net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 bool lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256
 257         return 0;
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_unregister);
 260
 261 /**
 262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 263  * @protocol : Protocol family or PF_UNSPEC
 264  *
 265  * Identical to calling rtnl_unregster() for all registered message types
 266  * of a certain protocol family.
 267  */
 268 void rtnl_unregister_all(int protocol)
 269 {
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271
 272         kfree(rtnl_msg_handlers[protocol]);
 273         rtnl_msg_handlers[protocol] = NULL;
 274 }
 275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 276
 277 static LIST_HEAD(link_ops);
 278
 279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 280 {
 281         const struct rtnl_link_ops *ops;
 282
 283         list_for_each_entry(ops, &link_ops, list) {
 284                 if (!strcmp(ops->kind, kind))
 285                         return ops;
 286         }
 287         return NULL;
 288 }
 289
 290 /**
 291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 292  * @ops: struct rtnl_link_ops * to register
 293  *
 294  * The caller must hold the rtnl_mutex. This function should be used
 295  * by drivers that create devices during module initialization. It
 296  * must be called before registering the devices.
 297  *
 298  * Returns 0 on success or a negative error code.
 299  */
 300 int __rtnl_link_register(struct rtnl_link_ops *ops)
 301 {
 302         if (rtnl_link_ops_get(ops->kind))
 303                 return -EEXIST;
 304
 305         /* The check for setup is here because if ops
 306          * does not have that filled up, it is not possible
 307          * to use the ops for creating device. So do not
 308          * fill up dellink as well. That disables rtnl_dellink.
 309          */
 310         if (ops->setup && !ops->dellink)
 311                 ops->dellink = unregister_netdevice_queue;
 312
 313         list_add_tail(&ops->list, &link_ops);
 314         return 0;
 315 }
 316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 317
 318 /**
 319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 320  * @ops: struct rtnl_link_ops * to register
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_link_register(struct rtnl_link_ops *ops)
 325 {
 326         int err;
 327
 328         rtnl_lock();
 329         err = __rtnl_link_register(ops);
 330         rtnl_unlock();
 331         return err;
 332 }
 333 EXPORT_SYMBOL_GPL(rtnl_link_register);
 334
 335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 336 {
 337         struct net_device *dev;
 338         LIST_HEAD(list_kill);
 339
 340         for_each_netdev(net, dev) {
 341                 if (dev->rtnl_link_ops == ops)
 342                         ops->dellink(dev, &list_kill);
 343         }
 344         unregister_netdevice_many(&list_kill);
 345 }
 346
 347 /**
 348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 349  * @ops: struct rtnl_link_ops * to unregister
 350  *
 351  * The caller must hold the rtnl_mutex.
 352  */
 353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 354 {
 355         struct net *net;
 356
 357         for_each_net(net) {
 358                 __rtnl_kill_links(net, ops);
 359         }
 360         list_del(&ops->list);
 361 }
 362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 363
 364 /* Return with the rtnl_lock held when there are no network
 365  * devices unregistering in any network namespace.
 366  */
 367 static void rtnl_lock_unregistering_all(void)
 368 {
 369         struct net *net;
 370         bool unregistering;
 371         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 372
 373         add_wait_queue(&netdev_unregistering_wq, &wait);
 374         for (;;) {
 375                 unregistering = false;
 376                 rtnl_lock();
 377                 for_each_net(net) {
 378                         if (net->dev_unreg_count > 0) {
 379                                 unregistering = true;
 380                                 break;
 381                         }
 382                 }
 383                 if (!unregistering)
 384                         break;
 385                 __rtnl_unlock();
 386
 387                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 388         }
 389         remove_wait_queue(&netdev_unregistering_wq, &wait);
 390 }
 391
 392 /**
 393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 394  * @ops: struct rtnl_link_ops * to unregister
 395  */
 396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 397 {
 398         /* Close the race with cleanup_net() */
 399         mutex_lock(&net_mutex);
 400         rtnl_lock_unregistering_all();
 401         __rtnl_link_unregister(ops);
 402         rtnl_unlock();
 403         mutex_unlock(&net_mutex);
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 406
 407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 408 {
 409         struct net_device *master_dev;
 410         const struct rtnl_link_ops *ops;
 411
 412         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 413         if (!master_dev)
 414                 return 0;
 415         ops = master_dev->rtnl_link_ops;
 416         if (!ops || !ops->get_slave_size)
 417                 return 0;
 418         /* IFLA_INFO_SLAVE_DATA + nested data */
 419         return nla_total_size(sizeof(struct nlattr)) +
 420                ops->get_slave_size(master_dev, dev);
 421 }
 422
 423 static size_t rtnl_link_get_size(const struct net_device *dev)
 424 {
 425         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 426         size_t size;
 427
 428         if (!ops)
 429                 return 0;
 430
 431         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 432                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 433
 434         if (ops->get_size)
 435                 /* IFLA_INFO_DATA + nested data */
 436                 size += nla_total_size(sizeof(struct nlattr)) +
 437                         ops->get_size(dev);
 438
 439         if (ops->get_xstats_size)
 440                 /* IFLA_INFO_XSTATS */
 441                 size += nla_total_size(ops->get_xstats_size(dev));
 442
 443         size += rtnl_link_get_slave_info_data_size(dev);
 444
 445         return size;
 446 }
 447
 448 static LIST_HEAD(rtnl_af_ops);
 449
 450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 451 {
 452         const struct rtnl_af_ops *ops;
 453
 454         list_for_each_entry(ops, &rtnl_af_ops, list) {
 455                 if (ops->family == family)
 456                         return ops;
 457         }
 458
 459         return NULL;
 460 }
 461
 462 /**
 463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 464  * @ops: struct rtnl_af_ops * to register
 465  *
 466  * Returns 0 on success or a negative error code.
 467  */
 468 void rtnl_af_register(struct rtnl_af_ops *ops)
 469 {
 470         rtnl_lock();
 471         list_add_tail(&ops->list, &rtnl_af_ops);
 472         rtnl_unlock();
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_af_register);
 475
 476 /**
 477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 478  * @ops: struct rtnl_af_ops * to unregister
 479  *
 480  * The caller must hold the rtnl_mutex.
 481  */
 482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 483 {
 484         list_del(&ops->list);
 485 }
 486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 487
 488 /**
 489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 490  * @ops: struct rtnl_af_ops * to unregister
 491  */
 492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 493 {
 494         rtnl_lock();
 495         __rtnl_af_unregister(ops);
 496         rtnl_unlock();
 497 }
 498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 499
 500 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 501                                     u32 ext_filter_mask)
 502 {
 503         struct rtnl_af_ops *af_ops;
 504         size_t size;
 505
 506         /* IFLA_AF_SPEC */
 507         size = nla_total_size(sizeof(struct nlattr));
 508
 509         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 510                 if (af_ops->get_link_af_size) {
 511                         /* AF_* + nested data */
 512                         size += nla_total_size(sizeof(struct nlattr)) +
 513                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 514                 }
 515         }
 516
 517         return size;
 518 }
 519
 520 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 521 {
 522         struct net_device *master_dev;
 523
 524         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 525         if (master_dev && master_dev->rtnl_link_ops)
 526                 return true;
 527         return false;
 528 }
 529
 530 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 531                                      const struct net_device *dev)
 532 {
 533         struct net_device *master_dev;
 534         const struct rtnl_link_ops *ops;
 535         struct nlattr *slave_data;
 536         int err;
 537
 538         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 539         if (!master_dev)
 540                 return 0;
 541         ops = master_dev->rtnl_link_ops;
 542         if (!ops)
 543                 return 0;
 544         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 545                 return -EMSGSIZE;
 546         if (ops->fill_slave_info) {
 547                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 548                 if (!slave_data)
 549                         return -EMSGSIZE;
 550                 err = ops->fill_slave_info(skb, master_dev, dev);
 551                 if (err < 0)
 552                         goto err_cancel_slave_data;
 553                 nla_nest_end(skb, slave_data);
 554         }
 555         return 0;
 556
 557 err_cancel_slave_data:
 558         nla_nest_cancel(skb, slave_data);
 559         return err;
 560 }
 561
 562 static int rtnl_link_info_fill(struct sk_buff *skb,
 563                                const struct net_device *dev)
 564 {
 565         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 566         struct nlattr *data;
 567         int err;
 568
 569         if (!ops)
 570                 return 0;
 571         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 572                 return -EMSGSIZE;
 573         if (ops->fill_xstats) {
 574                 err = ops->fill_xstats(skb, dev);
 575                 if (err < 0)
 576                         return err;
 577         }
 578         if (ops->fill_info) {
 579                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 580                 if (data == NULL)
 581                         return -EMSGSIZE;
 582                 err = ops->fill_info(skb, dev);
 583                 if (err < 0)
 584                         goto err_cancel_data;
 585                 nla_nest_end(skb, data);
 586         }
 587         return 0;
 588
 589 err_cancel_data:
 590         nla_nest_cancel(skb, data);
 591         return err;
 592 }
 593
 594 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 595 {
 596         struct nlattr *linkinfo;
 597         int err = -EMSGSIZE;
 598
 599         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 600         if (linkinfo == NULL)
 601                 goto out;
 602
 603         err = rtnl_link_info_fill(skb, dev);
 604         if (err < 0)
 605                 goto err_cancel_link;
 606
 607         err = rtnl_link_slave_info_fill(skb, dev);
 608         if (err < 0)
 609                 goto err_cancel_link;
 610
 611         nla_nest_end(skb, linkinfo);
 612         return 0;
 613
 614 err_cancel_link:
 615         nla_nest_cancel(skb, linkinfo);
 616 out:
 617         return err;
 618 }
 619
 620 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 621 {
 622         struct sock *rtnl = net->rtnl;
 623         int err = 0;
 624
 625         NETLINK_CB(skb).dst_group = group;
 626         if (echo)
 627                 atomic_inc(&skb->users);
 628         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 629         if (echo)
 630                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 631         return err;
 632 }
 633
 634 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 635 {
 636         struct sock *rtnl = net->rtnl;
 637
 638         return nlmsg_unicast(rtnl, skb, pid);
 639 }
 640 EXPORT_SYMBOL(rtnl_unicast);
 641
 642 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 643                  struct nlmsghdr *nlh, gfp_t flags)
 644 {
 645         struct sock *rtnl = net->rtnl;
 646         int report = 0;
 647
 648         if (nlh)
 649                 report = nlmsg_report(nlh);
 650
 651         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 652 }
 653 EXPORT_SYMBOL(rtnl_notify);
 654
 655 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 656 {
 657         struct sock *rtnl = net->rtnl;
 658
 659         netlink_set_err(rtnl, 0, group, error);
 660 }
 661 EXPORT_SYMBOL(rtnl_set_sk_err);
 662
 663 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 664 {
 665         struct nlattr *mx;
 666         int i, valid = 0;
 667
 668         mx = nla_nest_start(skb, RTA_METRICS);
 669         if (mx == NULL)
 670                 return -ENOBUFS;
 671
 672         for (i = 0; i < RTAX_MAX; i++) {
 673                 if (metrics[i]) {
 674                         if (i == RTAX_CC_ALGO - 1) {
 675                                 char tmp[TCP_CA_NAME_MAX], *name;
 676
 677                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 678                                 if (!name)
 679                                         continue;
 680                                 if (nla_put_string(skb, i + 1, name))
 681                                         goto nla_put_failure;
 682                         } else if (i == RTAX_FEATURES - 1) {
 683                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 684
 685                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 686                                 if (nla_put_u32(skb, i + 1, user_features))
 687                                         goto nla_put_failure;
 688                         } else {
 689                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 690                                         goto nla_put_failure;
 691                         }
 692                         valid++;
 693                 }
 694         }
 695
 696         if (!valid) {
 697                 nla_nest_cancel(skb, mx);
 698                 return 0;
 699         }
 700
 701         return nla_nest_end(skb, mx);
 702
 703 nla_put_failure:
 704         nla_nest_cancel(skb, mx);
 705         return -EMSGSIZE;
 706 }
 707 EXPORT_SYMBOL(rtnetlink_put_metrics);
 708
 709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 710                        long expires, u32 error)
 711 {
 712         struct rta_cacheinfo ci = {
 713                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 714                 .rta_used = dst->__use,
 715                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 716                 .rta_error = error,
 717                 .rta_id =  id,
 718         };
 719
 720         if (expires) {
 721                 unsigned long clock;
 722
 723                 clock = jiffies_to_clock_t(abs(expires));
 724                 clock = min_t(unsigned long, clock, INT_MAX);
 725                 ci.rta_expires = (expires > 0) ? clock : -clock;
 726         }
 727         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 728 }
 729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 730
 731 static void set_operstate(struct net_device *dev, unsigned char transition)
 732 {
 733         unsigned char operstate = dev->operstate;
 734
 735         switch (transition) {
 736         case IF_OPER_UP:
 737                 if ((operstate == IF_OPER_DORMANT ||
 738                      operstate == IF_OPER_UNKNOWN) &&
 739                     !netif_dormant(dev))
 740                         operstate = IF_OPER_UP;
 741                 break;
 742
 743         case IF_OPER_DORMANT:
 744                 if (operstate == IF_OPER_UP ||
 745                     operstate == IF_OPER_UNKNOWN)
 746                         operstate = IF_OPER_DORMANT;
 747                 break;
 748         }
 749
 750         if (dev->operstate != operstate) {
 751                 write_lock_bh(&dev_base_lock);
 752                 dev->operstate = operstate;
 753                 write_unlock_bh(&dev_base_lock);
 754                 netdev_state_change(dev);
 755         }
 756 }
 757
 758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 759 {
 760         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 761                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 762 }
 763
 764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 765                                            const struct ifinfomsg *ifm)
 766 {
 767         unsigned int flags = ifm->ifi_flags;
 768
 769         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 770         if (ifm->ifi_change)
 771                 flags = (flags & ifm->ifi_change) |
 772                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 773
 774         return flags;
 775 }
 776
 777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 778                                  const struct rtnl_link_stats64 *b)
 779 {
 780         a->rx_packets = b->rx_packets;
 781         a->tx_packets = b->tx_packets;
 782         a->rx_bytes = b->rx_bytes;
 783         a->tx_bytes = b->tx_bytes;
 784         a->rx_errors = b->rx_errors;
 785         a->tx_errors = b->tx_errors;
 786         a->rx_dropped = b->rx_dropped;
 787         a->tx_dropped = b->tx_dropped;
 788
 789         a->multicast = b->multicast;
 790         a->collisions = b->collisions;
 791
 792         a->rx_length_errors = b->rx_length_errors;
 793         a->rx_over_errors = b->rx_over_errors;
 794         a->rx_crc_errors = b->rx_crc_errors;
 795         a->rx_frame_errors = b->rx_frame_errors;
 796         a->rx_fifo_errors = b->rx_fifo_errors;
 797         a->rx_missed_errors = b->rx_missed_errors;
 798
 799         a->tx_aborted_errors = b->tx_aborted_errors;
 800         a->tx_carrier_errors = b->tx_carrier_errors;
 801         a->tx_fifo_errors = b->tx_fifo_errors;
 802         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 803         a->tx_window_errors = b->tx_window_errors;
 804
 805         a->rx_compressed = b->rx_compressed;
 806         a->tx_compressed = b->tx_compressed;
 807
 808         a->rx_nohandler = b->rx_nohandler;
 809 }
 810
 811 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 812 {
 813         memcpy(v, b, sizeof(*b));
 814 }
 815
 816 /* All VF info */
 817 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 818                                    u32 ext_filter_mask)
 819 {
 820         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 821             (ext_filter_mask & RTEXT_FILTER_VF)) {
 822                 int num_vfs = dev_num_vf(dev->dev.parent);
 823                 size_t size = nla_total_size(sizeof(struct nlattr));
 824                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 825                 size += num_vfs *
 826                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 827                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 828                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 829                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 830                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 831                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 832                          /* IFLA_VF_STATS_RX_PACKETS */
 833                          nla_total_size(sizeof(__u64)) +
 834                          /* IFLA_VF_STATS_TX_PACKETS */
 835                          nla_total_size(sizeof(__u64)) +
 836                          /* IFLA_VF_STATS_RX_BYTES */
 837                          nla_total_size(sizeof(__u64)) +
 838                          /* IFLA_VF_STATS_TX_BYTES */
 839                          nla_total_size(sizeof(__u64)) +
 840                          /* IFLA_VF_STATS_BROADCAST */
 841                          nla_total_size(sizeof(__u64)) +
 842                          /* IFLA_VF_STATS_MULTICAST */
 843                          nla_total_size(sizeof(__u64)) +
 844                          nla_total_size(sizeof(struct ifla_vf_trust)));
 845                 return size;
 846         } else
 847                 return 0;
 848 }
 849
 850 static size_t rtnl_port_size(const struct net_device *dev,
 851                              u32 ext_filter_mask)
 852 {
 853         size_t port_size = nla_total_size(4)            /* PORT_VF */
 854                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 855                 + nla_total_size(sizeof(struct ifla_port_vsi))
 856                                                         /* PORT_VSI_TYPE */
 857                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 858                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 859                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 860                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 861         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 862         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 863                 + port_size;
 864         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 865                 + port_size;
 866
 867         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 868             !(ext_filter_mask & RTEXT_FILTER_VF))
 869                 return 0;
 870         if (dev_num_vf(dev->dev.parent))
 871                 return port_self_size + vf_ports_size +
 872                         vf_port_size * dev_num_vf(dev->dev.parent);
 873         else
 874                 return port_self_size;
 875 }
 876
 877 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 878                                      u32 ext_filter_mask)
 879 {
 880         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 881                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 882                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 883                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 884                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 885                + nla_total_size(sizeof(struct rtnl_link_stats))
 886                + nla_total_size(sizeof(struct rtnl_link_stats64))
 887                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 888                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 889                + nla_total_size(4) /* IFLA_TXQLEN */
 890                + nla_total_size(4) /* IFLA_WEIGHT */
 891                + nla_total_size(4) /* IFLA_MTU */
 892                + nla_total_size(4) /* IFLA_LINK */
 893                + nla_total_size(4) /* IFLA_MASTER */
 894                + nla_total_size(1) /* IFLA_CARRIER */
 895                + nla_total_size(4) /* IFLA_PROMISCUITY */
 896                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 897                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 898                + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
 899                + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
 900                + nla_total_size(1) /* IFLA_OPERSTATE */
 901                + nla_total_size(1) /* IFLA_LINKMODE */
 902                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 903                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 904                + nla_total_size(ext_filter_mask
 905                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 906                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 907                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 908                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 909                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 910                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 911                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 912                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 913                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 914
 915 }
 916
 917 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 918 {
 919         struct nlattr *vf_ports;
 920         struct nlattr *vf_port;
 921         int vf;
 922         int err;
 923
 924         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 925         if (!vf_ports)
 926                 return -EMSGSIZE;
 927
 928         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 929                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 930                 if (!vf_port)
 931                         goto nla_put_failure;
 932                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 933                         goto nla_put_failure;
 934                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 935                 if (err == -EMSGSIZE)
 936                         goto nla_put_failure;
 937                 if (err) {
 938                         nla_nest_cancel(skb, vf_port);
 939                         continue;
 940                 }
 941                 nla_nest_end(skb, vf_port);
 942         }
 943
 944         nla_nest_end(skb, vf_ports);
 945
 946         return 0;
 947
 948 nla_put_failure:
 949         nla_nest_cancel(skb, vf_ports);
 950         return -EMSGSIZE;
 951 }
 952
 953 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 954 {
 955         struct nlattr *port_self;
 956         int err;
 957
 958         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 959         if (!port_self)
 960                 return -EMSGSIZE;
 961
 962         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 963         if (err) {
 964                 nla_nest_cancel(skb, port_self);
 965                 return (err == -EMSGSIZE) ? err : 0;
 966         }
 967
 968         nla_nest_end(skb, port_self);
 969
 970         return 0;
 971 }
 972
 973 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 974                           u32 ext_filter_mask)
 975 {
 976         int err;
 977
 978         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 979             !(ext_filter_mask & RTEXT_FILTER_VF))
 980                 return 0;
 981
 982         err = rtnl_port_self_fill(skb, dev);
 983         if (err)
 984                 return err;
 985
 986         if (dev_num_vf(dev->dev.parent)) {
 987                 err = rtnl_vf_ports_fill(skb, dev);
 988                 if (err)
 989                         return err;
 990         }
 991
 992         return 0;
 993 }
 994
 995 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 996 {
 997         int err;
 998         struct netdev_phys_item_id ppid;
 999
1000         err = dev_get_phys_port_id(dev, &ppid);
1001         if (err) {
1002                 if (err == -EOPNOTSUPP)
1003                         return 0;
1004                 return err;
1005         }
1006
1007         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1008                 return -EMSGSIZE;
1009
1010         return 0;
1011 }
1012
1013 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1014 {
1015         char name[IFNAMSIZ];
1016         int err;
1017
1018         err = dev_get_phys_port_name(dev, name, sizeof(name));
1019         if (err) {
1020                 if (err == -EOPNOTSUPP)
1021                         return 0;
1022                 return err;
1023         }
1024
1025         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1026                 return -EMSGSIZE;
1027
1028         return 0;
1029 }
1030
1031 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 {
1033         int err;
1034         struct switchdev_attr attr = {
1035                 .orig_dev = dev,
1036                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1037                 .flags = SWITCHDEV_F_NO_RECURSE,
1038         };
1039
1040         err = switchdev_port_attr_get(dev, &attr);
1041         if (err) {
1042                 if (err == -EOPNOTSUPP)
1043                         return 0;
1044                 return err;
1045         }
1046
1047         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1048                     attr.u.ppid.id))
1049                 return -EMSGSIZE;
1050
1051         return 0;
1052 }
1053
1054 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1055                                               struct net_device *dev)
1056 {
1057         const struct rtnl_link_stats64 *stats;
1058         struct rtnl_link_stats64 temp;
1059         struct nlattr *attr;
1060
1061         stats = dev_get_stats(dev, &temp);
1062
1063         attr = nla_reserve(skb, IFLA_STATS,
1064                            sizeof(struct rtnl_link_stats));
1065         if (!attr)
1066                 return -EMSGSIZE;
1067
1068         copy_rtnl_link_stats(nla_data(attr), stats);
1069
1070         attr = nla_reserve(skb, IFLA_STATS64,
1071                            sizeof(struct rtnl_link_stats64));
1072         if (!attr)
1073                 return -EMSGSIZE;
1074
1075         copy_rtnl_link_stats64(nla_data(attr), stats);
1076
1077         return 0;
1078 }
1079
1080 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1081                                                struct net_device *dev,
1082                                                int vfs_num,
1083                                                struct nlattr *vfinfo)
1084 {
1085         struct ifla_vf_rss_query_en vf_rss_query_en;
1086         struct ifla_vf_link_state vf_linkstate;
1087         struct ifla_vf_spoofchk vf_spoofchk;
1088         struct ifla_vf_tx_rate vf_tx_rate;
1089         struct ifla_vf_stats vf_stats;
1090         struct ifla_vf_trust vf_trust;
1091         struct ifla_vf_vlan vf_vlan;
1092         struct ifla_vf_rate vf_rate;
1093         struct nlattr *vf, *vfstats;
1094         struct ifla_vf_mac vf_mac;
1095         struct ifla_vf_info ivi;
1096
1097         /* Not all SR-IOV capable drivers support the
1098          * spoofcheck and "RSS query enable" query.  Preset to
1099          * -1 so the user space tool can detect that the driver
1100          * didn't report anything.
1101          */
1102         ivi.spoofchk = -1;
1103         ivi.rss_query_en = -1;
1104         ivi.trusted = -1;
1105         memset(ivi.mac, 0, sizeof(ivi.mac));
1106         /* The default value for VF link state is "auto"
1107          * IFLA_VF_LINK_STATE_AUTO which equals zero
1108          */
1109         ivi.linkstate = 0;
1110         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1111                 return 0;
1112
1113         vf_mac.vf =
1114                 vf_vlan.vf =
1115                 vf_rate.vf =
1116                 vf_tx_rate.vf =
1117                 vf_spoofchk.vf =
1118                 vf_linkstate.vf =
1119                 vf_rss_query_en.vf =
1120                 vf_trust.vf = ivi.vf;
1121
1122         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1123         vf_vlan.vlan = ivi.vlan;
1124         vf_vlan.qos = ivi.qos;
1125         vf_tx_rate.rate = ivi.max_tx_rate;
1126         vf_rate.min_tx_rate = ivi.min_tx_rate;
1127         vf_rate.max_tx_rate = ivi.max_tx_rate;
1128         vf_spoofchk.setting = ivi.spoofchk;
1129         vf_linkstate.link_state = ivi.linkstate;
1130         vf_rss_query_en.setting = ivi.rss_query_en;
1131         vf_trust.setting = ivi.trusted;
1132         vf = nla_nest_start(skb, IFLA_VF_INFO);
1133         if (!vf) {
1134                 nla_nest_cancel(skb, vfinfo);
1135                 return -EMSGSIZE;
1136         }
1137         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1138             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1139             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1140                     &vf_rate) ||
1141             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1142                     &vf_tx_rate) ||
1143             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1144                     &vf_spoofchk) ||
1145             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1146                     &vf_linkstate) ||
1147             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1148                     sizeof(vf_rss_query_en),
1149                     &vf_rss_query_en) ||
1150             nla_put(skb, IFLA_VF_TRUST,
1151                     sizeof(vf_trust), &vf_trust))
1152                 return -EMSGSIZE;
1153         memset(&vf_stats, 0, sizeof(vf_stats));
1154         if (dev->netdev_ops->ndo_get_vf_stats)
1155                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1156                                                 &vf_stats);
1157         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1158         if (!vfstats) {
1159                 nla_nest_cancel(skb, vf);
1160                 nla_nest_cancel(skb, vfinfo);
1161                 return -EMSGSIZE;
1162         }
1163         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1164                         vf_stats.rx_packets) ||
1165             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1166                         vf_stats.tx_packets) ||
1167             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1168                         vf_stats.rx_bytes) ||
1169             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1170                         vf_stats.tx_bytes) ||
1171             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1172                         vf_stats.broadcast) ||
1173             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1174                         vf_stats.multicast))
1175                 return -EMSGSIZE;
1176         nla_nest_end(skb, vfstats);
1177         nla_nest_end(skb, vf);
1178         return 0;
1179 }
1180
1181 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1182 {
1183         struct rtnl_link_ifmap map = {
1184                 .mem_start   = dev->mem_start,
1185                 .mem_end     = dev->mem_end,
1186                 .base_addr   = dev->base_addr,
1187                 .irq         = dev->irq,
1188                 .dma         = dev->dma,
1189                 .port        = dev->if_port,
1190         };
1191         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1192                 return -EMSGSIZE;
1193
1194         return 0;
1195 }
1196
1197 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1198                             int type, u32 pid, u32 seq, u32 change,
1199                             unsigned int flags, u32 ext_filter_mask)
1200 {
1201         struct ifinfomsg *ifm;
1202         struct nlmsghdr *nlh;
1203         struct nlattr *af_spec;
1204         struct rtnl_af_ops *af_ops;
1205         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1206
1207         ASSERT_RTNL();
1208         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1209         if (nlh == NULL)
1210                 return -EMSGSIZE;
1211
1212         ifm = nlmsg_data(nlh);
1213         ifm->ifi_family = AF_UNSPEC;
1214         ifm->__ifi_pad = 0;
1215         ifm->ifi_type = dev->type;
1216         ifm->ifi_index = dev->ifindex;
1217         ifm->ifi_flags = dev_get_flags(dev);
1218         ifm->ifi_change = change;
1219
1220         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1221             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1222             nla_put_u8(skb, IFLA_OPERSTATE,
1223                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1224             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1225             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1226             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1227             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1228             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1229             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1230             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1231 #ifdef CONFIG_RPS
1232             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1233 #endif
1234             (dev->ifindex != dev_get_iflink(dev) &&
1235              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1236             (upper_dev &&
1237              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1238             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1239             (dev->qdisc &&
1240              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1241             (dev->ifalias &&
1242              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1243             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1244                         atomic_read(&dev->carrier_changes)) ||
1245             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1246                 goto nla_put_failure;
1247
1248         if (rtnl_fill_link_ifmap(skb, dev))
1249                 goto nla_put_failure;
1250
1251         if (dev->addr_len) {
1252                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1253                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1254                         goto nla_put_failure;
1255         }
1256
1257         if (rtnl_phys_port_id_fill(skb, dev))
1258                 goto nla_put_failure;
1259
1260         if (rtnl_phys_port_name_fill(skb, dev))
1261                 goto nla_put_failure;
1262
1263         if (rtnl_phys_switch_id_fill(skb, dev))
1264                 goto nla_put_failure;
1265
1266         if (rtnl_fill_stats(skb, dev))
1267                 goto nla_put_failure;
1268
1269         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1270             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1271                 goto nla_put_failure;
1272
1273         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1274             ext_filter_mask & RTEXT_FILTER_VF) {
1275                 int i;
1276                 struct nlattr *vfinfo;
1277                 int num_vfs = dev_num_vf(dev->dev.parent);
1278
1279                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1280                 if (!vfinfo)
1281                         goto nla_put_failure;
1282                 for (i = 0; i < num_vfs; i++) {
1283                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1284                                 goto nla_put_failure;
1285                 }
1286
1287                 nla_nest_end(skb, vfinfo);
1288         }
1289
1290         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1291                 goto nla_put_failure;
1292
1293         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1294                 if (rtnl_link_fill(skb, dev) < 0)
1295                         goto nla_put_failure;
1296         }
1297
1298         if (dev->rtnl_link_ops &&
1299             dev->rtnl_link_ops->get_link_net) {
1300                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1301
1302                 if (!net_eq(dev_net(dev), link_net)) {
1303                         int id = peernet2id_alloc(dev_net(dev), link_net);
1304
1305                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1306                                 goto nla_put_failure;
1307                 }
1308         }
1309
1310         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1311                 goto nla_put_failure;
1312
1313         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1314                 if (af_ops->fill_link_af) {
1315                         struct nlattr *af;
1316                         int err;
1317
1318                         if (!(af = nla_nest_start(skb, af_ops->family)))
1319                                 goto nla_put_failure;
1320
1321                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1322
1323                         /*
1324                          * Caller may return ENODATA to indicate that there
1325                          * was no data to be dumped. This is not an error, it
1326                          * means we should trim the attribute header and
1327                          * continue.
1328                          */
1329                         if (err == -ENODATA)
1330                                 nla_nest_cancel(skb, af);
1331                         else if (err < 0)
1332                                 goto nla_put_failure;
1333
1334                         nla_nest_end(skb, af);
1335                 }
1336         }
1337
1338         nla_nest_end(skb, af_spec);
1339
1340         nlmsg_end(skb, nlh);
1341         return 0;
1342
1343 nla_put_failure:
1344         nlmsg_cancel(skb, nlh);
1345         return -EMSGSIZE;
1346 }
1347
1348 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1349         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1350         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1351         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1352         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1353         [IFLA_MTU]              = { .type = NLA_U32 },
1354         [IFLA_LINK]             = { .type = NLA_U32 },
1355         [IFLA_MASTER]           = { .type = NLA_U32 },
1356         [IFLA_CARRIER]          = { .type = NLA_U8 },
1357         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1358         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1359         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1360         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1361         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1362         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1363         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1364         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1365         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1366         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1367         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1368         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1369         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1370         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1371         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1372         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1373         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1374         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1375         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1376         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1377         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1378 };
1379
1380 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1381         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1382         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1383         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1384         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1385 };
1386
1387 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1388         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1389         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1390         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1391         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1392         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1393         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1394         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1395         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1396         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1397         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1398         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1399 };
1400
1401 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1402         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1403         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1404                                     .len = PORT_PROFILE_MAX },
1405         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1406                                     .len = sizeof(struct ifla_port_vsi)},
1407         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1408                                       .len = PORT_UUID_MAX },
1409         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1410                                     .len = PORT_UUID_MAX },
1411         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1412         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1413 };
1414
1415 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1416 {
1417         const struct rtnl_link_ops *ops = NULL;
1418         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1419
1420         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1421                 return NULL;
1422
1423         if (linfo[IFLA_INFO_KIND]) {
1424                 char kind[MODULE_NAME_LEN];
1425
1426                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1427                 ops = rtnl_link_ops_get(kind);
1428         }
1429
1430         return ops;
1431 }
1432
1433 static bool link_master_filtered(struct net_device *dev, int master_idx)
1434 {
1435         struct net_device *master;
1436
1437         if (!master_idx)
1438                 return false;
1439
1440         master = netdev_master_upper_dev_get(dev);
1441         if (!master || master->ifindex != master_idx)
1442                 return true;
1443
1444         return false;
1445 }
1446
1447 static bool link_kind_filtered(const struct net_device *dev,
1448                                const struct rtnl_link_ops *kind_ops)
1449 {
1450         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1451                 return true;
1452
1453         return false;
1454 }
1455
1456 static bool link_dump_filtered(struct net_device *dev,
1457                                int master_idx,
1458                                const struct rtnl_link_ops *kind_ops)
1459 {
1460         if (link_master_filtered(dev, master_idx) ||
1461             link_kind_filtered(dev, kind_ops))
1462                 return true;
1463
1464         return false;
1465 }
1466
1467 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1468 {
1469         struct net *net = sock_net(skb->sk);
1470         int h, s_h;
1471         int idx = 0, s_idx;
1472         struct net_device *dev;
1473         struct hlist_head *head;
1474         struct nlattr *tb[IFLA_MAX+1];
1475         u32 ext_filter_mask = 0;
1476         const struct rtnl_link_ops *kind_ops = NULL;
1477         unsigned int flags = NLM_F_MULTI;
1478         int master_idx = 0;
1479         int err;
1480         int hdrlen;
1481
1482         s_h = cb->args[0];
1483         s_idx = cb->args[1];
1484
1485         cb->seq = net->dev_base_seq;
1486
1487         /* A hack to preserve kernel<->userspace interface.
1488          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1489          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1490          * what iproute2 < v3.9.0 used.
1491          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1492          * attribute, its netlink message is shorter than struct ifinfomsg.
1493          */
1494         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1495                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1496
1497         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1498
1499                 if (tb[IFLA_EXT_MASK])
1500                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1501
1502                 if (tb[IFLA_MASTER])
1503                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1504
1505                 if (tb[IFLA_LINKINFO])
1506                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1507
1508                 if (master_idx || kind_ops)
1509                         flags |= NLM_F_DUMP_FILTERED;
1510         }
1511
1512         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1513                 idx = 0;
1514                 head = &net->dev_index_head[h];
1515                 hlist_for_each_entry(dev, head, index_hlist) {
1516                         if (link_dump_filtered(dev, master_idx, kind_ops))
1517                                 continue;
1518                         if (idx < s_idx)
1519                                 goto cont;
1520                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1521                                                NETLINK_CB(cb->skb).portid,
1522                                                cb->nlh->nlmsg_seq, 0,
1523                                                flags,
1524                                                ext_filter_mask);
1525                         /* If we ran out of room on the first message,
1526                          * we're in trouble
1527                          */
1528                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1529
1530                         if (err < 0)
1531                                 goto out;
1532
1533                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1534 cont:
1535                         idx++;
1536                 }
1537         }
1538 out:
1539         cb->args[1] = idx;
1540         cb->args[0] = h;
1541
1542         return skb->len;
1543 }
1544
1545 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1546 {
1547         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1548 }
1549 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1550
1551 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1552 {
1553         struct net *net;
1554         /* Examine the link attributes and figure out which
1555          * network namespace we are talking about.
1556          */
1557         if (tb[IFLA_NET_NS_PID])
1558                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1559         else if (tb[IFLA_NET_NS_FD])
1560                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1561         else
1562                 net = get_net(src_net);
1563         return net;
1564 }
1565 EXPORT_SYMBOL(rtnl_link_get_net);
1566
1567 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1568 {
1569         if (dev) {
1570                 if (tb[IFLA_ADDRESS] &&
1571                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1572                         return -EINVAL;
1573
1574                 if (tb[IFLA_BROADCAST] &&
1575                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1576                         return -EINVAL;
1577         }
1578
1579         if (tb[IFLA_AF_SPEC]) {
1580                 struct nlattr *af;
1581                 int rem, err;
1582
1583                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1584                         const struct rtnl_af_ops *af_ops;
1585
1586                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1587                                 return -EAFNOSUPPORT;
1588
1589                         if (!af_ops->set_link_af)
1590                                 return -EOPNOTSUPP;
1591
1592                         if (af_ops->validate_link_af) {
1593                                 err = af_ops->validate_link_af(dev, af);
1594                                 if (err < 0)
1595                                         return err;
1596                         }
1597                 }
1598         }
1599
1600         return 0;
1601 }
1602
1603 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1604                                   int guid_type)
1605 {
1606         const struct net_device_ops *ops = dev->netdev_ops;
1607
1608         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1609 }
1610
1611 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1612 {
1613         if (dev->type != ARPHRD_INFINIBAND)
1614                 return -EOPNOTSUPP;
1615
1616         return handle_infiniband_guid(dev, ivt, guid_type);
1617 }
1618
1619 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1620 {
1621         const struct net_device_ops *ops = dev->netdev_ops;
1622         int err = -EINVAL;
1623
1624         if (tb[IFLA_VF_MAC]) {
1625                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1626
1627                 err = -EOPNOTSUPP;
1628                 if (ops->ndo_set_vf_mac)
1629                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1630                                                   ivm->mac);
1631                 if (err < 0)
1632                         return err;
1633         }
1634
1635         if (tb[IFLA_VF_VLAN]) {
1636                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1637
1638                 err = -EOPNOTSUPP;
1639                 if (ops->ndo_set_vf_vlan)
1640                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1641                                                    ivv->qos);
1642                 if (err < 0)
1643                         return err;
1644         }
1645
1646         if (tb[IFLA_VF_TX_RATE]) {
1647                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1648                 struct ifla_vf_info ivf;
1649
1650                 err = -EOPNOTSUPP;
1651                 if (ops->ndo_get_vf_config)
1652                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1653                 if (err < 0)
1654                         return err;
1655
1656                 err = -EOPNOTSUPP;
1657                 if (ops->ndo_set_vf_rate)
1658                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1659                                                    ivf.min_tx_rate,
1660                                                    ivt->rate);
1661                 if (err < 0)
1662                         return err;
1663         }
1664
1665         if (tb[IFLA_VF_RATE]) {
1666                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1667
1668                 err = -EOPNOTSUPP;
1669                 if (ops->ndo_set_vf_rate)
1670                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1671                                                    ivt->min_tx_rate,
1672                                                    ivt->max_tx_rate);
1673                 if (err < 0)
1674                         return err;
1675         }
1676
1677         if (tb[IFLA_VF_SPOOFCHK]) {
1678                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1679
1680                 err = -EOPNOTSUPP;
1681                 if (ops->ndo_set_vf_spoofchk)
1682                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1683                                                        ivs->setting);
1684                 if (err < 0)
1685                         return err;
1686         }
1687
1688         if (tb[IFLA_VF_LINK_STATE]) {
1689                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1690
1691                 err = -EOPNOTSUPP;
1692                 if (ops->ndo_set_vf_link_state)
1693                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1694                                                          ivl->link_state);
1695                 if (err < 0)
1696                         return err;
1697         }
1698
1699         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1700                 struct ifla_vf_rss_query_en *ivrssq_en;
1701
1702                 err = -EOPNOTSUPP;
1703                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1704                 if (ops->ndo_set_vf_rss_query_en)
1705                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1706                                                            ivrssq_en->setting);
1707                 if (err < 0)
1708                         return err;
1709         }
1710
1711         if (tb[IFLA_VF_TRUST]) {
1712                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1713
1714                 err = -EOPNOTSUPP;
1715                 if (ops->ndo_set_vf_trust)
1716                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1717                 if (err < 0)
1718                         return err;
1719         }
1720
1721         if (tb[IFLA_VF_IB_NODE_GUID]) {
1722                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1723
1724                 if (!ops->ndo_set_vf_guid)
1725                         return -EOPNOTSUPP;
1726
1727                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1728         }
1729
1730         if (tb[IFLA_VF_IB_PORT_GUID]) {
1731                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1732
1733                 if (!ops->ndo_set_vf_guid)
1734                         return -EOPNOTSUPP;
1735
1736                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1737         }
1738
1739         return err;
1740 }
1741
1742 static int do_set_master(struct net_device *dev, int ifindex)
1743 {
1744         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1745         const struct net_device_ops *ops;
1746         int err;
1747
1748         if (upper_dev) {
1749                 if (upper_dev->ifindex == ifindex)
1750                         return 0;
1751                 ops = upper_dev->netdev_ops;
1752                 if (ops->ndo_del_slave) {
1753                         err = ops->ndo_del_slave(upper_dev, dev);
1754                         if (err)
1755                                 return err;
1756                 } else {
1757                         return -EOPNOTSUPP;
1758                 }
1759         }
1760
1761         if (ifindex) {
1762                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1763                 if (!upper_dev)
1764                         return -EINVAL;
1765                 ops = upper_dev->netdev_ops;
1766                 if (ops->ndo_add_slave) {
1767                         err = ops->ndo_add_slave(upper_dev, dev);
1768                         if (err)
1769                                 return err;
1770                 } else {
1771                         return -EOPNOTSUPP;
1772                 }
1773         }
1774         return 0;
1775 }
1776
1777 #define DO_SETLINK_MODIFIED     0x01
1778 /* notify flag means notify + modified. */
1779 #define DO_SETLINK_NOTIFY       0x03
1780 static int do_setlink(const struct sk_buff *skb,
1781                       struct net_device *dev, struct ifinfomsg *ifm,
1782                       struct nlattr **tb, char *ifname, int status)
1783 {
1784         const struct net_device_ops *ops = dev->netdev_ops;
1785         int err;
1786
1787         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1788                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1789                 if (IS_ERR(net)) {
1790                         err = PTR_ERR(net);
1791                         goto errout;
1792                 }
1793                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1794                         put_net(net);
1795                         err = -EPERM;
1796                         goto errout;
1797                 }
1798                 err = dev_change_net_namespace(dev, net, ifname);
1799                 put_net(net);
1800                 if (err)
1801                         goto errout;
1802                 status |= DO_SETLINK_MODIFIED;
1803         }
1804
1805         if (tb[IFLA_MAP]) {
1806                 struct rtnl_link_ifmap *u_map;
1807                 struct ifmap k_map;
1808
1809                 if (!ops->ndo_set_config) {
1810                         err = -EOPNOTSUPP;
1811                         goto errout;
1812                 }
1813
1814                 if (!netif_device_present(dev)) {
1815                         err = -ENODEV;
1816                         goto errout;
1817                 }
1818
1819                 u_map = nla_data(tb[IFLA_MAP]);
1820                 k_map.mem_start = (unsigned long) u_map->mem_start;
1821                 k_map.mem_end = (unsigned long) u_map->mem_end;
1822                 k_map.base_addr = (unsigned short) u_map->base_addr;
1823                 k_map.irq = (unsigned char) u_map->irq;
1824                 k_map.dma = (unsigned char) u_map->dma;
1825                 k_map.port = (unsigned char) u_map->port;
1826
1827                 err = ops->ndo_set_config(dev, &k_map);
1828                 if (err < 0)
1829                         goto errout;
1830
1831                 status |= DO_SETLINK_NOTIFY;
1832         }
1833
1834         if (tb[IFLA_ADDRESS]) {
1835                 struct sockaddr *sa;
1836                 int len;
1837
1838                 len = sizeof(sa_family_t) + dev->addr_len;
1839                 sa = kmalloc(len, GFP_KERNEL);
1840                 if (!sa) {
1841                         err = -ENOMEM;
1842                         goto errout;
1843                 }
1844                 sa->sa_family = dev->type;
1845                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1846                        dev->addr_len);
1847                 err = dev_set_mac_address(dev, sa);
1848                 kfree(sa);
1849                 if (err)
1850                         goto errout;
1851                 status |= DO_SETLINK_MODIFIED;
1852         }
1853
1854         if (tb[IFLA_MTU]) {
1855                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1856                 if (err < 0)
1857                         goto errout;
1858                 status |= DO_SETLINK_MODIFIED;
1859         }
1860
1861         if (tb[IFLA_GROUP]) {
1862                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1863                 status |= DO_SETLINK_NOTIFY;
1864         }
1865
1866         /*
1867          * Interface selected by interface index but interface
1868          * name provided implies that a name change has been
1869          * requested.
1870          */
1871         if (ifm->ifi_index > 0 && ifname[0]) {
1872                 err = dev_change_name(dev, ifname);
1873                 if (err < 0)
1874                         goto errout;
1875                 status |= DO_SETLINK_MODIFIED;
1876         }
1877
1878         if (tb[IFLA_IFALIAS]) {
1879                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1880                                     nla_len(tb[IFLA_IFALIAS]));
1881                 if (err < 0)
1882                         goto errout;
1883                 status |= DO_SETLINK_NOTIFY;
1884         }
1885
1886         if (tb[IFLA_BROADCAST]) {
1887                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1888                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1889         }
1890
1891         if (ifm->ifi_flags || ifm->ifi_change) {
1892                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1893                 if (err < 0)
1894                         goto errout;
1895         }
1896
1897         if (tb[IFLA_MASTER]) {
1898                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1899                 if (err)
1900                         goto errout;
1901                 status |= DO_SETLINK_MODIFIED;
1902         }
1903
1904         if (tb[IFLA_CARRIER]) {
1905                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1906                 if (err)
1907                         goto errout;
1908                 status |= DO_SETLINK_MODIFIED;
1909         }
1910
1911         if (tb[IFLA_TXQLEN]) {
1912                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1913
1914                 if (dev->tx_queue_len ^ value)
1915                         status |= DO_SETLINK_NOTIFY;
1916
1917                 dev->tx_queue_len = value;
1918         }
1919
1920         if (tb[IFLA_OPERSTATE])
1921                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1922
1923         if (tb[IFLA_LINKMODE]) {
1924                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1925
1926                 write_lock_bh(&dev_base_lock);
1927                 if (dev->link_mode ^ value)
1928                         status |= DO_SETLINK_NOTIFY;
1929                 dev->link_mode = value;
1930                 write_unlock_bh(&dev_base_lock);
1931         }
1932
1933         if (tb[IFLA_VFINFO_LIST]) {
1934                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1935                 struct nlattr *attr;
1936                 int rem;
1937
1938                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1939                         if (nla_type(attr) != IFLA_VF_INFO ||
1940                             nla_len(attr) < NLA_HDRLEN) {
1941                                 err = -EINVAL;
1942                                 goto errout;
1943                         }
1944                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1945                                                ifla_vf_policy);
1946                         if (err < 0)
1947                                 goto errout;
1948                         err = do_setvfinfo(dev, vfinfo);
1949                         if (err < 0)
1950                                 goto errout;
1951                         status |= DO_SETLINK_NOTIFY;
1952                 }
1953         }
1954         err = 0;
1955
1956         if (tb[IFLA_VF_PORTS]) {
1957                 struct nlattr *port[IFLA_PORT_MAX+1];
1958                 struct nlattr *attr;
1959                 int vf;
1960                 int rem;
1961
1962                 err = -EOPNOTSUPP;
1963                 if (!ops->ndo_set_vf_port)
1964                         goto errout;
1965
1966                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1967                         if (nla_type(attr) != IFLA_VF_PORT ||
1968                             nla_len(attr) < NLA_HDRLEN) {
1969                                 err = -EINVAL;
1970                                 goto errout;
1971                         }
1972                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1973                                                ifla_port_policy);
1974                         if (err < 0)
1975                                 goto errout;
1976                         if (!port[IFLA_PORT_VF]) {
1977                                 err = -EOPNOTSUPP;
1978                                 goto errout;
1979                         }
1980                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1981                         err = ops->ndo_set_vf_port(dev, vf, port);
1982                         if (err < 0)
1983                                 goto errout;
1984                         status |= DO_SETLINK_NOTIFY;
1985                 }
1986         }
1987         err = 0;
1988
1989         if (tb[IFLA_PORT_SELF]) {
1990                 struct nlattr *port[IFLA_PORT_MAX+1];
1991
1992                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1993                         tb[IFLA_PORT_SELF], ifla_port_policy);
1994                 if (err < 0)
1995                         goto errout;
1996
1997                 err = -EOPNOTSUPP;
1998                 if (ops->ndo_set_vf_port)
1999                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2000                 if (err < 0)
2001                         goto errout;
2002                 status |= DO_SETLINK_NOTIFY;
2003         }
2004
2005         if (tb[IFLA_AF_SPEC]) {
2006                 struct nlattr *af;
2007                 int rem;
2008
2009                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2010                         const struct rtnl_af_ops *af_ops;
2011
2012                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2013                                 BUG();
2014
2015                         err = af_ops->set_link_af(dev, af);
2016                         if (err < 0)
2017                                 goto errout;
2018
2019                         status |= DO_SETLINK_NOTIFY;
2020                 }
2021         }
2022         err = 0;
2023
2024         if (tb[IFLA_PROTO_DOWN]) {
2025                 err = dev_change_proto_down(dev,
2026                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2027                 if (err)
2028                         goto errout;
2029                 status |= DO_SETLINK_NOTIFY;
2030         }
2031
2032 errout:
2033         if (status & DO_SETLINK_MODIFIED) {
2034                 if (status & DO_SETLINK_NOTIFY)
2035                         netdev_state_change(dev);
2036
2037                 if (err < 0)
2038                         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",
2039                                              dev->name);
2040         }
2041
2042         return err;
2043 }
2044
2045 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2046 {
2047         struct net *net = sock_net(skb->sk);
2048         struct ifinfomsg *ifm;
2049         struct net_device *dev;
2050         int err;
2051         struct nlattr *tb[IFLA_MAX+1];
2052         char ifname[IFNAMSIZ];
2053
2054         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2055         if (err < 0)
2056                 goto errout;
2057
2058         if (tb[IFLA_IFNAME])
2059                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2060         else
2061                 ifname[0] = '\0';
2062
2063         err = -EINVAL;
2064         ifm = nlmsg_data(nlh);
2065         if (ifm->ifi_index > 0)
2066                 dev = __dev_get_by_index(net, ifm->ifi_index);
2067         else if (tb[IFLA_IFNAME])
2068                 dev = __dev_get_by_name(net, ifname);
2069         else
2070                 goto errout;
2071
2072         if (dev == NULL) {
2073                 err = -ENODEV;
2074                 goto errout;
2075         }
2076
2077         err = validate_linkmsg(dev, tb);
2078         if (err < 0)
2079                 goto errout;
2080
2081         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2082 errout:
2083         return err;
2084 }
2085
2086 static int rtnl_group_dellink(const struct net *net, int group)
2087 {
2088         struct net_device *dev, *aux;
2089         LIST_HEAD(list_kill);
2090         bool found = false;
2091
2092         if (!group)
2093                 return -EPERM;
2094
2095         for_each_netdev(net, dev) {
2096                 if (dev->group == group) {
2097                         const struct rtnl_link_ops *ops;
2098
2099                         found = true;
2100                         ops = dev->rtnl_link_ops;
2101                         if (!ops || !ops->dellink)
2102                                 return -EOPNOTSUPP;
2103                 }
2104         }
2105
2106         if (!found)
2107                 return -ENODEV;
2108
2109         for_each_netdev_safe(net, dev, aux) {
2110                 if (dev->group == group) {
2111                         const struct rtnl_link_ops *ops;
2112
2113                         ops = dev->rtnl_link_ops;
2114                         ops->dellink(dev, &list_kill);
2115                 }
2116         }
2117         unregister_netdevice_many(&list_kill);
2118
2119         return 0;
2120 }
2121
2122 int rtnl_delete_link(struct net_device *dev)
2123 {
2124         const struct rtnl_link_ops *ops;
2125         LIST_HEAD(list_kill);
2126
2127         ops = dev->rtnl_link_ops;
2128         if (!ops || !ops->dellink)
2129                 return -EOPNOTSUPP;
2130
2131         ops->dellink(dev, &list_kill);
2132         unregister_netdevice_many(&list_kill);
2133
2134         return 0;
2135 }
2136 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2137
2138 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2139 {
2140         struct net *net = sock_net(skb->sk);
2141         struct net_device *dev;
2142         struct ifinfomsg *ifm;
2143         char ifname[IFNAMSIZ];
2144         struct nlattr *tb[IFLA_MAX+1];
2145         int err;
2146
2147         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2148         if (err < 0)
2149                 return err;
2150
2151         if (tb[IFLA_IFNAME])
2152                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2153
2154         ifm = nlmsg_data(nlh);
2155         if (ifm->ifi_index > 0)
2156                 dev = __dev_get_by_index(net, ifm->ifi_index);
2157         else if (tb[IFLA_IFNAME])
2158                 dev = __dev_get_by_name(net, ifname);
2159         else if (tb[IFLA_GROUP])
2160                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2161         else
2162                 return -EINVAL;
2163
2164         if (!dev)
2165                 return -ENODEV;
2166
2167         return rtnl_delete_link(dev);
2168 }
2169
2170 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2171 {
2172         unsigned int old_flags;
2173         int err;
2174
2175         old_flags = dev->flags;
2176         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2177                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2178                 if (err < 0)
2179                         return err;
2180         }
2181
2182         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2183
2184         __dev_notify_flags(dev, old_flags, ~0U);
2185         return 0;
2186 }
2187 EXPORT_SYMBOL(rtnl_configure_link);
2188
2189 struct net_device *rtnl_create_link(struct net *net,
2190         const char *ifname, unsigned char name_assign_type,
2191         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2192 {
2193         int err;
2194         struct net_device *dev;
2195         unsigned int num_tx_queues = 1;
2196         unsigned int num_rx_queues = 1;
2197
2198         if (tb[IFLA_NUM_TX_QUEUES])
2199                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2200         else if (ops->get_num_tx_queues)
2201                 num_tx_queues = ops->get_num_tx_queues();
2202
2203         if (tb[IFLA_NUM_RX_QUEUES])
2204                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2205         else if (ops->get_num_rx_queues)
2206                 num_rx_queues = ops->get_num_rx_queues();
2207
2208         err = -ENOMEM;
2209         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2210                                ops->setup, num_tx_queues, num_rx_queues);
2211         if (!dev)
2212                 goto err;
2213
2214         dev_net_set(dev, net);
2215         dev->rtnl_link_ops = ops;
2216         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2217
2218         if (tb[IFLA_MTU])
2219                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2220         if (tb[IFLA_ADDRESS]) {
2221                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2222                                 nla_len(tb[IFLA_ADDRESS]));
2223                 dev->addr_assign_type = NET_ADDR_SET;
2224         }
2225         if (tb[IFLA_BROADCAST])
2226                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2227                                 nla_len(tb[IFLA_BROADCAST]));
2228         if (tb[IFLA_TXQLEN])
2229                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2230         if (tb[IFLA_OPERSTATE])
2231                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2232         if (tb[IFLA_LINKMODE])
2233                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2234         if (tb[IFLA_GROUP])
2235                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2236
2237         return dev;
2238
2239 err:
2240         return ERR_PTR(err);
2241 }
2242 EXPORT_SYMBOL(rtnl_create_link);
2243
2244 static int rtnl_group_changelink(const struct sk_buff *skb,
2245                 struct net *net, int group,
2246                 struct ifinfomsg *ifm,
2247                 struct nlattr **tb)
2248 {
2249         struct net_device *dev, *aux;
2250         int err;
2251
2252         for_each_netdev_safe(net, dev, aux) {
2253                 if (dev->group == group) {
2254                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2255                         if (err < 0)
2256                                 return err;
2257                 }
2258         }
2259
2260         return 0;
2261 }
2262
2263 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2264 {
2265         struct net *net = sock_net(skb->sk);
2266         const struct rtnl_link_ops *ops;
2267         const struct rtnl_link_ops *m_ops = NULL;
2268         struct net_device *dev;
2269         struct net_device *master_dev = NULL;
2270         struct ifinfomsg *ifm;
2271         char kind[MODULE_NAME_LEN];
2272         char ifname[IFNAMSIZ];
2273         struct nlattr *tb[IFLA_MAX+1];
2274         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2275         unsigned char name_assign_type = NET_NAME_USER;
2276         int err;
2277
2278 #ifdef CONFIG_MODULES
2279 replay:
2280 #endif
2281         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2282         if (err < 0)
2283                 return err;
2284
2285         if (tb[IFLA_IFNAME])
2286                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2287         else
2288                 ifname[0] = '\0';
2289
2290         ifm = nlmsg_data(nlh);
2291         if (ifm->ifi_index > 0)
2292                 dev = __dev_get_by_index(net, ifm->ifi_index);
2293         else {
2294                 if (ifname[0])
2295                         dev = __dev_get_by_name(net, ifname);
2296                 else
2297                         dev = NULL;
2298         }
2299
2300         if (dev) {
2301                 master_dev = netdev_master_upper_dev_get(dev);
2302                 if (master_dev)
2303                         m_ops = master_dev->rtnl_link_ops;
2304         }
2305
2306         err = validate_linkmsg(dev, tb);
2307         if (err < 0)
2308                 return err;
2309
2310         if (tb[IFLA_LINKINFO]) {
2311                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2312                                        tb[IFLA_LINKINFO], ifla_info_policy);
2313                 if (err < 0)
2314                         return err;
2315         } else
2316                 memset(linkinfo, 0, sizeof(linkinfo));
2317
2318         if (linkinfo[IFLA_INFO_KIND]) {
2319                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2320                 ops = rtnl_link_ops_get(kind);
2321         } else {
2322                 kind[0] = '\0';
2323                 ops = NULL;
2324         }
2325
2326         if (1) {
2327                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2328                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2329                 struct nlattr **data = NULL;
2330                 struct nlattr **slave_data = NULL;
2331                 struct net *dest_net, *link_net = NULL;
2332
2333                 if (ops) {
2334                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2335                                 err = nla_parse_nested(attr, ops->maxtype,
2336                                                        linkinfo[IFLA_INFO_DATA],
2337                                                        ops->policy);
2338                                 if (err < 0)
2339                                         return err;
2340                                 data = attr;
2341                         }
2342                         if (ops->validate) {
2343                                 err = ops->validate(tb, data);
2344                                 if (err < 0)
2345                                         return err;
2346                         }
2347                 }
2348
2349                 if (m_ops) {
2350                         if (m_ops->slave_maxtype &&
2351                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2352                                 err = nla_parse_nested(slave_attr,
2353                                                        m_ops->slave_maxtype,
2354                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2355                                                        m_ops->slave_policy);
2356                                 if (err < 0)
2357                                         return err;
2358                                 slave_data = slave_attr;
2359                         }
2360                         if (m_ops->slave_validate) {
2361                                 err = m_ops->slave_validate(tb, slave_data);
2362                                 if (err < 0)
2363                                         return err;
2364                         }
2365                 }
2366
2367                 if (dev) {
2368                         int status = 0;
2369
2370                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2371                                 return -EEXIST;
2372                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2373                                 return -EOPNOTSUPP;
2374
2375                         if (linkinfo[IFLA_INFO_DATA]) {
2376                                 if (!ops || ops != dev->rtnl_link_ops ||
2377                                     !ops->changelink)
2378                                         return -EOPNOTSUPP;
2379
2380                                 err = ops->changelink(dev, tb, data);
2381                                 if (err < 0)
2382                                         return err;
2383                                 status |= DO_SETLINK_NOTIFY;
2384                         }
2385
2386                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2387                                 if (!m_ops || !m_ops->slave_changelink)
2388                                         return -EOPNOTSUPP;
2389
2390                                 err = m_ops->slave_changelink(master_dev, dev,
2391                                                               tb, slave_data);
2392                                 if (err < 0)
2393                                         return err;
2394                                 status |= DO_SETLINK_NOTIFY;
2395                         }
2396
2397                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2398                 }
2399
2400                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2401                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2402                                 return rtnl_group_changelink(skb, net,
2403                                                 nla_get_u32(tb[IFLA_GROUP]),
2404                                                 ifm, tb);
2405                         return -ENODEV;
2406                 }
2407
2408                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2409                         return -EOPNOTSUPP;
2410
2411                 if (!ops) {
2412 #ifdef CONFIG_MODULES
2413                         if (kind[0]) {
2414                                 __rtnl_unlock();
2415                                 request_module("rtnl-link-%s", kind);
2416                                 rtnl_lock();
2417                                 ops = rtnl_link_ops_get(kind);
2418                                 if (ops)
2419                                         goto replay;
2420                         }
2421 #endif
2422                         return -EOPNOTSUPP;
2423                 }
2424
2425                 if (!ops->setup)
2426                         return -EOPNOTSUPP;
2427
2428                 if (!ifname[0]) {
2429                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2430                         name_assign_type = NET_NAME_ENUM;
2431                 }
2432
2433                 dest_net = rtnl_link_get_net(net, tb);
2434                 if (IS_ERR(dest_net))
2435                         return PTR_ERR(dest_net);
2436
2437                 err = -EPERM;
2438                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2439                         goto out;
2440
2441                 if (tb[IFLA_LINK_NETNSID]) {
2442                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2443
2444                         link_net = get_net_ns_by_id(dest_net, id);
2445                         if (!link_net) {
2446                                 err =  -EINVAL;
2447                                 goto out;
2448                         }
2449                         err = -EPERM;
2450                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2451                                 goto out;
2452                 }
2453
2454                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2455                                        name_assign_type, ops, tb);
2456                 if (IS_ERR(dev)) {
2457                         err = PTR_ERR(dev);
2458                         goto out;
2459                 }
2460
2461                 dev->ifindex = ifm->ifi_index;
2462
2463                 if (ops->newlink) {
2464                         err = ops->newlink(link_net ? : net, dev, tb, data);
2465                         /* Drivers should call free_netdev() in ->destructor
2466                          * and unregister it on failure after registration
2467                          * so that device could be finally freed in rtnl_unlock.
2468                          */
2469                         if (err < 0) {
2470                                 /* If device is not registered at all, free it now */
2471                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2472                                         free_netdev(dev);
2473                                 goto out;
2474                         }
2475                 } else {
2476                         err = register_netdevice(dev);
2477                         if (err < 0) {
2478                                 free_netdev(dev);
2479                                 goto out;
2480                         }
2481                 }
2482                 err = rtnl_configure_link(dev, ifm);
2483                 if (err < 0)
2484                         goto out_unregister;
2485                 if (link_net) {
2486                         err = dev_change_net_namespace(dev, dest_net, ifname);
2487                         if (err < 0)
2488                                 goto out_unregister;
2489                 }
2490 out:
2491                 if (link_net)
2492                         put_net(link_net);
2493                 put_net(dest_net);
2494                 return err;
2495 out_unregister:
2496                 if (ops->newlink) {
2497                         LIST_HEAD(list_kill);
2498
2499                         ops->dellink(dev, &list_kill);
2500                         unregister_netdevice_many(&list_kill);
2501                 } else {
2502                         unregister_netdevice(dev);
2503                 }
2504                 goto out;
2505         }
2506 }
2507
2508 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2509 {
2510         struct net *net = sock_net(skb->sk);
2511         struct ifinfomsg *ifm;
2512         char ifname[IFNAMSIZ];
2513         struct nlattr *tb[IFLA_MAX+1];
2514         struct net_device *dev = NULL;
2515         struct sk_buff *nskb;
2516         int err;
2517         u32 ext_filter_mask = 0;
2518
2519         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2520         if (err < 0)
2521                 return err;
2522
2523         if (tb[IFLA_IFNAME])
2524                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2525
2526         if (tb[IFLA_EXT_MASK])
2527                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2528
2529         ifm = nlmsg_data(nlh);
2530         if (ifm->ifi_index > 0)
2531                 dev = __dev_get_by_index(net, ifm->ifi_index);
2532         else if (tb[IFLA_IFNAME])
2533                 dev = __dev_get_by_name(net, ifname);
2534         else
2535                 return -EINVAL;
2536
2537         if (dev == NULL)
2538                 return -ENODEV;
2539
2540         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2541         if (nskb == NULL)
2542                 return -ENOBUFS;
2543
2544         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2545                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2546         if (err < 0) {
2547                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2548                 WARN_ON(err == -EMSGSIZE);
2549                 kfree_skb(nskb);
2550         } else
2551                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2552
2553         return err;
2554 }
2555
2556 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2557 {
2558         struct net *net = sock_net(skb->sk);
2559         struct net_device *dev;
2560         struct nlattr *tb[IFLA_MAX+1];
2561         u32 ext_filter_mask = 0;
2562         u16 min_ifinfo_dump_size = 0;
2563         int hdrlen;
2564
2565         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2566         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2567                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2568
2569         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2570                 if (tb[IFLA_EXT_MASK])
2571                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2572         }
2573
2574         if (!ext_filter_mask)
2575                 return NLMSG_GOODSIZE;
2576         /*
2577          * traverse the list of net devices and compute the minimum
2578          * buffer size based upon the filter mask.
2579          */
2580         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2581                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2582                                              if_nlmsg_size(dev,
2583                                                            ext_filter_mask));
2584         }
2585
2586         return min_ifinfo_dump_size;
2587 }
2588
2589 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2590 {
2591         int idx;
2592         int s_idx = cb->family;
2593
2594         if (s_idx == 0)
2595                 s_idx = 1;
2596         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2597                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2598                 if (idx < s_idx || idx == PF_PACKET)
2599                         continue;
2600                 if (rtnl_msg_handlers[idx] == NULL ||
2601                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2602                         continue;
2603                 if (idx > s_idx) {
2604                         memset(&cb->args[0], 0, sizeof(cb->args));
2605                         cb->prev_seq = 0;
2606                         cb->seq = 0;
2607                 }
2608                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2609                         break;
2610         }
2611         cb->family = idx;
2612
2613         return skb->len;
2614 }
2615
2616 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2617                                        unsigned int change, gfp_t flags)
2618 {
2619         struct net *net = dev_net(dev);
2620         struct sk_buff *skb;
2621         int err = -ENOBUFS;
2622         size_t if_info_size;
2623
2624         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2625         if (skb == NULL)
2626                 goto errout;
2627
2628         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2629         if (err < 0) {
2630                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2631                 WARN_ON(err == -EMSGSIZE);
2632                 kfree_skb(skb);
2633                 goto errout;
2634         }
2635         return skb;
2636 errout:
2637         if (err < 0)
2638                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2639         return NULL;
2640 }
2641
2642 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2643 {
2644         struct net *net = dev_net(dev);
2645
2646         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2647 }
2648
2649 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2650                   gfp_t flags)
2651 {
2652         struct sk_buff *skb;
2653
2654         if (dev->reg_state != NETREG_REGISTERED)
2655                 return;
2656
2657         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2658         if (skb)
2659                 rtmsg_ifinfo_send(skb, dev, flags);
2660 }
2661 EXPORT_SYMBOL(rtmsg_ifinfo);
2662
2663 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2664                                    struct net_device *dev,
2665                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2666                                    int type, unsigned int flags,
2667                                    int nlflags, u16 ndm_state)
2668 {
2669         struct nlmsghdr *nlh;
2670         struct ndmsg *ndm;
2671
2672         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2673         if (!nlh)
2674                 return -EMSGSIZE;
2675
2676         ndm = nlmsg_data(nlh);
2677         ndm->ndm_family  = AF_BRIDGE;
2678         ndm->ndm_pad1    = 0;
2679         ndm->ndm_pad2    = 0;
2680         ndm->ndm_flags   = flags;
2681         ndm->ndm_type    = 0;
2682         ndm->ndm_ifindex = dev->ifindex;
2683         ndm->ndm_state   = ndm_state;
2684
2685         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2686                 goto nla_put_failure;
2687         if (vid)
2688                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2689                         goto nla_put_failure;
2690
2691         nlmsg_end(skb, nlh);
2692         return 0;
2693
2694 nla_put_failure:
2695         nlmsg_cancel(skb, nlh);
2696         return -EMSGSIZE;
2697 }
2698
2699 static inline size_t rtnl_fdb_nlmsg_size(void)
2700 {
2701         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2702 }
2703
2704 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2705                             u16 ndm_state)
2706 {
2707         struct net *net = dev_net(dev);
2708         struct sk_buff *skb;
2709         int err = -ENOBUFS;
2710
2711         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2712         if (!skb)
2713                 goto errout;
2714
2715         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2716                                       0, 0, type, NTF_SELF, 0, ndm_state);
2717         if (err < 0) {
2718                 kfree_skb(skb);
2719                 goto errout;
2720         }
2721
2722         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2723         return;
2724 errout:
2725         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2726 }
2727
2728 /**
2729  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2730  */
2731 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2732                      struct nlattr *tb[],
2733                      struct net_device *dev,
2734                      const unsigned char *addr, u16 vid,
2735                      u16 flags)
2736 {
2737         int err = -EINVAL;
2738
2739         /* If aging addresses are supported device will need to
2740          * implement its own handler for this.
2741          */
2742         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2743                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2744                 return err;
2745         }
2746
2747         if (vid) {
2748                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2749                 return err;
2750         }
2751
2752         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2753                 err = dev_uc_add_excl(dev, addr);
2754         else if (is_multicast_ether_addr(addr))
2755                 err = dev_mc_add_excl(dev, addr);
2756
2757         /* Only return duplicate errors if NLM_F_EXCL is set */
2758         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2759                 err = 0;
2760
2761         return err;
2762 }
2763 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2764
2765 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2766 {
2767         u16 vid = 0;
2768
2769         if (vlan_attr) {
2770                 if (nla_len(vlan_attr) != sizeof(u16)) {
2771                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2772                         return -EINVAL;
2773                 }
2774
2775                 vid = nla_get_u16(vlan_attr);
2776
2777                 if (!vid || vid >= VLAN_VID_MASK) {
2778                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2779                                 vid);
2780                         return -EINVAL;
2781                 }
2782         }
2783         *p_vid = vid;
2784         return 0;
2785 }
2786
2787 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2788 {
2789         struct net *net = sock_net(skb->sk);
2790         struct ndmsg *ndm;
2791         struct nlattr *tb[NDA_MAX+1];
2792         struct net_device *dev;
2793         u8 *addr;
2794         u16 vid;
2795         int err;
2796
2797         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2798         if (err < 0)
2799                 return err;
2800
2801         ndm = nlmsg_data(nlh);
2802         if (ndm->ndm_ifindex == 0) {
2803                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2804                 return -EINVAL;
2805         }
2806
2807         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2808         if (dev == NULL) {
2809                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2810                 return -ENODEV;
2811         }
2812
2813         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2814                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2815                 return -EINVAL;
2816         }
2817
2818         addr = nla_data(tb[NDA_LLADDR]);
2819
2820         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2821         if (err)
2822                 return err;
2823
2824         err = -EOPNOTSUPP;
2825
2826         /* Support fdb on master device the net/bridge default case */
2827         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2828             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2829                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2830                 const struct net_device_ops *ops = br_dev->netdev_ops;
2831
2832                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2833                                        nlh->nlmsg_flags);
2834                 if (err)
2835                         goto out;
2836                 else
2837                         ndm->ndm_flags &= ~NTF_MASTER;
2838         }
2839
2840         /* Embedded bridge, macvlan, and any other device support */
2841         if ((ndm->ndm_flags & NTF_SELF)) {
2842                 if (dev->netdev_ops->ndo_fdb_add)
2843                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2844                                                            vid,
2845                                                            nlh->nlmsg_flags);
2846                 else
2847                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2848                                                nlh->nlmsg_flags);
2849
2850                 if (!err) {
2851                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2852                                         ndm->ndm_state);
2853                         ndm->ndm_flags &= ~NTF_SELF;
2854                 }
2855         }
2856 out:
2857         return err;
2858 }
2859
2860 /**
2861  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2862  */
2863 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2864                      struct nlattr *tb[],
2865                      struct net_device *dev,
2866                      const unsigned char *addr, u16 vid)
2867 {
2868         int err = -EINVAL;
2869
2870         /* If aging addresses are supported device will need to
2871          * implement its own handler for this.
2872          */
2873         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2874                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2875                 return err;
2876         }
2877
2878         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2879                 err = dev_uc_del(dev, addr);
2880         else if (is_multicast_ether_addr(addr))
2881                 err = dev_mc_del(dev, addr);
2882
2883         return err;
2884 }
2885 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2886
2887 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2888 {
2889         struct net *net = sock_net(skb->sk);
2890         struct ndmsg *ndm;
2891         struct nlattr *tb[NDA_MAX+1];
2892         struct net_device *dev;
2893         int err = -EINVAL;
2894         __u8 *addr;
2895         u16 vid;
2896
2897         if (!netlink_capable(skb, CAP_NET_ADMIN))
2898                 return -EPERM;
2899
2900         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2901         if (err < 0)
2902                 return err;
2903
2904         ndm = nlmsg_data(nlh);
2905         if (ndm->ndm_ifindex == 0) {
2906                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2907                 return -EINVAL;
2908         }
2909
2910         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2911         if (dev == NULL) {
2912                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2913                 return -ENODEV;
2914         }
2915
2916         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2917                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2918                 return -EINVAL;
2919         }
2920
2921         addr = nla_data(tb[NDA_LLADDR]);
2922
2923         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2924         if (err)
2925                 return err;
2926
2927         err = -EOPNOTSUPP;
2928
2929         /* Support fdb on master device the net/bridge default case */
2930         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2931             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2932                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2933                 const struct net_device_ops *ops = br_dev->netdev_ops;
2934
2935                 if (ops->ndo_fdb_del)
2936                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2937
2938                 if (err)
2939                         goto out;
2940                 else
2941                         ndm->ndm_flags &= ~NTF_MASTER;
2942         }
2943
2944         /* Embedded bridge, macvlan, and any other device support */
2945         if (ndm->ndm_flags & NTF_SELF) {
2946                 if (dev->netdev_ops->ndo_fdb_del)
2947                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2948                                                            vid);
2949                 else
2950                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2951
2952                 if (!err) {
2953                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2954                                         ndm->ndm_state);
2955                         ndm->ndm_flags &= ~NTF_SELF;
2956                 }
2957         }
2958 out:
2959         return err;
2960 }
2961
2962 static int nlmsg_populate_fdb(struct sk_buff *skb,
2963                               struct netlink_callback *cb,
2964                               struct net_device *dev,
2965                               int *idx,
2966                               struct netdev_hw_addr_list *list)
2967 {
2968         struct netdev_hw_addr *ha;
2969         int err;
2970         u32 portid, seq;
2971
2972         portid = NETLINK_CB(cb->skb).portid;
2973         seq = cb->nlh->nlmsg_seq;
2974
2975         list_for_each_entry(ha, &list->list, list) {
2976                 if (*idx < cb->args[0])
2977                         goto skip;
2978
2979                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2980                                               portid, seq,
2981                                               RTM_NEWNEIGH, NTF_SELF,
2982                                               NLM_F_MULTI, NUD_PERMANENT);
2983                 if (err < 0)
2984                         return err;
2985 skip:
2986                 *idx += 1;
2987         }
2988         return 0;
2989 }
2990
2991 /**
2992  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2993  * @nlh: netlink message header
2994  * @dev: netdevice
2995  *
2996  * Default netdevice operation to dump the existing unicast address list.
2997  * Returns number of addresses from list put in skb.
2998  */
2999 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3000                       struct netlink_callback *cb,
3001                       struct net_device *dev,
3002                       struct net_device *filter_dev,
3003                       int idx)
3004 {
3005         int err;
3006
3007         netif_addr_lock_bh(dev);
3008         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
3009         if (err)
3010                 goto out;
3011         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
3012 out:
3013         netif_addr_unlock_bh(dev);
3014         cb->args[1] = err;
3015         return idx;
3016 }
3017 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3018
3019 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3020 {
3021         struct net_device *dev;
3022         struct nlattr *tb[IFLA_MAX+1];
3023         struct net_device *br_dev = NULL;
3024         const struct net_device_ops *ops = NULL;
3025         const struct net_device_ops *cops = NULL;
3026         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3027         struct net *net = sock_net(skb->sk);
3028         int brport_idx = 0;
3029         int br_idx = 0;
3030         int idx = 0;
3031
3032         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3033                         ifla_policy) == 0) {
3034                 if (tb[IFLA_MASTER])
3035                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3036         }
3037
3038         brport_idx = ifm->ifi_index;
3039
3040         if (br_idx) {
3041                 br_dev = __dev_get_by_index(net, br_idx);
3042                 if (!br_dev)
3043                         return -ENODEV;
3044
3045                 ops = br_dev->netdev_ops;
3046         }
3047
3048         cb->args[1] = 0;
3049         for_each_netdev(net, dev) {
3050                 if (brport_idx && (dev->ifindex != brport_idx))
3051                         continue;
3052
3053                 if (!br_idx) { /* user did not specify a specific bridge */
3054                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3055                                 br_dev = netdev_master_upper_dev_get(dev);
3056                                 cops = br_dev->netdev_ops;
3057                         }
3058
3059                 } else {
3060                         if (dev != br_dev &&
3061                             !(dev->priv_flags & IFF_BRIDGE_PORT))
3062                                 continue;
3063
3064                         if (br_dev != netdev_master_upper_dev_get(dev) &&
3065                             !(dev->priv_flags & IFF_EBRIDGE))
3066                                 continue;
3067
3068                         cops = ops;
3069                 }
3070
3071                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3072                         if (cops && cops->ndo_fdb_dump)
3073                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3074                                                          idx);
3075                 }
3076                 if (cb->args[1] == -EMSGSIZE)
3077                         break;
3078
3079                 if (dev->netdev_ops->ndo_fdb_dump)
3080                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3081                                                             idx);
3082                 else
3083                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3084                 if (cb->args[1] == -EMSGSIZE)
3085                         break;
3086
3087                 cops = NULL;
3088         }
3089
3090         cb->args[0] = idx;
3091         return skb->len;
3092 }
3093
3094 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3095                                unsigned int attrnum, unsigned int flag)
3096 {
3097         if (mask & flag)
3098                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3099         return 0;
3100 }
3101
3102 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3103                             struct net_device *dev, u16 mode,
3104                             u32 flags, u32 mask, int nlflags,
3105                             u32 filter_mask,
3106                             int (*vlan_fill)(struct sk_buff *skb,
3107                                              struct net_device *dev,
3108                                              u32 filter_mask))
3109 {
3110         struct nlmsghdr *nlh;
3111         struct ifinfomsg *ifm;
3112         struct nlattr *br_afspec;
3113         struct nlattr *protinfo;
3114         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3115         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3116         int err = 0;
3117
3118         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3119         if (nlh == NULL)
3120                 return -EMSGSIZE;
3121
3122         ifm = nlmsg_data(nlh);
3123         ifm->ifi_family = AF_BRIDGE;
3124         ifm->__ifi_pad = 0;
3125         ifm->ifi_type = dev->type;
3126         ifm->ifi_index = dev->ifindex;
3127         ifm->ifi_flags = dev_get_flags(dev);
3128         ifm->ifi_change = 0;
3129
3130
3131         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3132             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3133             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3134             (br_dev &&
3135              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3136             (dev->addr_len &&
3137              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3138             (dev->ifindex != dev_get_iflink(dev) &&
3139              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3140                 goto nla_put_failure;
3141
3142         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3143         if (!br_afspec)
3144                 goto nla_put_failure;
3145
3146         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3147                 nla_nest_cancel(skb, br_afspec);
3148                 goto nla_put_failure;
3149         }
3150
3151         if (mode != BRIDGE_MODE_UNDEF) {
3152                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3153                         nla_nest_cancel(skb, br_afspec);
3154                         goto nla_put_failure;
3155                 }
3156         }
3157         if (vlan_fill) {
3158                 err = vlan_fill(skb, dev, filter_mask);
3159                 if (err) {
3160                         nla_nest_cancel(skb, br_afspec);
3161                         goto nla_put_failure;
3162                 }
3163         }
3164         nla_nest_end(skb, br_afspec);
3165
3166         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3167         if (!protinfo)
3168                 goto nla_put_failure;
3169
3170         if (brport_nla_put_flag(skb, flags, mask,
3171                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3172             brport_nla_put_flag(skb, flags, mask,
3173                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3174             brport_nla_put_flag(skb, flags, mask,
3175                                 IFLA_BRPORT_FAST_LEAVE,
3176                                 BR_MULTICAST_FAST_LEAVE) ||
3177             brport_nla_put_flag(skb, flags, mask,
3178                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3179             brport_nla_put_flag(skb, flags, mask,
3180                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3181             brport_nla_put_flag(skb, flags, mask,
3182                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3183             brport_nla_put_flag(skb, flags, mask,
3184                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3185             brport_nla_put_flag(skb, flags, mask,
3186                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3187                 nla_nest_cancel(skb, protinfo);
3188                 goto nla_put_failure;
3189         }
3190
3191         nla_nest_end(skb, protinfo);
3192
3193         nlmsg_end(skb, nlh);
3194         return 0;
3195 nla_put_failure:
3196         nlmsg_cancel(skb, nlh);
3197         return err ? err : -EMSGSIZE;
3198 }
3199 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3200
3201 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3202 {
3203         struct net *net = sock_net(skb->sk);
3204         struct net_device *dev;
3205         int idx = 0;
3206         u32 portid = NETLINK_CB(cb->skb).portid;
3207         u32 seq = cb->nlh->nlmsg_seq;
3208         u32 filter_mask = 0;
3209         int err;
3210
3211         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3212                 struct nlattr *extfilt;
3213
3214                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3215                                           IFLA_EXT_MASK);
3216                 if (extfilt) {
3217                         if (nla_len(extfilt) < sizeof(filter_mask))
3218                                 return -EINVAL;
3219
3220                         filter_mask = nla_get_u32(extfilt);
3221                 }
3222         }
3223
3224         rcu_read_lock();
3225         for_each_netdev_rcu(net, dev) {
3226                 const struct net_device_ops *ops = dev->netdev_ops;
3227                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3228
3229                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3230                         if (idx >= cb->args[0]) {
3231                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3232                                                 skb, portid, seq, dev,
3233                                                 filter_mask, NLM_F_MULTI);
3234                                 if (err < 0 && err != -EOPNOTSUPP)
3235                                         break;
3236                         }
3237                         idx++;
3238                 }
3239
3240                 if (ops->ndo_bridge_getlink) {
3241                         if (idx >= cb->args[0]) {
3242                                 err = ops->ndo_bridge_getlink(skb, portid,
3243                                                               seq, dev,
3244                                                               filter_mask,
3245                                                               NLM_F_MULTI);
3246                                 if (err < 0 && err != -EOPNOTSUPP)
3247                                         break;
3248                         }
3249                         idx++;
3250                 }
3251         }
3252         rcu_read_unlock();
3253         cb->args[0] = idx;
3254
3255         return skb->len;
3256 }
3257
3258 static inline size_t bridge_nlmsg_size(void)
3259 {
3260         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3261                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3262                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3263                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3264                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3265                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3266                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3267                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3268                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3269                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3270                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3271 }
3272
3273 static int rtnl_bridge_notify(struct net_device *dev)
3274 {
3275         struct net *net = dev_net(dev);
3276         struct sk_buff *skb;
3277         int err = -EOPNOTSUPP;
3278
3279         if (!dev->netdev_ops->ndo_bridge_getlink)
3280                 return 0;
3281
3282         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3283         if (!skb) {
3284                 err = -ENOMEM;
3285                 goto errout;
3286         }
3287
3288         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3289         if (err < 0)
3290                 goto errout;
3291
3292         if (!skb->len)
3293                 goto errout;
3294
3295         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3296         return 0;
3297 errout:
3298         WARN_ON(err == -EMSGSIZE);
3299         kfree_skb(skb);
3300         if (err)
3301                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3302         return err;
3303 }
3304
3305 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3306 {
3307         struct net *net = sock_net(skb->sk);
3308         struct ifinfomsg *ifm;
3309         struct net_device *dev;
3310         struct nlattr *br_spec, *attr = NULL;
3311         int rem, err = -EOPNOTSUPP;
3312         u16 flags = 0;
3313         bool have_flags = false;
3314
3315         if (nlmsg_len(nlh) < sizeof(*ifm))
3316                 return -EINVAL;
3317
3318         ifm = nlmsg_data(nlh);
3319         if (ifm->ifi_family != AF_BRIDGE)
3320                 return -EPFNOSUPPORT;
3321
3322         dev = __dev_get_by_index(net, ifm->ifi_index);
3323         if (!dev) {
3324                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3325                 return -ENODEV;
3326         }
3327
3328         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3329         if (br_spec) {
3330                 nla_for_each_nested(attr, br_spec, rem) {
3331                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3332                                 if (nla_len(attr) < sizeof(flags))
3333                                         return -EINVAL;
3334
3335                                 have_flags = true;
3336                                 flags = nla_get_u16(attr);
3337                                 break;
3338                         }
3339                 }
3340         }
3341
3342         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3343                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3344
3345                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3346                         err = -EOPNOTSUPP;
3347                         goto out;
3348                 }
3349
3350                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3351                 if (err)
3352                         goto out;
3353
3354                 flags &= ~BRIDGE_FLAGS_MASTER;
3355         }
3356
3357         if ((flags & BRIDGE_FLAGS_SELF)) {
3358                 if (!dev->netdev_ops->ndo_bridge_setlink)
3359                         err = -EOPNOTSUPP;
3360                 else
3361                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3362                                                                   flags);
3363                 if (!err) {
3364                         flags &= ~BRIDGE_FLAGS_SELF;
3365
3366                         /* Generate event to notify upper layer of bridge
3367                          * change
3368                          */
3369                         err = rtnl_bridge_notify(dev);
3370                 }
3371         }
3372
3373         if (have_flags)
3374                 memcpy(nla_data(attr), &flags, sizeof(flags));
3375 out:
3376         return err;
3377 }
3378
3379 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3380 {
3381         struct net *net = sock_net(skb->sk);
3382         struct ifinfomsg *ifm;
3383         struct net_device *dev;
3384         struct nlattr *br_spec, *attr = NULL;
3385         int rem, err = -EOPNOTSUPP;
3386         u16 flags = 0;
3387         bool have_flags = false;
3388
3389         if (nlmsg_len(nlh) < sizeof(*ifm))
3390                 return -EINVAL;
3391
3392         ifm = nlmsg_data(nlh);
3393         if (ifm->ifi_family != AF_BRIDGE)
3394                 return -EPFNOSUPPORT;
3395
3396         dev = __dev_get_by_index(net, ifm->ifi_index);
3397         if (!dev) {
3398                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3399                 return -ENODEV;
3400         }
3401
3402         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3403         if (br_spec) {
3404                 nla_for_each_nested(attr, br_spec, rem) {
3405                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3406                                 if (nla_len(attr) < sizeof(flags))
3407                                         return -EINVAL;
3408
3409                                 have_flags = true;
3410                                 flags = nla_get_u16(attr);
3411                                 break;
3412                         }
3413                 }
3414         }
3415
3416         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3417                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3418
3419                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3420                         err = -EOPNOTSUPP;
3421                         goto out;
3422                 }
3423
3424                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3425                 if (err)
3426                         goto out;
3427
3428                 flags &= ~BRIDGE_FLAGS_MASTER;
3429         }
3430
3431         if ((flags & BRIDGE_FLAGS_SELF)) {
3432                 if (!dev->netdev_ops->ndo_bridge_dellink)
3433                         err = -EOPNOTSUPP;
3434                 else
3435                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3436                                                                   flags);
3437
3438                 if (!err) {
3439                         flags &= ~BRIDGE_FLAGS_SELF;
3440
3441                         /* Generate event to notify upper layer of bridge
3442                          * change
3443                          */
3444                         err = rtnl_bridge_notify(dev);
3445                 }
3446         }
3447
3448         if (have_flags)
3449                 memcpy(nla_data(attr), &flags, sizeof(flags));
3450 out:
3451         return err;
3452 }
3453
3454 /* Process one rtnetlink message. */
3455
3456 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3457 {
3458         struct net *net = sock_net(skb->sk);
3459         rtnl_doit_func doit;
3460         int kind;
3461         int family;
3462         int type;
3463         int err;
3464
3465         type = nlh->nlmsg_type;
3466         if (type > RTM_MAX)
3467                 return -EOPNOTSUPP;
3468
3469         type -= RTM_BASE;
3470
3471         /* All the messages must have at least 1 byte length */
3472         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3473                 return 0;
3474
3475         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3476         kind = type&3;
3477
3478         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3479                 return -EPERM;
3480
3481         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3482                 struct sock *rtnl;
3483                 rtnl_dumpit_func dumpit;
3484                 rtnl_calcit_func calcit;
3485                 u16 min_dump_alloc = 0;
3486
3487                 dumpit = rtnl_get_dumpit(family, type);
3488                 if (dumpit == NULL)
3489                         return -EOPNOTSUPP;
3490                 calcit = rtnl_get_calcit(family, type);
3491                 if (calcit)
3492                         min_dump_alloc = calcit(skb, nlh);
3493
3494                 __rtnl_unlock();
3495                 rtnl = net->rtnl;
3496                 {
3497                         struct netlink_dump_control c = {
3498                                 .dump           = dumpit,
3499                                 .min_dump_alloc = min_dump_alloc,
3500                         };
3501                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3502                 }
3503                 rtnl_lock();
3504                 return err;
3505         }
3506
3507         doit = rtnl_get_doit(family, type);
3508         if (doit == NULL)
3509                 return -EOPNOTSUPP;
3510
3511         return doit(skb, nlh);
3512 }
3513
3514 static void rtnetlink_rcv(struct sk_buff *skb)
3515 {
3516         rtnl_lock();
3517         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3518         rtnl_unlock();
3519 }
3520
3521 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3522 {
3523         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3524
3525         switch (event) {
3526         case NETDEV_UP:
3527         case NETDEV_DOWN:
3528         case NETDEV_PRE_UP:
3529         case NETDEV_POST_INIT:
3530         case NETDEV_REGISTER:
3531         case NETDEV_CHANGE:
3532         case NETDEV_PRE_TYPE_CHANGE:
3533         case NETDEV_GOING_DOWN:
3534         case NETDEV_UNREGISTER:
3535         case NETDEV_UNREGISTER_FINAL:
3536         case NETDEV_RELEASE:
3537         case NETDEV_JOIN:
3538         case NETDEV_BONDING_INFO:
3539                 break;
3540         default:
3541                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3542                 break;
3543         }
3544         return NOTIFY_DONE;
3545 }
3546
3547 static struct notifier_block rtnetlink_dev_notifier = {
3548         .notifier_call  = rtnetlink_event,
3549 };
3550
3551
3552 static int __net_init rtnetlink_net_init(struct net *net)
3553 {
3554         struct sock *sk;
3555         struct netlink_kernel_cfg cfg = {
3556                 .groups         = RTNLGRP_MAX,
3557                 .input          = rtnetlink_rcv,
3558                 .cb_mutex       = &rtnl_mutex,
3559                 .flags          = NL_CFG_F_NONROOT_RECV,
3560         };
3561
3562         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3563         if (!sk)
3564                 return -ENOMEM;
3565         net->rtnl = sk;
3566         return 0;
3567 }
3568
3569 static void __net_exit rtnetlink_net_exit(struct net *net)
3570 {
3571         netlink_kernel_release(net->rtnl);
3572         net->rtnl = NULL;
3573 }
3574
3575 static struct pernet_operations rtnetlink_net_ops = {
3576         .init = rtnetlink_net_init,
3577         .exit = rtnetlink_net_exit,
3578 };
3579
3580 void __init rtnetlink_init(void)
3581 {
3582         if (register_pernet_subsys(&rtnetlink_net_ops))
3583                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3584
3585         register_netdevice_notifier(&rtnetlink_dev_notifier);
3586
3587         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3588                       rtnl_dump_ifinfo, rtnl_calcit);
3589         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3590         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3591         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3592
3593         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3594         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3595
3596         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3597         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3598         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3599
3600         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3601         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3602         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3603 }