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