA_CARRIER]) {
1805                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1806                 if (err)
1807                         goto errout;
1808                 status |= DO_SETLINK_MODIFIED;
1809         }
1810
1811         if (tb[IFLA_TXQLEN]) {
1812                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1813
1814                 if (dev->tx_queue_len ^ value)
1815                         status |= DO_SETLINK_NOTIFY;
1816
1817                 dev->tx_queue_len = value;
1818         }
1819
1820         if (tb[IFLA_OPERSTATE])
1821                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1822
1823         if (tb[IFLA_LINKMODE]) {
1824                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1825
1826                 write_lock_bh(&dev_base_lock);
1827                 if (dev->link_mode ^ value)
1828                         status |= DO_SETLINK_NOTIFY;
1829                 dev->link_mode = value;
1830                 write_unlock_bh(&dev_base_lock);
1831         }
1832
1833         if (tb[IFLA_VFINFO_LIST]) {
1834                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1835                 struct nlattr *attr;
1836                 int rem;
1837
1838                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1839                         if (nla_type(attr) != IFLA_VF_INFO ||
1840                             nla_len(attr) < NLA_HDRLEN) {
1841                                 err = -EINVAL;
1842                                 goto errout;
1843                         }
1844                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1845                                                ifla_vf_policy);
1846                         if (err < 0)
1847                                 goto errout;
1848                         err = do_setvfinfo(dev, vfinfo);
1849                         if (err < 0)
1850                                 goto errout;
1851                         status |= DO_SETLINK_NOTIFY;
1852                 }
1853         }
1854         err = 0;
1855
1856         if (tb[IFLA_VF_PORTS]) {
1857                 struct nlattr *port[IFLA_PORT_MAX+1];
1858                 struct nlattr *attr;
1859                 int vf;
1860                 int rem;
1861
1862                 err = -EOPNOTSUPP;
1863                 if (!ops->ndo_set_vf_port)
1864                         goto errout;
1865
1866                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1867                         if (nla_type(attr) != IFLA_VF_PORT ||
1868                             nla_len(attr) < NLA_HDRLEN) {
1869                                 err = -EINVAL;
1870                                 goto errout;
1871                         }
1872                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1873                                                ifla_port_policy);
1874                         if (err < 0)
1875                                 goto errout;
1876                         if (!port[IFLA_PORT_VF]) {
1877                                 err = -EOPNOTSUPP;
1878                                 goto errout;
1879                         }
1880                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1881                         err = ops->ndo_set_vf_port(dev, vf, port);
1882                         if (err < 0)
1883                                 goto errout;
1884                         status |= DO_SETLINK_NOTIFY;
1885                 }
1886         }
1887         err = 0;
1888
1889         if (tb[IFLA_PORT_SELF]) {
1890                 struct nlattr *port[IFLA_PORT_MAX+1];
1891
1892                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1893                         tb[IFLA_PORT_SELF], ifla_port_policy);
1894                 if (err < 0)
1895                         goto errout;
1896
1897                 err = -EOPNOTSUPP;
1898                 if (ops->ndo_set_vf_port)
1899                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1900                 if (err < 0)
1901                         goto errout;
1902                 status |= DO_SETLINK_NOTIFY;
1903         }
1904
1905         if (tb[IFLA_AF_SPEC]) {
1906                 struct nlattr *af;
1907                 int rem;
1908
1909                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1910                         const struct rtnl_af_ops *af_ops;
1911
1912                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1913                                 BUG();
1914
1915                         err = af_ops->set_link_af(dev, af);
1916                         if (err < 0)
1917                                 goto errout;
1918
1919                         status |= DO_SETLINK_NOTIFY;
1920                 }
1921         }
1922         err = 0;
1923
1924         if (tb[IFLA_PROTO_DOWN]) {
1925                 err = dev_change_proto_down(dev,
1926                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
1927                 if (err)
1928                         goto errout;
1929                 status |= DO_SETLINK_NOTIFY;
1930         }
1931
1932 errout:
1933         if (status & DO_SETLINK_MODIFIED) {
1934                 if (status & DO_SETLINK_NOTIFY)
1935                         netdev_state_change(dev);
1936
1937                 if (err < 0)
1938                         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",
1939                                              dev->name);
1940         }
1941
1942         return err;
1943 }
1944
1945 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1946 {
1947         struct net *net = sock_net(skb->sk);
1948         struct ifinfomsg *ifm;
1949         struct net_device *dev;
1950         int err;
1951         struct nlattr *tb[IFLA_MAX+1];
1952         char ifname[IFNAMSIZ];
1953
1954         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1955         if (err < 0)
1956                 goto errout;
1957
1958         if (tb[IFLA_IFNAME])
1959                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1960         else
1961                 ifname[0] = '\0';
1962
1963         err = -EINVAL;
1964         ifm = nlmsg_data(nlh);
1965         if (ifm->ifi_index > 0)
1966                 dev = __dev_get_by_index(net, ifm->ifi_index);
1967         else if (tb[IFLA_IFNAME])
1968                 dev = __dev_get_by_name(net, ifname);
1969         else
1970                 goto errout;
1971
1972         if (dev == NULL) {
1973                 err = -ENODEV;
1974                 goto errout;
1975         }
1976
1977         err = validate_linkmsg(dev, tb);
1978         if (err < 0)
1979                 goto errout;
1980
1981         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1982 errout:
1983         return err;
1984 }
1985
1986 static int rtnl_group_dellink(const struct net *net, int group)
1987 {
1988         struct net_device *dev, *aux;
1989         LIST_HEAD(list_kill);
1990         bool found = false;
1991
1992         if (!group)
1993                 return -EPERM;
1994
1995         for_each_netdev(net, dev) {
1996                 if (dev->group == group) {
1997                         const struct rtnl_link_ops *ops;
1998
1999                         found = true;
2000                         ops = dev->rtnl_link_ops;
2001                         if (!ops || !ops->dellink)
2002                                 return -EOPNOTSUPP;
2003                 }
2004         }
2005
2006         if (!found)
2007                 return -ENODEV;
2008
2009         for_each_netdev_safe(net, dev, aux) {
2010                 if (dev->group == group) {
2011                         const struct rtnl_link_ops *ops;
2012
2013                         ops = dev->rtnl_link_ops;
2014                         ops->dellink(dev, &list_kill);
2015                 }
2016         }
2017         unregister_netdevice_many(&list_kill);
2018
2019         return 0;
2020 }
2021
2022 int rtnl_delete_link(struct net_device *dev)
2023 {
2024         const struct rtnl_link_ops *ops;
2025         LIST_HEAD(list_kill);
2026
2027         ops = dev->rtnl_link_ops;
2028         if (!ops || !ops->dellink)
2029                 return -EOPNOTSUPP;
2030
2031         ops->dellink(dev, &list_kill);
2032         unregister_netdevice_many(&list_kill);
2033
2034         return 0;
2035 }
2036 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2037
2038 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2039 {
2040         struct net *net = sock_net(skb->sk);
2041         struct net_device *dev;
2042         struct ifinfomsg *ifm;
2043         char ifname[IFNAMSIZ];
2044         struct nlattr *tb[IFLA_MAX+1];
2045         int err;
2046
2047         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2048         if (err < 0)
2049                 return err;
2050
2051         if (tb[IFLA_IFNAME])
2052                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2053
2054         ifm = nlmsg_data(nlh);
2055         if (ifm->ifi_index > 0)
2056                 dev = __dev_get_by_index(net, ifm->ifi_index);
2057         else if (tb[IFLA_IFNAME])
2058                 dev = __dev_get_by_name(net, ifname);
2059         else if (tb[IFLA_GROUP])
2060                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2061         else
2062                 return -EINVAL;
2063
2064         if (!dev)
2065                 return -ENODEV;
2066
2067         return rtnl_delete_link(dev);
2068 }
2069
2070 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2071 {
2072         unsigned int old_flags;
2073         int err;
2074
2075         old_flags = dev->flags;
2076         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2077                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2078                 if (err < 0)
2079                         return err;
2080         }
2081
2082         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2083
2084         __dev_notify_flags(dev, old_flags, ~0U);
2085         return 0;
2086 }
2087 EXPORT_SYMBOL(rtnl_configure_link);
2088
2089 struct net_device *rtnl_create_link(struct net *net,
2090         const char *ifname, unsigned char name_assign_type,
2091         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2092 {
2093         int err;
2094         struct net_device *dev;
2095         unsigned int num_tx_queues = 1;
2096         unsigned int num_rx_queues = 1;
2097
2098         if (tb[IFLA_NUM_TX_QUEUES])
2099                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2100         else if (ops->get_num_tx_queues)
2101                 num_tx_queues = ops->get_num_tx_queues();
2102
2103         if (tb[IFLA_NUM_RX_QUEUES])
2104                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2105         else if (ops->get_num_rx_queues)
2106                 num_rx_queues = ops->get_num_rx_queues();
2107
2108         err = -ENOMEM;
2109         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2110                                ops->setup, num_tx_queues, num_rx_queues);
2111         if (!dev)
2112                 goto err;
2113
2114         dev_net_set(dev, net);
2115         dev->rtnl_link_ops = ops;
2116         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2117
2118         if (tb[IFLA_MTU])
2119                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2120         if (tb[IFLA_ADDRESS]) {
2121                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2122                                 nla_len(tb[IFLA_ADDRESS]));
2123                 dev->addr_assign_type = NET_ADDR_SET;
2124         }
2125         if (tb[IFLA_BROADCAST])
2126                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2127                                 nla_len(tb[IFLA_BROADCAST]));
2128         if (tb[IFLA_TXQLEN])
2129                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2130         if (tb[IFLA_OPERSTATE])
2131                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2132         if (tb[IFLA_LINKMODE])
2133                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2134         if (tb[IFLA_GROUP])
2135                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2136
2137         return dev;
2138
2139 err:
2140         return ERR_PTR(err);
2141 }
2142 EXPORT_SYMBOL(rtnl_create_link);
2143
2144 static int rtnl_group_changelink(const struct sk_buff *skb,
2145                 struct net *net, int group,
2146                 struct ifinfomsg *ifm,
2147                 struct nlattr **tb)
2148 {
2149         struct net_device *dev, *aux;
2150         int err;
2151
2152         for_each_netdev_safe(net, dev, aux) {
2153                 if (dev->group == group) {
2154                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2155                         if (err < 0)
2156                                 return err;
2157                 }
2158         }
2159
2160         return 0;
2161 }
2162
2163 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2164 {
2165         struct net *net = sock_net(skb->sk);
2166         const struct rtnl_link_ops *ops;
2167         const struct rtnl_link_ops *m_ops = NULL;
2168         struct net_device *dev;
2169         struct net_device *master_dev = NULL;
2170         struct ifinfomsg *ifm;
2171         char kind[MODULE_NAME_LEN];
2172         char ifname[IFNAMSIZ];
2173         struct nlattr *tb[IFLA_MAX+1];
2174         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2175         unsigned char name_assign_type = NET_NAME_USER;
2176         int err;
2177
2178 #ifdef CONFIG_MODULES
2179 replay:
2180 #endif
2181         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2182         if (err < 0)
2183                 return err;
2184
2185         if (tb[IFLA_IFNAME])
2186                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2187         else
2188                 ifname[0] = '\0';
2189
2190         ifm = nlmsg_data(nlh);
2191         if (ifm->ifi_index > 0)
2192                 dev = __dev_get_by_index(net, ifm->ifi_index);
2193         else {
2194                 if (ifname[0])
2195                         dev = __dev_get_by_name(net, ifname);
2196                 else
2197                         dev = NULL;
2198         }
2199
2200         if (dev) {
2201                 master_dev = netdev_master_upper_dev_get(dev);
2202                 if (master_dev)
2203                         m_ops = master_dev->rtnl_link_ops;
2204         }
2205
2206         err = validate_linkmsg(dev, tb);
2207         if (err < 0)
2208                 return err;
2209
2210         if (tb[IFLA_LINKINFO]) {
2211                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2212                                        tb[IFLA_LINKINFO], ifla_info_policy);
2213                 if (err < 0)
2214                         return err;
2215         } else
2216                 memset(linkinfo, 0, sizeof(linkinfo));
2217
2218         if (linkinfo[IFLA_INFO_KIND]) {
2219                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2220                 ops = rtnl_link_ops_get(kind);
2221         } else {
2222                 kind[0] = '\0';
2223                 ops = NULL;
2224         }
2225
2226         if (1) {
2227                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2228                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2229                 struct nlattr **data = NULL;
2230                 struct nlattr **slave_data = NULL;
2231                 struct net *dest_net, *link_net = NULL;
2232
2233                 if (ops) {
2234                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2235                                 err = nla_parse_nested(attr, ops->maxtype,
2236                                                        linkinfo[IFLA_INFO_DATA],
2237                                                        ops->policy);
2238                                 if (err < 0)
2239                                         return err;
2240                                 data = attr;
2241                         }
2242                         if (ops->validate) {
2243                                 err = ops->validate(tb, data);
2244                                 if (err < 0)
2245                                         return err;
2246                         }
2247                 }
2248
2249                 if (m_ops) {
2250                         if (m_ops->slave_maxtype &&
2251                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2252                                 err = nla_parse_nested(slave_attr,
2253                                                        m_ops->slave_maxtype,
2254                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2255                                                        m_ops->slave_policy);
2256                                 if (err < 0)
2257                                         return err;
2258                                 slave_data = slave_attr;
2259                         }
2260                         if (m_ops->slave_validate) {
2261                                 err = m_ops->slave_validate(tb, slave_data);
2262                                 if (err < 0)
2263                                         return err;
2264                         }
2265                 }
2266
2267                 if (dev) {
2268                         int status = 0;
2269
2270                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2271                                 return -EEXIST;
2272                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2273                                 return -EOPNOTSUPP;
2274
2275                         if (linkinfo[IFLA_INFO_DATA]) {
2276                                 if (!ops || ops != dev->rtnl_link_ops ||
2277                                     !ops->changelink)
2278                                         return -EOPNOTSUPP;
2279
2280                                 err = ops->changelink(dev, tb, data);
2281                                 if (err < 0)
2282                                         return err;
2283                                 status |= DO_SETLINK_NOTIFY;
2284                         }
2285
2286                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2287                                 if (!m_ops || !m_ops->slave_changelink)
2288                                         return -EOPNOTSUPP;
2289
2290                                 err = m_ops->slave_changelink(master_dev, dev,
2291                                                               tb, slave_data);
2292                                 if (err < 0)
2293                                         return err;
2294                                 status |= DO_SETLINK_NOTIFY;
2295                         }
2296
2297                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2298                 }
2299
2300                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2301                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2302                                 return rtnl_group_changelink(skb, net,
2303                                                 nla_get_u32(tb[IFLA_GROUP]),
2304                                                 ifm, tb);
2305                         return -ENODEV;
2306                 }
2307
2308                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2309                         return -EOPNOTSUPP;
2310
2311                 if (!ops) {
2312 #ifdef CONFIG_MODULES
2313                         if (kind[0]) {
2314                                 __rtnl_unlock();
2315                                 request_module("rtnl-link-%s", kind);
2316                                 rtnl_lock();
2317                                 ops = rtnl_link_ops_get(kind);
2318                                 if (ops)
2319                                         goto replay;
2320                         }
2321 #endif
2322                         return -EOPNOTSUPP;
2323                 }
2324
2325                 if (!ops->setup)
2326                         return -EOPNOTSUPP;
2327
2328                 if (!ifname[0]) {
2329                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2330                         name_assign_type = NET_NAME_ENUM;
2331                 }
2332
2333                 dest_net = rtnl_link_get_net(net, tb);
2334                 if (IS_ERR(dest_net))
2335                         return PTR_ERR(dest_net);
2336
2337                 err = -EPERM;
2338                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2339                         goto out;
2340
2341                 if (tb[IFLA_LINK_NETNSID]) {
2342                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2343
2344                         link_net = get_net_ns_by_id(dest_net, id);
2345                         if (!link_net) {
2346                                 err =  -EINVAL;
2347                                 goto out;
2348                         }
2349                         err = -EPERM;
2350                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2351                                 goto out;
2352                 }
2353
2354                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2355                                        name_assign_type, ops, tb);
2356                 if (IS_ERR(dev)) {
2357                         err = PTR_ERR(dev);
2358                         goto out;
2359                 }
2360
2361                 dev->ifindex = ifm->ifi_index;
2362
2363                 if (ops->newlink) {
2364                         err = ops->newlink(link_net ? : net, dev, tb, data);
2365                         /* Drivers should call free_netdev() in ->destructor
2366                          * and unregister it on failure after registration
2367                          * so that device could be finally freed in rtnl_unlock.
2368                          */
2369                         if (err < 0) {
2370                                 /* If device is not registered at all, free it now */
2371                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2372                                         free_netdev(dev);
2373                                 goto out;
2374                         }
2375                 } else {
2376                         err = register_netdevice(dev);
2377                         if (err < 0) {
2378                                 free_netdev(dev);
2379                                 goto out;
2380                         }
2381                 }
2382                 err = rtnl_configure_link(dev, ifm);
2383                 if (err < 0)
2384                         goto out_unregister;
2385                 if (link_net) {
2386                         err = dev_change_net_namespace(dev, dest_net, ifname);
2387                         if (err < 0)
2388                                 goto out_unregister;
2389                 }
2390 out:
2391                 if (link_net)
2392                         put_net(link_net);
2393                 put_net(dest_net);
2394                 return err;
2395 out_unregister:
2396                 if (ops->newlink) {
2397                         LIST_HEAD(list_kill);
2398
2399                         ops->dellink(dev, &list_kill);
2400                         unregister_netdevice_many(&list_kill);
2401                 } else {
2402                         unregister_netdevice(dev);
2403                 }
2404                 goto out;
2405         }
2406 }
2407
2408 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2409 {
2410         struct net *net = sock_net(skb->sk);
2411         struct ifinfomsg *ifm;
2412         char ifname[IFNAMSIZ];
2413         struct nlattr *tb[IFLA_MAX+1];
2414         struct net_device *dev = NULL;
2415         struct sk_buff *nskb;
2416         int err;
2417         u32 ext_filter_mask = 0;
2418
2419         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2420         if (err < 0)
2421                 return err;
2422
2423         if (tb[IFLA_IFNAME])
2424                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2425
2426         if (tb[IFLA_EXT_MASK])
2427                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2428
2429         ifm = nlmsg_data(nlh);
2430         if (ifm->ifi_index > 0)
2431                 dev = __dev_get_by_index(net, ifm->ifi_index);
2432         else if (tb[IFLA_IFNAME])
2433                 dev = __dev_get_by_name(net, ifname);
2434         else
2435                 return -EINVAL;
2436
2437         if (dev == NULL)
2438                 return -ENODEV;
2439
2440         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2441         if (nskb == NULL)
2442                 return -ENOBUFS;
2443
2444         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2445                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2446         if (err < 0) {
2447                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2448                 WARN_ON(err == -EMSGSIZE);
2449                 kfree_skb(nskb);
2450         } else
2451                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2452
2453         return err;
2454 }
2455
2456 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2457 {
2458         struct net *net = sock_net(skb->sk);
2459         struct net_device *dev;
2460         struct nlattr *tb[IFLA_MAX+1];
2461         u32 ext_filter_mask = 0;
2462         u16 min_ifinfo_dump_size = 0;
2463         int hdrlen;
2464
2465         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2466         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2467                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2468
2469         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2470                 if (tb[IFLA_EXT_MASK])
2471                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2472         }
2473
2474         if (!ext_filter_mask)
2475                 return NLMSG_GOODSIZE;
2476         /*
2477          * traverse the list of net devices and compute the minimum
2478          * buffer size based upon the filter mask.
2479          */
2480         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2481                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2482                                              if_nlmsg_size(dev,
2483                                                            ext_filter_mask));
2484         }
2485
2486         return min_ifinfo_dump_size;
2487 }
2488
2489 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2490 {
2491         int idx;
2492         int s_idx = cb->family;
2493
2494         if (s_idx == 0)
2495                 s_idx = 1;
2496         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2497                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2498                 if (idx < s_idx || idx == PF_PACKET)
2499                         continue;
2500                 if (rtnl_msg_handlers[idx] == NULL ||
2501                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2502                         continue;
2503                 if (idx > s_idx) {
2504                         memset(&cb->args[0], 0, sizeof(cb->args));
2505                         cb->prev_seq = 0;
2506                         cb->seq = 0;
2507                 }
2508                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2509                         break;
2510         }
2511         cb->family = idx;
2512
2513         return skb->len;
2514 }
2515
2516 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2517                                        unsigned int change, gfp_t flags)
2518 {
2519         struct net *net = dev_net(dev);
2520         struct sk_buff *skb;
2521         int err = -ENOBUFS;
2522         size_t if_info_size;
2523
2524         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2525         if (skb == NULL)
2526                 goto errout;
2527
2528         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2529         if (err < 0) {
2530                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2531                 WARN_ON(err == -EMSGSIZE);
2532                 kfree_skb(skb);
2533                 goto errout;
2534         }
2535         return skb;
2536 errout:
2537         if (err < 0)
2538                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2539         return NULL;
2540 }
2541
2542 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2543 {
2544         struct net *net = dev_net(dev);
2545
2546         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2547 }
2548
2549 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2550                   gfp_t flags)
2551 {
2552         struct sk_buff *skb;
2553
2554         if (dev->reg_state != NETREG_REGISTERED)
2555                 return;
2556
2557         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2558         if (skb)
2559                 rtmsg_ifinfo_send(skb, dev, flags);
2560 }
2561 EXPORT_SYMBOL(rtmsg_ifinfo);
2562
2563 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2564                                    struct net_device *dev,
2565                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2566                                    int type, unsigned int flags,
2567                                    int nlflags, u16 ndm_state)
2568 {
2569         struct nlmsghdr *nlh;
2570         struct ndmsg *ndm;
2571
2572         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2573         if (!nlh)
2574                 return -EMSGSIZE;
2575
2576         ndm = nlmsg_data(nlh);
2577         ndm->ndm_family  = AF_BRIDGE;
2578         ndm->ndm_pad1    = 0;
2579         ndm->ndm_pad2    = 0;
2580         ndm->ndm_flags   = flags;
2581         ndm->ndm_type    = 0;
2582         ndm->ndm_ifindex = dev->ifindex;
2583         ndm->ndm_state   = ndm_state;
2584
2585         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2586                 goto nla_put_failure;
2587         if (vid)
2588                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2589                         goto nla_put_failure;
2590
2591         nlmsg_end(skb, nlh);
2592         return 0;
2593
2594 nla_put_failure:
2595         nlmsg_cancel(skb, nlh);
2596         return -EMSGSIZE;
2597 }
2598
2599 static inline size_t rtnl_fdb_nlmsg_size(void)
2600 {
2601         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2602 }
2603
2604 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2605                             u16 ndm_state)
2606 {
2607         struct net *net = dev_net(dev);
2608         struct sk_buff *skb;
2609         int err = -ENOBUFS;
2610
2611         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2612         if (!skb)
2613                 goto errout;
2614
2615         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2616                                       0, 0, type, NTF_SELF, 0, ndm_state);
2617         if (err < 0) {
2618                 kfree_skb(skb);
2619                 goto errout;
2620         }
2621
2622         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2623         return;
2624 errout:
2625         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2626 }
2627
2628 /**
2629  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2630  */
2631 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2632                      struct nlattr *tb[],
2633                      struct net_device *dev,
2634                      const unsigned char *addr, u16 vid,
2635                      u16 flags)
2636 {
2637         int err = -EINVAL;
2638
2639         /* If aging addresses are supported device will need to
2640          * implement its own handler for this.
2641          */
2642         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2643                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2644                 return err;
2645         }
2646
2647         if (vid) {
2648                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2649                 return err;
2650         }
2651
2652         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2653                 err = dev_uc_add_excl(dev, addr);
2654         else if (is_multicast_ether_addr(addr))
2655                 err = dev_mc_add_excl(dev, addr);
2656
2657         /* Only return duplicate errors if NLM_F_EXCL is set */
2658         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2659                 err = 0;
2660
2661         return err;
2662 }
2663 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2664
2665 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2666 {
2667         u16 vid = 0;
2668
2669         if (vlan_attr) {
2670                 if (nla_len(vlan_attr) != sizeof(u16)) {
2671                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2672                         return -EINVAL;
2673                 }
2674
2675                 vid = nla_get_u16(vlan_attr);
2676
2677                 if (!vid || vid >= VLAN_VID_MASK) {
2678                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2679                                 vid);
2680                         return -EINVAL;
2681                 }
2682         }
2683         *p_vid = vid;
2684         return 0;
2685 }
2686
2687 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2688 {
2689         struct net *net = sock_net(skb->sk);
2690         struct ndmsg *ndm;
2691         struct nlattr *tb[NDA_MAX+1];
2692         struct net_device *dev;
2693         u8 *addr;
2694         u16 vid;
2695         int err;
2696
2697         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2698         if (err < 0)
2699                 return err;
2700
2701         ndm = nlmsg_data(nlh);
2702         if (ndm->ndm_ifindex == 0) {
2703                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2704                 return -EINVAL;
2705         }
2706
2707         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2708         if (dev == NULL) {
2709                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2710                 return -ENODEV;
2711         }
2712
2713         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2714                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2715                 return -EINVAL;
2716         }
2717
2718         addr = nla_data(tb[NDA_LLADDR]);
2719
2720         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2721         if (err)
2722                 return err;
2723
2724         err = -EOPNOTSUPP;
2725
2726         /* Support fdb on master device the net/bridge default case */
2727         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2728             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2729                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2730                 const struct net_device_ops *ops = br_dev->netdev_ops;
2731
2732                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2733                                        nlh->nlmsg_flags);
2734                 if (err)
2735                         goto out;
2736                 else
2737                         ndm->ndm_flags &= ~NTF_MASTER;
2738         }
2739
2740         /* Embedded bridge, macvlan, and any other device support */
2741         if ((ndm->ndm_flags & NTF_SELF)) {
2742                 if (dev->netdev_ops->ndo_fdb_add)
2743                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2744                                                            vid,
2745                                                            nlh->nlmsg_flags);
2746                 else
2747                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2748                                                nlh->nlmsg_flags);
2749
2750                 if (!err) {
2751                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2752                                         ndm->ndm_state);
2753                         ndm->ndm_flags &= ~NTF_SELF;
2754                 }
2755         }
2756 out:
2757         return err;
2758 }
2759
2760 /**
2761  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2762  */
2763 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2764                      struct nlattr *tb[],
2765                      struct net_device *dev,
2766                      const unsigned char *addr, u16 vid)
2767 {
2768         int err = -EINVAL;
2769
2770         /* If aging addresses are supported device will need to
2771          * implement its own handler for this.
2772          */
2773         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2774                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2775                 return err;
2776         }
2777
2778         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2779                 err = dev_uc_del(dev, addr);
2780         else if (is_multicast_ether_addr(addr))
2781                 err = dev_mc_del(dev, addr);
2782
2783         return err;
2784 }
2785 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2786
2787 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2788 {
2789         struct net *net = sock_net(skb->sk);
2790         struct ndmsg *ndm;
2791         struct nlattr *tb[NDA_MAX+1];
2792         struct net_device *dev;
2793         int err = -EINVAL;
2794         __u8 *addr;
2795         u16 vid;
2796
2797         if (!netlink_capable(skb, CAP_NET_ADMIN))
2798                 return -EPERM;
2799
2800         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2801         if (err < 0)
2802                 return err;
2803
2804         ndm = nlmsg_data(nlh);
2805         if (ndm->ndm_ifindex == 0) {
2806                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2807                 return -EINVAL;
2808         }
2809
2810         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2811         if (dev == NULL) {
2812                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2813                 return -ENODEV;
2814         }
2815
2816         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2817                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2818                 return -EINVAL;
2819         }
2820
2821         addr = nla_data(tb[NDA_LLADDR]);
2822
2823         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2824         if (err)
2825                 return err;
2826
2827         err = -EOPNOTSUPP;
2828
2829         /* Support fdb on master device the net/bridge default case */
2830         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2831             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2832                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2833                 const struct net_device_ops *ops = br_dev->netdev_ops;
2834
2835                 if (ops->ndo_fdb_del)
2836                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2837
2838                 if (err)
2839                         goto out;
2840                 else
2841                         ndm->ndm_flags &= ~NTF_MASTER;
2842         }
2843
2844         /* Embedded bridge, macvlan, and any other device support */
2845         if (ndm->ndm_flags & NTF_SELF) {
2846                 if (dev->netdev_ops->ndo_fdb_del)
2847                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2848                                                            vid);
2849                 else
2850                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2851
2852                 if (!err) {
2853                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2854                                         ndm->ndm_state);
2855                         ndm->ndm_flags &= ~NTF_SELF;
2856                 }
2857         }
2858 out:
2859         return err;
2860 }
2861
2862 static int nlmsg_populate_fdb(struct sk_buff *skb,
2863                               struct netlink_callback *cb,
2864                               struct net_device *dev,
2865                               int *idx,
2866                               struct netdev_hw_addr_list *list)
2867 {
2868         struct netdev_hw_addr *ha;
2869         int err;
2870         u32 portid, seq;
2871
2872         portid = NETLINK_CB(cb->skb).portid;
2873         seq = cb->nlh->nlmsg_seq;
2874
2875         list_for_each_entry(ha, &list->list, list) {
2876                 if (*idx < cb->args[0])
2877                         goto skip;
2878
2879                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2880                                               portid, seq,
2881                                               RTM_NEWNEIGH, NTF_SELF,
2882                                               NLM_F_MULTI, NUD_PERMANENT);
2883                 if (err < 0)
2884                         return err;
2885 skip:
2886                 *idx += 1;
2887         }
2888         return 0;
2889 }
2890
2891 /**
2892  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2893  * @nlh: netlink message header
2894  * @dev: netdevice
2895  *
2896  * Default netdevice operation to dump the existing unicast address list.
2897  * Returns number of addresses from list put in skb.
2898  */
2899 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2900                       struct netlink_callback *cb,
2901                       struct net_device *dev,
2902                       struct net_device *filter_dev,
2903                       int idx)
2904 {
2905         int err;
2906
2907         netif_addr_lock_bh(dev);
2908         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2909         if (err)
2910                 goto out;
2911         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2912 out:
2913         netif_addr_unlock_bh(dev);
2914         return idx;
2915 }
2916 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2917
2918 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2919 {
2920         struct net_device *dev;
2921         struct nlattr *tb[IFLA_MAX+1];
2922         struct net_device *br_dev = NULL;
2923         const struct net_device_ops *ops = NULL;
2924         const struct net_device_ops *cops = NULL;
2925         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2926         struct net *net = sock_net(skb->sk);
2927         int brport_idx = 0;
2928         int br_idx = 0;
2929         int idx = 0;
2930
2931         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2932                         ifla_policy) == 0) {
2933                 if (tb[IFLA_MASTER])
2934                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2935         }
2936
2937         brport_idx = ifm->ifi_index;
2938
2939         if (br_idx) {
2940                 br_dev = __dev_get_by_index(net, br_idx);
2941                 if (!br_dev)
2942                         return -ENODEV;
2943
2944                 ops = br_dev->netdev_ops;
2945         }
2946
2947         for_each_netdev(net, dev) {
2948                 if (brport_idx && (dev->ifindex != brport_idx))
2949                         continue;
2950
2951                 if (!br_idx) { /* user did not specify a specific bridge */
2952                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2953                                 br_dev = netdev_master_upper_dev_get(dev);
2954                                 cops = br_dev->netdev_ops;
2955                         }
2956
2957                 } else {
2958                         if (dev != br_dev &&
2959                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2960                                 continue;
2961
2962                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2963                             !(dev->priv_flags & IFF_EBRIDGE))
2964                                 continue;
2965
2966                         cops = ops;
2967                 }
2968
2969                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2970                         if (cops && cops->ndo_fdb_dump)
2971                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2972                                                          idx);
2973                 }
2974
2975                 if (dev->netdev_ops->ndo_fdb_dump)
2976                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2977                                                             idx);
2978                 else
2979                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2980
2981                 cops = NULL;
2982         }
2983
2984         cb->args[0] = idx;
2985         return skb->len;
2986 }
2987
2988 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2989                                unsigned int attrnum, unsigned int flag)
2990 {
2991         if (mask & flag)
2992                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2993         return 0;
2994 }
2995
2996 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2997                             struct net_device *dev, u16 mode,
2998                             u32 flags, u32 mask, int nlflags,
2999                             u32 filter_mask,
3000                             int (*vlan_fill)(struct sk_buff *skb,
3001                                              struct net_device *dev,
3002                                              u32 filter_mask))
3003 {
3004         struct nlmsghdr *nlh;
3005         struct ifinfomsg *ifm;
3006         struct nlattr *br_afspec;
3007         struct nlattr *protinfo;
3008         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3009         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3010         int err = 0;
3011
3012         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3013         if (nlh == NULL)
3014                 return -EMSGSIZE;
3015
3016         ifm = nlmsg_data(nlh);
3017         ifm->ifi_family = AF_BRIDGE;
3018         ifm->__ifi_pad = 0;
3019         ifm->ifi_type = dev->type;
3020         ifm->ifi_index = dev->ifindex;
3021         ifm->ifi_flags = dev_get_flags(dev);
3022         ifm->ifi_change = 0;
3023
3024
3025         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3026             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3027             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3028             (br_dev &&
3029              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3030             (dev->addr_len &&
3031              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3032             (dev->ifindex != dev_get_iflink(dev) &&
3033              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3034                 goto nla_put_failure;
3035
3036         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3037         if (!br_afspec)
3038                 goto nla_put_failure;
3039
3040         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3041                 nla_nest_cancel(skb, br_afspec);
3042                 goto nla_put_failure;
3043         }
3044
3045         if (mode != BRIDGE_MODE_UNDEF) {
3046                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3047                         nla_nest_cancel(skb, br_afspec);
3048                         goto nla_put_failure;
3049                 }
3050         }
3051         if (vlan_fill) {
3052                 err = vlan_fill(skb, dev, filter_mask);
3053                 if (err) {
3054                         nla_nest_cancel(skb, br_afspec);
3055                         goto nla_put_failure;
3056                 }
3057         }
3058         nla_nest_end(skb, br_afspec);
3059
3060         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3061         if (!protinfo)
3062                 goto nla_put_failure;
3063
3064         if (brport_nla_put_flag(skb, flags, mask,
3065                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3066             brport_nla_put_flag(skb, flags, mask,
3067                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3068             brport_nla_put_flag(skb, flags, mask,
3069                                 IFLA_BRPORT_FAST_LEAVE,
3070                                 BR_MULTICAST_FAST_LEAVE) ||
3071             brport_nla_put_flag(skb, flags, mask,
3072                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3073             brport_nla_put_flag(skb, flags, mask,
3074                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3075             brport_nla_put_flag(skb, flags, mask,
3076                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3077             brport_nla_put_flag(skb, flags, mask,
3078                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3079             brport_nla_put_flag(skb, flags, mask,
3080                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3081                 nla_nest_cancel(skb, protinfo);
3082                 goto nla_put_failure;
3083         }
3084
3085         nla_nest_end(skb, protinfo);
3086
3087         nlmsg_end(skb, nlh);
3088         return 0;
3089 nla_put_failure:
3090         nlmsg_cancel(skb, nlh);
3091         return err ? err : -EMSGSIZE;
3092 }
3093 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3094
3095 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3096 {
3097         struct net *net = sock_net(skb->sk);
3098         struct net_device *dev;
3099         int idx = 0;
3100         u32 portid = NETLINK_CB(cb->skb).portid;
3101         u32 seq = cb->nlh->nlmsg_seq;
3102         u32 filter_mask = 0;
3103         int err;
3104
3105         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3106                 struct nlattr *extfilt;
3107
3108                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3109                                           IFLA_EXT_MASK);
3110                 if (extfilt) {
3111                         if (nla_len(extfilt) < sizeof(filter_mask))
3112                                 return -EINVAL;
3113
3114                         filter_mask = nla_get_u32(extfilt);
3115                 }
3116         }
3117
3118         rcu_read_lock();
3119         for_each_netdev_rcu(net, dev) {
3120                 const struct net_device_ops *ops = dev->netdev_ops;
3121                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3122
3123                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3124                         if (idx >= cb->args[0]) {
3125                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3126                                                 skb, portid, seq, dev,
3127                                                 filter_mask, NLM_F_MULTI);
3128                                 if (err < 0 && err != -EOPNOTSUPP)
3129                                         break;
3130                         }
3131                         idx++;
3132                 }
3133
3134                 if (ops->ndo_bridge_getlink) {
3135                         if (idx >= cb->args[0]) {
3136                                 err = ops->ndo_bridge_getlink(skb, portid,
3137                                                               seq, dev,
3138                                                               filter_mask,
3139                                                               NLM_F_MULTI);
3140                                 if (err < 0 && err != -EOPNOTSUPP)
3141                                         break;
3142                         }
3143                         idx++;
3144                 }
3145         }
3146         rcu_read_unlock();
3147         cb->args[0] = idx;
3148
3149         return skb->len;
3150 }
3151
3152 static inline size_t bridge_nlmsg_size(void)
3153 {
3154         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3155                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3156                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3157                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3158                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3159                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3160                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3161                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3162                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3163                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3164                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3165 }
3166
3167 static int rtnl_bridge_notify(struct net_device *dev)
3168 {
3169         struct net *net = dev_net(dev);
3170         struct sk_buff *skb;
3171         int err = -EOPNOTSUPP;
3172
3173         if (!dev->netdev_ops->ndo_bridge_getlink)
3174                 return 0;
3175
3176         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3177         if (!skb) {
3178                 err = -ENOMEM;
3179                 goto errout;
3180         }
3181
3182         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3183         if (err < 0)
3184                 goto errout;
3185
3186         if (!skb->len)
3187                 goto errout;
3188
3189         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3190         return 0;
3191 errout:
3192         WARN_ON(err == -EMSGSIZE);
3193         kfree_skb(skb);
3194         if (err)
3195                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3196         return err;
3197 }
3198
3199 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3200 {
3201         struct net *net = sock_net(skb->sk);
3202         struct ifinfomsg *ifm;
3203         struct net_device *dev;
3204         struct nlattr *br_spec, *attr = NULL;
3205         int rem, err = -EOPNOTSUPP;
3206         u16 flags = 0;
3207         bool have_flags = false;
3208
3209         if (nlmsg_len(nlh) < sizeof(*ifm))
3210                 return -EINVAL;
3211
3212         ifm = nlmsg_data(nlh);
3213         if (ifm->ifi_family != AF_BRIDGE)
3214                 return -EPFNOSUPPORT;
3215
3216         dev = __dev_get_by_index(net, ifm->ifi_index);
3217         if (!dev) {
3218                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3219                 return -ENODEV;
3220         }
3221
3222         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3223         if (br_spec) {
3224                 nla_for_each_nested(attr, br_spec, rem) {
3225                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3226                                 if (nla_len(attr) < sizeof(flags))
3227                                         return -EINVAL;
3228
3229                                 have_flags = true;
3230                                 flags = nla_get_u16(attr);
3231                                 break;
3232                         }
3233                 }
3234         }
3235
3236         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3237                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3238
3239                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3240                         err = -EOPNOTSUPP;
3241                         goto out;
3242                 }
3243
3244                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3245                 if (err)
3246                         goto out;
3247
3248                 flags &= ~BRIDGE_FLAGS_MASTER;
3249         }
3250
3251         if ((flags & BRIDGE_FLAGS_SELF)) {
3252                 if (!dev->netdev_ops->ndo_bridge_setlink)
3253                         err = -EOPNOTSUPP;
3254                 else
3255                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3256                                                                   flags);
3257                 if (!err) {
3258                         flags &= ~BRIDGE_FLAGS_SELF;
3259
3260                         /* Generate event to notify upper layer of bridge
3261                          * change
3262                          */
3263                         err = rtnl_bridge_notify(dev);
3264                 }
3265         }
3266
3267         if (have_flags)
3268                 memcpy(nla_data(attr), &flags, sizeof(flags));
3269 out:
3270         return err;
3271 }
3272
3273 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3274 {
3275         struct net *net = sock_net(skb->sk);
3276         struct ifinfomsg *ifm;
3277         struct net_device *dev;
3278         struct nlattr *br_spec, *attr = NULL;
3279         int rem, err = -EOPNOTSUPP;
3280         u16 flags = 0;
3281         bool have_flags = false;
3282
3283         if (nlmsg_len(nlh) < sizeof(*ifm))
3284                 return -EINVAL;
3285
3286         ifm = nlmsg_data(nlh);
3287         if (ifm->ifi_family != AF_BRIDGE)
3288                 return -EPFNOSUPPORT;
3289
3290         dev = __dev_get_by_index(net, ifm->ifi_index);
3291         if (!dev) {
3292                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3293                 return -ENODEV;
3294         }
3295
3296         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3297         if (br_spec) {
3298                 nla_for_each_nested(attr, br_spec, rem) {
3299                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3300                                 if (nla_len(attr) < sizeof(flags))
3301                                         return -EINVAL;
3302
3303                                 have_flags = true;
3304                                 flags = nla_get_u16(attr);
3305                                 break;
3306                         }
3307                 }
3308         }
3309
3310         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3311                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3312
3313                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3314                         err = -EOPNOTSUPP;
3315                         goto out;
3316                 }
3317
3318                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3319                 if (err)
3320                         goto out;
3321
3322                 flags &= ~BRIDGE_FLAGS_MASTER;
3323         }
3324
3325         if ((flags & BRIDGE_FLAGS_SELF)) {
3326                 if (!dev->netdev_ops->ndo_bridge_dellink)
3327                         err = -EOPNOTSUPP;
3328                 else
3329                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3330                                                                   flags);
3331
3332                 if (!err) {
3333                         flags &= ~BRIDGE_FLAGS_SELF;
3334
3335                         /* Generate event to notify upper layer of bridge
3336                          * change
3337                          */
3338                         err = rtnl_bridge_notify(dev);
3339                 }
3340         }
3341
3342         if (have_flags)
3343                 memcpy(nla_data(attr), &flags, sizeof(flags));
3344 out:
3345         return err;
3346 }
3347
3348 /* Process one rtnetlink message. */
3349
3350 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3351 {
3352         struct net *net = sock_net(skb->sk);
3353         rtnl_doit_func doit;
3354         int kind;
3355         int family;
3356         int type;
3357         int err;
3358
3359         type = nlh->nlmsg_type;
3360         if (type > RTM_MAX)
3361                 return -EOPNOTSUPP;
3362
3363         type -= RTM_BASE;
3364
3365         /* All the messages must have at least 1 byte length */
3366         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3367                 return 0;
3368
3369         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3370         kind = type&3;
3371
3372         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3373                 return -EPERM;
3374
3375         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3376                 struct sock *rtnl;
3377                 rtnl_dumpit_func dumpit;
3378                 rtnl_calcit_func calcit;
3379                 u16 min_dump_alloc = 0;
3380
3381                 dumpit = rtnl_get_dumpit(family, type);
3382                 if (dumpit == NULL)
3383                         return -EOPNOTSUPP;
3384                 calcit = rtnl_get_calcit(family, type);
3385                 if (calcit)
3386                         min_dump_alloc = calcit(skb, nlh);
3387
3388                 __rtnl_unlock();
3389                 rtnl = net->rtnl;
3390                 {
3391                         struct netlink_dump_control c = {
3392                                 .dump           = dumpit,
3393                                 .min_dump_alloc = min_dump_alloc,
3394                         };
3395                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3396                 }
3397                 rtnl_lock();
3398                 return err;
3399         }
3400
3401         doit = rtnl_get_doit(family, type);
3402         if (doit == NULL)
3403                 return -EOPNOTSUPP;
3404
3405         return doit(skb, nlh);
3406 }
3407
3408 static void rtnetlink_rcv(struct sk_buff *skb)
3409 {
3410         rtnl_lock();
3411         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3412         rtnl_unlock();
3413 }
3414
3415 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3416 {
3417         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3418
3419         switch (event) {
3420         case NETDEV_UP:
3421         case NETDEV_DOWN:
3422         case NETDEV_PRE_UP:
3423         case NETDEV_POST_INIT:
3424         case NETDEV_REGISTER:
3425         case NETDEV_CHANGE:
3426         case NETDEV_PRE_TYPE_CHANGE:
3427         case NETDEV_GOING_DOWN:
3428         case NETDEV_UNREGISTER:
3429         case NETDEV_UNREGISTER_FINAL:
3430         case NETDEV_RELEASE:
3431         case NETDEV_JOIN:
3432         case NETDEV_BONDING_INFO:
3433                 break;
3434         default:
3435                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3436                 break;
3437         }
3438         return NOTIFY_DONE;
3439 }
3440
3441 static struct notifier_block rtnetlink_dev_notifier = {
3442         .notifier_call  = rtnetlink_event,
3443 };
3444
3445
3446 static int __net_init rtnetlink_net_init(struct net *net)
3447 {
3448         struct sock *sk;
3449         struct netlink_kernel_cfg cfg = {
3450                 .groups         = RTNLGRP_MAX,
3451                 .input          = rtnetlink_rcv,
3452                 .cb_mutex       = &rtnl_mutex,
3453                 .flags          = NL_CFG_F_NONROOT_RECV,
3454         };
3455
3456         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3457         if (!sk)
3458                 return -ENOMEM;
3459         net->rtnl = sk;
3460         return 0;
3461 }
3462
3463 static void __net_exit rtnetlink_net_exit(struct net *net)
3464 {
3465         netlink_kernel_release(net->rtnl);
3466         net->rtnl = NULL;
3467 }
3468
3469 static struct pernet_operations rtnetlink_net_ops = {
3470         .init = rtnetlink_net_init,
3471         .exit = rtnetlink_net_exit,
3472 };
3473
3474 void __init rtnetlink_init(void)
3475 {
3476         if (register_pernet_subsys(&rtnetlink_net_ops))
3477                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3478
3479         register_netdevice_notifier(&rtnetlink_dev_notifier);
3480
3481         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3482                       rtnl_dump_ifinfo, rtnl_calcit);
3483         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3484         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3485         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3486
3487         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3488         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3489
3490         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3491         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3492         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3493
3494         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3495         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3496         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3497 }