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                 return size;
 823         } else
 824                 return 0;
 825 }
 826
 827 static size_t rtnl_port_size(const struct net_device *dev,
 828                              u32 ext_filter_mask)
 829 {
 830         size_t port_size = nla_total_size(4)            /* PORT_VF */
 831                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 832                 + nla_total_size(sizeof(struct ifla_port_vsi))
 833                                                         /* PORT_VSI_TYPE */
 834                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 835                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 836                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 837                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 838         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 839         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 840                 + port_size;
 841         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 842                 + port_size;
 843
 844         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 845             !(ext_filter_mask & RTEXT_FILTER_VF))
 846                 return 0;
 847         if (dev_num_vf(dev->dev.parent))
 848                 return port_self_size + vf_ports_size +
 849                         vf_port_size * dev_num_vf(dev->dev.parent);
 850         else
 851                 return port_self_size;
 852 }
 853
 854 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 855                                      u32 ext_filter_mask)
 856 {
 857         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 858                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 859                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 860                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 861                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 862                + nla_total_size(sizeof(struct rtnl_link_stats))
 863                + nla_total_size(sizeof(struct rtnl_link_stats64))
 864                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 865                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 866                + nla_total_size(4) /* IFLA_TXQLEN */
 867                + nla_total_size(4) /* IFLA_WEIGHT */
 868                + nla_total_size(4) /* IFLA_MTU */
 869                + nla_total_size(4) /* IFLA_LINK */
 870                + nla_total_size(4) /* IFLA_MASTER */
 871                + nla_total_size(1) /* IFLA_CARRIER */
 872                + nla_total_size(4) /* IFLA_PROMISCUITY */
 873                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 874                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 875                + nla_total_size(1) /* IFLA_OPERSTATE */
 876                + nla_total_size(1) /* IFLA_LINKMODE */
 877                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 878                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 879                + nla_total_size(ext_filter_mask
 880                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 881                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 882                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 883                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 884                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 885                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 886                + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
 887 }
 888
 889 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 890 {
 891         struct nlattr *vf_ports;
 892         struct nlattr *vf_port;
 893         int vf;
 894         int err;
 895
 896         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 897         if (!vf_ports)
 898                 return -EMSGSIZE;
 899
 900         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 901                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 902                 if (!vf_port)
 903                         goto nla_put_failure;
 904                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 905                         goto nla_put_failure;
 906                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 907                 if (err == -EMSGSIZE)
 908                         goto nla_put_failure;
 909                 if (err) {
 910                         nla_nest_cancel(skb, vf_port);
 911                         continue;
 912                 }
 913                 nla_nest_end(skb, vf_port);
 914         }
 915
 916         nla_nest_end(skb, vf_ports);
 917
 918         return 0;
 919
 920 nla_put_failure:
 921         nla_nest_cancel(skb, vf_ports);
 922         return -EMSGSIZE;
 923 }
 924
 925 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 926 {
 927         struct nlattr *port_self;
 928         int err;
 929
 930         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 931         if (!port_self)
 932                 return -EMSGSIZE;
 933
 934         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 935         if (err) {
 936                 nla_nest_cancel(skb, port_self);
 937                 return (err == -EMSGSIZE) ? err : 0;
 938         }
 939
 940         nla_nest_end(skb, port_self);
 941
 942         return 0;
 943 }
 944
 945 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 946                           u32 ext_filter_mask)
 947 {
 948         int err;
 949
 950         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 951             !(ext_filter_mask & RTEXT_FILTER_VF))
 952                 return 0;
 953
 954         err = rtnl_port_self_fill(skb, dev);
 955         if (err)
 956                 return err;
 957
 958         if (dev_num_vf(dev->dev.parent)) {
 959                 err = rtnl_vf_ports_fill(skb, dev);
 960                 if (err)
 961                         return err;
 962         }
 963
 964         return 0;
 965 }
 966
 967 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 968 {
 969         int err;
 970         struct netdev_phys_item_id ppid;
 971
 972         err = dev_get_phys_port_id(dev, &ppid);
 973         if (err) {
 974                 if (err == -EOPNOTSUPP)
 975                         return 0;
 976                 return err;
 977         }
 978
 979         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 980                 return -EMSGSIZE;
 981
 982         return 0;
 983 }
 984
 985 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
 986 {
 987         int err;
 988         struct netdev_phys_item_id psid;
 989
 990         err = netdev_switch_parent_id_get(dev, &psid);
 991         if (err) {
 992                 if (err == -EOPNOTSUPP)
 993                         return 0;
 994                 return err;
 995         }
 996
 997         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
 998                 return -EMSGSIZE;
 999
1000         return 0;
1001 }
1002
1003 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1004                             int type, u32 pid, u32 seq, u32 change,
1005                             unsigned int flags, u32 ext_filter_mask)
1006 {
1007         struct ifinfomsg *ifm;
1008         struct nlmsghdr *nlh;
1009         struct rtnl_link_stats64 temp;
1010         const struct rtnl_link_stats64 *stats;
1011         struct nlattr *attr, *af_spec;
1012         struct rtnl_af_ops *af_ops;
1013         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1014
1015         ASSERT_RTNL();
1016         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1017         if (nlh == NULL)
1018                 return -EMSGSIZE;
1019
1020         ifm = nlmsg_data(nlh);
1021         ifm->ifi_family = AF_UNSPEC;
1022         ifm->__ifi_pad = 0;
1023         ifm->ifi_type = dev->type;
1024         ifm->ifi_index = dev->ifindex;
1025         ifm->ifi_flags = dev_get_flags(dev);
1026         ifm->ifi_change = change;
1027
1028         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1029             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1030             nla_put_u8(skb, IFLA_OPERSTATE,
1031                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1032             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1033             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1034             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1035             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1036             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1037 #ifdef CONFIG_RPS
1038             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1039 #endif
1040             (dev->ifindex != dev->iflink &&
1041              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1042             (upper_dev &&
1043              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1044             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1045             (dev->qdisc &&
1046              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1047             (dev->ifalias &&
1048              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1049             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1050                         atomic_read(&dev->carrier_changes)))
1051                 goto nla_put_failure;
1052
1053         if (1) {
1054                 struct rtnl_link_ifmap map = {
1055                         .mem_start   = dev->mem_start,
1056                         .mem_end     = dev->mem_end,
1057                         .base_addr   = dev->base_addr,
1058                         .irq         = dev->irq,
1059                         .dma         = dev->dma,
1060                         .port        = dev->if_port,
1061                 };
1062                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1063                         goto nla_put_failure;
1064         }
1065
1066         if (dev->addr_len) {
1067                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1068                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1069                         goto nla_put_failure;
1070         }
1071
1072         if (rtnl_phys_port_id_fill(skb, dev))
1073                 goto nla_put_failure;
1074
1075         if (rtnl_phys_switch_id_fill(skb, dev))
1076                 goto nla_put_failure;
1077
1078         attr = nla_reserve(skb, IFLA_STATS,
1079                         sizeof(struct rtnl_link_stats));
1080         if (attr == NULL)
1081                 goto nla_put_failure;
1082
1083         stats = dev_get_stats(dev, &temp);
1084         copy_rtnl_link_stats(nla_data(attr), stats);
1085
1086         attr = nla_reserve(skb, IFLA_STATS64,
1087                         sizeof(struct rtnl_link_stats64));
1088         if (attr == NULL)
1089                 goto nla_put_failure;
1090         copy_rtnl_link_stats64(nla_data(attr), stats);
1091
1092         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1093             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1094                 goto nla_put_failure;
1095
1096         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1097             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1098                 int i;
1099
1100                 struct nlattr *vfinfo, *vf;
1101                 int num_vfs = dev_num_vf(dev->dev.parent);
1102
1103                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1104                 if (!vfinfo)
1105                         goto nla_put_failure;
1106                 for (i = 0; i < num_vfs; i++) {
1107                         struct ifla_vf_info ivi;
1108                         struct ifla_vf_mac vf_mac;
1109                         struct ifla_vf_vlan vf_vlan;
1110                         struct ifla_vf_rate vf_rate;
1111                         struct ifla_vf_tx_rate vf_tx_rate;
1112                         struct ifla_vf_spoofchk vf_spoofchk;
1113                         struct ifla_vf_link_state vf_linkstate;
1114
1115                         /*
1116                          * Not all SR-IOV capable drivers support the
1117                          * spoofcheck query.  Preset to -1 so the user
1118                          * space tool can detect that the driver didn't
1119                          * report anything.
1120                          */
1121                         ivi.spoofchk = -1;
1122                         memset(ivi.mac, 0, sizeof(ivi.mac));
1123                         /* The default value for VF link state is "auto"
1124                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1125                          */
1126                         ivi.linkstate = 0;
1127                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1128                                 break;
1129                         vf_mac.vf =
1130                                 vf_vlan.vf =
1131                                 vf_rate.vf =
1132                                 vf_tx_rate.vf =
1133                                 vf_spoofchk.vf =
1134                                 vf_linkstate.vf = ivi.vf;
1135
1136                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1137                         vf_vlan.vlan = ivi.vlan;
1138                         vf_vlan.qos = ivi.qos;
1139                         vf_tx_rate.rate = ivi.max_tx_rate;
1140                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1141                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1142                         vf_spoofchk.setting = ivi.spoofchk;
1143                         vf_linkstate.link_state = ivi.linkstate;
1144                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1145                         if (!vf) {
1146                                 nla_nest_cancel(skb, vfinfo);
1147                                 goto nla_put_failure;
1148                         }
1149                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1150                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1151                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1152                                     &vf_rate) ||
1153                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1154                                     &vf_tx_rate) ||
1155                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1156                                     &vf_spoofchk) ||
1157                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1158                                     &vf_linkstate))
1159                                 goto nla_put_failure;
1160                         nla_nest_end(skb, vf);
1161                 }
1162                 nla_nest_end(skb, vfinfo);
1163         }
1164
1165         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1166                 goto nla_put_failure;
1167
1168         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1169                 if (rtnl_link_fill(skb, dev) < 0)
1170                         goto nla_put_failure;
1171         }
1172
1173         if (dev->rtnl_link_ops &&
1174             dev->rtnl_link_ops->get_link_net) {
1175                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1176
1177                 if (!net_eq(dev_net(dev), link_net)) {
1178                         int id = peernet2id(dev_net(dev), link_net);
1179
1180                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1181                                 goto nla_put_failure;
1182                 }
1183         }
1184
1185         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1186                 goto nla_put_failure;
1187
1188         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1189                 if (af_ops->fill_link_af) {
1190                         struct nlattr *af;
1191                         int err;
1192
1193                         if (!(af = nla_nest_start(skb, af_ops->family)))
1194                                 goto nla_put_failure;
1195
1196                         err = af_ops->fill_link_af(skb, dev);
1197
1198                         /*
1199                          * Caller may return ENODATA to indicate that there
1200                          * was no data to be dumped. This is not an error, it
1201                          * means we should trim the attribute header and
1202                          * continue.
1203                          */
1204                         if (err == -ENODATA)
1205                                 nla_nest_cancel(skb, af);
1206                         else if (err < 0)
1207                                 goto nla_put_failure;
1208
1209                         nla_nest_end(skb, af);
1210                 }
1211         }
1212
1213         nla_nest_end(skb, af_spec);
1214
1215         nlmsg_end(skb, nlh);
1216         return 0;
1217
1218 nla_put_failure:
1219         nlmsg_cancel(skb, nlh);
1220         return -EMSGSIZE;
1221 }
1222
1223 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1224         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1225         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1226         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1227         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1228         [IFLA_MTU]              = { .type = NLA_U32 },
1229         [IFLA_LINK]             = { .type = NLA_U32 },
1230         [IFLA_MASTER]           = { .type = NLA_U32 },
1231         [IFLA_CARRIER]          = { .type = NLA_U8 },
1232         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1233         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1234         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1235         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1236         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1237         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1238         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1239         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1240         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1241         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1242         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1243         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1244         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1245         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1246         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1247         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1248         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1249         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1250         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1251         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1252 };
1253
1254 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1255         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1256         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1257         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1258         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1259 };
1260
1261 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1262         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1263 };
1264
1265 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1266         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1267         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1268         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1269         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1270         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1271         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1272 };
1273
1274 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1275         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1276         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1277                                     .len = PORT_PROFILE_MAX },
1278         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1279                                     .len = sizeof(struct ifla_port_vsi)},
1280         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1281                                       .len = PORT_UUID_MAX },
1282         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1283                                     .len = PORT_UUID_MAX },
1284         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1285         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1286 };
1287
1288 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1289 {
1290         struct net *net = sock_net(skb->sk);
1291         int h, s_h;
1292         int idx = 0, s_idx;
1293         struct net_device *dev;
1294         struct hlist_head *head;
1295         struct nlattr *tb[IFLA_MAX+1];
1296         u32 ext_filter_mask = 0;
1297         int err;
1298         int hdrlen;
1299
1300         s_h = cb->args[0];
1301         s_idx = cb->args[1];
1302
1303         cb->seq = net->dev_base_seq;
1304
1305         /* A hack to preserve kernel<->userspace interface.
1306          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1307          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1308          * what iproute2 < v3.9.0 used.
1309          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1310          * attribute, its netlink message is shorter than struct ifinfomsg.
1311          */
1312         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1313                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1314
1315         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1316
1317                 if (tb[IFLA_EXT_MASK])
1318                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1319         }
1320
1321         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1322                 idx = 0;
1323                 head = &net->dev_index_head[h];
1324                 hlist_for_each_entry(dev, head, index_hlist) {
1325                         if (idx < s_idx)
1326                                 goto cont;
1327                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1328                                                NETLINK_CB(cb->skb).portid,
1329                                                cb->nlh->nlmsg_seq, 0,
1330                                                NLM_F_MULTI,
1331                                                ext_filter_mask);
1332                         /* If we ran out of room on the first message,
1333                          * we're in trouble
1334                          */
1335                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1336
1337                         if (err < 0)
1338                                 goto out;
1339
1340                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1341 cont:
1342                         idx++;
1343                 }
1344         }
1345 out:
1346         cb->args[1] = idx;
1347         cb->args[0] = h;
1348
1349         return skb->len;
1350 }
1351
1352 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1353 {
1354         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1355 }
1356 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1357
1358 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1359 {
1360         struct net *net;
1361         /* Examine the link attributes and figure out which
1362          * network namespace we are talking about.
1363          */
1364         if (tb[IFLA_NET_NS_PID])
1365                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1366         else if (tb[IFLA_NET_NS_FD])
1367                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1368         else
1369                 net = get_net(src_net);
1370         return net;
1371 }
1372 EXPORT_SYMBOL(rtnl_link_get_net);
1373
1374 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1375 {
1376         if (dev) {
1377                 if (tb[IFLA_ADDRESS] &&
1378                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1379                         return -EINVAL;
1380
1381                 if (tb[IFLA_BROADCAST] &&
1382                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1383                         return -EINVAL;
1384         }
1385
1386         if (tb[IFLA_AF_SPEC]) {
1387                 struct nlattr *af;
1388                 int rem, err;
1389
1390                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1391                         const struct rtnl_af_ops *af_ops;
1392
1393                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1394                                 return -EAFNOSUPPORT;
1395
1396                         if (!af_ops->set_link_af)
1397                                 return -EOPNOTSUPP;
1398
1399                         if (af_ops->validate_link_af) {
1400                                 err = af_ops->validate_link_af(dev, af);
1401                                 if (err < 0)
1402                                         return err;
1403                         }
1404                 }
1405         }
1406
1407         return 0;
1408 }
1409
1410 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1411 {
1412         int rem, err = -EINVAL;
1413         struct nlattr *vf;
1414         const struct net_device_ops *ops = dev->netdev_ops;
1415
1416         nla_for_each_nested(vf, attr, rem) {
1417                 switch (nla_type(vf)) {
1418                 case IFLA_VF_MAC: {
1419                         struct ifla_vf_mac *ivm;
1420                         ivm = nla_data(vf);
1421                         err = -EOPNOTSUPP;
1422                         if (ops->ndo_set_vf_mac)
1423                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1424                                                           ivm->mac);
1425                         break;
1426                 }
1427                 case IFLA_VF_VLAN: {
1428                         struct ifla_vf_vlan *ivv;
1429                         ivv = nla_data(vf);
1430                         err = -EOPNOTSUPP;
1431                         if (ops->ndo_set_vf_vlan)
1432                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1433                                                            ivv->vlan,
1434                                                            ivv->qos);
1435                         break;
1436                 }
1437                 case IFLA_VF_TX_RATE: {
1438                         struct ifla_vf_tx_rate *ivt;
1439                         struct ifla_vf_info ivf;
1440                         ivt = nla_data(vf);
1441                         err = -EOPNOTSUPP;
1442                         if (ops->ndo_get_vf_config)
1443                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1444                                                              &ivf);
1445                         if (err)
1446                                 break;
1447                         err = -EOPNOTSUPP;
1448                         if (ops->ndo_set_vf_rate)
1449                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1450                                                            ivf.min_tx_rate,
1451                                                            ivt->rate);
1452                         break;
1453                 }
1454                 case IFLA_VF_RATE: {
1455                         struct ifla_vf_rate *ivt;
1456                         ivt = nla_data(vf);
1457                         err = -EOPNOTSUPP;
1458                         if (ops->ndo_set_vf_rate)
1459                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1460                                                            ivt->min_tx_rate,
1461                                                            ivt->max_tx_rate);
1462                         break;
1463                 }
1464                 case IFLA_VF_SPOOFCHK: {
1465                         struct ifla_vf_spoofchk *ivs;
1466                         ivs = nla_data(vf);
1467                         err = -EOPNOTSUPP;
1468                         if (ops->ndo_set_vf_spoofchk)
1469                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1470                                                                ivs->setting);
1471                         break;
1472                 }
1473                 case IFLA_VF_LINK_STATE: {
1474                         struct ifla_vf_link_state *ivl;
1475                         ivl = nla_data(vf);
1476                         err = -EOPNOTSUPP;
1477                         if (ops->ndo_set_vf_link_state)
1478                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1479                                                                  ivl->link_state);
1480                         break;
1481                 }
1482                 default:
1483                         err = -EINVAL;
1484                         break;
1485                 }
1486                 if (err)
1487                         break;
1488         }
1489         return err;
1490 }
1491
1492 static int do_set_master(struct net_device *dev, int ifindex)
1493 {
1494         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1495         const struct net_device_ops *ops;
1496         int err;
1497
1498         if (upper_dev) {
1499                 if (upper_dev->ifindex == ifindex)
1500                         return 0;
1501                 ops = upper_dev->netdev_ops;
1502                 if (ops->ndo_del_slave) {
1503                         err = ops->ndo_del_slave(upper_dev, dev);
1504                         if (err)
1505                                 return err;
1506                 } else {
1507                         return -EOPNOTSUPP;
1508                 }
1509         }
1510
1511         if (ifindex) {
1512                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1513                 if (!upper_dev)
1514                         return -EINVAL;
1515                 ops = upper_dev->netdev_ops;
1516                 if (ops->ndo_add_slave) {
1517                         err = ops->ndo_add_slave(upper_dev, dev);
1518                         if (err)
1519                                 return err;
1520                 } else {
1521                         return -EOPNOTSUPP;
1522                 }
1523         }
1524         return 0;
1525 }
1526
1527 #define DO_SETLINK_MODIFIED     0x01
1528 /* notify flag means notify + modified. */
1529 #define DO_SETLINK_NOTIFY       0x03
1530 static int do_setlink(const struct sk_buff *skb,
1531                       struct net_device *dev, struct ifinfomsg *ifm,
1532                       struct nlattr **tb, char *ifname, int status)
1533 {
1534         const struct net_device_ops *ops = dev->netdev_ops;
1535         int err;
1536
1537         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1538                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1539                 if (IS_ERR(net)) {
1540                         err = PTR_ERR(net);
1541                         goto errout;
1542                 }
1543                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1544                         put_net(net);
1545                         err = -EPERM;
1546                         goto errout;
1547                 }
1548                 err = dev_change_net_namespace(dev, net, ifname);
1549                 put_net(net);
1550                 if (err)
1551                         goto errout;
1552                 status |= DO_SETLINK_MODIFIED;
1553         }
1554
1555         if (tb[IFLA_MAP]) {
1556                 struct rtnl_link_ifmap *u_map;
1557                 struct ifmap k_map;
1558
1559                 if (!ops->ndo_set_config) {
1560                         err = -EOPNOTSUPP;
1561                         goto errout;
1562                 }
1563
1564                 if (!netif_device_present(dev)) {
1565                         err = -ENODEV;
1566                         goto errout;
1567                 }
1568
1569                 u_map = nla_data(tb[IFLA_MAP]);
1570                 k_map.mem_start = (unsigned long) u_map->mem_start;
1571                 k_map.mem_end = (unsigned long) u_map->mem_end;
1572                 k_map.base_addr = (unsigned short) u_map->base_addr;
1573                 k_map.irq = (unsigned char) u_map->irq;
1574                 k_map.dma = (unsigned char) u_map->dma;
1575                 k_map.port = (unsigned char) u_map->port;
1576
1577                 err = ops->ndo_set_config(dev, &k_map);
1578                 if (err < 0)
1579                         goto errout;
1580
1581                 status |= DO_SETLINK_NOTIFY;
1582         }
1583
1584         if (tb[IFLA_ADDRESS]) {
1585                 struct sockaddr *sa;
1586                 int len;
1587
1588                 len = sizeof(sa_family_t) + dev->addr_len;
1589                 sa = kmalloc(len, GFP_KERNEL);
1590                 if (!sa) {
1591                         err = -ENOMEM;
1592                         goto errout;
1593                 }
1594                 sa->sa_family = dev->type;
1595                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1596                        dev->addr_len);
1597                 err = dev_set_mac_address(dev, sa);
1598                 kfree(sa);
1599                 if (err)
1600                         goto errout;
1601                 status |= DO_SETLINK_MODIFIED;
1602         }
1603
1604         if (tb[IFLA_MTU]) {
1605                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1606                 if (err < 0)
1607                         goto errout;
1608                 status |= DO_SETLINK_MODIFIED;
1609         }
1610
1611         if (tb[IFLA_GROUP]) {
1612                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1613                 status |= DO_SETLINK_NOTIFY;
1614         }
1615
1616         /*
1617          * Interface selected by interface index but interface
1618          * name provided implies that a name change has been
1619          * requested.
1620          */
1621         if (ifm->ifi_index > 0 && ifname[0]) {
1622                 err = dev_change_name(dev, ifname);
1623                 if (err < 0)
1624                         goto errout;
1625                 status |= DO_SETLINK_MODIFIED;
1626         }
1627
1628         if (tb[IFLA_IFALIAS]) {
1629                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1630                                     nla_len(tb[IFLA_IFALIAS]));
1631                 if (err < 0)
1632                         goto errout;
1633                 status |= DO_SETLINK_NOTIFY;
1634         }
1635
1636         if (tb[IFLA_BROADCAST]) {
1637                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1638                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1639         }
1640
1641         if (ifm->ifi_flags || ifm->ifi_change) {
1642                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1643                 if (err < 0)
1644                         goto errout;
1645         }
1646
1647         if (tb[IFLA_MASTER]) {
1648                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1649                 if (err)
1650                         goto errout;
1651                 status |= DO_SETLINK_MODIFIED;
1652         }
1653
1654         if (tb[IFLA_CARRIER]) {
1655                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1656                 if (err)
1657                         goto errout;
1658                 status |= DO_SETLINK_MODIFIED;
1659         }
1660
1661         if (tb[IFLA_TXQLEN]) {
1662                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1663
1664                 if (dev->tx_queue_len ^ value)
1665                         status |= DO_SETLINK_NOTIFY;
1666
1667                 dev->tx_queue_len = value;
1668         }
1669
1670         if (tb[IFLA_OPERSTATE])
1671                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1672
1673         if (tb[IFLA_LINKMODE]) {
1674                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1675
1676                 write_lock_bh(&dev_base_lock);
1677                 if (dev->link_mode ^ value)
1678                         status |= DO_SETLINK_NOTIFY;
1679                 dev->link_mode = value;
1680                 write_unlock_bh(&dev_base_lock);
1681         }
1682
1683         if (tb[IFLA_VFINFO_LIST]) {
1684                 struct nlattr *attr;
1685                 int rem;
1686                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1687                         if (nla_type(attr) != IFLA_VF_INFO) {
1688                                 err = -EINVAL;
1689                                 goto errout;
1690                         }
1691                         err = do_setvfinfo(dev, attr);
1692                         if (err < 0)
1693                                 goto errout;
1694                         status |= DO_SETLINK_NOTIFY;
1695                 }
1696         }
1697         err = 0;
1698
1699         if (tb[IFLA_VF_PORTS]) {
1700                 struct nlattr *port[IFLA_PORT_MAX+1];
1701                 struct nlattr *attr;
1702                 int vf;
1703                 int rem;
1704
1705                 err = -EOPNOTSUPP;
1706                 if (!ops->ndo_set_vf_port)
1707                         goto errout;
1708
1709                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1710                         if (nla_type(attr) != IFLA_VF_PORT)
1711                                 continue;
1712                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1713                                 attr, ifla_port_policy);
1714                         if (err < 0)
1715                                 goto errout;
1716                         if (!port[IFLA_PORT_VF]) {
1717                                 err = -EOPNOTSUPP;
1718                                 goto errout;
1719                         }
1720                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1721                         err = ops->ndo_set_vf_port(dev, vf, port);
1722                         if (err < 0)
1723                                 goto errout;
1724                         status |= DO_SETLINK_NOTIFY;
1725                 }
1726         }
1727         err = 0;
1728
1729         if (tb[IFLA_PORT_SELF]) {
1730                 struct nlattr *port[IFLA_PORT_MAX+1];
1731
1732                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1733                         tb[IFLA_PORT_SELF], ifla_port_policy);
1734                 if (err < 0)
1735                         goto errout;
1736
1737                 err = -EOPNOTSUPP;
1738                 if (ops->ndo_set_vf_port)
1739                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1740                 if (err < 0)
1741                         goto errout;
1742                 status |= DO_SETLINK_NOTIFY;
1743         }
1744
1745         if (tb[IFLA_AF_SPEC]) {
1746                 struct nlattr *af;
1747                 int rem;
1748
1749                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1750                         const struct rtnl_af_ops *af_ops;
1751
1752                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1753                                 BUG();
1754
1755                         err = af_ops->set_link_af(dev, af);
1756                         if (err < 0)
1757                                 goto errout;
1758
1759                         status |= DO_SETLINK_NOTIFY;
1760                 }
1761         }
1762         err = 0;
1763
1764 errout:
1765         if (status & DO_SETLINK_MODIFIED) {
1766                 if (status & DO_SETLINK_NOTIFY)
1767                         netdev_state_change(dev);
1768
1769                 if (err < 0)
1770                         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",
1771                                              dev->name);
1772         }
1773
1774         return err;
1775 }
1776
1777 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1778 {
1779         struct net *net = sock_net(skb->sk);
1780         struct ifinfomsg *ifm;
1781         struct net_device *dev;
1782         int err;
1783         struct nlattr *tb[IFLA_MAX+1];
1784         char ifname[IFNAMSIZ];
1785
1786         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1787         if (err < 0)
1788                 goto errout;
1789
1790         if (tb[IFLA_IFNAME])
1791                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1792         else
1793                 ifname[0] = '\0';
1794
1795         err = -EINVAL;
1796         ifm = nlmsg_data(nlh);
1797         if (ifm->ifi_index > 0)
1798                 dev = __dev_get_by_index(net, ifm->ifi_index);
1799         else if (tb[IFLA_IFNAME])
1800                 dev = __dev_get_by_name(net, ifname);
1801         else
1802                 goto errout;
1803
1804         if (dev == NULL) {
1805                 err = -ENODEV;
1806                 goto errout;
1807         }
1808
1809         err = validate_linkmsg(dev, tb);
1810         if (err < 0)
1811                 goto errout;
1812
1813         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1814 errout:
1815         return err;
1816 }
1817
1818 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1819 {
1820         struct net *net = sock_net(skb->sk);
1821         const struct rtnl_link_ops *ops;
1822         struct net_device *dev;
1823         struct ifinfomsg *ifm;
1824         char ifname[IFNAMSIZ];
1825         struct nlattr *tb[IFLA_MAX+1];
1826         int err;
1827         LIST_HEAD(list_kill);
1828
1829         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1830         if (err < 0)
1831                 return err;
1832
1833         if (tb[IFLA_IFNAME])
1834                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1835
1836         ifm = nlmsg_data(nlh);
1837         if (ifm->ifi_index > 0)
1838                 dev = __dev_get_by_index(net, ifm->ifi_index);
1839         else if (tb[IFLA_IFNAME])
1840                 dev = __dev_get_by_name(net, ifname);
1841         else
1842                 return -EINVAL;
1843
1844         if (!dev)
1845                 return -ENODEV;
1846
1847         ops = dev->rtnl_link_ops;
1848         if (!ops || !ops->dellink)
1849                 return -EOPNOTSUPP;
1850
1851         ops->dellink(dev, &list_kill);
1852         unregister_netdevice_many(&list_kill);
1853         return 0;
1854 }
1855
1856 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1857 {
1858         unsigned int old_flags;
1859         int err;
1860
1861         old_flags = dev->flags;
1862         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1863                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1864                 if (err < 0)
1865                         return err;
1866         }
1867
1868         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1869
1870         __dev_notify_flags(dev, old_flags, ~0U);
1871         return 0;
1872 }
1873 EXPORT_SYMBOL(rtnl_configure_link);
1874
1875 struct net_device *rtnl_create_link(struct net *net,
1876         char *ifname, unsigned char name_assign_type,
1877         const struct rtnl_link_ops *ops, struct nlattr *tb[])
1878 {
1879         int err;
1880         struct net_device *dev;
1881         unsigned int num_tx_queues = 1;
1882         unsigned int num_rx_queues = 1;
1883
1884         if (tb[IFLA_NUM_TX_QUEUES])
1885                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1886         else if (ops->get_num_tx_queues)
1887                 num_tx_queues = ops->get_num_tx_queues();
1888
1889         if (tb[IFLA_NUM_RX_QUEUES])
1890                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1891         else if (ops->get_num_rx_queues)
1892                 num_rx_queues = ops->get_num_rx_queues();
1893
1894         err = -ENOMEM;
1895         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1896                                ops->setup, num_tx_queues, num_rx_queues);
1897         if (!dev)
1898                 goto err;
1899
1900         dev_net_set(dev, net);
1901         dev->rtnl_link_ops = ops;
1902         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1903
1904         if (tb[IFLA_MTU])
1905                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1906         if (tb[IFLA_ADDRESS]) {
1907                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1908                                 nla_len(tb[IFLA_ADDRESS]));
1909                 dev->addr_assign_type = NET_ADDR_SET;
1910         }
1911         if (tb[IFLA_BROADCAST])
1912                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1913                                 nla_len(tb[IFLA_BROADCAST]));
1914         if (tb[IFLA_TXQLEN])
1915                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1916         if (tb[IFLA_OPERSTATE])
1917                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1918         if (tb[IFLA_LINKMODE])
1919                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1920         if (tb[IFLA_GROUP])
1921                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1922
1923         return dev;
1924
1925 err:
1926         return ERR_PTR(err);
1927 }
1928 EXPORT_SYMBOL(rtnl_create_link);
1929
1930 static int rtnl_group_changelink(const struct sk_buff *skb,
1931                 struct net *net, int group,
1932                 struct ifinfomsg *ifm,
1933                 struct nlattr **tb)
1934 {
1935         struct net_device *dev;
1936         int err;
1937
1938         for_each_netdev(net, dev) {
1939                 if (dev->group == group) {
1940                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1941                         if (err < 0)
1942                                 return err;
1943                 }
1944         }
1945
1946         return 0;
1947 }
1948
1949 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1950 {
1951         struct net *net = sock_net(skb->sk);
1952         const struct rtnl_link_ops *ops;
1953         const struct rtnl_link_ops *m_ops = NULL;
1954         struct net_device *dev;
1955         struct net_device *master_dev = NULL;
1956         struct ifinfomsg *ifm;
1957         char kind[MODULE_NAME_LEN];
1958         char ifname[IFNAMSIZ];
1959         struct nlattr *tb[IFLA_MAX+1];
1960         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1961         unsigned char name_assign_type = NET_NAME_USER;
1962         int err;
1963
1964 #ifdef CONFIG_MODULES
1965 replay:
1966 #endif
1967         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1968         if (err < 0)
1969                 return err;
1970
1971         if (tb[IFLA_IFNAME])
1972                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1973         else
1974                 ifname[0] = '\0';
1975
1976         ifm = nlmsg_data(nlh);
1977         if (ifm->ifi_index > 0)
1978                 dev = __dev_get_by_index(net, ifm->ifi_index);
1979         else {
1980                 if (ifname[0])
1981                         dev = __dev_get_by_name(net, ifname);
1982                 else
1983                         dev = NULL;
1984         }
1985
1986         if (dev) {
1987                 master_dev = netdev_master_upper_dev_get(dev);
1988                 if (master_dev)
1989                         m_ops = master_dev->rtnl_link_ops;
1990         }
1991
1992         err = validate_linkmsg(dev, tb);
1993         if (err < 0)
1994                 return err;
1995
1996         if (tb[IFLA_LINKINFO]) {
1997                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1998                                        tb[IFLA_LINKINFO], ifla_info_policy);
1999                 if (err < 0)
2000                         return err;
2001         } else
2002                 memset(linkinfo, 0, sizeof(linkinfo));
2003
2004         if (linkinfo[IFLA_INFO_KIND]) {
2005                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2006                 ops = rtnl_link_ops_get(kind);
2007         } else {
2008                 kind[0] = '\0';
2009                 ops = NULL;
2010         }
2011
2012         if (1) {
2013                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2014                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2015                 struct nlattr **data = NULL;
2016                 struct nlattr **slave_data = NULL;
2017                 struct net *dest_net, *link_net = NULL;
2018
2019                 if (ops) {
2020                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2021                                 err = nla_parse_nested(attr, ops->maxtype,
2022                                                        linkinfo[IFLA_INFO_DATA],
2023                                                        ops->policy);
2024                                 if (err < 0)
2025                                         return err;
2026                                 data = attr;
2027                         }
2028                         if (ops->validate) {
2029                                 err = ops->validate(tb, data);
2030                                 if (err < 0)
2031                                         return err;
2032                         }
2033                 }
2034
2035                 if (m_ops) {
2036                         if (m_ops->slave_maxtype &&
2037                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2038                                 err = nla_parse_nested(slave_attr,
2039                                                        m_ops->slave_maxtype,
2040                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2041                                                        m_ops->slave_policy);
2042                                 if (err < 0)
2043                                         return err;
2044                                 slave_data = slave_attr;
2045                         }
2046                         if (m_ops->slave_validate) {
2047                                 err = m_ops->slave_validate(tb, slave_data);
2048                                 if (err < 0)
2049                                         return err;
2050                         }
2051                 }
2052
2053                 if (dev) {
2054                         int status = 0;
2055
2056                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2057                                 return -EEXIST;
2058                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2059                                 return -EOPNOTSUPP;
2060
2061                         if (linkinfo[IFLA_INFO_DATA]) {
2062                                 if (!ops || ops != dev->rtnl_link_ops ||
2063                                     !ops->changelink)
2064                                         return -EOPNOTSUPP;
2065
2066                                 err = ops->changelink(dev, tb, data);
2067                                 if (err < 0)
2068                                         return err;
2069                                 status |= DO_SETLINK_NOTIFY;
2070                         }
2071
2072                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2073                                 if (!m_ops || !m_ops->slave_changelink)
2074                                         return -EOPNOTSUPP;
2075
2076                                 err = m_ops->slave_changelink(master_dev, dev,
2077                                                               tb, slave_data);
2078                                 if (err < 0)
2079                                         return err;
2080                                 status |= DO_SETLINK_NOTIFY;
2081                         }
2082
2083                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2084                 }
2085
2086                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2087                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2088                                 return rtnl_group_changelink(skb, net,
2089                                                 nla_get_u32(tb[IFLA_GROUP]),
2090                                                 ifm, tb);
2091                         return -ENODEV;
2092                 }
2093
2094                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2095                         return -EOPNOTSUPP;
2096
2097                 if (!ops) {
2098 #ifdef CONFIG_MODULES
2099                         if (kind[0]) {
2100                                 __rtnl_unlock();
2101                                 request_module("rtnl-link-%s", kind);
2102                                 rtnl_lock();
2103                                 ops = rtnl_link_ops_get(kind);
2104                                 if (ops)
2105                                         goto replay;
2106                         }
2107 #endif
2108                         return -EOPNOTSUPP;
2109                 }
2110
2111                 if (!ops->setup)
2112                         return -EOPNOTSUPP;
2113
2114                 if (!ifname[0]) {
2115                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2116                         name_assign_type = NET_NAME_ENUM;
2117                 }
2118
2119                 dest_net = rtnl_link_get_net(net, tb);
2120                 if (IS_ERR(dest_net))
2121                         return PTR_ERR(dest_net);
2122
2123                 err = -EPERM;
2124                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2125                         goto out;
2126
2127                 if (tb[IFLA_LINK_NETNSID]) {
2128                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2129
2130                         link_net = get_net_ns_by_id(dest_net, id);
2131                         if (!link_net) {
2132                                 err =  -EINVAL;
2133                                 goto out;
2134                         }
2135                         err = -EPERM;
2136                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2137                                 goto out;
2138                 }
2139
2140                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2141                                        name_assign_type, ops, tb);
2142                 if (IS_ERR(dev)) {
2143                         err = PTR_ERR(dev);
2144                         goto out;
2145                 }
2146
2147                 dev->ifindex = ifm->ifi_index;
2148
2149                 if (ops->newlink) {
2150                         err = ops->newlink(link_net ? : net, dev, tb, data);
2151                         /* Drivers should call free_netdev() in ->destructor
2152                          * and unregister it on failure after registration
2153                          * so that device could be finally freed in rtnl_unlock.
2154                          */
2155                         if (err < 0) {
2156                                 /* If device is not registered at all, free it now */
2157                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2158                                         free_netdev(dev);
2159                                 goto out;
2160                         }
2161                 } else {
2162                         err = register_netdevice(dev);
2163                         if (err < 0) {
2164                                 free_netdev(dev);
2165                                 goto out;
2166                         }
2167                 }
2168                 err = rtnl_configure_link(dev, ifm);
2169                 if (err < 0)
2170                         goto out_unregister;
2171                 if (link_net) {
2172                         err = dev_change_net_namespace(dev, dest_net, ifname);
2173                         if (err < 0)
2174                                 goto out_unregister;
2175                 }
2176 out:
2177                 if (link_net)
2178                         put_net(link_net);
2179                 put_net(dest_net);
2180                 return err;
2181 out_unregister:
2182                 if (ops->newlink) {
2183                         LIST_HEAD(list_kill);
2184
2185                         ops->dellink(dev, &list_kill);
2186                         unregister_netdevice_many(&list_kill);
2187                 } else {
2188                         unregister_netdevice(dev);
2189                 }
2190                 goto out;
2191         }
2192 }
2193
2194 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2195 {
2196         struct net *net = sock_net(skb->sk);
2197         struct ifinfomsg *ifm;
2198         char ifname[IFNAMSIZ];
2199         struct nlattr *tb[IFLA_MAX+1];
2200         struct net_device *dev = NULL;
2201         struct sk_buff *nskb;
2202         int err;
2203         u32 ext_filter_mask = 0;
2204
2205         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2206         if (err < 0)
2207                 return err;
2208
2209         if (tb[IFLA_IFNAME])
2210                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2211
2212         if (tb[IFLA_EXT_MASK])
2213                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2214
2215         ifm = nlmsg_data(nlh);
2216         if (ifm->ifi_index > 0)
2217                 dev = __dev_get_by_index(net, ifm->ifi_index);
2218         else if (tb[IFLA_IFNAME])
2219                 dev = __dev_get_by_name(net, ifname);
2220         else
2221                 return -EINVAL;
2222
2223         if (dev == NULL)
2224                 return -ENODEV;
2225
2226         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2227         if (nskb == NULL)
2228                 return -ENOBUFS;
2229
2230         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2231                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2232         if (err < 0) {
2233                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2234                 WARN_ON(err == -EMSGSIZE);
2235                 kfree_skb(nskb);
2236         } else
2237                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2238
2239         return err;
2240 }
2241
2242 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2243 {
2244         struct net *net = sock_net(skb->sk);
2245         struct net_device *dev;
2246         struct nlattr *tb[IFLA_MAX+1];
2247         u32 ext_filter_mask = 0;
2248         u16 min_ifinfo_dump_size = 0;
2249         int hdrlen;
2250
2251         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2252         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2253                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2254
2255         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2256                 if (tb[IFLA_EXT_MASK])
2257                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2258         }
2259
2260         if (!ext_filter_mask)
2261                 return NLMSG_GOODSIZE;
2262         /*
2263          * traverse the list of net devices and compute the minimum
2264          * buffer size based upon the filter mask.
2265          */
2266         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2267                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2268                                              if_nlmsg_size(dev,
2269                                                            ext_filter_mask));
2270         }
2271
2272         return min_ifinfo_dump_size;
2273 }
2274
2275 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2276 {
2277         int idx;
2278         int s_idx = cb->family;
2279
2280         if (s_idx == 0)
2281                 s_idx = 1;
2282         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2283                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2284                 if (idx < s_idx || idx == PF_PACKET)
2285                         continue;
2286                 if (rtnl_msg_handlers[idx] == NULL ||
2287                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2288                         continue;
2289                 if (idx > s_idx) {
2290                         memset(&cb->args[0], 0, sizeof(cb->args));
2291                         cb->prev_seq = 0;
2292                         cb->seq = 0;
2293                 }
2294                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2295                         break;
2296         }
2297         cb->family = idx;
2298
2299         return skb->len;
2300 }
2301
2302 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2303                                        unsigned int change, gfp_t flags)
2304 {
2305         struct net *net = dev_net(dev);
2306         struct sk_buff *skb;
2307         int err = -ENOBUFS;
2308         size_t if_info_size;
2309
2310         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2311         if (skb == NULL)
2312                 goto errout;
2313
2314         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2315         if (err < 0) {
2316                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2317                 WARN_ON(err == -EMSGSIZE);
2318                 kfree_skb(skb);
2319                 goto errout;
2320         }
2321         return skb;
2322 errout:
2323         if (err < 0)
2324                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2325         return NULL;
2326 }
2327
2328 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2329 {
2330         struct net *net = dev_net(dev);
2331
2332         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2333 }
2334
2335 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2336                   gfp_t flags)
2337 {
2338         struct sk_buff *skb;
2339
2340         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2341         if (skb)
2342                 rtmsg_ifinfo_send(skb, dev, flags);
2343 }
2344 EXPORT_SYMBOL(rtmsg_ifinfo);
2345
2346 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2347                                    struct net_device *dev,
2348                                    u8 *addr, u32 pid, u32 seq,
2349                                    int type, unsigned int flags,
2350                                    int nlflags)
2351 {
2352         struct nlmsghdr *nlh;
2353         struct ndmsg *ndm;
2354
2355         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2356         if (!nlh)
2357                 return -EMSGSIZE;
2358
2359         ndm = nlmsg_data(nlh);
2360         ndm->ndm_family  = AF_BRIDGE;
2361         ndm->ndm_pad1    = 0;
2362         ndm->ndm_pad2    = 0;
2363         ndm->ndm_flags   = flags;
2364         ndm->ndm_type    = 0;
2365         ndm->ndm_ifindex = dev->ifindex;
2366         ndm->ndm_state   = NUD_PERMANENT;
2367
2368         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2369                 goto nla_put_failure;
2370
2371         nlmsg_end(skb, nlh);
2372         return 0;
2373
2374 nla_put_failure:
2375         nlmsg_cancel(skb, nlh);
2376         return -EMSGSIZE;
2377 }
2378
2379 static inline size_t rtnl_fdb_nlmsg_size(void)
2380 {
2381         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2382 }
2383
2384 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2385 {
2386         struct net *net = dev_net(dev);
2387         struct sk_buff *skb;
2388         int err = -ENOBUFS;
2389
2390         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2391         if (!skb)
2392                 goto errout;
2393
2394         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2395         if (err < 0) {
2396                 kfree_skb(skb);
2397                 goto errout;
2398         }
2399
2400         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2401         return;
2402 errout:
2403         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2404 }
2405
2406 /**
2407  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2408  */
2409 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2410                      struct nlattr *tb[],
2411                      struct net_device *dev,
2412                      const unsigned char *addr, u16 vid,
2413                      u16 flags)
2414 {
2415         int err = -EINVAL;
2416
2417         /* If aging addresses are supported device will need to
2418          * implement its own handler for this.
2419          */
2420         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2421                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2422                 return err;
2423         }
2424
2425         if (vid) {
2426                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2427                 return err;
2428         }
2429
2430         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2431                 err = dev_uc_add_excl(dev, addr);
2432         else if (is_multicast_ether_addr(addr))
2433                 err = dev_mc_add_excl(dev, addr);
2434
2435         /* Only return duplicate errors if NLM_F_EXCL is set */
2436         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2437                 err = 0;
2438
2439         return err;
2440 }
2441 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2442
2443 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2444 {
2445         u16 vid = 0;
2446
2447         if (vlan_attr) {
2448                 if (nla_len(vlan_attr) != sizeof(u16)) {
2449                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2450                         return -EINVAL;
2451                 }
2452
2453                 vid = nla_get_u16(vlan_attr);
2454
2455                 if (!vid || vid >= VLAN_VID_MASK) {
2456                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2457                                 vid);
2458                         return -EINVAL;
2459                 }
2460         }
2461         *p_vid = vid;
2462         return 0;
2463 }
2464
2465 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2466 {
2467         struct net *net = sock_net(skb->sk);
2468         struct ndmsg *ndm;
2469         struct nlattr *tb[NDA_MAX+1];
2470         struct net_device *dev;
2471         u8 *addr;
2472         u16 vid;
2473         int err;
2474
2475         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2476         if (err < 0)
2477                 return err;
2478
2479         ndm = nlmsg_data(nlh);
2480         if (ndm->ndm_ifindex == 0) {
2481                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2482                 return -EINVAL;
2483         }
2484
2485         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2486         if (dev == NULL) {
2487                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2488                 return -ENODEV;
2489         }
2490
2491         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2492                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2493                 return -EINVAL;
2494         }
2495
2496         addr = nla_data(tb[NDA_LLADDR]);
2497
2498         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2499         if (err)
2500                 return err;
2501
2502         err = -EOPNOTSUPP;
2503
2504         /* Support fdb on master device the net/bridge default case */
2505         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2506             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2507                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2508                 const struct net_device_ops *ops = br_dev->netdev_ops;
2509
2510                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2511                                        nlh->nlmsg_flags);
2512                 if (err)
2513                         goto out;
2514                 else
2515                         ndm->ndm_flags &= ~NTF_MASTER;
2516         }
2517
2518         /* Embedded bridge, macvlan, and any other device support */
2519         if ((ndm->ndm_flags & NTF_SELF)) {
2520                 if (dev->netdev_ops->ndo_fdb_add)
2521                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2522                                                            vid,
2523                                                            nlh->nlmsg_flags);
2524                 else
2525                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2526                                                nlh->nlmsg_flags);
2527
2528                 if (!err) {
2529                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2530                         ndm->ndm_flags &= ~NTF_SELF;
2531                 }
2532         }
2533 out:
2534         return err;
2535 }
2536
2537 /**
2538  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2539  */
2540 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2541                      struct nlattr *tb[],
2542                      struct net_device *dev,
2543                      const unsigned char *addr, u16 vid)
2544 {
2545         int err = -EINVAL;
2546
2547         /* If aging addresses are supported device will need to
2548          * implement its own handler for this.
2549          */
2550         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2551                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2552                 return err;
2553         }
2554
2555         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2556                 err = dev_uc_del(dev, addr);
2557         else if (is_multicast_ether_addr(addr))
2558                 err = dev_mc_del(dev, addr);
2559
2560         return err;
2561 }
2562 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2563
2564 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2565 {
2566         struct net *net = sock_net(skb->sk);
2567         struct ndmsg *ndm;
2568         struct nlattr *tb[NDA_MAX+1];
2569         struct net_device *dev;
2570         int err = -EINVAL;
2571         __u8 *addr;
2572         u16 vid;
2573
2574         if (!netlink_capable(skb, CAP_NET_ADMIN))
2575                 return -EPERM;
2576
2577         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2578         if (err < 0)
2579                 return err;
2580
2581         ndm = nlmsg_data(nlh);
2582         if (ndm->ndm_ifindex == 0) {
2583                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2584                 return -EINVAL;
2585         }
2586
2587         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2588         if (dev == NULL) {
2589                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2590                 return -ENODEV;
2591         }
2592
2593         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2594                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2595                 return -EINVAL;
2596         }
2597
2598         addr = nla_data(tb[NDA_LLADDR]);
2599
2600         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2601         if (err)
2602                 return err;
2603
2604         err = -EOPNOTSUPP;
2605
2606         /* Support fdb on master device the net/bridge default case */
2607         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2608             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2609                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2610                 const struct net_device_ops *ops = br_dev->netdev_ops;
2611
2612                 if (ops->ndo_fdb_del)
2613                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2614
2615                 if (err)
2616                         goto out;
2617                 else
2618                         ndm->ndm_flags &= ~NTF_MASTER;
2619         }
2620
2621         /* Embedded bridge, macvlan, and any other device support */
2622         if (ndm->ndm_flags & NTF_SELF) {
2623                 if (dev->netdev_ops->ndo_fdb_del)
2624                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2625                                                            vid);
2626                 else
2627                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2628
2629                 if (!err) {
2630                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2631                         ndm->ndm_flags &= ~NTF_SELF;
2632                 }
2633         }
2634 out:
2635         return err;
2636 }
2637
2638 static int nlmsg_populate_fdb(struct sk_buff *skb,
2639                               struct netlink_callback *cb,
2640                               struct net_device *dev,
2641                               int *idx,
2642                               struct netdev_hw_addr_list *list)
2643 {
2644         struct netdev_hw_addr *ha;
2645         int err;
2646         u32 portid, seq;
2647
2648         portid = NETLINK_CB(cb->skb).portid;
2649         seq = cb->nlh->nlmsg_seq;
2650
2651         list_for_each_entry(ha, &list->list, list) {
2652                 if (*idx < cb->args[0])
2653                         goto skip;
2654
2655                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2656                                               portid, seq,
2657                                               RTM_NEWNEIGH, NTF_SELF,
2658                                               NLM_F_MULTI);
2659                 if (err < 0)
2660                         return err;
2661 skip:
2662                 *idx += 1;
2663         }
2664         return 0;
2665 }
2666
2667 /**
2668  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2669  * @nlh: netlink message header
2670  * @dev: netdevice
2671  *
2672  * Default netdevice operation to dump the existing unicast address list.
2673  * Returns number of addresses from list put in skb.
2674  */
2675 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2676                       struct netlink_callback *cb,
2677                       struct net_device *dev,
2678                       struct net_device *filter_dev,
2679                       int idx)
2680 {
2681         int err;
2682
2683         netif_addr_lock_bh(dev);
2684         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2685         if (err)
2686                 goto out;
2687         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2688 out:
2689         netif_addr_unlock_bh(dev);
2690         return idx;
2691 }
2692 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2693
2694 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2695 {
2696         struct net_device *dev;
2697         struct nlattr *tb[IFLA_MAX+1];
2698         struct net_device *bdev = NULL;
2699         struct net_device *br_dev = NULL;
2700         const struct net_device_ops *ops = NULL;
2701         const struct net_device_ops *cops = NULL;
2702         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2703         struct net *net = sock_net(skb->sk);
2704         int brport_idx = 0;
2705         int br_idx = 0;
2706         int idx = 0;
2707
2708         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2709                         ifla_policy) == 0) {
2710                 if (tb[IFLA_MASTER])
2711                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2712         }
2713
2714         brport_idx = ifm->ifi_index;
2715
2716         if (br_idx) {
2717                 br_dev = __dev_get_by_index(net, br_idx);
2718                 if (!br_dev)
2719                         return -ENODEV;
2720
2721                 ops = br_dev->netdev_ops;
2722                 bdev = br_dev;
2723         }
2724
2725         for_each_netdev(net, dev) {
2726                 if (brport_idx && (dev->ifindex != brport_idx))
2727                         continue;
2728
2729                 if (!br_idx) { /* user did not specify a specific bridge */
2730                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2731                                 br_dev = netdev_master_upper_dev_get(dev);
2732                                 cops = br_dev->netdev_ops;
2733                         }
2734
2735                         bdev = dev;
2736                 } else {
2737                         if (dev != br_dev &&
2738                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2739                                 continue;
2740
2741                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2742                             !(dev->priv_flags & IFF_EBRIDGE))
2743                                 continue;
2744
2745                         bdev = br_dev;
2746                         cops = ops;
2747                 }
2748
2749                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2750                         if (cops && cops->ndo_fdb_dump)
2751                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2752                                                          idx);
2753                 }
2754
2755                 if (dev->netdev_ops->ndo_fdb_dump)
2756                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2757                                                             idx);
2758                 else
2759                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2760
2761                 cops = NULL;
2762         }
2763
2764         cb->args[0] = idx;
2765         return skb->len;
2766 }
2767
2768 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2769                                unsigned int attrnum, unsigned int flag)
2770 {
2771         if (mask & flag)
2772                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2773         return 0;
2774 }
2775
2776 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2777                             struct net_device *dev, u16 mode,
2778                             u32 flags, u32 mask)
2779 {
2780         struct nlmsghdr *nlh;
2781         struct ifinfomsg *ifm;
2782         struct nlattr *br_afspec;
2783         struct nlattr *protinfo;
2784         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2785         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2786
2787         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2788         if (nlh == NULL)
2789                 return -EMSGSIZE;
2790
2791         ifm = nlmsg_data(nlh);
2792         ifm->ifi_family = AF_BRIDGE;
2793         ifm->__ifi_pad = 0;
2794         ifm->ifi_type = dev->type;
2795         ifm->ifi_index = dev->ifindex;
2796         ifm->ifi_flags = dev_get_flags(dev);
2797         ifm->ifi_change = 0;
2798
2799
2800         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2801             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2802             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2803             (br_dev &&
2804              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2805             (dev->addr_len &&
2806              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2807             (dev->ifindex != dev->iflink &&
2808              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2809                 goto nla_put_failure;
2810
2811         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2812         if (!br_afspec)
2813                 goto nla_put_failure;
2814
2815         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2816                 nla_nest_cancel(skb, br_afspec);
2817                 goto nla_put_failure;
2818         }
2819
2820         if (mode != BRIDGE_MODE_UNDEF) {
2821                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2822                         nla_nest_cancel(skb, br_afspec);
2823                         goto nla_put_failure;
2824                 }
2825         }
2826         nla_nest_end(skb, br_afspec);
2827
2828         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2829         if (!protinfo)
2830                 goto nla_put_failure;
2831
2832         if (brport_nla_put_flag(skb, flags, mask,
2833                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2834             brport_nla_put_flag(skb, flags, mask,
2835                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2836             brport_nla_put_flag(skb, flags, mask,
2837                                 IFLA_BRPORT_FAST_LEAVE,
2838                                 BR_MULTICAST_FAST_LEAVE) ||
2839             brport_nla_put_flag(skb, flags, mask,
2840                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2841             brport_nla_put_flag(skb, flags, mask,
2842                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2843             brport_nla_put_flag(skb, flags, mask,
2844                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2845             brport_nla_put_flag(skb, flags, mask,
2846                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2847             brport_nla_put_flag(skb, flags, mask,
2848                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2849                 nla_nest_cancel(skb, protinfo);
2850                 goto nla_put_failure;
2851         }
2852
2853         nla_nest_end(skb, protinfo);
2854
2855         nlmsg_end(skb, nlh);
2856         return 0;
2857 nla_put_failure:
2858         nlmsg_cancel(skb, nlh);
2859         return -EMSGSIZE;
2860 }
2861 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2862
2863 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2864 {
2865         struct net *net = sock_net(skb->sk);
2866         struct net_device *dev;
2867         int idx = 0;
2868         u32 portid = NETLINK_CB(cb->skb).portid;
2869         u32 seq = cb->nlh->nlmsg_seq;
2870         u32 filter_mask = 0;
2871
2872         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2873                 struct nlattr *extfilt;
2874
2875                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2876                                           IFLA_EXT_MASK);
2877                 if (extfilt) {
2878                         if (nla_len(extfilt) < sizeof(filter_mask))
2879                                 return -EINVAL;
2880
2881                         filter_mask = nla_get_u32(extfilt);
2882                 }
2883         }
2884
2885         rcu_read_lock();
2886         for_each_netdev_rcu(net, dev) {
2887                 const struct net_device_ops *ops = dev->netdev_ops;
2888                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2889
2890                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2891                         if (idx >= cb->args[0] &&
2892                             br_dev->netdev_ops->ndo_bridge_getlink(
2893                                     skb, portid, seq, dev, filter_mask) < 0)
2894                                 break;
2895                         idx++;
2896                 }
2897
2898                 if (ops->ndo_bridge_getlink) {
2899                         if (idx >= cb->args[0] &&
2900                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2901                                                     filter_mask) < 0)
2902                                 break;
2903                         idx++;
2904                 }
2905         }
2906         rcu_read_unlock();
2907         cb->args[0] = idx;
2908
2909         return skb->len;
2910 }
2911
2912 static inline size_t bridge_nlmsg_size(void)
2913 {
2914         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2915                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2916                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2917                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2918                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2919                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2920                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2921                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2922                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2923                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2924                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2925 }
2926
2927 static int rtnl_bridge_notify(struct net_device *dev)
2928 {
2929         struct net *net = dev_net(dev);
2930         struct sk_buff *skb;
2931         int err = -EOPNOTSUPP;
2932
2933         if (!dev->netdev_ops->ndo_bridge_getlink)
2934                 return 0;
2935
2936         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2937         if (!skb) {
2938                 err = -ENOMEM;
2939                 goto errout;
2940         }
2941
2942         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2943         if (err < 0)
2944                 goto errout;
2945
2946         if (!skb->len)
2947                 goto errout;
2948
2949         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2950         return 0;
2951 errout:
2952         WARN_ON(err == -EMSGSIZE);
2953         kfree_skb(skb);
2954         if (err)
2955                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2956         return err;
2957 }
2958
2959 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2960 {
2961         struct net *net = sock_net(skb->sk);
2962         struct ifinfomsg *ifm;
2963         struct net_device *dev;
2964         struct nlattr *br_spec, *attr = NULL;
2965         int rem, err = -EOPNOTSUPP;
2966         u16 flags = 0;
2967         bool have_flags = false;
2968
2969         if (nlmsg_len(nlh) < sizeof(*ifm))
2970                 return -EINVAL;
2971
2972         ifm = nlmsg_data(nlh);
2973         if (ifm->ifi_family != AF_BRIDGE)
2974                 return -EPFNOSUPPORT;
2975
2976         dev = __dev_get_by_index(net, ifm->ifi_index);
2977         if (!dev) {
2978                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2979                 return -ENODEV;
2980         }
2981
2982         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2983         if (br_spec) {
2984                 nla_for_each_nested(attr, br_spec, rem) {
2985                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2986                                 if (nla_len(attr) < sizeof(flags))
2987                                         return -EINVAL;
2988
2989                                 have_flags = true;
2990                                 flags = nla_get_u16(attr);
2991                                 break;
2992                         }
2993                 }
2994         }
2995
2996         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2997                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2998
2999                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3000                         err = -EOPNOTSUPP;
3001                         goto out;
3002                 }
3003
3004                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3005                 if (err)
3006                         goto out;
3007
3008                 flags &= ~BRIDGE_FLAGS_MASTER;
3009         }
3010
3011         if ((flags & BRIDGE_FLAGS_SELF)) {
3012                 if (!dev->netdev_ops->ndo_bridge_setlink)
3013                         err = -EOPNOTSUPP;
3014                 else
3015                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3016                                                                   flags);
3017                 if (!err) {
3018                         flags &= ~BRIDGE_FLAGS_SELF;
3019
3020                         /* Generate event to notify upper layer of bridge
3021                          * change
3022                          */
3023                         err = rtnl_bridge_notify(dev);
3024                 }
3025         }
3026
3027         if (have_flags)
3028                 memcpy(nla_data(attr), &flags, sizeof(flags));
3029 out:
3030         return err;
3031 }
3032
3033 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3034 {
3035         struct net *net = sock_net(skb->sk);
3036         struct ifinfomsg *ifm;
3037         struct net_device *dev;
3038         struct nlattr *br_spec, *attr = NULL;
3039         int rem, err = -EOPNOTSUPP;
3040         u16 flags = 0;
3041         bool have_flags = false;
3042
3043         if (nlmsg_len(nlh) < sizeof(*ifm))
3044                 return -EINVAL;
3045
3046         ifm = nlmsg_data(nlh);
3047         if (ifm->ifi_family != AF_BRIDGE)
3048                 return -EPFNOSUPPORT;
3049
3050         dev = __dev_get_by_index(net, ifm->ifi_index);
3051         if (!dev) {
3052                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3053                 return -ENODEV;
3054         }
3055
3056         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3057         if (br_spec) {
3058                 nla_for_each_nested(attr, br_spec, rem) {
3059                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3060                                 if (nla_len(attr) < sizeof(flags))
3061                                         return -EINVAL;
3062
3063                                 have_flags = true;
3064                                 flags = nla_get_u16(attr);
3065                                 break;
3066                         }
3067                 }
3068         }
3069
3070         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3071                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3072
3073                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3074                         err = -EOPNOTSUPP;
3075                         goto out;
3076                 }
3077
3078                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3079                 if (err)
3080                         goto out;
3081
3082                 flags &= ~BRIDGE_FLAGS_MASTER;
3083         }
3084
3085         if ((flags & BRIDGE_FLAGS_SELF)) {
3086                 if (!dev->netdev_ops->ndo_bridge_dellink)
3087                         err = -EOPNOTSUPP;
3088                 else
3089                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3090                                                                   flags);
3091
3092                 if (!err) {
3093                         flags &= ~BRIDGE_FLAGS_SELF;
3094
3095                         /* Generate event to notify upper layer of bridge
3096                          * change
3097                          */
3098                         err = rtnl_bridge_notify(dev);
3099                 }
3100         }
3101
3102         if (have_flags)
3103                 memcpy(nla_data(attr), &flags, sizeof(flags));
3104 out:
3105         return err;
3106 }
3107
3108 /* Process one rtnetlink message. */
3109
3110 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3111 {
3112         struct net *net = sock_net(skb->sk);
3113         rtnl_doit_func doit;
3114         int sz_idx, kind;
3115         int family;
3116         int type;
3117         int err;
3118
3119         type = nlh->nlmsg_type;
3120         if (type > RTM_MAX)
3121                 return -EOPNOTSUPP;
3122
3123         type -= RTM_BASE;
3124
3125         /* All the messages must have at least 1 byte length */
3126         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3127                 return 0;
3128
3129         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3130         sz_idx = type>>2;
3131         kind = type&3;
3132
3133         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3134                 return -EPERM;
3135
3136         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3137                 struct sock *rtnl;
3138                 rtnl_dumpit_func dumpit;
3139                 rtnl_calcit_func calcit;
3140                 u16 min_dump_alloc = 0;
3141
3142                 dumpit = rtnl_get_dumpit(family, type);
3143                 if (dumpit == NULL)
3144                         return -EOPNOTSUPP;
3145                 calcit = rtnl_get_calcit(family, type);
3146                 if (calcit)
3147                         min_dump_alloc = calcit(skb, nlh);
3148
3149                 __rtnl_unlock();
3150                 rtnl = net->rtnl;
3151                 {
3152                         struct netlink_dump_control c = {
3153                                 .dump           = dumpit,
3154                                 .min_dump_alloc = min_dump_alloc,
3155                         };
3156                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3157                 }
3158                 rtnl_lock();
3159                 return err;
3160         }
3161
3162         doit = rtnl_get_doit(family, type);
3163         if (doit == NULL)
3164                 return -EOPNOTSUPP;
3165
3166         return doit(skb, nlh);
3167 }
3168
3169 static void rtnetlink_rcv(struct sk_buff *skb)
3170 {
3171         rtnl_lock();
3172         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3173         rtnl_unlock();
3174 }
3175
3176 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3177 {
3178         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3179
3180         switch (event) {
3181         case NETDEV_UP:
3182         case NETDEV_DOWN:
3183         case NETDEV_PRE_UP:
3184         case NETDEV_POST_INIT:
3185         case NETDEV_REGISTER:
3186         case NETDEV_CHANGE:
3187         case NETDEV_PRE_TYPE_CHANGE:
3188         case NETDEV_GOING_DOWN:
3189         case NETDEV_UNREGISTER:
3190         case NETDEV_UNREGISTER_FINAL:
3191         case NETDEV_RELEASE:
3192         case NETDEV_JOIN:
3193         case NETDEV_BONDING_INFO:
3194                 break;
3195         default:
3196                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3197                 break;
3198         }
3199         return NOTIFY_DONE;
3200 }
3201
3202 static struct notifier_block rtnetlink_dev_notifier = {
3203         .notifier_call  = rtnetlink_event,
3204 };
3205
3206
3207 static int __net_init rtnetlink_net_init(struct net *net)
3208 {
3209         struct sock *sk;
3210         struct netlink_kernel_cfg cfg = {
3211                 .groups         = RTNLGRP_MAX,
3212                 .input          = rtnetlink_rcv,
3213                 .cb_mutex       = &rtnl_mutex,
3214                 .flags          = NL_CFG_F_NONROOT_RECV,
3215         };
3216
3217         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3218         if (!sk)
3219                 return -ENOMEM;
3220         net->rtnl = sk;
3221         return 0;
3222 }
3223
3224 static void __net_exit rtnetlink_net_exit(struct net *net)
3225 {
3226         netlink_kernel_release(net->rtnl);
3227         net->rtnl = NULL;
3228 }
3229
3230 static struct pernet_operations rtnetlink_net_ops = {
3231         .init = rtnetlink_net_init,
3232         .exit = rtnetlink_net_exit,
3233 };
3234
3235 void __init rtnetlink_init(void)
3236 {
3237         if (register_pernet_subsys(&rtnetlink_net_ops))
3238                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3239
3240         register_netdevice_notifier(&rtnetlink_dev_notifier);
3241
3242         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3243                       rtnl_dump_ifinfo, rtnl_calcit);
3244         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3245         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3246         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3247
3248         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3249         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3250
3251         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3252         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3253         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3254
3255         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3256         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3257         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3258 }
3259