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 int 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 {
 502         struct rtnl_af_ops *af_ops;
 503         size_t size;
 504
 505         /* IFLA_AF_SPEC */
 506         size = nla_total_size(sizeof(struct nlattr));
 507
 508         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 509                 if (af_ops->get_link_af_size) {
 510                         /* AF_* + nested data */
 511                         size += nla_total_size(sizeof(struct nlattr)) +
 512                                 af_ops->get_link_af_size(dev);
 513                 }
 514         }
 515
 516         return size;
 517 }
 518
 519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 520 {
 521         struct net_device *master_dev;
 522
 523         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 524         if (master_dev && master_dev->rtnl_link_ops)
 525                 return true;
 526         return false;
 527 }
 528
 529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 530                                      const struct net_device *dev)
 531 {
 532         struct net_device *master_dev;
 533         const struct rtnl_link_ops *ops;
 534         struct nlattr *slave_data;
 535         int err;
 536
 537         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 538         if (!master_dev)
 539                 return 0;
 540         ops = master_dev->rtnl_link_ops;
 541         if (!ops)
 542                 return 0;
 543         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 544                 return -EMSGSIZE;
 545         if (ops->fill_slave_info) {
 546                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 547                 if (!slave_data)
 548                         return -EMSGSIZE;
 549                 err = ops->fill_slave_info(skb, master_dev, dev);
 550                 if (err < 0)
 551                         goto err_cancel_slave_data;
 552                 nla_nest_end(skb, slave_data);
 553         }
 554         return 0;
 555
 556 err_cancel_slave_data:
 557         nla_nest_cancel(skb, slave_data);
 558         return err;
 559 }
 560
 561 static int rtnl_link_info_fill(struct sk_buff *skb,
 562                                const struct net_device *dev)
 563 {
 564         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 565         struct nlattr *data;
 566         int err;
 567
 568         if (!ops)
 569                 return 0;
 570         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 571                 return -EMSGSIZE;
 572         if (ops->fill_xstats) {
 573                 err = ops->fill_xstats(skb, dev);
 574                 if (err < 0)
 575                         return err;
 576         }
 577         if (ops->fill_info) {
 578                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 579                 if (data == NULL)
 580                         return -EMSGSIZE;
 581                 err = ops->fill_info(skb, dev);
 582                 if (err < 0)
 583                         goto err_cancel_data;
 584                 nla_nest_end(skb, data);
 585         }
 586         return 0;
 587
 588 err_cancel_data:
 589         nla_nest_cancel(skb, data);
 590         return err;
 591 }
 592
 593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 594 {
 595         struct nlattr *linkinfo;
 596         int err = -EMSGSIZE;
 597
 598         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 599         if (linkinfo == NULL)
 600                 goto out;
 601
 602         err = rtnl_link_info_fill(skb, dev);
 603         if (err < 0)
 604                 goto err_cancel_link;
 605
 606         err = rtnl_link_slave_info_fill(skb, dev);
 607         if (err < 0)
 608                 goto err_cancel_link;
 609
 610         nla_nest_end(skb, linkinfo);
 611         return 0;
 612
 613 err_cancel_link:
 614         nla_nest_cancel(skb, linkinfo);
 615 out:
 616         return err;
 617 }
 618
 619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 620 {
 621         struct sock *rtnl = net->rtnl;
 622         int err = 0;
 623
 624         NETLINK_CB(skb).dst_group = group;
 625         if (echo)
 626                 atomic_inc(&skb->users);
 627         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 628         if (echo)
 629                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 630         return err;
 631 }
 632
 633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 634 {
 635         struct sock *rtnl = net->rtnl;
 636
 637         return nlmsg_unicast(rtnl, skb, pid);
 638 }
 639 EXPORT_SYMBOL(rtnl_unicast);
 640
 641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 642                  struct nlmsghdr *nlh, gfp_t flags)
 643 {
 644         struct sock *rtnl = net->rtnl;
 645         int report = 0;
 646
 647         if (nlh)
 648                 report = nlmsg_report(nlh);
 649
 650         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 651 }
 652 EXPORT_SYMBOL(rtnl_notify);
 653
 654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 655 {
 656         struct sock *rtnl = net->rtnl;
 657
 658         netlink_set_err(rtnl, 0, group, error);
 659 }
 660 EXPORT_SYMBOL(rtnl_set_sk_err);
 661
 662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 663 {
 664         struct nlattr *mx;
 665         int i, valid = 0;
 666
 667         mx = nla_nest_start(skb, RTA_METRICS);
 668         if (mx == NULL)
 669                 return -ENOBUFS;
 670
 671         for (i = 0; i < RTAX_MAX; i++) {
 672                 if (metrics[i]) {
 673                         if (i == RTAX_CC_ALGO - 1) {
 674                                 char tmp[TCP_CA_NAME_MAX], *name;
 675
 676                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 677                                 if (!name)
 678                                         continue;
 679                                 if (nla_put_string(skb, i + 1, name))
 680                                         goto nla_put_failure;
 681                         } else {
 682                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 683                                         goto nla_put_failure;
 684                         }
 685                         valid++;
 686                 }
 687         }
 688
 689         if (!valid) {
 690                 nla_nest_cancel(skb, mx);
 691                 return 0;
 692         }
 693
 694         return nla_nest_end(skb, mx);
 695
 696 nla_put_failure:
 697         nla_nest_cancel(skb, mx);
 698         return -EMSGSIZE;
 699 }
 700 EXPORT_SYMBOL(rtnetlink_put_metrics);
 701
 702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 703                        long expires, u32 error)
 704 {
 705         struct rta_cacheinfo ci = {
 706                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 707                 .rta_used = dst->__use,
 708                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 709                 .rta_error = error,
 710                 .rta_id =  id,
 711         };
 712
 713         if (expires) {
 714                 unsigned long clock;
 715
 716                 clock = jiffies_to_clock_t(abs(expires));
 717                 clock = min_t(unsigned long, clock, INT_MAX);
 718                 ci.rta_expires = (expires > 0) ? clock : -clock;
 719         }
 720         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 721 }
 722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 723
 724 static void set_operstate(struct net_device *dev, unsigned char transition)
 725 {
 726         unsigned char operstate = dev->operstate;
 727
 728         switch (transition) {
 729         case IF_OPER_UP:
 730                 if ((operstate == IF_OPER_DORMANT ||
 731                      operstate == IF_OPER_UNKNOWN) &&
 732                     !netif_dormant(dev))
 733                         operstate = IF_OPER_UP;
 734                 break;
 735
 736         case IF_OPER_DORMANT:
 737                 if (operstate == IF_OPER_UP ||
 738                     operstate == IF_OPER_UNKNOWN)
 739                         operstate = IF_OPER_DORMANT;
 740                 break;
 741         }
 742
 743         if (dev->operstate != operstate) {
 744                 write_lock_bh(&dev_base_lock);
 745                 dev->operstate = operstate;
 746                 write_unlock_bh(&dev_base_lock);
 747                 netdev_state_change(dev);
 748         }
 749 }
 750
 751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 752 {
 753         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 754                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 755 }
 756
 757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 758                                            const struct ifinfomsg *ifm)
 759 {
 760         unsigned int flags = ifm->ifi_flags;
 761
 762         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 763         if (ifm->ifi_change)
 764                 flags = (flags & ifm->ifi_change) |
 765                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 766
 767         return flags;
 768 }
 769
 770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 771                                  const struct rtnl_link_stats64 *b)
 772 {
 773         a->rx_packets = b->rx_packets;
 774         a->tx_packets = b->tx_packets;
 775         a->rx_bytes = b->rx_bytes;
 776         a->tx_bytes = b->tx_bytes;
 777         a->rx_errors = b->rx_errors;
 778         a->tx_errors = b->tx_errors;
 779         a->rx_dropped = b->rx_dropped;
 780         a->tx_dropped = b->tx_dropped;
 781
 782         a->multicast = b->multicast;
 783         a->collisions = b->collisions;
 784
 785         a->rx_length_errors = b->rx_length_errors;
 786         a->rx_over_errors = b->rx_over_errors;
 787         a->rx_crc_errors = b->rx_crc_errors;
 788         a->rx_frame_errors = b->rx_frame_errors;
 789         a->rx_fifo_errors = b->rx_fifo_errors;
 790         a->rx_missed_errors = b->rx_missed_errors;
 791
 792         a->tx_aborted_errors = b->tx_aborted_errors;
 793         a->tx_carrier_errors = b->tx_carrier_errors;
 794         a->tx_fifo_errors = b->tx_fifo_errors;
 795         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 796         a->tx_window_errors = b->tx_window_errors;
 797
 798         a->rx_compressed = b->rx_compressed;
 799         a->tx_compressed = b->tx_compressed;
 800 }
 801
 802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 803 {
 804         memcpy(v, b, sizeof(*b));
 805 }
 806
 807 /* All VF info */
 808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 809                                    u32 ext_filter_mask)
 810 {
 811         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 812             (ext_filter_mask & RTEXT_FILTER_VF)) {
 813                 int num_vfs = dev_num_vf(dev->dev.parent);
 814                 size_t size = nla_total_size(sizeof(struct nlattr));
 815                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 816                 size += num_vfs *
 817                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 818                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 819                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 820                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 821                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 822                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 823                          /* IFLA_VF_STATS_RX_PACKETS */
 824                          nla_total_size(sizeof(__u64)) +
 825                          /* IFLA_VF_STATS_TX_PACKETS */
 826                          nla_total_size(sizeof(__u64)) +
 827                          /* IFLA_VF_STATS_RX_BYTES */
 828                          nla_total_size(sizeof(__u64)) +
 829                          /* IFLA_VF_STATS_TX_BYTES */
 830                          nla_total_size(sizeof(__u64)) +
 831                          /* IFLA_VF_STATS_BROADCAST */
 832                          nla_total_size(sizeof(__u64)) +
 833                          /* IFLA_VF_STATS_MULTICAST */
 834                          nla_total_size(sizeof(__u64)));
 835                 return size;
 836         } else
 837                 return 0;
 838 }
 839
 840 static size_t rtnl_port_size(const struct net_device *dev,
 841                              u32 ext_filter_mask)
 842 {
 843         size_t port_size = nla_total_size(4)            /* PORT_VF */
 844                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 845                 + nla_total_size(sizeof(struct ifla_port_vsi))
 846                                                         /* PORT_VSI_TYPE */
 847                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 848                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 849                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 850                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 851         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 852         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 853                 + port_size;
 854         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 855                 + port_size;
 856
 857         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 858             !(ext_filter_mask & RTEXT_FILTER_VF))
 859                 return 0;
 860         if (dev_num_vf(dev->dev.parent))
 861                 return port_self_size + vf_ports_size +
 862                         vf_port_size * dev_num_vf(dev->dev.parent);
 863         else
 864                 return port_self_size;
 865 }
 866
 867 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 868                                      u32 ext_filter_mask)
 869 {
 870         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 871                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 872                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 873                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 874                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 875                + nla_total_size(sizeof(struct rtnl_link_stats))
 876                + nla_total_size(sizeof(struct rtnl_link_stats64))
 877                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 878                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 879                + nla_total_size(4) /* IFLA_TXQLEN */
 880                + nla_total_size(4) /* IFLA_WEIGHT */
 881                + nla_total_size(4) /* IFLA_MTU */
 882                + nla_total_size(4) /* IFLA_LINK */
 883                + nla_total_size(4) /* IFLA_MASTER */
 884                + nla_total_size(1) /* IFLA_CARRIER */
 885                + nla_total_size(4) /* IFLA_PROMISCUITY */
 886                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 887                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 888                + nla_total_size(1) /* IFLA_OPERSTATE */
 889                + nla_total_size(1) /* IFLA_LINKMODE */
 890                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 891                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 892                + nla_total_size(ext_filter_mask
 893                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 894                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 895                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 896                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 897                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 898                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 899                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 900 }
 901
 902 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 903 {
 904         struct nlattr *vf_ports;
 905         struct nlattr *vf_port;
 906         int vf;
 907         int err;
 908
 909         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 910         if (!vf_ports)
 911                 return -EMSGSIZE;
 912
 913         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 914                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 915                 if (!vf_port)
 916                         goto nla_put_failure;
 917                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 918                         goto nla_put_failure;
 919                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 920                 if (err == -EMSGSIZE)
 921                         goto nla_put_failure;
 922                 if (err) {
 923                         nla_nest_cancel(skb, vf_port);
 924                         continue;
 925                 }
 926                 nla_nest_end(skb, vf_port);
 927         }
 928
 929         nla_nest_end(skb, vf_ports);
 930
 931         return 0;
 932
 933 nla_put_failure:
 934         nla_nest_cancel(skb, vf_ports);
 935         return -EMSGSIZE;
 936 }
 937
 938 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 939 {
 940         struct nlattr *port_self;
 941         int err;
 942
 943         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 944         if (!port_self)
 945                 return -EMSGSIZE;
 946
 947         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 948         if (err) {
 949                 nla_nest_cancel(skb, port_self);
 950                 return (err == -EMSGSIZE) ? err : 0;
 951         }
 952
 953         nla_nest_end(skb, port_self);
 954
 955         return 0;
 956 }
 957
 958 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 959                           u32 ext_filter_mask)
 960 {
 961         int err;
 962
 963         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 964             !(ext_filter_mask & RTEXT_FILTER_VF))
 965                 return 0;
 966
 967         err = rtnl_port_self_fill(skb, dev);
 968         if (err)
 969                 return err;
 970
 971         if (dev_num_vf(dev->dev.parent)) {
 972                 err = rtnl_vf_ports_fill(skb, dev);
 973                 if (err)
 974                         return err;
 975         }
 976
 977         return 0;
 978 }
 979
 980 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 981 {
 982         int err;
 983         struct netdev_phys_item_id ppid;
 984
 985         err = dev_get_phys_port_id(dev, &ppid);
 986         if (err) {
 987                 if (err == -EOPNOTSUPP)
 988                         return 0;
 989                 return err;
 990         }
 991
 992         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 993                 return -EMSGSIZE;
 994
 995         return 0;
 996 }
 997
 998 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
 999 {
1000         char name[IFNAMSIZ];
1001         int err;
1002
1003         err = dev_get_phys_port_name(dev, name, sizeof(name));
1004         if (err) {
1005                 if (err == -EOPNOTSUPP)
1006                         return 0;
1007                 return err;
1008         }
1009
1010         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1011                 return -EMSGSIZE;
1012
1013         return 0;
1014 }
1015
1016 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1017 {
1018         int err;
1019         struct switchdev_attr attr = {
1020                 .id = SWITCHDEV_ATTR_PORT_PARENT_ID,
1021                 .flags = SWITCHDEV_F_NO_RECURSE,
1022         };
1023
1024         err = switchdev_port_attr_get(dev, &attr);
1025         if (err) {
1026                 if (err == -EOPNOTSUPP)
1027                         return 0;
1028                 return err;
1029         }
1030
1031         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1032                     attr.u.ppid.id))
1033                 return -EMSGSIZE;
1034
1035         return 0;
1036 }
1037
1038 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1039                             int type, u32 pid, u32 seq, u32 change,
1040                             unsigned int flags, u32 ext_filter_mask)
1041 {
1042         struct ifinfomsg *ifm;
1043         struct nlmsghdr *nlh;
1044         struct rtnl_link_stats64 temp;
1045         const struct rtnl_link_stats64 *stats;
1046         struct nlattr *attr, *af_spec;
1047         struct rtnl_af_ops *af_ops;
1048         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1049
1050         ASSERT_RTNL();
1051         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1052         if (nlh == NULL)
1053                 return -EMSGSIZE;
1054
1055         ifm = nlmsg_data(nlh);
1056         ifm->ifi_family = AF_UNSPEC;
1057         ifm->__ifi_pad = 0;
1058         ifm->ifi_type = dev->type;
1059         ifm->ifi_index = dev->ifindex;
1060         ifm->ifi_flags = dev_get_flags(dev);
1061         ifm->ifi_change = change;
1062
1063         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1064             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1065             nla_put_u8(skb, IFLA_OPERSTATE,
1066                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1067             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1068             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1069             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1070             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1071             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1072 #ifdef CONFIG_RPS
1073             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1074 #endif
1075             (dev->ifindex != dev_get_iflink(dev) &&
1076              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1077             (upper_dev &&
1078              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1079             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1080             (dev->qdisc &&
1081              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1082             (dev->ifalias &&
1083              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1084             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1085                         atomic_read(&dev->carrier_changes)))
1086                 goto nla_put_failure;
1087
1088         if (1) {
1089                 struct rtnl_link_ifmap map = {
1090                         .mem_start   = dev->mem_start,
1091                         .mem_end     = dev->mem_end,
1092                         .base_addr   = dev->base_addr,
1093                         .irq         = dev->irq,
1094                         .dma         = dev->dma,
1095                         .port        = dev->if_port,
1096                 };
1097                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1098                         goto nla_put_failure;
1099         }
1100
1101         if (dev->addr_len) {
1102                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1103                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1104                         goto nla_put_failure;
1105         }
1106
1107         if (rtnl_phys_port_id_fill(skb, dev))
1108                 goto nla_put_failure;
1109
1110         if (rtnl_phys_port_name_fill(skb, dev))
1111                 goto nla_put_failure;
1112
1113         if (rtnl_phys_switch_id_fill(skb, dev))
1114                 goto nla_put_failure;
1115
1116         attr = nla_reserve(skb, IFLA_STATS,
1117                         sizeof(struct rtnl_link_stats));
1118         if (attr == NULL)
1119                 goto nla_put_failure;
1120
1121         stats = dev_get_stats(dev, &temp);
1122         copy_rtnl_link_stats(nla_data(attr), stats);
1123
1124         attr = nla_reserve(skb, IFLA_STATS64,
1125                         sizeof(struct rtnl_link_stats64));
1126         if (attr == NULL)
1127                 goto nla_put_failure;
1128         copy_rtnl_link_stats64(nla_data(attr), stats);
1129
1130         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1131             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1132                 goto nla_put_failure;
1133
1134         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1135             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1136                 int i;
1137
1138                 struct nlattr *vfinfo, *vf, *vfstats;
1139                 int num_vfs = dev_num_vf(dev->dev.parent);
1140
1141                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1142                 if (!vfinfo)
1143                         goto nla_put_failure;
1144                 for (i = 0; i < num_vfs; i++) {
1145                         struct ifla_vf_info ivi;
1146                         struct ifla_vf_mac vf_mac;
1147                         struct ifla_vf_vlan vf_vlan;
1148                         struct ifla_vf_rate vf_rate;
1149                         struct ifla_vf_tx_rate vf_tx_rate;
1150                         struct ifla_vf_spoofchk vf_spoofchk;
1151                         struct ifla_vf_link_state vf_linkstate;
1152                         struct ifla_vf_rss_query_en vf_rss_query_en;
1153                         struct ifla_vf_stats vf_stats;
1154
1155                         /*
1156                          * Not all SR-IOV capable drivers support the
1157                          * spoofcheck and "RSS query enable" query.  Preset to
1158                          * -1 so the user space tool can detect that the driver
1159                          * didn't report anything.
1160                          */
1161                         ivi.spoofchk = -1;
1162                         ivi.rss_query_en = -1;
1163                         memset(ivi.mac, 0, sizeof(ivi.mac));
1164                         /* The default value for VF link state is "auto"
1165                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1166                          */
1167                         ivi.linkstate = 0;
1168                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1169                                 break;
1170                         vf_mac.vf =
1171                                 vf_vlan.vf =
1172                                 vf_rate.vf =
1173                                 vf_tx_rate.vf =
1174                                 vf_spoofchk.vf =
1175                                 vf_linkstate.vf =
1176                                 vf_rss_query_en.vf = ivi.vf;
1177
1178                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1179                         vf_vlan.vlan = ivi.vlan;
1180                         vf_vlan.qos = ivi.qos;
1181                         vf_tx_rate.rate = ivi.max_tx_rate;
1182                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1183                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1184                         vf_spoofchk.setting = ivi.spoofchk;
1185                         vf_linkstate.link_state = ivi.linkstate;
1186                         vf_rss_query_en.setting = ivi.rss_query_en;
1187                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1188                         if (!vf) {
1189                                 nla_nest_cancel(skb, vfinfo);
1190                                 goto nla_put_failure;
1191                         }
1192                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1193                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1194                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1195                                     &vf_rate) ||
1196                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1197                                     &vf_tx_rate) ||
1198                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1199                                     &vf_spoofchk) ||
1200                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1201                                     &vf_linkstate) ||
1202                             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1203                                     sizeof(vf_rss_query_en),
1204                                     &vf_rss_query_en))
1205                                 goto nla_put_failure;
1206                         memset(&vf_stats, 0, sizeof(vf_stats));
1207                         if (dev->netdev_ops->ndo_get_vf_stats)
1208                                 dev->netdev_ops->ndo_get_vf_stats(dev, i,
1209                                                                   &vf_stats);
1210                         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1211                         if (!vfstats) {
1212                                 nla_nest_cancel(skb, vf);
1213                                 nla_nest_cancel(skb, vfinfo);
1214                                 goto nla_put_failure;
1215                         }
1216                         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1217                                         vf_stats.rx_packets) ||
1218                             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1219                                         vf_stats.tx_packets) ||
1220                             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1221                                         vf_stats.rx_bytes) ||
1222                             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1223                                         vf_stats.tx_bytes) ||
1224                             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1225                                         vf_stats.broadcast) ||
1226                             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1227                                         vf_stats.multicast))
1228                                 goto nla_put_failure;
1229                         nla_nest_end(skb, vfstats);
1230                         nla_nest_end(skb, vf);
1231                 }
1232                 nla_nest_end(skb, vfinfo);
1233         }
1234
1235         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1236                 goto nla_put_failure;
1237
1238         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1239                 if (rtnl_link_fill(skb, dev) < 0)
1240                         goto nla_put_failure;
1241         }
1242
1243         if (dev->rtnl_link_ops &&
1244             dev->rtnl_link_ops->get_link_net) {
1245                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1246
1247                 if (!net_eq(dev_net(dev), link_net)) {
1248                         int id = peernet2id_alloc(dev_net(dev), link_net);
1249
1250                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1251                                 goto nla_put_failure;
1252                 }
1253         }
1254
1255         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1256                 goto nla_put_failure;
1257
1258         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1259                 if (af_ops->fill_link_af) {
1260                         struct nlattr *af;
1261                         int err;
1262
1263                         if (!(af = nla_nest_start(skb, af_ops->family)))
1264                                 goto nla_put_failure;
1265
1266                         err = af_ops->fill_link_af(skb, dev);
1267
1268                         /*
1269                          * Caller may return ENODATA to indicate that there
1270                          * was no data to be dumped. This is not an error, it
1271                          * means we should trim the attribute header and
1272                          * continue.
1273                          */
1274                         if (err == -ENODATA)
1275                                 nla_nest_cancel(skb, af);
1276                         else if (err < 0)
1277                                 goto nla_put_failure;
1278
1279                         nla_nest_end(skb, af);
1280                 }
1281         }
1282
1283         nla_nest_end(skb, af_spec);
1284
1285         nlmsg_end(skb, nlh);
1286         return 0;
1287
1288 nla_put_failure:
1289         nlmsg_cancel(skb, nlh);
1290         return -EMSGSIZE;
1291 }
1292
1293 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1294         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1295         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1296         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1297         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1298         [IFLA_MTU]              = { .type = NLA_U32 },
1299         [IFLA_LINK]             = { .type = NLA_U32 },
1300         [IFLA_MASTER]           = { .type = NLA_U32 },
1301         [IFLA_CARRIER]          = { .type = NLA_U8 },
1302         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1303         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1304         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1305         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1306         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1307         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1308         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1309         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1310         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1311         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1312         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1313         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1314         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1315         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1316         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1317         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1318         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1319         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1320         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1321         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1322 };
1323
1324 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1325         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1326         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1327         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1328         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1329 };
1330
1331 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1332         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1333         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1334         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1335         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1336         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1337         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1338         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1339         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1340 };
1341
1342 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1343         [IFLA_VF_STATS_RX_PACKETS]      = { .type = NLA_U64 },
1344         [IFLA_VF_STATS_TX_PACKETS]      = { .type = NLA_U64 },
1345         [IFLA_VF_STATS_RX_BYTES]        = { .type = NLA_U64 },
1346         [IFLA_VF_STATS_TX_BYTES]        = { .type = NLA_U64 },
1347         [IFLA_VF_STATS_BROADCAST]       = { .type = NLA_U64 },
1348         [IFLA_VF_STATS_MULTICAST]       = { .type = NLA_U64 },
1349 };
1350
1351 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1352         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1353         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1354                                     .len = PORT_PROFILE_MAX },
1355         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1356                                     .len = sizeof(struct ifla_port_vsi)},
1357         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1358                                       .len = PORT_UUID_MAX },
1359         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1360                                     .len = PORT_UUID_MAX },
1361         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1362         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1363 };
1364
1365 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1366 {
1367         struct net *net = sock_net(skb->sk);
1368         int h, s_h;
1369         int idx = 0, s_idx;
1370         struct net_device *dev;
1371         struct hlist_head *head;
1372         struct nlattr *tb[IFLA_MAX+1];
1373         u32 ext_filter_mask = 0;
1374         int err;
1375         int hdrlen;
1376
1377         s_h = cb->args[0];
1378         s_idx = cb->args[1];
1379
1380         cb->seq = net->dev_base_seq;
1381
1382         /* A hack to preserve kernel<->userspace interface.
1383          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1384          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1385          * what iproute2 < v3.9.0 used.
1386          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1387          * attribute, its netlink message is shorter than struct ifinfomsg.
1388          */
1389         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1390                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1391
1392         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1393
1394                 if (tb[IFLA_EXT_MASK])
1395                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1396         }
1397
1398         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1399                 idx = 0;
1400                 head = &net->dev_index_head[h];
1401                 hlist_for_each_entry(dev, head, index_hlist) {
1402                         if (idx < s_idx)
1403                                 goto cont;
1404                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1405                                                NETLINK_CB(cb->skb).portid,
1406                                                cb->nlh->nlmsg_seq, 0,
1407                                                NLM_F_MULTI,
1408                                                ext_filter_mask);
1409                         /* If we ran out of room on the first message,
1410                          * we're in trouble
1411                          */
1412                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1413
1414                         if (err < 0)
1415                                 goto out;
1416
1417                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1418 cont:
1419                         idx++;
1420                 }
1421         }
1422 out:
1423         cb->args[1] = idx;
1424         cb->args[0] = h;
1425
1426         return skb->len;
1427 }
1428
1429 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1430 {
1431         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1432 }
1433 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1434
1435 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1436 {
1437         struct net *net;
1438         /* Examine the link attributes and figure out which
1439          * network namespace we are talking about.
1440          */
1441         if (tb[IFLA_NET_NS_PID])
1442                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1443         else if (tb[IFLA_NET_NS_FD])
1444                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1445         else
1446                 net = get_net(src_net);
1447         return net;
1448 }
1449 EXPORT_SYMBOL(rtnl_link_get_net);
1450
1451 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1452 {
1453         if (dev) {
1454                 if (tb[IFLA_ADDRESS] &&
1455                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1456                         return -EINVAL;
1457
1458                 if (tb[IFLA_BROADCAST] &&
1459                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1460                         return -EINVAL;
1461         }
1462
1463         if (tb[IFLA_AF_SPEC]) {
1464                 struct nlattr *af;
1465                 int rem, err;
1466
1467                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1468                         const struct rtnl_af_ops *af_ops;
1469
1470                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1471                                 return -EAFNOSUPPORT;
1472
1473                         if (!af_ops->set_link_af)
1474                                 return -EOPNOTSUPP;
1475
1476                         if (af_ops->validate_link_af) {
1477                                 err = af_ops->validate_link_af(dev, af);
1478                                 if (err < 0)
1479                                         return err;
1480                         }
1481                 }
1482         }
1483
1484         return 0;
1485 }
1486
1487 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1488 {
1489         const struct net_device_ops *ops = dev->netdev_ops;
1490         int err = -EINVAL;
1491
1492         if (tb[IFLA_VF_MAC]) {
1493                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1494
1495                 err = -EOPNOTSUPP;
1496                 if (ops->ndo_set_vf_mac)
1497                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1498                                                   ivm->mac);
1499                 if (err < 0)
1500                         return err;
1501         }
1502
1503         if (tb[IFLA_VF_VLAN]) {
1504                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1505
1506                 err = -EOPNOTSUPP;
1507                 if (ops->ndo_set_vf_vlan)
1508                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1509                                                    ivv->qos);
1510                 if (err < 0)
1511                         return err;
1512         }
1513
1514         if (tb[IFLA_VF_TX_RATE]) {
1515                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1516                 struct ifla_vf_info ivf;
1517
1518                 err = -EOPNOTSUPP;
1519                 if (ops->ndo_get_vf_config)
1520                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1521                 if (err < 0)
1522                         return err;
1523
1524                 err = -EOPNOTSUPP;
1525                 if (ops->ndo_set_vf_rate)
1526                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1527                                                    ivf.min_tx_rate,
1528                                                    ivt->rate);
1529                 if (err < 0)
1530                         return err;
1531         }
1532
1533         if (tb[IFLA_VF_RATE]) {
1534                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1535
1536                 err = -EOPNOTSUPP;
1537                 if (ops->ndo_set_vf_rate)
1538                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1539                                                    ivt->min_tx_rate,
1540                                                    ivt->max_tx_rate);
1541                 if (err < 0)
1542                         return err;
1543         }
1544
1545         if (tb[IFLA_VF_SPOOFCHK]) {
1546                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1547
1548                 err = -EOPNOTSUPP;
1549                 if (ops->ndo_set_vf_spoofchk)
1550                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1551                                                        ivs->setting);
1552                 if (err < 0)
1553                         return err;
1554         }
1555
1556         if (tb[IFLA_VF_LINK_STATE]) {
1557                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1558
1559                 err = -EOPNOTSUPP;
1560                 if (ops->ndo_set_vf_link_state)
1561                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1562                                                          ivl->link_state);
1563                 if (err < 0)
1564                         return err;
1565         }
1566
1567         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1568                 struct ifla_vf_rss_query_en *ivrssq_en;
1569
1570                 err = -EOPNOTSUPP;
1571                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1572                 if (ops->ndo_set_vf_rss_query_en)
1573                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1574                                                            ivrssq_en->setting);
1575                 if (err < 0)
1576                         return err;
1577         }
1578
1579         return err;
1580 }
1581
1582 static int do_set_master(struct net_device *dev, int ifindex)
1583 {
1584         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1585         const struct net_device_ops *ops;
1586         int err;
1587
1588         if (upper_dev) {
1589                 if (upper_dev->ifindex == ifindex)
1590                         return 0;
1591                 ops = upper_dev->netdev_ops;
1592                 if (ops->ndo_del_slave) {
1593                         err = ops->ndo_del_slave(upper_dev, dev);
1594                         if (err)
1595                                 return err;
1596                 } else {
1597                         return -EOPNOTSUPP;
1598                 }
1599         }
1600
1601         if (ifindex) {
1602                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1603                 if (!upper_dev)
1604                         return -EINVAL;
1605                 ops = upper_dev->netdev_ops;
1606                 if (ops->ndo_add_slave) {
1607                         err = ops->ndo_add_slave(upper_dev, dev);
1608                         if (err)
1609                                 return err;
1610                 } else {
1611                         return -EOPNOTSUPP;
1612                 }
1613         }
1614         return 0;
1615 }
1616
1617 #define DO_SETLINK_MODIFIED     0x01
1618 /* notify flag means notify + modified. */
1619 #define DO_SETLINK_NOTIFY       0x03
1620 static int do_setlink(const struct sk_buff *skb,
1621                       struct net_device *dev, struct ifinfomsg *ifm,
1622                       struct nlattr **tb, char *ifname, int status)
1623 {
1624         const struct net_device_ops *ops = dev->netdev_ops;
1625         int err;
1626
1627         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1628                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1629                 if (IS_ERR(net)) {
1630                         err = PTR_ERR(net);
1631                         goto errout;
1632                 }
1633                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1634                         put_net(net);
1635                         err = -EPERM;
1636                         goto errout;
1637                 }
1638                 err = dev_change_net_namespace(dev, net, ifname);
1639                 put_net(net);
1640                 if (err)
1641                         goto errout;
1642                 status |= DO_SETLINK_MODIFIED;
1643         }
1644
1645         if (tb[IFLA_MAP]) {
1646                 struct rtnl_link_ifmap *u_map;
1647                 struct ifmap k_map;
1648
1649                 if (!ops->ndo_set_config) {
1650                         err = -EOPNOTSUPP;
1651                         goto errout;
1652                 }
1653
1654                 if (!netif_device_present(dev)) {
1655                         err = -ENODEV;
1656                         goto errout;
1657                 }
1658
1659                 u_map = nla_data(tb[IFLA_MAP]);
1660                 k_map.mem_start = (unsigned long) u_map->mem_start;
1661                 k_map.mem_end = (unsigned long) u_map->mem_end;
1662                 k_map.base_addr = (unsigned short) u_map->base_addr;
1663                 k_map.irq = (unsigned char) u_map->irq;
1664                 k_map.dma = (unsigned char) u_map->dma;
1665                 k_map.port = (unsigned char) u_map->port;
1666
1667                 err = ops->ndo_set_config(dev, &k_map);
1668                 if (err < 0)
1669                         goto errout;
1670
1671                 status |= DO_SETLINK_NOTIFY;
1672         }
1673
1674         if (tb[IFLA_ADDRESS]) {
1675                 struct sockaddr *sa;
1676                 int len;
1677
1678                 len = sizeof(sa_family_t) + dev->addr_len;
1679                 sa = kmalloc(len, GFP_KERNEL);
1680                 if (!sa) {
1681                         err = -ENOMEM;
1682                         goto errout;
1683                 }
1684                 sa->sa_family = dev->type;
1685                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1686                        dev->addr_len);
1687                 err = dev_set_mac_address(dev, sa);
1688                 kfree(sa);
1689                 if (err)
1690                         goto errout;
1691                 status |= DO_SETLINK_MODIFIED;
1692         }
1693
1694         if (tb[IFLA_MTU]) {
1695                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1696                 if (err < 0)
1697                         goto errout;
1698                 status |= DO_SETLINK_MODIFIED;
1699         }
1700
1701         if (tb[IFLA_GROUP]) {
1702                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1703                 status |= DO_SETLINK_NOTIFY;
1704         }
1705
1706         /*
1707          * Interface selected by interface index but interface
1708          * name provided implies that a name change has been
1709          * requested.
1710          */
1711         if (ifm->ifi_index > 0 && ifname[0]) {
1712                 err = dev_change_name(dev, ifname);
1713                 if (err < 0)
1714                         goto errout;
1715                 status |= DO_SETLINK_MODIFIED;
1716         }
1717
1718         if (tb[IFLA_IFALIAS]) {
1719                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1720                                     nla_len(tb[IFLA_IFALIAS]));
1721                 if (err < 0)
1722                         goto errout;
1723                 status |= DO_SETLINK_NOTIFY;
1724         }
1725
1726         if (tb[IFLA_BROADCAST]) {
1727                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1728                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1729         }
1730
1731         if (ifm->ifi_flags || ifm->ifi_change) {
1732                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1733                 if (err < 0)
1734                         goto errout;
1735         }
1736
1737         if (tb[IFLA_MASTER]) {
1738                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1739                 if (err)
1740                         goto errout;
1741                 status |= DO_SETLINK_MODIFIED;
1742         }
1743
1744         if (tb[IFLA_CARRIER]) {
1745                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1746                 if (err)
1747                         goto errout;
1748                 status |= DO_SETLINK_MODIFIED;
1749         }
1750
1751         if (tb[IFLA_TXQLEN]) {
1752                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1753
1754                 if (dev->tx_queue_len ^ value)
1755                         status |= DO_SETLINK_NOTIFY;
1756
1757                 dev->tx_queue_len = value;
1758         }
1759
1760         if (tb[IFLA_OPERSTATE])
1761                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1762
1763         if (tb[IFLA_LINKMODE]) {
1764                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1765
1766                 write_lock_bh(&dev_base_lock);
1767                 if (dev->link_mode ^ value)
1768                         status |= DO_SETLINK_NOTIFY;
1769                 dev->link_mode = value;
1770                 write_unlock_bh(&dev_base_lock);
1771         }
1772
1773         if (tb[IFLA_VFINFO_LIST]) {
1774                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1775                 struct nlattr *attr;
1776                 int rem;
1777
1778                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1779                         if (nla_type(attr) != IFLA_VF_INFO ||
1780                             nla_len(attr) < NLA_HDRLEN) {
1781                                 err = -EINVAL;
1782                                 goto errout;
1783                         }
1784                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1785                                                ifla_vf_policy);
1786                         if (err < 0)
1787                                 goto errout;
1788                         err = do_setvfinfo(dev, vfinfo);
1789                         if (err < 0)
1790                                 goto errout;
1791                         status |= DO_SETLINK_NOTIFY;
1792                 }
1793         }
1794         err = 0;
1795
1796         if (tb[IFLA_VF_PORTS]) {
1797                 struct nlattr *port[IFLA_PORT_MAX+1];
1798                 struct nlattr *attr;
1799                 int vf;
1800                 int rem;
1801
1802                 err = -EOPNOTSUPP;
1803                 if (!ops->ndo_set_vf_port)
1804                         goto errout;
1805
1806                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1807                         if (nla_type(attr) != IFLA_VF_PORT ||
1808                             nla_len(attr) < NLA_HDRLEN) {
1809                                 err = -EINVAL;
1810                                 goto errout;
1811                         }
1812                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1813                                                ifla_port_policy);
1814                         if (err < 0)
1815                                 goto errout;
1816                         if (!port[IFLA_PORT_VF]) {
1817                                 err = -EOPNOTSUPP;
1818                                 goto errout;
1819                         }
1820                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1821                         err = ops->ndo_set_vf_port(dev, vf, port);
1822                         if (err < 0)
1823                                 goto errout;
1824                         status |= DO_SETLINK_NOTIFY;
1825                 }
1826         }
1827         err = 0;
1828
1829         if (tb[IFLA_PORT_SELF]) {
1830                 struct nlattr *port[IFLA_PORT_MAX+1];
1831
1832                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1833                         tb[IFLA_PORT_SELF], ifla_port_policy);
1834                 if (err < 0)
1835                         goto errout;
1836
1837                 err = -EOPNOTSUPP;
1838                 if (ops->ndo_set_vf_port)
1839                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1840                 if (err < 0)
1841                         goto errout;
1842                 status |= DO_SETLINK_NOTIFY;
1843         }
1844
1845         if (tb[IFLA_AF_SPEC]) {
1846                 struct nlattr *af;
1847                 int rem;
1848
1849                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1850                         const struct rtnl_af_ops *af_ops;
1851
1852                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1853                                 BUG();
1854
1855                         err = af_ops->set_link_af(dev, af);
1856                         if (err < 0)
1857                                 goto errout;
1858
1859                         status |= DO_SETLINK_NOTIFY;
1860                 }
1861         }
1862         err = 0;
1863
1864 errout:
1865         if (status & DO_SETLINK_MODIFIED) {
1866                 if (status & DO_SETLINK_NOTIFY)
1867                         netdev_state_change(dev);
1868
1869                 if (err < 0)
1870                         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",
1871                                              dev->name);
1872         }
1873
1874         return err;
1875 }
1876
1877 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1878 {
1879         struct net *net = sock_net(skb->sk);
1880         struct ifinfomsg *ifm;
1881         struct net_device *dev;
1882         int err;
1883         struct nlattr *tb[IFLA_MAX+1];
1884         char ifname[IFNAMSIZ];
1885
1886         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1887         if (err < 0)
1888                 goto errout;
1889
1890         if (tb[IFLA_IFNAME])
1891                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1892         else
1893                 ifname[0] = '\0';
1894
1895         err = -EINVAL;
1896         ifm = nlmsg_data(nlh);
1897         if (ifm->ifi_index > 0)
1898                 dev = __dev_get_by_index(net, ifm->ifi_index);
1899         else if (tb[IFLA_IFNAME])
1900                 dev = __dev_get_by_name(net, ifname);
1901         else
1902                 goto errout;
1903
1904         if (dev == NULL) {
1905                 err = -ENODEV;
1906                 goto errout;
1907         }
1908
1909         err = validate_linkmsg(dev, tb);
1910         if (err < 0)
1911                 goto errout;
1912
1913         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1914 errout:
1915         return err;
1916 }
1917
1918 static int rtnl_group_dellink(const struct net *net, int group)
1919 {
1920         struct net_device *dev, *aux;
1921         LIST_HEAD(list_kill);
1922         bool found = false;
1923
1924         if (!group)
1925                 return -EPERM;
1926
1927         for_each_netdev(net, dev) {
1928                 if (dev->group == group) {
1929                         const struct rtnl_link_ops *ops;
1930
1931                         found = true;
1932                         ops = dev->rtnl_link_ops;
1933                         if (!ops || !ops->dellink)
1934                                 return -EOPNOTSUPP;
1935                 }
1936         }
1937
1938         if (!found)
1939                 return -ENODEV;
1940
1941         for_each_netdev_safe(net, dev, aux) {
1942                 if (dev->group == group) {
1943                         const struct rtnl_link_ops *ops;
1944
1945                         ops = dev->rtnl_link_ops;
1946                         ops->dellink(dev, &list_kill);
1947                 }
1948         }
1949         unregister_netdevice_many(&list_kill);
1950
1951         return 0;
1952 }
1953
1954 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1955 {
1956         struct net *net = sock_net(skb->sk);
1957         const struct rtnl_link_ops *ops;
1958         struct net_device *dev;
1959         struct ifinfomsg *ifm;
1960         char ifname[IFNAMSIZ];
1961         struct nlattr *tb[IFLA_MAX+1];
1962         int err;
1963         LIST_HEAD(list_kill);
1964
1965         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1966         if (err < 0)
1967                 return err;
1968
1969         if (tb[IFLA_IFNAME])
1970                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1971
1972         ifm = nlmsg_data(nlh);
1973         if (ifm->ifi_index > 0)
1974                 dev = __dev_get_by_index(net, ifm->ifi_index);
1975         else if (tb[IFLA_IFNAME])
1976                 dev = __dev_get_by_name(net, ifname);
1977         else if (tb[IFLA_GROUP])
1978                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1979         else
1980                 return -EINVAL;
1981
1982         if (!dev)
1983                 return -ENODEV;
1984
1985         ops = dev->rtnl_link_ops;
1986         if (!ops || !ops->dellink)
1987                 return -EOPNOTSUPP;
1988
1989         ops->dellink(dev, &list_kill);
1990         unregister_netdevice_many(&list_kill);
1991         return 0;
1992 }
1993
1994 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1995 {
1996         unsigned int old_flags;
1997         int err;
1998
1999         old_flags = dev->flags;
2000         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2001                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2002                 if (err < 0)
2003                         return err;
2004         }
2005
2006         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2007
2008         __dev_notify_flags(dev, old_flags, ~0U);
2009         return 0;
2010 }
2011 EXPORT_SYMBOL(rtnl_configure_link);
2012
2013 struct net_device *rtnl_create_link(struct net *net,
2014         const char *ifname, unsigned char name_assign_type,
2015         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2016 {
2017         int err;
2018         struct net_device *dev;
2019         unsigned int num_tx_queues = 1;
2020         unsigned int num_rx_queues = 1;
2021
2022         if (tb[IFLA_NUM_TX_QUEUES])
2023                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2024         else if (ops->get_num_tx_queues)
2025                 num_tx_queues = ops->get_num_tx_queues();
2026
2027         if (tb[IFLA_NUM_RX_QUEUES])
2028                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2029         else if (ops->get_num_rx_queues)
2030                 num_rx_queues = ops->get_num_rx_queues();
2031
2032         err = -ENOMEM;
2033         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2034                                ops->setup, num_tx_queues, num_rx_queues);
2035         if (!dev)
2036                 goto err;
2037
2038         dev_net_set(dev, net);
2039         dev->rtnl_link_ops = ops;
2040         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2041
2042         if (tb[IFLA_MTU])
2043                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2044         if (tb[IFLA_ADDRESS]) {
2045                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2046                                 nla_len(tb[IFLA_ADDRESS]));
2047                 dev->addr_assign_type = NET_ADDR_SET;
2048         }
2049         if (tb[IFLA_BROADCAST])
2050                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2051                                 nla_len(tb[IFLA_BROADCAST]));
2052         if (tb[IFLA_TXQLEN])
2053                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2054         if (tb[IFLA_OPERSTATE])
2055                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2056         if (tb[IFLA_LINKMODE])
2057                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2058         if (tb[IFLA_GROUP])
2059                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2060
2061         return dev;
2062
2063 err:
2064         return ERR_PTR(err);
2065 }
2066 EXPORT_SYMBOL(rtnl_create_link);
2067
2068 static int rtnl_group_changelink(const struct sk_buff *skb,
2069                 struct net *net, int group,
2070                 struct ifinfomsg *ifm,
2071                 struct nlattr **tb)
2072 {
2073         struct net_device *dev, *aux;
2074         int err;
2075
2076         for_each_netdev_safe(net, dev, aux) {
2077                 if (dev->group == group) {
2078                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2079                         if (err < 0)
2080                                 return err;
2081                 }
2082         }
2083
2084         return 0;
2085 }
2086
2087 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2088 {
2089         struct net *net = sock_net(skb->sk);
2090         const struct rtnl_link_ops *ops;
2091         const struct rtnl_link_ops *m_ops = NULL;
2092         struct net_device *dev;
2093         struct net_device *master_dev = NULL;
2094         struct ifinfomsg *ifm;
2095         char kind[MODULE_NAME_LEN];
2096         char ifname[IFNAMSIZ];
2097         struct nlattr *tb[IFLA_MAX+1];
2098         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2099         unsigned char name_assign_type = NET_NAME_USER;
2100         int err;
2101
2102 #ifdef CONFIG_MODULES
2103 replay:
2104 #endif
2105         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2106         if (err < 0)
2107                 return err;
2108
2109         if (tb[IFLA_IFNAME])
2110                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2111         else
2112                 ifname[0] = '\0';
2113
2114         ifm = nlmsg_data(nlh);
2115         if (ifm->ifi_index > 0)
2116                 dev = __dev_get_by_index(net, ifm->ifi_index);
2117         else {
2118                 if (ifname[0])
2119                         dev = __dev_get_by_name(net, ifname);
2120                 else
2121                         dev = NULL;
2122         }
2123
2124         if (dev) {
2125                 master_dev = netdev_master_upper_dev_get(dev);
2126                 if (master_dev)
2127                         m_ops = master_dev->rtnl_link_ops;
2128         }
2129
2130         err = validate_linkmsg(dev, tb);
2131         if (err < 0)
2132                 return err;
2133
2134         if (tb[IFLA_LINKINFO]) {
2135                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2136                                        tb[IFLA_LINKINFO], ifla_info_policy);
2137                 if (err < 0)
2138                         return err;
2139         } else
2140                 memset(linkinfo, 0, sizeof(linkinfo));
2141
2142         if (linkinfo[IFLA_INFO_KIND]) {
2143                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2144                 ops = rtnl_link_ops_get(kind);
2145         } else {
2146                 kind[0] = '\0';
2147                 ops = NULL;
2148         }
2149
2150         if (1) {
2151                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2152                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2153                 struct nlattr **data = NULL;
2154                 struct nlattr **slave_data = NULL;
2155                 struct net *dest_net, *link_net = NULL;
2156
2157                 if (ops) {
2158                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2159                                 err = nla_parse_nested(attr, ops->maxtype,
2160                                                        linkinfo[IFLA_INFO_DATA],
2161                                                        ops->policy);
2162                                 if (err < 0)
2163                                         return err;
2164                                 data = attr;
2165                         }
2166                         if (ops->validate) {
2167                                 err = ops->validate(tb, data);
2168                                 if (err < 0)
2169                                         return err;
2170                         }
2171                 }
2172
2173                 if (m_ops) {
2174                         if (m_ops->slave_maxtype &&
2175                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2176                                 err = nla_parse_nested(slave_attr,
2177                                                        m_ops->slave_maxtype,
2178                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2179                                                        m_ops->slave_policy);
2180                                 if (err < 0)
2181                                         return err;
2182                                 slave_data = slave_attr;
2183                         }
2184                         if (m_ops->slave_validate) {
2185                                 err = m_ops->slave_validate(tb, slave_data);
2186                                 if (err < 0)
2187                                         return err;
2188                         }
2189                 }
2190
2191                 if (dev) {
2192                         int status = 0;
2193
2194                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2195                                 return -EEXIST;
2196                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2197                                 return -EOPNOTSUPP;
2198
2199                         if (linkinfo[IFLA_INFO_DATA]) {
2200                                 if (!ops || ops != dev->rtnl_link_ops ||
2201                                     !ops->changelink)
2202                                         return -EOPNOTSUPP;
2203
2204                                 err = ops->changelink(dev, tb, data);
2205                                 if (err < 0)
2206                                         return err;
2207                                 status |= DO_SETLINK_NOTIFY;
2208                         }
2209
2210                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2211                                 if (!m_ops || !m_ops->slave_changelink)
2212                                         return -EOPNOTSUPP;
2213
2214                                 err = m_ops->slave_changelink(master_dev, dev,
2215                                                               tb, slave_data);
2216                                 if (err < 0)
2217                                         return err;
2218                                 status |= DO_SETLINK_NOTIFY;
2219                         }
2220
2221                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2222                 }
2223
2224                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2225                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2226                                 return rtnl_group_changelink(skb, net,
2227                                                 nla_get_u32(tb[IFLA_GROUP]),
2228                                                 ifm, tb);
2229                         return -ENODEV;
2230                 }
2231
2232                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2233                         return -EOPNOTSUPP;
2234
2235                 if (!ops) {
2236 #ifdef CONFIG_MODULES
2237                         if (kind[0]) {
2238                                 __rtnl_unlock();
2239                                 request_module("rtnl-link-%s", kind);
2240                                 rtnl_lock();
2241                                 ops = rtnl_link_ops_get(kind);
2242                                 if (ops)
2243                                         goto replay;
2244                         }
2245 #endif
2246                         return -EOPNOTSUPP;
2247                 }
2248
2249                 if (!ops->setup)
2250                         return -EOPNOTSUPP;
2251
2252                 if (!ifname[0]) {
2253                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2254                         name_assign_type = NET_NAME_ENUM;
2255                 }
2256
2257                 dest_net = rtnl_link_get_net(net, tb);
2258                 if (IS_ERR(dest_net))
2259                         return PTR_ERR(dest_net);
2260
2261                 err = -EPERM;
2262                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2263                         goto out;
2264
2265                 if (tb[IFLA_LINK_NETNSID]) {
2266                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2267
2268                         link_net = get_net_ns_by_id(dest_net, id);
2269                         if (!link_net) {
2270                                 err =  -EINVAL;
2271                                 goto out;
2272                         }
2273                         err = -EPERM;
2274                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2275                                 goto out;
2276                 }
2277
2278                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2279                                        name_assign_type, ops, tb);
2280                 if (IS_ERR(dev)) {
2281                         err = PTR_ERR(dev);
2282                         goto out;
2283                 }
2284
2285                 dev->ifindex = ifm->ifi_index;
2286
2287                 if (ops->newlink) {
2288                         err = ops->newlink(link_net ? : net, dev, tb, data);
2289                         /* Drivers should call free_netdev() in ->destructor
2290                          * and unregister it on failure after registration
2291                          * so that device could be finally freed in rtnl_unlock.
2292                          */
2293                         if (err < 0) {
2294                                 /* If device is not registered at all, free it now */
2295                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2296                                         free_netdev(dev);
2297                                 goto out;
2298                         }
2299                 } else {
2300                         err = register_netdevice(dev);
2301                         if (err < 0) {
2302                                 free_netdev(dev);
2303                                 goto out;
2304                         }
2305                 }
2306                 err = rtnl_configure_link(dev, ifm);
2307                 if (err < 0)
2308                         goto out_unregister;
2309                 if (link_net) {
2310                         err = dev_change_net_namespace(dev, dest_net, ifname);
2311                         if (err < 0)
2312                                 goto out_unregister;
2313                 }
2314 out:
2315                 if (link_net)
2316                         put_net(link_net);
2317                 put_net(dest_net);
2318                 return err;
2319 out_unregister:
2320                 if (ops->newlink) {
2321                         LIST_HEAD(list_kill);
2322
2323                         ops->dellink(dev, &list_kill);
2324                         unregister_netdevice_many(&list_kill);
2325                 } else {
2326                         unregister_netdevice(dev);
2327                 }
2328                 goto out;
2329         }
2330 }
2331
2332 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2333 {
2334         struct net *net = sock_net(skb->sk);
2335         struct ifinfomsg *ifm;
2336         char ifname[IFNAMSIZ];
2337         struct nlattr *tb[IFLA_MAX+1];
2338         struct net_device *dev = NULL;
2339         struct sk_buff *nskb;
2340         int err;
2341         u32 ext_filter_mask = 0;
2342
2343         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2344         if (err < 0)
2345                 return err;
2346
2347         if (tb[IFLA_IFNAME])
2348                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2349
2350         if (tb[IFLA_EXT_MASK])
2351                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2352
2353         ifm = nlmsg_data(nlh);
2354         if (ifm->ifi_index > 0)
2355                 dev = __dev_get_by_index(net, ifm->ifi_index);
2356         else if (tb[IFLA_IFNAME])
2357                 dev = __dev_get_by_name(net, ifname);
2358         else
2359                 return -EINVAL;
2360
2361         if (dev == NULL)
2362                 return -ENODEV;
2363
2364         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2365         if (nskb == NULL)
2366                 return -ENOBUFS;
2367
2368         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2369                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2370         if (err < 0) {
2371                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2372                 WARN_ON(err == -EMSGSIZE);
2373                 kfree_skb(nskb);
2374         } else
2375                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2376
2377         return err;
2378 }
2379
2380 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2381 {
2382         struct net *net = sock_net(skb->sk);
2383         struct net_device *dev;
2384         struct nlattr *tb[IFLA_MAX+1];
2385         u32 ext_filter_mask = 0;
2386         u16 min_ifinfo_dump_size = 0;
2387         int hdrlen;
2388
2389         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2390         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2391                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2392
2393         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2394                 if (tb[IFLA_EXT_MASK])
2395                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2396         }
2397
2398         if (!ext_filter_mask)
2399                 return NLMSG_GOODSIZE;
2400         /*
2401          * traverse the list of net devices and compute the minimum
2402          * buffer size based upon the filter mask.
2403          */
2404         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2405                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2406                                              if_nlmsg_size(dev,
2407                                                            ext_filter_mask));
2408         }
2409
2410         return min_ifinfo_dump_size;
2411 }
2412
2413 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2414 {
2415         int idx;
2416         int s_idx = cb->family;
2417
2418         if (s_idx == 0)
2419                 s_idx = 1;
2420         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2421                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2422                 if (idx < s_idx || idx == PF_PACKET)
2423                         continue;
2424                 if (rtnl_msg_handlers[idx] == NULL ||
2425                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2426                         continue;
2427                 if (idx > s_idx) {
2428                         memset(&cb->args[0], 0, sizeof(cb->args));
2429                         cb->prev_seq = 0;
2430                         cb->seq = 0;
2431                 }
2432                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2433                         break;
2434         }
2435         cb->family = idx;
2436
2437         return skb->len;
2438 }
2439
2440 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2441                                        unsigned int change, gfp_t flags)
2442 {
2443         struct net *net = dev_net(dev);
2444         struct sk_buff *skb;
2445         int err = -ENOBUFS;
2446         size_t if_info_size;
2447
2448         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2449         if (skb == NULL)
2450                 goto errout;
2451
2452         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2453         if (err < 0) {
2454                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2455                 WARN_ON(err == -EMSGSIZE);
2456                 kfree_skb(skb);
2457                 goto errout;
2458         }
2459         return skb;
2460 errout:
2461         if (err < 0)
2462                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2463         return NULL;
2464 }
2465
2466 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2467 {
2468         struct net *net = dev_net(dev);
2469
2470         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2471 }
2472
2473 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2474                   gfp_t flags)
2475 {
2476         struct sk_buff *skb;
2477
2478         if (dev->reg_state != NETREG_REGISTERED)
2479                 return;
2480
2481         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2482         if (skb)
2483                 rtmsg_ifinfo_send(skb, dev, flags);
2484 }
2485 EXPORT_SYMBOL(rtmsg_ifinfo);
2486
2487 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2488                                    struct net_device *dev,
2489                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2490                                    int type, unsigned int flags,
2491                                    int nlflags)
2492 {
2493         struct nlmsghdr *nlh;
2494         struct ndmsg *ndm;
2495
2496         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2497         if (!nlh)
2498                 return -EMSGSIZE;
2499
2500         ndm = nlmsg_data(nlh);
2501         ndm->ndm_family  = AF_BRIDGE;
2502         ndm->ndm_pad1    = 0;
2503         ndm->ndm_pad2    = 0;
2504         ndm->ndm_flags   = flags;
2505         ndm->ndm_type    = 0;
2506         ndm->ndm_ifindex = dev->ifindex;
2507         ndm->ndm_state   = NUD_PERMANENT;
2508
2509         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2510                 goto nla_put_failure;
2511         if (vid)
2512                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2513                         goto nla_put_failure;
2514
2515         nlmsg_end(skb, nlh);
2516         return 0;
2517
2518 nla_put_failure:
2519         nlmsg_cancel(skb, nlh);
2520         return -EMSGSIZE;
2521 }
2522
2523 static inline size_t rtnl_fdb_nlmsg_size(void)
2524 {
2525         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2526 }
2527
2528 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2529 {
2530         struct net *net = dev_net(dev);
2531         struct sk_buff *skb;
2532         int err = -ENOBUFS;
2533
2534         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2535         if (!skb)
2536                 goto errout;
2537
2538         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2539                                       0, 0, type, NTF_SELF, 0);
2540         if (err < 0) {
2541                 kfree_skb(skb);
2542                 goto errout;
2543         }
2544
2545         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2546         return;
2547 errout:
2548         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2549 }
2550
2551 /**
2552  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2553  */
2554 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2555                      struct nlattr *tb[],
2556                      struct net_device *dev,
2557                      const unsigned char *addr, u16 vid,
2558                      u16 flags)
2559 {
2560         int err = -EINVAL;
2561
2562         /* If aging addresses are supported device will need to
2563          * implement its own handler for this.
2564          */
2565         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2566                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2567                 return err;
2568         }
2569
2570         if (vid) {
2571                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2572                 return err;
2573         }
2574
2575         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2576                 err = dev_uc_add_excl(dev, addr);
2577         else if (is_multicast_ether_addr(addr))
2578                 err = dev_mc_add_excl(dev, addr);
2579
2580         /* Only return duplicate errors if NLM_F_EXCL is set */
2581         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2582                 err = 0;
2583
2584         return err;
2585 }
2586 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2587
2588 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2589 {
2590         u16 vid = 0;
2591
2592         if (vlan_attr) {
2593                 if (nla_len(vlan_attr) != sizeof(u16)) {
2594                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2595                         return -EINVAL;
2596                 }
2597
2598                 vid = nla_get_u16(vlan_attr);
2599
2600                 if (!vid || vid >= VLAN_VID_MASK) {
2601                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2602                                 vid);
2603                         return -EINVAL;
2604                 }
2605         }
2606         *p_vid = vid;
2607         return 0;
2608 }
2609
2610 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2611 {
2612         struct net *net = sock_net(skb->sk);
2613         struct ndmsg *ndm;
2614         struct nlattr *tb[NDA_MAX+1];
2615         struct net_device *dev;
2616         u8 *addr;
2617         u16 vid;
2618         int err;
2619
2620         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2621         if (err < 0)
2622                 return err;
2623
2624         ndm = nlmsg_data(nlh);
2625         if (ndm->ndm_ifindex == 0) {
2626                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2627                 return -EINVAL;
2628         }
2629
2630         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2631         if (dev == NULL) {
2632                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2633                 return -ENODEV;
2634         }
2635
2636         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2637                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2638                 return -EINVAL;
2639         }
2640
2641         addr = nla_data(tb[NDA_LLADDR]);
2642
2643         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2644         if (err)
2645                 return err;
2646
2647         err = -EOPNOTSUPP;
2648
2649         /* Support fdb on master device the net/bridge default case */
2650         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2651             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2652                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2653                 const struct net_device_ops *ops = br_dev->netdev_ops;
2654
2655                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2656                                        nlh->nlmsg_flags);
2657                 if (err)
2658                         goto out;
2659                 else
2660                         ndm->ndm_flags &= ~NTF_MASTER;
2661         }
2662
2663         /* Embedded bridge, macvlan, and any other device support */
2664         if ((ndm->ndm_flags & NTF_SELF)) {
2665                 if (dev->netdev_ops->ndo_fdb_add)
2666                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2667                                                            vid,
2668                                                            nlh->nlmsg_flags);
2669                 else
2670                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2671                                                nlh->nlmsg_flags);
2672
2673                 if (!err) {
2674                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2675                         ndm->ndm_flags &= ~NTF_SELF;
2676                 }
2677         }
2678 out:
2679         return err;
2680 }
2681
2682 /**
2683  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2684  */
2685 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2686                      struct nlattr *tb[],
2687                      struct net_device *dev,
2688                      const unsigned char *addr, u16 vid)
2689 {
2690         int err = -EINVAL;
2691
2692         /* If aging addresses are supported device will need to
2693          * implement its own handler for this.
2694          */
2695         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2696                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2697                 return err;
2698         }
2699
2700         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2701                 err = dev_uc_del(dev, addr);
2702         else if (is_multicast_ether_addr(addr))
2703                 err = dev_mc_del(dev, addr);
2704
2705         return err;
2706 }
2707 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2708
2709 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2710 {
2711         struct net *net = sock_net(skb->sk);
2712         struct ndmsg *ndm;
2713         struct nlattr *tb[NDA_MAX+1];
2714         struct net_device *dev;
2715         int err = -EINVAL;
2716         __u8 *addr;
2717         u16 vid;
2718
2719         if (!netlink_capable(skb, CAP_NET_ADMIN))
2720                 return -EPERM;
2721
2722         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2723         if (err < 0)
2724                 return err;
2725
2726         ndm = nlmsg_data(nlh);
2727         if (ndm->ndm_ifindex == 0) {
2728                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2729                 return -EINVAL;
2730         }
2731
2732         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2733         if (dev == NULL) {
2734                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2735                 return -ENODEV;
2736         }
2737
2738         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2739                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2740                 return -EINVAL;
2741         }
2742
2743         addr = nla_data(tb[NDA_LLADDR]);
2744
2745         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2746         if (err)
2747                 return err;
2748
2749         err = -EOPNOTSUPP;
2750
2751         /* Support fdb on master device the net/bridge default case */
2752         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2753             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2754                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2755                 const struct net_device_ops *ops = br_dev->netdev_ops;
2756
2757                 if (ops->ndo_fdb_del)
2758                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2759
2760                 if (err)
2761                         goto out;
2762                 else
2763                         ndm->ndm_flags &= ~NTF_MASTER;
2764         }
2765
2766         /* Embedded bridge, macvlan, and any other device support */
2767         if (ndm->ndm_flags & NTF_SELF) {
2768                 if (dev->netdev_ops->ndo_fdb_del)
2769                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2770                                                            vid);
2771                 else
2772                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2773
2774                 if (!err) {
2775                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2776                         ndm->ndm_flags &= ~NTF_SELF;
2777                 }
2778         }
2779 out:
2780         return err;
2781 }
2782
2783 static int nlmsg_populate_fdb(struct sk_buff *skb,
2784                               struct netlink_callback *cb,
2785                               struct net_device *dev,
2786                               int *idx,
2787                               struct netdev_hw_addr_list *list)
2788 {
2789         struct netdev_hw_addr *ha;
2790         int err;
2791         u32 portid, seq;
2792
2793         portid = NETLINK_CB(cb->skb).portid;
2794         seq = cb->nlh->nlmsg_seq;
2795
2796         list_for_each_entry(ha, &list->list, list) {
2797                 if (*idx < cb->args[0])
2798                         goto skip;
2799
2800                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2801                                               portid, seq,
2802                                               RTM_NEWNEIGH, NTF_SELF,
2803                                               NLM_F_MULTI);
2804                 if (err < 0)
2805                         return err;
2806 skip:
2807                 *idx += 1;
2808         }
2809         return 0;
2810 }
2811
2812 /**
2813  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2814  * @nlh: netlink message header
2815  * @dev: netdevice
2816  *
2817  * Default netdevice operation to dump the existing unicast address list.
2818  * Returns number of addresses from list put in skb.
2819  */
2820 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2821                       struct netlink_callback *cb,
2822                       struct net_device *dev,
2823                       struct net_device *filter_dev,
2824                       int idx)
2825 {
2826         int err;
2827
2828         netif_addr_lock_bh(dev);
2829         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2830         if (err)
2831                 goto out;
2832         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2833 out:
2834         netif_addr_unlock_bh(dev);
2835         return idx;
2836 }
2837 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2838
2839 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2840 {
2841         struct net_device *dev;
2842         struct nlattr *tb[IFLA_MAX+1];
2843         struct net_device *br_dev = NULL;
2844         const struct net_device_ops *ops = NULL;
2845         const struct net_device_ops *cops = NULL;
2846         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2847         struct net *net = sock_net(skb->sk);
2848         int brport_idx = 0;
2849         int br_idx = 0;
2850         int idx = 0;
2851
2852         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2853                         ifla_policy) == 0) {
2854                 if (tb[IFLA_MASTER])
2855                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2856         }
2857
2858         brport_idx = ifm->ifi_index;
2859
2860         if (br_idx) {
2861                 br_dev = __dev_get_by_index(net, br_idx);
2862                 if (!br_dev)
2863                         return -ENODEV;
2864
2865                 ops = br_dev->netdev_ops;
2866         }
2867
2868         for_each_netdev(net, dev) {
2869                 if (brport_idx && (dev->ifindex != brport_idx))
2870                         continue;
2871
2872                 if (!br_idx) { /* user did not specify a specific bridge */
2873                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2874                                 br_dev = netdev_master_upper_dev_get(dev);
2875                                 cops = br_dev->netdev_ops;
2876                         }
2877
2878                 } else {
2879                         if (dev != br_dev &&
2880                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2881                                 continue;
2882
2883                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2884                             !(dev->priv_flags & IFF_EBRIDGE))
2885                                 continue;
2886
2887                         cops = ops;
2888                 }
2889
2890                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2891                         if (cops && cops->ndo_fdb_dump)
2892                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2893                                                          idx);
2894                 }
2895
2896                 if (dev->netdev_ops->ndo_fdb_dump)
2897                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2898                                                             idx);
2899                 else
2900                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2901
2902                 cops = NULL;
2903         }
2904
2905         cb->args[0] = idx;
2906         return skb->len;
2907 }
2908
2909 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2910                                unsigned int attrnum, unsigned int flag)
2911 {
2912         if (mask & flag)
2913                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2914         return 0;
2915 }
2916
2917 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2918                             struct net_device *dev, u16 mode,
2919                             u32 flags, u32 mask, int nlflags,
2920                             u32 filter_mask,
2921                             int (*vlan_fill)(struct sk_buff *skb,
2922                                              struct net_device *dev,
2923                                              u32 filter_mask))
2924 {
2925         struct nlmsghdr *nlh;
2926         struct ifinfomsg *ifm;
2927         struct nlattr *br_afspec;
2928         struct nlattr *protinfo;
2929         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2930         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2931         int err = 0;
2932
2933         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2934         if (nlh == NULL)
2935                 return -EMSGSIZE;
2936
2937         ifm = nlmsg_data(nlh);
2938         ifm->ifi_family = AF_BRIDGE;
2939         ifm->__ifi_pad = 0;
2940         ifm->ifi_type = dev->type;
2941         ifm->ifi_index = dev->ifindex;
2942         ifm->ifi_flags = dev_get_flags(dev);
2943         ifm->ifi_change = 0;
2944
2945
2946         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2947             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2948             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2949             (br_dev &&
2950              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2951             (dev->addr_len &&
2952              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2953             (dev->ifindex != dev_get_iflink(dev) &&
2954              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2955                 goto nla_put_failure;
2956
2957         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2958         if (!br_afspec)
2959                 goto nla_put_failure;
2960
2961         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2962                 nla_nest_cancel(skb, br_afspec);
2963                 goto nla_put_failure;
2964         }
2965
2966         if (mode != BRIDGE_MODE_UNDEF) {
2967                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2968                         nla_nest_cancel(skb, br_afspec);
2969                         goto nla_put_failure;
2970                 }
2971         }
2972         if (vlan_fill) {
2973                 err = vlan_fill(skb, dev, filter_mask);
2974                 if (err) {
2975                         nla_nest_cancel(skb, br_afspec);
2976                         goto nla_put_failure;
2977                 }
2978         }
2979         nla_nest_end(skb, br_afspec);
2980
2981         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2982         if (!protinfo)
2983                 goto nla_put_failure;
2984
2985         if (brport_nla_put_flag(skb, flags, mask,
2986                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2987             brport_nla_put_flag(skb, flags, mask,
2988                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2989             brport_nla_put_flag(skb, flags, mask,
2990                                 IFLA_BRPORT_FAST_LEAVE,
2991                                 BR_MULTICAST_FAST_LEAVE) ||
2992             brport_nla_put_flag(skb, flags, mask,
2993                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2994             brport_nla_put_flag(skb, flags, mask,
2995                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2996             brport_nla_put_flag(skb, flags, mask,
2997                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2998             brport_nla_put_flag(skb, flags, mask,
2999                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3000             brport_nla_put_flag(skb, flags, mask,
3001                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3002                 nla_nest_cancel(skb, protinfo);
3003                 goto nla_put_failure;
3004         }
3005
3006         nla_nest_end(skb, protinfo);
3007
3008         nlmsg_end(skb, nlh);
3009         return 0;
3010 nla_put_failure:
3011         nlmsg_cancel(skb, nlh);
3012         return err ? err : -EMSGSIZE;
3013 }
3014 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3015
3016 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3017 {
3018         struct net *net = sock_net(skb->sk);
3019         struct net_device *dev;
3020         int idx = 0;
3021         u32 portid = NETLINK_CB(cb->skb).portid;
3022         u32 seq = cb->nlh->nlmsg_seq;
3023         u32 filter_mask = 0;
3024
3025         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3026                 struct nlattr *extfilt;
3027
3028                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3029                                           IFLA_EXT_MASK);
3030                 if (extfilt) {
3031                         if (nla_len(extfilt) < sizeof(filter_mask))
3032                                 return -EINVAL;
3033
3034                         filter_mask = nla_get_u32(extfilt);
3035                 }
3036         }
3037
3038         rcu_read_lock();
3039         for_each_netdev_rcu(net, dev) {
3040                 const struct net_device_ops *ops = dev->netdev_ops;
3041                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3042
3043                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3044                         if (idx >= cb->args[0] &&
3045                             br_dev->netdev_ops->ndo_bridge_getlink(
3046                                     skb, portid, seq, dev, filter_mask,
3047                                     NLM_F_MULTI) < 0)
3048                                 break;
3049                         idx++;
3050                 }
3051
3052                 if (ops->ndo_bridge_getlink) {
3053                         if (idx >= cb->args[0] &&
3054                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
3055                                                     filter_mask,
3056                                                     NLM_F_MULTI) < 0)
3057                                 break;
3058                         idx++;
3059                 }
3060         }
3061         rcu_read_unlock();
3062         cb->args[0] = idx;
3063
3064         return skb->len;
3065 }
3066
3067 static inline size_t bridge_nlmsg_size(void)
3068 {
3069         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3070                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3071                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3072                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3073                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3074                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3075                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3076                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3077                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3078                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3079                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3080 }
3081
3082 static int rtnl_bridge_notify(struct net_device *dev)
3083 {
3084         struct net *net = dev_net(dev);
3085         struct sk_buff *skb;
3086         int err = -EOPNOTSUPP;
3087
3088         if (!dev->netdev_ops->ndo_bridge_getlink)
3089                 return 0;
3090
3091         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3092         if (!skb) {
3093                 err = -ENOMEM;
3094                 goto errout;
3095         }
3096
3097         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3098         if (err < 0)
3099                 goto errout;
3100
3101         if (!skb->len)
3102                 goto errout;
3103
3104         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3105         return 0;
3106 errout:
3107         WARN_ON(err == -EMSGSIZE);
3108         kfree_skb(skb);
3109         if (err)
3110                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3111         return err;
3112 }
3113
3114 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3115 {
3116         struct net *net = sock_net(skb->sk);
3117         struct ifinfomsg *ifm;
3118         struct net_device *dev;
3119         struct nlattr *br_spec, *attr = NULL;
3120         int rem, err = -EOPNOTSUPP;
3121         u16 flags = 0;
3122         bool have_flags = false;
3123
3124         if (nlmsg_len(nlh) < sizeof(*ifm))
3125                 return -EINVAL;
3126
3127         ifm = nlmsg_data(nlh);
3128         if (ifm->ifi_family != AF_BRIDGE)
3129                 return -EPFNOSUPPORT;
3130
3131         dev = __dev_get_by_index(net, ifm->ifi_index);
3132         if (!dev) {
3133                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3134                 return -ENODEV;
3135         }
3136
3137         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3138         if (br_spec) {
3139                 nla_for_each_nested(attr, br_spec, rem) {
3140                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3141                                 if (nla_len(attr) < sizeof(flags))
3142                                         return -EINVAL;
3143
3144                                 have_flags = true;
3145                                 flags = nla_get_u16(attr);
3146                                 break;
3147                         }
3148                 }
3149         }
3150
3151         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3152                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3153
3154                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3155                         err = -EOPNOTSUPP;
3156                         goto out;
3157                 }
3158
3159                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3160                 if (err)
3161                         goto out;
3162
3163                 flags &= ~BRIDGE_FLAGS_MASTER;
3164         }
3165
3166         if ((flags & BRIDGE_FLAGS_SELF)) {
3167                 if (!dev->netdev_ops->ndo_bridge_setlink)
3168                         err = -EOPNOTSUPP;
3169                 else
3170                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3171                                                                   flags);
3172                 if (!err) {
3173                         flags &= ~BRIDGE_FLAGS_SELF;
3174
3175                         /* Generate event to notify upper layer of bridge
3176                          * change
3177                          */
3178                         err = rtnl_bridge_notify(dev);
3179                 }
3180         }
3181
3182         if (have_flags)
3183                 memcpy(nla_data(attr), &flags, sizeof(flags));
3184 out:
3185         return err;
3186 }
3187
3188 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3189 {
3190         struct net *net = sock_net(skb->sk);
3191         struct ifinfomsg *ifm;
3192         struct net_device *dev;
3193         struct nlattr *br_spec, *attr = NULL;
3194         int rem, err = -EOPNOTSUPP;
3195         u16 flags = 0;
3196         bool have_flags = false;
3197
3198         if (nlmsg_len(nlh) < sizeof(*ifm))
3199                 return -EINVAL;
3200
3201         ifm = nlmsg_data(nlh);
3202         if (ifm->ifi_family != AF_BRIDGE)
3203                 return -EPFNOSUPPORT;
3204
3205         dev = __dev_get_by_index(net, ifm->ifi_index);
3206         if (!dev) {
3207                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3208                 return -ENODEV;
3209         }
3210
3211         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3212         if (br_spec) {
3213                 nla_for_each_nested(attr, br_spec, rem) {
3214                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3215                                 if (nla_len(attr) < sizeof(flags))
3216                                         return -EINVAL;
3217
3218                                 have_flags = true;
3219                                 flags = nla_get_u16(attr);
3220                                 break;
3221                         }
3222                 }
3223         }
3224
3225         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3226                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3227
3228                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3229                         err = -EOPNOTSUPP;
3230                         goto out;
3231                 }
3232
3233                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3234                 if (err)
3235                         goto out;
3236
3237                 flags &= ~BRIDGE_FLAGS_MASTER;
3238         }
3239
3240         if ((flags & BRIDGE_FLAGS_SELF)) {
3241                 if (!dev->netdev_ops->ndo_bridge_dellink)
3242                         err = -EOPNOTSUPP;
3243                 else
3244                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3245                                                                   flags);
3246
3247                 if (!err) {
3248                         flags &= ~BRIDGE_FLAGS_SELF;
3249
3250                         /* Generate event to notify upper layer of bridge
3251                          * change
3252                          */
3253                         err = rtnl_bridge_notify(dev);
3254                 }
3255         }
3256
3257         if (have_flags)
3258                 memcpy(nla_data(attr), &flags, sizeof(flags));
3259 out:
3260         return err;
3261 }
3262
3263 /* Process one rtnetlink message. */
3264
3265 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3266 {
3267         struct net *net = sock_net(skb->sk);
3268         rtnl_doit_func doit;
3269         int sz_idx, kind;
3270         int family;
3271         int type;
3272         int err;
3273
3274         type = nlh->nlmsg_type;
3275         if (type > RTM_MAX)
3276                 return -EOPNOTSUPP;
3277
3278         type -= RTM_BASE;
3279
3280         /* All the messages must have at least 1 byte length */
3281         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3282                 return 0;
3283
3284         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3285         sz_idx = type>>2;
3286         kind = type&3;
3287
3288         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3289                 return -EPERM;
3290
3291         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3292                 struct sock *rtnl;
3293                 rtnl_dumpit_func dumpit;
3294                 rtnl_calcit_func calcit;
3295                 u16 min_dump_alloc = 0;
3296
3297                 dumpit = rtnl_get_dumpit(family, type);
3298                 if (dumpit == NULL)
3299                         return -EOPNOTSUPP;
3300                 calcit = rtnl_get_calcit(family, type);
3301                 if (calcit)
3302                         min_dump_alloc = calcit(skb, nlh);
3303
3304                 __rtnl_unlock();
3305                 rtnl = net->rtnl;
3306                 {
3307                         struct netlink_dump_control c = {
3308                                 .dump           = dumpit,
3309                                 .min_dump_alloc = min_dump_alloc,
3310                         };
3311                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3312                 }
3313                 rtnl_lock();
3314                 return err;
3315         }
3316
3317         doit = rtnl_get_doit(family, type);
3318         if (doit == NULL)
3319                 return -EOPNOTSUPP;
3320
3321         return doit(skb, nlh);
3322 }
3323
3324 static void rtnetlink_rcv(struct sk_buff *skb)
3325 {
3326         rtnl_lock();
3327         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3328         rtnl_unlock();
3329 }
3330
3331 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3332 {
3333         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3334
3335         switch (event) {
3336         case NETDEV_UP:
3337         case NETDEV_DOWN:
3338         case NETDEV_PRE_UP:
3339         case NETDEV_POST_INIT:
3340         case NETDEV_REGISTER:
3341         case NETDEV_CHANGE:
3342         case NETDEV_PRE_TYPE_CHANGE:
3343         case NETDEV_GOING_DOWN:
3344         case NETDEV_UNREGISTER:
3345         case NETDEV_UNREGISTER_FINAL:
3346         case NETDEV_RELEASE:
3347         case NETDEV_JOIN:
3348         case NETDEV_BONDING_INFO:
3349                 break;
3350         default:
3351                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3352                 break;
3353         }
3354         return NOTIFY_DONE;
3355 }
3356
3357 static struct notifier_block rtnetlink_dev_notifier = {
3358         .notifier_call  = rtnetlink_event,
3359 };
3360
3361
3362 static int __net_init rtnetlink_net_init(struct net *net)
3363 {
3364         struct sock *sk;
3365         struct netlink_kernel_cfg cfg = {
3366                 .groups         = RTNLGRP_MAX,
3367                 .input          = rtnetlink_rcv,
3368                 .cb_mutex       = &rtnl_mutex,
3369                 .flags          = NL_CFG_F_NONROOT_RECV,
3370         };
3371
3372         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3373         if (!sk)
3374                 return -ENOMEM;
3375         net->rtnl = sk;
3376         return 0;
3377 }
3378
3379 static void __net_exit rtnetlink_net_exit(struct net *net)
3380 {
3381         netlink_kernel_release(net->rtnl);
3382         net->rtnl = NULL;
3383 }
3384
3385 static struct pernet_operations rtnetlink_net_ops = {
3386         .init = rtnetlink_net_init,
3387         .exit = rtnetlink_net_exit,
3388 };
3389
3390 void __init rtnetlink_init(void)
3391 {
3392         if (register_pernet_subsys(&rtnetlink_net_ops))
3393                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3394
3395         register_netdevice_notifier(&rtnetlink_dev_notifier);
3396
3397         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3398                       rtnl_dump_ifinfo, rtnl_calcit);
3399         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3400         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3401         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3402
3403         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3404         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3405
3406         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3407         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3408         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3409
3410         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3411         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3412         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3413 }
3414