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 if (i == RTAX_FEATURES - 1) {
 682                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 683
 684                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 685                                 if (nla_put_u32(skb, i + 1, user_features))
 686                                         goto nla_put_failure;
 687                         } else {
 688                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 689                                         goto nla_put_failure;
 690                         }
 691                         valid++;
 692                 }
 693         }
 694
 695         if (!valid) {
 696                 nla_nest_cancel(skb, mx);
 697                 return 0;
 698         }
 699
 700         return nla_nest_end(skb, mx);
 701
 702 nla_put_failure:
 703         nla_nest_cancel(skb, mx);
 704         return -EMSGSIZE;
 705 }
 706 EXPORT_SYMBOL(rtnetlink_put_metrics);
 707
 708 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 709                        long expires, u32 error)
 710 {
 711         struct rta_cacheinfo ci = {
 712                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 713                 .rta_used = dst->__use,
 714                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 715                 .rta_error = error,
 716                 .rta_id =  id,
 717         };
 718
 719         if (expires) {
 720                 unsigned long clock;
 721
 722                 clock = jiffies_to_clock_t(abs(expires));
 723                 clock = min_t(unsigned long, clock, INT_MAX);
 724                 ci.rta_expires = (expires > 0) ? clock : -clock;
 725         }
 726         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 727 }
 728 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 729
 730 static void set_operstate(struct net_device *dev, unsigned char transition)
 731 {
 732         unsigned char operstate = dev->operstate;
 733
 734         switch (transition) {
 735         case IF_OPER_UP:
 736                 if ((operstate == IF_OPER_DORMANT ||
 737                      operstate == IF_OPER_UNKNOWN) &&
 738                     !netif_dormant(dev))
 739                         operstate = IF_OPER_UP;
 740                 break;
 741
 742         case IF_OPER_DORMANT:
 743                 if (operstate == IF_OPER_UP ||
 744                     operstate == IF_OPER_UNKNOWN)
 745                         operstate = IF_OPER_DORMANT;
 746                 break;
 747         }
 748
 749         if (dev->operstate != operstate) {
 750                 write_lock_bh(&dev_base_lock);
 751                 dev->operstate = operstate;
 752                 write_unlock_bh(&dev_base_lock);
 753                 netdev_state_change(dev);
 754         }
 755 }
 756
 757 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 758 {
 759         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 760                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 761 }
 762
 763 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 764                                            const struct ifinfomsg *ifm)
 765 {
 766         unsigned int flags = ifm->ifi_flags;
 767
 768         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 769         if (ifm->ifi_change)
 770                 flags = (flags & ifm->ifi_change) |
 771                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 772
 773         return flags;
 774 }
 775
 776 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 777                                  const struct rtnl_link_stats64 *b)
 778 {
 779         a->rx_packets = b->rx_packets;
 780         a->tx_packets = b->tx_packets;
 781         a->rx_bytes = b->rx_bytes;
 782         a->tx_bytes = b->tx_bytes;
 783         a->rx_errors = b->rx_errors;
 784         a->tx_errors = b->tx_errors;
 785         a->rx_dropped = b->rx_dropped;
 786         a->tx_dropped = b->tx_dropped;
 787
 788         a->multicast = b->multicast;
 789         a->collisions = b->collisions;
 790
 791         a->rx_length_errors = b->rx_length_errors;
 792         a->rx_over_errors = b->rx_over_errors;
 793         a->rx_crc_errors = b->rx_crc_errors;
 794         a->rx_frame_errors = b->rx_frame_errors;
 795         a->rx_fifo_errors = b->rx_fifo_errors;
 796         a->rx_missed_errors = b->rx_missed_errors;
 797
 798         a->tx_aborted_errors = b->tx_aborted_errors;
 799         a->tx_carrier_errors = b->tx_carrier_errors;
 800         a->tx_fifo_errors = b->tx_fifo_errors;
 801         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 802         a->tx_window_errors = b->tx_window_errors;
 803
 804         a->rx_compressed = b->rx_compressed;
 805         a->tx_compressed = b->tx_compressed;
 806 }
 807
 808 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 809 {
 810         memcpy(v, b, sizeof(*b));
 811 }
 812
 813 /* All VF info */
 814 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 815                                    u32 ext_filter_mask)
 816 {
 817         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 818             (ext_filter_mask & RTEXT_FILTER_VF)) {
 819                 int num_vfs = dev_num_vf(dev->dev.parent);
 820                 size_t size = nla_total_size(sizeof(struct nlattr));
 821                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 822                 size += num_vfs *
 823                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 824                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 825                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 826                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 827                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 828                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 829                          /* IFLA_VF_STATS_RX_PACKETS */
 830                          nla_total_size(sizeof(__u64)) +
 831                          /* IFLA_VF_STATS_TX_PACKETS */
 832                          nla_total_size(sizeof(__u64)) +
 833                          /* IFLA_VF_STATS_RX_BYTES */
 834                          nla_total_size(sizeof(__u64)) +
 835                          /* IFLA_VF_STATS_TX_BYTES */
 836                          nla_total_size(sizeof(__u64)) +
 837                          /* IFLA_VF_STATS_BROADCAST */
 838                          nla_total_size(sizeof(__u64)) +
 839                          /* IFLA_VF_STATS_MULTICAST */
 840                          nla_total_size(sizeof(__u64)));
 841                 return size;
 842         } else
 843                 return 0;
 844 }
 845
 846 static size_t rtnl_port_size(const struct net_device *dev,
 847                              u32 ext_filter_mask)
 848 {
 849         size_t port_size = nla_total_size(4)            /* PORT_VF */
 850                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 851                 + nla_total_size(sizeof(struct ifla_port_vsi))
 852                                                         /* PORT_VSI_TYPE */
 853                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 854                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 855                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 856                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 857         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 858         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 859                 + port_size;
 860         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 861                 + port_size;
 862
 863         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 864             !(ext_filter_mask & RTEXT_FILTER_VF))
 865                 return 0;
 866         if (dev_num_vf(dev->dev.parent))
 867                 return port_self_size + vf_ports_size +
 868                         vf_port_size * dev_num_vf(dev->dev.parent);
 869         else
 870                 return port_self_size;
 871 }
 872
 873 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 874                                      u32 ext_filter_mask)
 875 {
 876         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 877                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 878                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 879                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 880                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 881                + nla_total_size(sizeof(struct rtnl_link_stats))
 882                + nla_total_size(sizeof(struct rtnl_link_stats64))
 883                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 884                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 885                + nla_total_size(4) /* IFLA_TXQLEN */
 886                + nla_total_size(4) /* IFLA_WEIGHT */
 887                + nla_total_size(4) /* IFLA_MTU */
 888                + nla_total_size(4) /* IFLA_LINK */
 889                + nla_total_size(4) /* IFLA_MASTER */
 890                + nla_total_size(1) /* IFLA_CARRIER */
 891                + nla_total_size(4) /* IFLA_PROMISCUITY */
 892                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 893                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 894                + nla_total_size(1) /* IFLA_OPERSTATE */
 895                + nla_total_size(1) /* IFLA_LINKMODE */
 896                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 897                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 898                + nla_total_size(ext_filter_mask
 899                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 900                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 901                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 902                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 903                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 904                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 905                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 906                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 907
 908 }
 909
 910 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 911 {
 912         struct nlattr *vf_ports;
 913         struct nlattr *vf_port;
 914         int vf;
 915         int err;
 916
 917         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 918         if (!vf_ports)
 919                 return -EMSGSIZE;
 920
 921         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 922                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 923                 if (!vf_port)
 924                         goto nla_put_failure;
 925                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 926                         goto nla_put_failure;
 927                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 928                 if (err == -EMSGSIZE)
 929                         goto nla_put_failure;
 930                 if (err) {
 931                         nla_nest_cancel(skb, vf_port);
 932                         continue;
 933                 }
 934                 nla_nest_end(skb, vf_port);
 935         }
 936
 937         nla_nest_end(skb, vf_ports);
 938
 939         return 0;
 940
 941 nla_put_failure:
 942         nla_nest_cancel(skb, vf_ports);
 943         return -EMSGSIZE;
 944 }
 945
 946 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 947 {
 948         struct nlattr *port_self;
 949         int err;
 950
 951         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 952         if (!port_self)
 953                 return -EMSGSIZE;
 954
 955         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 956         if (err) {
 957                 nla_nest_cancel(skb, port_self);
 958                 return (err == -EMSGSIZE) ? err : 0;
 959         }
 960
 961         nla_nest_end(skb, port_self);
 962
 963         return 0;
 964 }
 965
 966 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 967                           u32 ext_filter_mask)
 968 {
 969         int err;
 970
 971         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 972             !(ext_filter_mask & RTEXT_FILTER_VF))
 973                 return 0;
 974
 975         err = rtnl_port_self_fill(skb, dev);
 976         if (err)
 977                 return err;
 978
 979         if (dev_num_vf(dev->dev.parent)) {
 980                 err = rtnl_vf_ports_fill(skb, dev);
 981                 if (err)
 982                         return err;
 983         }
 984
 985         return 0;
 986 }
 987
 988 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 989 {
 990         int err;
 991         struct netdev_phys_item_id ppid;
 992
 993         err = dev_get_phys_port_id(dev, &ppid);
 994         if (err) {
 995                 if (err == -EOPNOTSUPP)
 996                         return 0;
 997                 return err;
 998         }
 999
1000         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1001                 return -EMSGSIZE;
1002
1003         return 0;
1004 }
1005
1006 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1007 {
1008         char name[IFNAMSIZ];
1009         int err;
1010
1011         err = dev_get_phys_port_name(dev, name, sizeof(name));
1012         if (err) {
1013                 if (err == -EOPNOTSUPP)
1014                         return 0;
1015                 return err;
1016         }
1017
1018         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1019                 return -EMSGSIZE;
1020
1021         return 0;
1022 }
1023
1024 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1025 {
1026         int err;
1027         struct switchdev_attr attr = {
1028                 .id = SWITCHDEV_ATTR_PORT_PARENT_ID,
1029                 .flags = SWITCHDEV_F_NO_RECURSE,
1030         };
1031
1032         err = switchdev_port_attr_get(dev, &attr);
1033         if (err) {
1034                 if (err == -EOPNOTSUPP)
1035                         return 0;
1036                 return err;
1037         }
1038
1039         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1040                     attr.u.ppid.id))
1041                 return -EMSGSIZE;
1042
1043         return 0;
1044 }
1045
1046 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1047                             int type, u32 pid, u32 seq, u32 change,
1048                             unsigned int flags, u32 ext_filter_mask)
1049 {
1050         struct ifinfomsg *ifm;
1051         struct nlmsghdr *nlh;
1052         struct rtnl_link_stats64 temp;
1053         const struct rtnl_link_stats64 *stats;
1054         struct nlattr *attr, *af_spec;
1055         struct rtnl_af_ops *af_ops;
1056         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1057
1058         ASSERT_RTNL();
1059         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1060         if (nlh == NULL)
1061                 return -EMSGSIZE;
1062
1063         ifm = nlmsg_data(nlh);
1064         ifm->ifi_family = AF_UNSPEC;
1065         ifm->__ifi_pad = 0;
1066         ifm->ifi_type = dev->type;
1067         ifm->ifi_index = dev->ifindex;
1068         ifm->ifi_flags = dev_get_flags(dev);
1069         ifm->ifi_change = change;
1070
1071         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1072             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1073             nla_put_u8(skb, IFLA_OPERSTATE,
1074                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1075             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1076             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1077             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1078             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1079             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1080 #ifdef CONFIG_RPS
1081             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1082 #endif
1083             (dev->ifindex != dev_get_iflink(dev) &&
1084              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1085             (upper_dev &&
1086              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1087             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1088             (dev->qdisc &&
1089              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1090             (dev->ifalias &&
1091              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1092             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1093                         atomic_read(&dev->carrier_changes)) ||
1094             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1095                 goto nla_put_failure;
1096
1097         if (1) {
1098                 struct rtnl_link_ifmap map = {
1099                         .mem_start   = dev->mem_start,
1100                         .mem_end     = dev->mem_end,
1101                         .base_addr   = dev->base_addr,
1102                         .irq         = dev->irq,
1103                         .dma         = dev->dma,
1104                         .port        = dev->if_port,
1105                 };
1106                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1107                         goto nla_put_failure;
1108         }
1109
1110         if (dev->addr_len) {
1111                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1112                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1113                         goto nla_put_failure;
1114         }
1115
1116         if (rtnl_phys_port_id_fill(skb, dev))
1117                 goto nla_put_failure;
1118
1119         if (rtnl_phys_port_name_fill(skb, dev))
1120                 goto nla_put_failure;
1121
1122         if (rtnl_phys_switch_id_fill(skb, dev))
1123                 goto nla_put_failure;
1124
1125         attr = nla_reserve(skb, IFLA_STATS,
1126                         sizeof(struct rtnl_link_stats));
1127         if (attr == NULL)
1128                 goto nla_put_failure;
1129
1130         stats = dev_get_stats(dev, &temp);
1131         copy_rtnl_link_stats(nla_data(attr), stats);
1132
1133         attr = nla_reserve(skb, IFLA_STATS64,
1134                         sizeof(struct rtnl_link_stats64));
1135         if (attr == NULL)
1136                 goto nla_put_failure;
1137         copy_rtnl_link_stats64(nla_data(attr), stats);
1138
1139         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1140             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1141                 goto nla_put_failure;
1142
1143         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1144             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1145                 int i;
1146
1147                 struct nlattr *vfinfo, *vf, *vfstats;
1148                 int num_vfs = dev_num_vf(dev->dev.parent);
1149
1150                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1151                 if (!vfinfo)
1152                         goto nla_put_failure;
1153                 for (i = 0; i < num_vfs; i++) {
1154                         struct ifla_vf_info ivi;
1155                         struct ifla_vf_mac vf_mac;
1156                         struct ifla_vf_vlan vf_vlan;
1157                         struct ifla_vf_rate vf_rate;
1158                         struct ifla_vf_tx_rate vf_tx_rate;
1159                         struct ifla_vf_spoofchk vf_spoofchk;
1160                         struct ifla_vf_link_state vf_linkstate;
1161                         struct ifla_vf_rss_query_en vf_rss_query_en;
1162                         struct ifla_vf_stats vf_stats;
1163
1164                         /*
1165                          * Not all SR-IOV capable drivers support the
1166                          * spoofcheck and "RSS query enable" query.  Preset to
1167                          * -1 so the user space tool can detect that the driver
1168                          * didn't report anything.
1169                          */
1170                         ivi.spoofchk = -1;
1171                         ivi.rss_query_en = -1;
1172                         memset(ivi.mac, 0, sizeof(ivi.mac));
1173                         /* The default value for VF link state is "auto"
1174                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1175                          */
1176                         ivi.linkstate = 0;
1177                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1178                                 break;
1179                         vf_mac.vf =
1180                                 vf_vlan.vf =
1181                                 vf_rate.vf =
1182                                 vf_tx_rate.vf =
1183                                 vf_spoofchk.vf =
1184                                 vf_linkstate.vf =
1185                                 vf_rss_query_en.vf = ivi.vf;
1186
1187                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1188                         vf_vlan.vlan = ivi.vlan;
1189                         vf_vlan.qos = ivi.qos;
1190                         vf_tx_rate.rate = ivi.max_tx_rate;
1191                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1192                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1193                         vf_spoofchk.setting = ivi.spoofchk;
1194                         vf_linkstate.link_state = ivi.linkstate;
1195                         vf_rss_query_en.setting = ivi.rss_query_en;
1196                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1197                         if (!vf) {
1198                                 nla_nest_cancel(skb, vfinfo);
1199                                 goto nla_put_failure;
1200                         }
1201                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1202                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1203                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1204                                     &vf_rate) ||
1205                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1206                                     &vf_tx_rate) ||
1207                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1208                                     &vf_spoofchk) ||
1209                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1210                                     &vf_linkstate) ||
1211                             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1212                                     sizeof(vf_rss_query_en),
1213                                     &vf_rss_query_en))
1214                                 goto nla_put_failure;
1215                         memset(&vf_stats, 0, sizeof(vf_stats));
1216                         if (dev->netdev_ops->ndo_get_vf_stats)
1217                                 dev->netdev_ops->ndo_get_vf_stats(dev, i,
1218                                                                   &vf_stats);
1219                         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1220                         if (!vfstats) {
1221                                 nla_nest_cancel(skb, vf);
1222                                 nla_nest_cancel(skb, vfinfo);
1223                                 goto nla_put_failure;
1224                         }
1225                         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1226                                         vf_stats.rx_packets) ||
1227                             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1228                                         vf_stats.tx_packets) ||
1229                             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1230                                         vf_stats.rx_bytes) ||
1231                             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1232                                         vf_stats.tx_bytes) ||
1233                             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1234                                         vf_stats.broadcast) ||
1235                             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1236                                         vf_stats.multicast))
1237                                 goto nla_put_failure;
1238                         nla_nest_end(skb, vfstats);
1239                         nla_nest_end(skb, vf);
1240                 }
1241                 nla_nest_end(skb, vfinfo);
1242         }
1243
1244         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1245                 goto nla_put_failure;
1246
1247         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1248                 if (rtnl_link_fill(skb, dev) < 0)
1249                         goto nla_put_failure;
1250         }
1251
1252         if (dev->rtnl_link_ops &&
1253             dev->rtnl_link_ops->get_link_net) {
1254                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1255
1256                 if (!net_eq(dev_net(dev), link_net)) {
1257                         int id = peernet2id_alloc(dev_net(dev), link_net);
1258
1259                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1260                                 goto nla_put_failure;
1261                 }
1262         }
1263
1264         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1265                 goto nla_put_failure;
1266
1267         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1268                 if (af_ops->fill_link_af) {
1269                         struct nlattr *af;
1270                         int err;
1271
1272                         if (!(af = nla_nest_start(skb, af_ops->family)))
1273                                 goto nla_put_failure;
1274
1275                         err = af_ops->fill_link_af(skb, dev);
1276
1277                         /*
1278                          * Caller may return ENODATA to indicate that there
1279                          * was no data to be dumped. This is not an error, it
1280                          * means we should trim the attribute header and
1281                          * continue.
1282                          */
1283                         if (err == -ENODATA)
1284                                 nla_nest_cancel(skb, af);
1285                         else if (err < 0)
1286                                 goto nla_put_failure;
1287
1288                         nla_nest_end(skb, af);
1289                 }
1290         }
1291
1292         nla_nest_end(skb, af_spec);
1293
1294         nlmsg_end(skb, nlh);
1295         return 0;
1296
1297 nla_put_failure:
1298         nlmsg_cancel(skb, nlh);
1299         return -EMSGSIZE;
1300 }
1301
1302 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1303         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1304         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1305         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1306         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1307         [IFLA_MTU]              = { .type = NLA_U32 },
1308         [IFLA_LINK]             = { .type = NLA_U32 },
1309         [IFLA_MASTER]           = { .type = NLA_U32 },
1310         [IFLA_CARRIER]          = { .type = NLA_U8 },
1311         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1312         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1313         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1314         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1315         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1316         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1317         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1318         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1319         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1320         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1321         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1322         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1323         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1324         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1325         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1326         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1327         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1328         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1329         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1330         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1331         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1332 };
1333
1334 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1335         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1336         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1337         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1338         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1339 };
1340
1341 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1342         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1343         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1344         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1345         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1346         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1347         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1348         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1349         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1350 };
1351
1352 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1353         [IFLA_VF_STATS_RX_PACKETS]      = { .type = NLA_U64 },
1354         [IFLA_VF_STATS_TX_PACKETS]      = { .type = NLA_U64 },
1355         [IFLA_VF_STATS_RX_BYTES]        = { .type = NLA_U64 },
1356         [IFLA_VF_STATS_TX_BYTES]        = { .type = NLA_U64 },
1357         [IFLA_VF_STATS_BROADCAST]       = { .type = NLA_U64 },
1358         [IFLA_VF_STATS_MULTICAST]       = { .type = NLA_U64 },
1359 };
1360
1361 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1362         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1363         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1364                                     .len = PORT_PROFILE_MAX },
1365         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1366                                     .len = sizeof(struct ifla_port_vsi)},
1367         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1368                                       .len = PORT_UUID_MAX },
1369         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1370                                     .len = PORT_UUID_MAX },
1371         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1372         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1373 };
1374
1375 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1376 {
1377         struct net *net = sock_net(skb->sk);
1378         int h, s_h;
1379         int idx = 0, s_idx;
1380         struct net_device *dev;
1381         struct hlist_head *head;
1382         struct nlattr *tb[IFLA_MAX+1];
1383         u32 ext_filter_mask = 0;
1384         int err;
1385         int hdrlen;
1386
1387         s_h = cb->args[0];
1388         s_idx = cb->args[1];
1389
1390         cb->seq = net->dev_base_seq;
1391
1392         /* A hack to preserve kernel<->userspace interface.
1393          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1394          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1395          * what iproute2 < v3.9.0 used.
1396          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1397          * attribute, its netlink message is shorter than struct ifinfomsg.
1398          */
1399         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1400                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1401
1402         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1403
1404                 if (tb[IFLA_EXT_MASK])
1405                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1406         }
1407
1408         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1409                 idx = 0;
1410                 head = &net->dev_index_head[h];
1411                 hlist_for_each_entry(dev, head, index_hlist) {
1412                         if (idx < s_idx)
1413                                 goto cont;
1414                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1415                                                NETLINK_CB(cb->skb).portid,
1416                                                cb->nlh->nlmsg_seq, 0,
1417                                                NLM_F_MULTI,
1418                                                ext_filter_mask);
1419                         /* If we ran out of room on the first message,
1420                          * we're in trouble
1421                          */
1422                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1423
1424                         if (err < 0)
1425                                 goto out;
1426
1427                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1428 cont:
1429                         idx++;
1430                 }
1431         }
1432 out:
1433         cb->args[1] = idx;
1434         cb->args[0] = h;
1435
1436         return skb->len;
1437 }
1438
1439 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1440 {
1441         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1442 }
1443 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1444
1445 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1446 {
1447         struct net *net;
1448         /* Examine the link attributes and figure out which
1449          * network namespace we are talking about.
1450          */
1451         if (tb[IFLA_NET_NS_PID])
1452                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1453         else if (tb[IFLA_NET_NS_FD])
1454                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1455         else
1456                 net = get_net(src_net);
1457         return net;
1458 }
1459 EXPORT_SYMBOL(rtnl_link_get_net);
1460
1461 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1462 {
1463         if (dev) {
1464                 if (tb[IFLA_ADDRESS] &&
1465                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1466                         return -EINVAL;
1467
1468                 if (tb[IFLA_BROADCAST] &&
1469                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1470                         return -EINVAL;
1471         }
1472
1473         if (tb[IFLA_AF_SPEC]) {
1474                 struct nlattr *af;
1475                 int rem, err;
1476
1477                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1478                         const struct rtnl_af_ops *af_ops;
1479
1480                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1481                                 return -EAFNOSUPPORT;
1482
1483                         if (!af_ops->set_link_af)
1484                                 return -EOPNOTSUPP;
1485
1486                         if (af_ops->validate_link_af) {
1487                                 err = af_ops->validate_link_af(dev, af);
1488                                 if (err < 0)
1489                                         return err;
1490                         }
1491                 }
1492         }
1493
1494         return 0;
1495 }
1496
1497 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1498 {
1499         const struct net_device_ops *ops = dev->netdev_ops;
1500         int err = -EINVAL;
1501
1502         if (tb[IFLA_VF_MAC]) {
1503                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1504
1505                 err = -EOPNOTSUPP;
1506                 if (ops->ndo_set_vf_mac)
1507                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1508                                                   ivm->mac);
1509                 if (err < 0)
1510                         return err;
1511         }
1512
1513         if (tb[IFLA_VF_VLAN]) {
1514                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1515
1516                 err = -EOPNOTSUPP;
1517                 if (ops->ndo_set_vf_vlan)
1518                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1519                                                    ivv->qos);
1520                 if (err < 0)
1521                         return err;
1522         }
1523
1524         if (tb[IFLA_VF_TX_RATE]) {
1525                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1526                 struct ifla_vf_info ivf;
1527
1528                 err = -EOPNOTSUPP;
1529                 if (ops->ndo_get_vf_config)
1530                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1531                 if (err < 0)
1532                         return err;
1533
1534                 err = -EOPNOTSUPP;
1535                 if (ops->ndo_set_vf_rate)
1536                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1537                                                    ivf.min_tx_rate,
1538                                                    ivt->rate);
1539                 if (err < 0)
1540                         return err;
1541         }
1542
1543         if (tb[IFLA_VF_RATE]) {
1544                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1545
1546                 err = -EOPNOTSUPP;
1547                 if (ops->ndo_set_vf_rate)
1548                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1549                                                    ivt->min_tx_rate,
1550                                                    ivt->max_tx_rate);
1551                 if (err < 0)
1552                         return err;
1553         }
1554
1555         if (tb[IFLA_VF_SPOOFCHK]) {
1556                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1557
1558                 err = -EOPNOTSUPP;
1559                 if (ops->ndo_set_vf_spoofchk)
1560                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1561                                                        ivs->setting);
1562                 if (err < 0)
1563                         return err;
1564         }
1565
1566         if (tb[IFLA_VF_LINK_STATE]) {
1567                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1568
1569                 err = -EOPNOTSUPP;
1570                 if (ops->ndo_set_vf_link_state)
1571                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1572                                                          ivl->link_state);
1573                 if (err < 0)
1574                         return err;
1575         }
1576
1577         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1578                 struct ifla_vf_rss_query_en *ivrssq_en;
1579
1580                 err = -EOPNOTSUPP;
1581                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1582                 if (ops->ndo_set_vf_rss_query_en)
1583                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1584                                                            ivrssq_en->setting);
1585                 if (err < 0)
1586                         return err;
1587         }
1588
1589         return err;
1590 }
1591
1592 static int do_set_master(struct net_device *dev, int ifindex)
1593 {
1594         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1595         const struct net_device_ops *ops;
1596         int err;
1597
1598         if (upper_dev) {
1599                 if (upper_dev->ifindex == ifindex)
1600                         return 0;
1601                 ops = upper_dev->netdev_ops;
1602                 if (ops->ndo_del_slave) {
1603                         err = ops->ndo_del_slave(upper_dev, dev);
1604                         if (err)
1605                                 return err;
1606                 } else {
1607                         return -EOPNOTSUPP;
1608                 }
1609         }
1610
1611         if (ifindex) {
1612                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1613                 if (!upper_dev)
1614                         return -EINVAL;
1615                 ops = upper_dev->netdev_ops;
1616                 if (ops->ndo_add_slave) {
1617                         err = ops->ndo_add_slave(upper_dev, dev);
1618                         if (err)
1619                                 return err;
1620                 } else {
1621                         return -EOPNOTSUPP;
1622                 }
1623         }
1624         return 0;
1625 }
1626
1627 #define DO_SETLINK_MODIFIED     0x01
1628 /* notify flag means notify + modified. */
1629 #define DO_SETLINK_NOTIFY       0x03
1630 static int do_setlink(const struct sk_buff *skb,
1631                       struct net_device *dev, struct ifinfomsg *ifm,
1632                       struct nlattr **tb, char *ifname, int status)
1633 {
1634         const struct net_device_ops *ops = dev->netdev_ops;
1635         int err;
1636
1637         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1638                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1639                 if (IS_ERR(net)) {
1640                         err = PTR_ERR(net);
1641                         goto errout;
1642                 }
1643                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1644                         put_net(net);
1645                         err = -EPERM;
1646                         goto errout;
1647                 }
1648                 err = dev_change_net_namespace(dev, net, ifname);
1649                 put_net(net);
1650                 if (err)
1651                         goto errout;
1652                 status |= DO_SETLINK_MODIFIED;
1653         }
1654
1655         if (tb[IFLA_MAP]) {
1656                 struct rtnl_link_ifmap *u_map;
1657                 struct ifmap k_map;
1658
1659                 if (!ops->ndo_set_config) {
1660                         err = -EOPNOTSUPP;
1661                         goto errout;
1662                 }
1663
1664                 if (!netif_device_present(dev)) {
1665                         err = -ENODEV;
1666                         goto errout;
1667                 }
1668
1669                 u_map = nla_data(tb[IFLA_MAP]);
1670                 k_map.mem_start = (unsigned long) u_map->mem_start;
1671                 k_map.mem_end = (unsigned long) u_map->mem_end;
1672                 k_map.base_addr = (unsigned short) u_map->base_addr;
1673                 k_map.irq = (unsigned char) u_map->irq;
1674                 k_map.dma = (unsigned char) u_map->dma;
1675                 k_map.port = (unsigned char) u_map->port;
1676
1677                 err = ops->ndo_set_config(dev, &k_map);
1678                 if (err < 0)
1679                         goto errout;
1680
1681                 status |= DO_SETLINK_NOTIFY;
1682         }
1683
1684         if (tb[IFLA_ADDRESS]) {
1685                 struct sockaddr *sa;
1686                 int len;
1687
1688                 len = sizeof(sa_family_t) + dev->addr_len;
1689                 sa = kmalloc(len, GFP_KERNEL);
1690                 if (!sa) {
1691                         err = -ENOMEM;
1692                         goto errout;
1693                 }
1694                 sa->sa_family = dev->type;
1695                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1696                        dev->addr_len);
1697                 err = dev_set_mac_address(dev, sa);
1698                 kfree(sa);
1699                 if (err)
1700                         goto errout;
1701                 status |= DO_SETLINK_MODIFIED;
1702         }
1703
1704         if (tb[IFLA_MTU]) {
1705                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1706                 if (err < 0)
1707                         goto errout;
1708                 status |= DO_SETLINK_MODIFIED;
1709         }
1710
1711         if (tb[IFLA_GROUP]) {
1712                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1713                 status |= DO_SETLINK_NOTIFY;
1714         }
1715
1716         /*
1717          * Interface selected by interface index but interface
1718          * name provided implies that a name change has been
1719          * requested.
1720          */
1721         if (ifm->ifi_index > 0 && ifname[0]) {
1722                 err = dev_change_name(dev, ifname);
1723                 if (err < 0)
1724                         goto errout;
1725                 status |= DO_SETLINK_MODIFIED;
1726         }
1727
1728         if (tb[IFLA_IFALIAS]) {
1729                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1730                                     nla_len(tb[IFLA_IFALIAS]));
1731                 if (err < 0)
1732                         goto errout;
1733                 status |= DO_SETLINK_NOTIFY;
1734         }
1735
1736         if (tb[IFLA_BROADCAST]) {
1737                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1738                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1739         }
1740
1741         if (ifm->ifi_flags || ifm->ifi_change) {
1742                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1743                 if (err < 0)
1744                         goto errout;
1745         }
1746
1747         if (tb[IFLA_MASTER]) {
1748                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1749                 if (err)
1750                         goto errout;
1751                 status |= DO_SETLINK_MODIFIED;
1752         }
1753
1754         if (tb[IFLA_CARRIER]) {
1755                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1756                 if (err)
1757                         goto errout;
1758                 status |= DO_SETLINK_MODIFIED;
1759         }
1760
1761         if (tb[IFLA_TXQLEN]) {
1762                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1763
1764                 if (dev->tx_queue_len ^ value)
1765                         status |= DO_SETLINK_NOTIFY;
1766
1767                 dev->tx_queue_len = value;
1768         }
1769
1770         if (tb[IFLA_OPERSTATE])
1771                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1772
1773         if (tb[IFLA_LINKMODE]) {
1774                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1775
1776                 write_lock_bh(&dev_base_lock);
1777                 if (dev->link_mode ^ value)
1778                         status |= DO_SETLINK_NOTIFY;
1779                 dev->link_mode = value;
1780                 write_unlock_bh(&dev_base_lock);
1781         }
1782
1783         if (tb[IFLA_VFINFO_LIST]) {
1784                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1785                 struct nlattr *attr;
1786                 int rem;
1787
1788                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1789                         if (nla_type(attr) != IFLA_VF_INFO ||
1790                             nla_len(attr) < NLA_HDRLEN) {
1791                                 err = -EINVAL;
1792                                 goto errout;
1793                         }
1794                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1795                                                ifla_vf_policy);
1796                         if (err < 0)
1797                                 goto errout;
1798                         err = do_setvfinfo(dev, vfinfo);
1799                         if (err < 0)
1800                                 goto errout;
1801                         status |= DO_SETLINK_NOTIFY;
1802                 }
1803         }
1804         err = 0;
1805
1806         if (tb[IFLA_VF_PORTS]) {
1807                 struct nlattr *port[IFLA_PORT_MAX+1];
1808                 struct nlattr *attr;
1809                 int vf;
1810                 int rem;
1811
1812                 err = -EOPNOTSUPP;
1813                 if (!ops->ndo_set_vf_port)
1814                         goto errout;
1815
1816                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1817                         if (nla_type(attr) != IFLA_VF_PORT ||
1818                             nla_len(attr) < NLA_HDRLEN) {
1819                                 err = -EINVAL;
1820                                 goto errout;
1821                         }
1822                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1823                                                ifla_port_policy);
1824                         if (err < 0)
1825                                 goto errout;
1826                         if (!port[IFLA_PORT_VF]) {
1827                                 err = -EOPNOTSUPP;
1828                                 goto errout;
1829                         }
1830                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1831                         err = ops->ndo_set_vf_port(dev, vf, port);
1832                         if (err < 0)
1833                                 goto errout;
1834                         status |= DO_SETLINK_NOTIFY;
1835                 }
1836         }
1837         err = 0;
1838
1839         if (tb[IFLA_PORT_SELF]) {
1840                 struct nlattr *port[IFLA_PORT_MAX+1];
1841
1842                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1843                         tb[IFLA_PORT_SELF], ifla_port_policy);
1844                 if (err < 0)
1845                         goto errout;
1846
1847                 err = -EOPNOTSUPP;
1848                 if (ops->ndo_set_vf_port)
1849                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1850                 if (err < 0)
1851                         goto errout;
1852                 status |= DO_SETLINK_NOTIFY;
1853         }
1854
1855         if (tb[IFLA_AF_SPEC]) {
1856                 struct nlattr *af;
1857                 int rem;
1858
1859                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1860                         const struct rtnl_af_ops *af_ops;
1861
1862                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1863                                 BUG();
1864
1865                         err = af_ops->set_link_af(dev, af);
1866                         if (err < 0)
1867                                 goto errout;
1868
1869                         status |= DO_SETLINK_NOTIFY;
1870                 }
1871         }
1872         err = 0;
1873
1874         if (tb[IFLA_PROTO_DOWN]) {
1875                 err = dev_change_proto_down(dev,
1876                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
1877                 if (err)
1878                         goto errout;
1879                 status |= DO_SETLINK_NOTIFY;
1880         }
1881
1882 errout:
1883         if (status & DO_SETLINK_MODIFIED) {
1884                 if (status & DO_SETLINK_NOTIFY)
1885                         netdev_state_change(dev);
1886
1887                 if (err < 0)
1888                         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",
1889                                              dev->name);
1890         }
1891
1892         return err;
1893 }
1894
1895 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1896 {
1897         struct net *net = sock_net(skb->sk);
1898         struct ifinfomsg *ifm;
1899         struct net_device *dev;
1900         int err;
1901         struct nlattr *tb[IFLA_MAX+1];
1902         char ifname[IFNAMSIZ];
1903
1904         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1905         if (err < 0)
1906                 goto errout;
1907
1908         if (tb[IFLA_IFNAME])
1909                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1910         else
1911                 ifname[0] = '\0';
1912
1913         err = -EINVAL;
1914         ifm = nlmsg_data(nlh);
1915         if (ifm->ifi_index > 0)
1916                 dev = __dev_get_by_index(net, ifm->ifi_index);
1917         else if (tb[IFLA_IFNAME])
1918                 dev = __dev_get_by_name(net, ifname);
1919         else
1920                 goto errout;
1921
1922         if (dev == NULL) {
1923                 err = -ENODEV;
1924                 goto errout;
1925         }
1926
1927         err = validate_linkmsg(dev, tb);
1928         if (err < 0)
1929                 goto errout;
1930
1931         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1932 errout:
1933         return err;
1934 }
1935
1936 static int rtnl_group_dellink(const struct net *net, int group)
1937 {
1938         struct net_device *dev, *aux;
1939         LIST_HEAD(list_kill);
1940         bool found = false;
1941
1942         if (!group)
1943                 return -EPERM;
1944
1945         for_each_netdev(net, dev) {
1946                 if (dev->group == group) {
1947                         const struct rtnl_link_ops *ops;
1948
1949                         found = true;
1950                         ops = dev->rtnl_link_ops;
1951                         if (!ops || !ops->dellink)
1952                                 return -EOPNOTSUPP;
1953                 }
1954         }
1955
1956         if (!found)
1957                 return -ENODEV;
1958
1959         for_each_netdev_safe(net, dev, aux) {
1960                 if (dev->group == group) {
1961                         const struct rtnl_link_ops *ops;
1962
1963                         ops = dev->rtnl_link_ops;
1964                         ops->dellink(dev, &list_kill);
1965                 }
1966         }
1967         unregister_netdevice_many(&list_kill);
1968
1969         return 0;
1970 }
1971
1972 int rtnl_delete_link(struct net_device *dev)
1973 {
1974         const struct rtnl_link_ops *ops;
1975         LIST_HEAD(list_kill);
1976
1977         ops = dev->rtnl_link_ops;
1978         if (!ops || !ops->dellink)
1979                 return -EOPNOTSUPP;
1980
1981         ops->dellink(dev, &list_kill);
1982         unregister_netdevice_many(&list_kill);
1983
1984         return 0;
1985 }
1986 EXPORT_SYMBOL_GPL(rtnl_delete_link);
1987
1988 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1989 {
1990         struct net *net = sock_net(skb->sk);
1991         struct net_device *dev;
1992         struct ifinfomsg *ifm;
1993         char ifname[IFNAMSIZ];
1994         struct nlattr *tb[IFLA_MAX+1];
1995         int err;
1996
1997         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1998         if (err < 0)
1999                 return err;
2000
2001         if (tb[IFLA_IFNAME])
2002                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2003
2004         ifm = nlmsg_data(nlh);
2005         if (ifm->ifi_index > 0)
2006                 dev = __dev_get_by_index(net, ifm->ifi_index);
2007         else if (tb[IFLA_IFNAME])
2008                 dev = __dev_get_by_name(net, ifname);
2009         else if (tb[IFLA_GROUP])
2010                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2011         else
2012                 return -EINVAL;
2013
2014         if (!dev)
2015                 return -ENODEV;
2016
2017         return rtnl_delete_link(dev);
2018 }
2019
2020 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2021 {
2022         unsigned int old_flags;
2023         int err;
2024
2025         old_flags = dev->flags;
2026         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2027                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2028                 if (err < 0)
2029                         return err;
2030         }
2031
2032         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2033
2034         __dev_notify_flags(dev, old_flags, ~0U);
2035         return 0;
2036 }
2037 EXPORT_SYMBOL(rtnl_configure_link);
2038
2039 struct net_device *rtnl_create_link(struct net *net,
2040         const char *ifname, unsigned char name_assign_type,
2041         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2042 {
2043         int err;
2044         struct net_device *dev;
2045         unsigned int num_tx_queues = 1;
2046         unsigned int num_rx_queues = 1;
2047
2048         if (tb[IFLA_NUM_TX_QUEUES])
2049                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2050         else if (ops->get_num_tx_queues)
2051                 num_tx_queues = ops->get_num_tx_queues();
2052
2053         if (tb[IFLA_NUM_RX_QUEUES])
2054                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2055         else if (ops->get_num_rx_queues)
2056                 num_rx_queues = ops->get_num_rx_queues();
2057
2058         err = -ENOMEM;
2059         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2060                                ops->setup, num_tx_queues, num_rx_queues);
2061         if (!dev)
2062                 goto err;
2063
2064         dev_net_set(dev, net);
2065         dev->rtnl_link_ops = ops;
2066         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2067
2068         if (tb[IFLA_MTU])
2069                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2070         if (tb[IFLA_ADDRESS]) {
2071                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2072                                 nla_len(tb[IFLA_ADDRESS]));
2073                 dev->addr_assign_type = NET_ADDR_SET;
2074         }
2075         if (tb[IFLA_BROADCAST])
2076                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2077                                 nla_len(tb[IFLA_BROADCAST]));
2078         if (tb[IFLA_TXQLEN])
2079                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2080         if (tb[IFLA_OPERSTATE])
2081                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2082         if (tb[IFLA_LINKMODE])
2083                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2084         if (tb[IFLA_GROUP])
2085                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2086
2087         return dev;
2088
2089 err:
2090         return ERR_PTR(err);
2091 }
2092 EXPORT_SYMBOL(rtnl_create_link);
2093
2094 static int rtnl_group_changelink(const struct sk_buff *skb,
2095                 struct net *net, int group,
2096                 struct ifinfomsg *ifm,
2097                 struct nlattr **tb)
2098 {
2099         struct net_device *dev, *aux;
2100         int err;
2101
2102         for_each_netdev_safe(net, dev, aux) {
2103                 if (dev->group == group) {
2104                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2105                         if (err < 0)
2106                                 return err;
2107                 }
2108         }
2109
2110         return 0;
2111 }
2112
2113 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2114 {
2115         struct net *net = sock_net(skb->sk);
2116         const struct rtnl_link_ops *ops;
2117         const struct rtnl_link_ops *m_ops = NULL;
2118         struct net_device *dev;
2119         struct net_device *master_dev = NULL;
2120         struct ifinfomsg *ifm;
2121         char kind[MODULE_NAME_LEN];
2122         char ifname[IFNAMSIZ];
2123         struct nlattr *tb[IFLA_MAX+1];
2124         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2125         unsigned char name_assign_type = NET_NAME_USER;
2126         int err;
2127
2128 #ifdef CONFIG_MODULES
2129 replay:
2130 #endif
2131         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2132         if (err < 0)
2133                 return err;
2134
2135         if (tb[IFLA_IFNAME])
2136                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2137         else
2138                 ifname[0] = '\0';
2139
2140         ifm = nlmsg_data(nlh);
2141         if (ifm->ifi_index > 0)
2142                 dev = __dev_get_by_index(net, ifm->ifi_index);
2143         else {
2144                 if (ifname[0])
2145                         dev = __dev_get_by_name(net, ifname);
2146                 else
2147                         dev = NULL;
2148         }
2149
2150         if (dev) {
2151                 master_dev = netdev_master_upper_dev_get(dev);
2152                 if (master_dev)
2153                         m_ops = master_dev->rtnl_link_ops;
2154         }
2155
2156         err = validate_linkmsg(dev, tb);
2157         if (err < 0)
2158                 return err;
2159
2160         if (tb[IFLA_LINKINFO]) {
2161                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2162                                        tb[IFLA_LINKINFO], ifla_info_policy);
2163                 if (err < 0)
2164                         return err;
2165         } else
2166                 memset(linkinfo, 0, sizeof(linkinfo));
2167
2168         if (linkinfo[IFLA_INFO_KIND]) {
2169                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2170                 ops = rtnl_link_ops_get(kind);
2171         } else {
2172                 kind[0] = '\0';
2173                 ops = NULL;
2174         }
2175
2176         if (1) {
2177                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2178                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2179                 struct nlattr **data = NULL;
2180                 struct nlattr **slave_data = NULL;
2181                 struct net *dest_net, *link_net = NULL;
2182
2183                 if (ops) {
2184                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2185                                 err = nla_parse_nested(attr, ops->maxtype,
2186                                                        linkinfo[IFLA_INFO_DATA],
2187                                                        ops->policy);
2188                                 if (err < 0)
2189                                         return err;
2190                                 data = attr;
2191                         }
2192                         if (ops->validate) {
2193                                 err = ops->validate(tb, data);
2194                                 if (err < 0)
2195                                         return err;
2196                         }
2197                 }
2198
2199                 if (m_ops) {
2200                         if (m_ops->slave_maxtype &&
2201                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2202                                 err = nla_parse_nested(slave_attr,
2203                                                        m_ops->slave_maxtype,
2204                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2205                                                        m_ops->slave_policy);
2206                                 if (err < 0)
2207                                         return err;
2208                                 slave_data = slave_attr;
2209                         }
2210                         if (m_ops->slave_validate) {
2211                                 err = m_ops->slave_validate(tb, slave_data);
2212                                 if (err < 0)
2213                                         return err;
2214                         }
2215                 }
2216
2217                 if (dev) {
2218                         int status = 0;
2219
2220                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2221                                 return -EEXIST;
2222                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2223                                 return -EOPNOTSUPP;
2224
2225                         if (linkinfo[IFLA_INFO_DATA]) {
2226                                 if (!ops || ops != dev->rtnl_link_ops ||
2227                                     !ops->changelink)
2228                                         return -EOPNOTSUPP;
2229
2230                                 err = ops->changelink(dev, tb, data);
2231                                 if (err < 0)
2232                                         return err;
2233                                 status |= DO_SETLINK_NOTIFY;
2234                         }
2235
2236                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2237                                 if (!m_ops || !m_ops->slave_changelink)
2238                                         return -EOPNOTSUPP;
2239
2240                                 err = m_ops->slave_changelink(master_dev, dev,
2241                                                               tb, slave_data);
2242                                 if (err < 0)
2243                                         return err;
2244                                 status |= DO_SETLINK_NOTIFY;
2245                         }
2246
2247                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2248                 }
2249
2250                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2251                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2252                                 return rtnl_group_changelink(skb, net,
2253                                                 nla_get_u32(tb[IFLA_GROUP]),
2254                                                 ifm, tb);
2255                         return -ENODEV;
2256                 }
2257
2258                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2259                         return -EOPNOTSUPP;
2260
2261                 if (!ops) {
2262 #ifdef CONFIG_MODULES
2263                         if (kind[0]) {
2264                                 __rtnl_unlock();
2265                                 request_module("rtnl-link-%s", kind);
2266                                 rtnl_lock();
2267                                 ops = rtnl_link_ops_get(kind);
2268                                 if (ops)
2269                                         goto replay;
2270                         }
2271 #endif
2272                         return -EOPNOTSUPP;
2273                 }
2274
2275                 if (!ops->setup)
2276                         return -EOPNOTSUPP;
2277
2278                 if (!ifname[0]) {
2279                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2280                         name_assign_type = NET_NAME_ENUM;
2281                 }
2282
2283                 dest_net = rtnl_link_get_net(net, tb);
2284                 if (IS_ERR(dest_net))
2285                         return PTR_ERR(dest_net);
2286
2287                 err = -EPERM;
2288                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2289                         goto out;
2290
2291                 if (tb[IFLA_LINK_NETNSID]) {
2292                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2293
2294                         link_net = get_net_ns_by_id(dest_net, id);
2295                         if (!link_net) {
2296                                 err =  -EINVAL;
2297                                 goto out;
2298                         }
2299                         err = -EPERM;
2300                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2301                                 goto out;
2302                 }
2303
2304                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2305                                        name_assign_type, ops, tb);
2306                 if (IS_ERR(dev)) {
2307                         err = PTR_ERR(dev);
2308                         goto out;
2309                 }
2310
2311                 dev->ifindex = ifm->ifi_index;
2312
2313                 if (ops->newlink) {
2314                         err = ops->newlink(link_net ? : net, dev, tb, data);
2315                         /* Drivers should call free_netdev() in ->destructor
2316                          * and unregister it on failure after registration
2317                          * so that device could be finally freed in rtnl_unlock.
2318                          */
2319                         if (err < 0) {
2320                                 /* If device is not registered at all, free it now */
2321                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2322                                         free_netdev(dev);
2323                                 goto out;
2324                         }
2325                 } else {
2326                         err = register_netdevice(dev);
2327                         if (err < 0) {
2328                                 free_netdev(dev);
2329                                 goto out;
2330                         }
2331                 }
2332                 err = rtnl_configure_link(dev, ifm);
2333                 if (err < 0)
2334                         goto out_unregister;
2335                 if (link_net) {
2336                         err = dev_change_net_namespace(dev, dest_net, ifname);
2337                         if (err < 0)
2338                                 goto out_unregister;
2339                 }
2340 out:
2341                 if (link_net)
2342                         put_net(link_net);
2343                 put_net(dest_net);
2344                 return err;
2345 out_unregister:
2346                 if (ops->newlink) {
2347                         LIST_HEAD(list_kill);
2348
2349                         ops->dellink(dev, &list_kill);
2350                         unregister_netdevice_many(&list_kill);
2351                 } else {
2352                         unregister_netdevice(dev);
2353                 }
2354                 goto out;
2355         }
2356 }
2357
2358 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2359 {
2360         struct net *net = sock_net(skb->sk);
2361         struct ifinfomsg *ifm;
2362         char ifname[IFNAMSIZ];
2363         struct nlattr *tb[IFLA_MAX+1];
2364         struct net_device *dev = NULL;
2365         struct sk_buff *nskb;
2366         int err;
2367         u32 ext_filter_mask = 0;
2368
2369         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2370         if (err < 0)
2371                 return err;
2372
2373         if (tb[IFLA_IFNAME])
2374                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2375
2376         if (tb[IFLA_EXT_MASK])
2377                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2378
2379         ifm = nlmsg_data(nlh);
2380         if (ifm->ifi_index > 0)
2381                 dev = __dev_get_by_index(net, ifm->ifi_index);
2382         else if (tb[IFLA_IFNAME])
2383                 dev = __dev_get_by_name(net, ifname);
2384         else
2385                 return -EINVAL;
2386
2387         if (dev == NULL)
2388                 return -ENODEV;
2389
2390         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2391         if (nskb == NULL)
2392                 return -ENOBUFS;
2393
2394         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2395                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2396         if (err < 0) {
2397                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2398                 WARN_ON(err == -EMSGSIZE);
2399                 kfree_skb(nskb);
2400         } else
2401                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2402
2403         return err;
2404 }
2405
2406 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2407 {
2408         struct net *net = sock_net(skb->sk);
2409         struct net_device *dev;
2410         struct nlattr *tb[IFLA_MAX+1];
2411         u32 ext_filter_mask = 0;
2412         u16 min_ifinfo_dump_size = 0;
2413         int hdrlen;
2414
2415         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2416         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2417                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2418
2419         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2420                 if (tb[IFLA_EXT_MASK])
2421                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2422         }
2423
2424         if (!ext_filter_mask)
2425                 return NLMSG_GOODSIZE;
2426         /*
2427          * traverse the list of net devices and compute the minimum
2428          * buffer size based upon the filter mask.
2429          */
2430         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2431                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2432                                              if_nlmsg_size(dev,
2433                                                            ext_filter_mask));
2434         }
2435
2436         return min_ifinfo_dump_size;
2437 }
2438
2439 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2440 {
2441         int idx;
2442         int s_idx = cb->family;
2443
2444         if (s_idx == 0)
2445                 s_idx = 1;
2446         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2447                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2448                 if (idx < s_idx || idx == PF_PACKET)
2449                         continue;
2450                 if (rtnl_msg_handlers[idx] == NULL ||
2451                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2452                         continue;
2453                 if (idx > s_idx) {
2454                         memset(&cb->args[0], 0, sizeof(cb->args));
2455                         cb->prev_seq = 0;
2456                         cb->seq = 0;
2457                 }
2458                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2459                         break;
2460         }
2461         cb->family = idx;
2462
2463         return skb->len;
2464 }
2465
2466 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2467                                        unsigned int change, gfp_t flags)
2468 {
2469         struct net *net = dev_net(dev);
2470         struct sk_buff *skb;
2471         int err = -ENOBUFS;
2472         size_t if_info_size;
2473
2474         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2475         if (skb == NULL)
2476                 goto errout;
2477
2478         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2479         if (err < 0) {
2480                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2481                 WARN_ON(err == -EMSGSIZE);
2482                 kfree_skb(skb);
2483                 goto errout;
2484         }
2485         return skb;
2486 errout:
2487         if (err < 0)
2488                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2489         return NULL;
2490 }
2491
2492 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2493 {
2494         struct net *net = dev_net(dev);
2495
2496         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2497 }
2498
2499 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2500                   gfp_t flags)
2501 {
2502         struct sk_buff *skb;
2503
2504         if (dev->reg_state != NETREG_REGISTERED)
2505                 return;
2506
2507         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2508         if (skb)
2509                 rtmsg_ifinfo_send(skb, dev, flags);
2510 }
2511 EXPORT_SYMBOL(rtmsg_ifinfo);
2512
2513 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2514                                    struct net_device *dev,
2515                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2516                                    int type, unsigned int flags,
2517                                    int nlflags)
2518 {
2519         struct nlmsghdr *nlh;
2520         struct ndmsg *ndm;
2521
2522         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2523         if (!nlh)
2524                 return -EMSGSIZE;
2525
2526         ndm = nlmsg_data(nlh);
2527         ndm->ndm_family  = AF_BRIDGE;
2528         ndm->ndm_pad1    = 0;
2529         ndm->ndm_pad2    = 0;
2530         ndm->ndm_flags   = flags;
2531         ndm->ndm_type    = 0;
2532         ndm->ndm_ifindex = dev->ifindex;
2533         ndm->ndm_state   = NUD_PERMANENT;
2534
2535         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2536                 goto nla_put_failure;
2537         if (vid)
2538                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2539                         goto nla_put_failure;
2540
2541         nlmsg_end(skb, nlh);
2542         return 0;
2543
2544 nla_put_failure:
2545         nlmsg_cancel(skb, nlh);
2546         return -EMSGSIZE;
2547 }
2548
2549 static inline size_t rtnl_fdb_nlmsg_size(void)
2550 {
2551         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2552 }
2553
2554 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2555 {
2556         struct net *net = dev_net(dev);
2557         struct sk_buff *skb;
2558         int err = -ENOBUFS;
2559
2560         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2561         if (!skb)
2562                 goto errout;
2563
2564         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2565                                       0, 0, type, NTF_SELF, 0);
2566         if (err < 0) {
2567                 kfree_skb(skb);
2568                 goto errout;
2569         }
2570
2571         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2572         return;
2573 errout:
2574         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2575 }
2576
2577 /**
2578  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2579  */
2580 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2581                      struct nlattr *tb[],
2582                      struct net_device *dev,
2583                      const unsigned char *addr, u16 vid,
2584                      u16 flags)
2585 {
2586         int err = -EINVAL;
2587
2588         /* If aging addresses are supported device will need to
2589          * implement its own handler for this.
2590          */
2591         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2592                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2593                 return err;
2594         }
2595
2596         if (vid) {
2597                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2598                 return err;
2599         }
2600
2601         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2602                 err = dev_uc_add_excl(dev, addr);
2603         else if (is_multicast_ether_addr(addr))
2604                 err = dev_mc_add_excl(dev, addr);
2605
2606         /* Only return duplicate errors if NLM_F_EXCL is set */
2607         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2608                 err = 0;
2609
2610         return err;
2611 }
2612 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2613
2614 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2615 {
2616         u16 vid = 0;
2617
2618         if (vlan_attr) {
2619                 if (nla_len(vlan_attr) != sizeof(u16)) {
2620                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2621                         return -EINVAL;
2622                 }
2623
2624                 vid = nla_get_u16(vlan_attr);
2625
2626                 if (!vid || vid >= VLAN_VID_MASK) {
2627                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2628                                 vid);
2629                         return -EINVAL;
2630                 }
2631         }
2632         *p_vid = vid;
2633         return 0;
2634 }
2635
2636 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2637 {
2638         struct net *net = sock_net(skb->sk);
2639         struct ndmsg *ndm;
2640         struct nlattr *tb[NDA_MAX+1];
2641         struct net_device *dev;
2642         u8 *addr;
2643         u16 vid;
2644         int err;
2645
2646         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2647         if (err < 0)
2648                 return err;
2649
2650         ndm = nlmsg_data(nlh);
2651         if (ndm->ndm_ifindex == 0) {
2652                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2653                 return -EINVAL;
2654         }
2655
2656         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2657         if (dev == NULL) {
2658                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2659                 return -ENODEV;
2660         }
2661
2662         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2663                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2664                 return -EINVAL;
2665         }
2666
2667         addr = nla_data(tb[NDA_LLADDR]);
2668
2669         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2670         if (err)
2671                 return err;
2672
2673         err = -EOPNOTSUPP;
2674
2675         /* Support fdb on master device the net/bridge default case */
2676         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2677             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2678                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2679                 const struct net_device_ops *ops = br_dev->netdev_ops;
2680
2681                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2682                                        nlh->nlmsg_flags);
2683                 if (err)
2684                         goto out;
2685                 else
2686                         ndm->ndm_flags &= ~NTF_MASTER;
2687         }
2688
2689         /* Embedded bridge, macvlan, and any other device support */
2690         if ((ndm->ndm_flags & NTF_SELF)) {
2691                 if (dev->netdev_ops->ndo_fdb_add)
2692                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2693                                                            vid,
2694                                                            nlh->nlmsg_flags);
2695                 else
2696                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2697                                                nlh->nlmsg_flags);
2698
2699                 if (!err) {
2700                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2701                         ndm->ndm_flags &= ~NTF_SELF;
2702                 }
2703         }
2704 out:
2705         return err;
2706 }
2707
2708 /**
2709  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2710  */
2711 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2712                      struct nlattr *tb[],
2713                      struct net_device *dev,
2714                      const unsigned char *addr, u16 vid)
2715 {
2716         int err = -EINVAL;
2717
2718         /* If aging addresses are supported device will need to
2719          * implement its own handler for this.
2720          */
2721         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2722                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2723                 return err;
2724         }
2725
2726         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2727                 err = dev_uc_del(dev, addr);
2728         else if (is_multicast_ether_addr(addr))
2729                 err = dev_mc_del(dev, addr);
2730
2731         return err;
2732 }
2733 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2734
2735 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2736 {
2737         struct net *net = sock_net(skb->sk);
2738         struct ndmsg *ndm;
2739         struct nlattr *tb[NDA_MAX+1];
2740         struct net_device *dev;
2741         int err = -EINVAL;
2742         __u8 *addr;
2743         u16 vid;
2744
2745         if (!netlink_capable(skb, CAP_NET_ADMIN))
2746                 return -EPERM;
2747
2748         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2749         if (err < 0)
2750                 return err;
2751
2752         ndm = nlmsg_data(nlh);
2753         if (ndm->ndm_ifindex == 0) {
2754                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2755                 return -EINVAL;
2756         }
2757
2758         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2759         if (dev == NULL) {
2760                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2761                 return -ENODEV;
2762         }
2763
2764         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2765                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2766                 return -EINVAL;
2767         }
2768
2769         addr = nla_data(tb[NDA_LLADDR]);
2770
2771         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2772         if (err)
2773                 return err;
2774
2775         err = -EOPNOTSUPP;
2776
2777         /* Support fdb on master device the net/bridge default case */
2778         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2779             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2780                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2781                 const struct net_device_ops *ops = br_dev->netdev_ops;
2782
2783                 if (ops->ndo_fdb_del)
2784                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2785
2786                 if (err)
2787                         goto out;
2788                 else
2789                         ndm->ndm_flags &= ~NTF_MASTER;
2790         }
2791
2792         /* Embedded bridge, macvlan, and any other device support */
2793         if (ndm->ndm_flags & NTF_SELF) {
2794                 if (dev->netdev_ops->ndo_fdb_del)
2795                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2796                                                            vid);
2797                 else
2798                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2799
2800                 if (!err) {
2801                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2802                         ndm->ndm_flags &= ~NTF_SELF;
2803                 }
2804         }
2805 out:
2806         return err;
2807 }
2808
2809 static int nlmsg_populate_fdb(struct sk_buff *skb,
2810                               struct netlink_callback *cb,
2811                               struct net_device *dev,
2812                               int *idx,
2813                               struct netdev_hw_addr_list *list)
2814 {
2815         struct netdev_hw_addr *ha;
2816         int err;
2817         u32 portid, seq;
2818
2819         portid = NETLINK_CB(cb->skb).portid;
2820         seq = cb->nlh->nlmsg_seq;
2821
2822         list_for_each_entry(ha, &list->list, list) {
2823                 if (*idx < cb->args[0])
2824                         goto skip;
2825
2826                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2827                                               portid, seq,
2828                                               RTM_NEWNEIGH, NTF_SELF,
2829                                               NLM_F_MULTI);
2830                 if (err < 0)
2831                         return err;
2832 skip:
2833                 *idx += 1;
2834         }
2835         return 0;
2836 }
2837
2838 /**
2839  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2840  * @nlh: netlink message header
2841  * @dev: netdevice
2842  *
2843  * Default netdevice operation to dump the existing unicast address list.
2844  * Returns number of addresses from list put in skb.
2845  */
2846 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2847                       struct netlink_callback *cb,
2848                       struct net_device *dev,
2849                       struct net_device *filter_dev,
2850                       int idx)
2851 {
2852         int err;
2853
2854         netif_addr_lock_bh(dev);
2855         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2856         if (err)
2857                 goto out;
2858         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2859 out:
2860         netif_addr_unlock_bh(dev);
2861         return idx;
2862 }
2863 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2864
2865 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2866 {
2867         struct net_device *dev;
2868         struct nlattr *tb[IFLA_MAX+1];
2869         struct net_device *br_dev = NULL;
2870         const struct net_device_ops *ops = NULL;
2871         const struct net_device_ops *cops = NULL;
2872         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2873         struct net *net = sock_net(skb->sk);
2874         int brport_idx = 0;
2875         int br_idx = 0;
2876         int idx = 0;
2877
2878         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2879                         ifla_policy) == 0) {
2880                 if (tb[IFLA_MASTER])
2881                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2882         }
2883
2884         brport_idx = ifm->ifi_index;
2885
2886         if (br_idx) {
2887                 br_dev = __dev_get_by_index(net, br_idx);
2888                 if (!br_dev)
2889                         return -ENODEV;
2890
2891                 ops = br_dev->netdev_ops;
2892         }
2893
2894         for_each_netdev(net, dev) {
2895                 if (brport_idx && (dev->ifindex != brport_idx))
2896                         continue;
2897
2898                 if (!br_idx) { /* user did not specify a specific bridge */
2899                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2900                                 br_dev = netdev_master_upper_dev_get(dev);
2901                                 cops = br_dev->netdev_ops;
2902                         }
2903
2904                 } else {
2905                         if (dev != br_dev &&
2906                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2907                                 continue;
2908
2909                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2910                             !(dev->priv_flags & IFF_EBRIDGE))
2911                                 continue;
2912
2913                         cops = ops;
2914                 }
2915
2916                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2917                         if (cops && cops->ndo_fdb_dump)
2918                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2919                                                          idx);
2920                 }
2921
2922                 if (dev->netdev_ops->ndo_fdb_dump)
2923                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2924                                                             idx);
2925                 else
2926                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2927
2928                 cops = NULL;
2929         }
2930
2931         cb->args[0] = idx;
2932         return skb->len;
2933 }
2934
2935 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2936                                unsigned int attrnum, unsigned int flag)
2937 {
2938         if (mask & flag)
2939                 return nla_put_u8(skb, attrnum, !!(flags & flag));
2940         return 0;
2941 }
2942
2943 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2944                             struct net_device *dev, u16 mode,
2945                             u32 flags, u32 mask, int nlflags,
2946                             u32 filter_mask,
2947                             int (*vlan_fill)(struct sk_buff *skb,
2948                                              struct net_device *dev,
2949                                              u32 filter_mask))
2950 {
2951         struct nlmsghdr *nlh;
2952         struct ifinfomsg *ifm;
2953         struct nlattr *br_afspec;
2954         struct nlattr *protinfo;
2955         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2956         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2957         int err = 0;
2958
2959         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2960         if (nlh == NULL)
2961                 return -EMSGSIZE;
2962
2963         ifm = nlmsg_data(nlh);
2964         ifm->ifi_family = AF_BRIDGE;
2965         ifm->__ifi_pad = 0;
2966         ifm->ifi_type = dev->type;
2967         ifm->ifi_index = dev->ifindex;
2968         ifm->ifi_flags = dev_get_flags(dev);
2969         ifm->ifi_change = 0;
2970
2971
2972         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2973             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2974             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2975             (br_dev &&
2976              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2977             (dev->addr_len &&
2978              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2979             (dev->ifindex != dev_get_iflink(dev) &&
2980              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2981                 goto nla_put_failure;
2982
2983         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2984         if (!br_afspec)
2985                 goto nla_put_failure;
2986
2987         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2988                 nla_nest_cancel(skb, br_afspec);
2989                 goto nla_put_failure;
2990         }
2991
2992         if (mode != BRIDGE_MODE_UNDEF) {
2993                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2994                         nla_nest_cancel(skb, br_afspec);
2995                         goto nla_put_failure;
2996                 }
2997         }
2998         if (vlan_fill) {
2999                 err = vlan_fill(skb, dev, filter_mask);
3000                 if (err) {
3001                         nla_nest_cancel(skb, br_afspec);
3002                         goto nla_put_failure;
3003                 }
3004         }
3005         nla_nest_end(skb, br_afspec);
3006
3007         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3008         if (!protinfo)
3009                 goto nla_put_failure;
3010
3011         if (brport_nla_put_flag(skb, flags, mask,
3012                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3013             brport_nla_put_flag(skb, flags, mask,
3014                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3015             brport_nla_put_flag(skb, flags, mask,
3016                                 IFLA_BRPORT_FAST_LEAVE,
3017                                 BR_MULTICAST_FAST_LEAVE) ||
3018             brport_nla_put_flag(skb, flags, mask,
3019                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3020             brport_nla_put_flag(skb, flags, mask,
3021                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3022             brport_nla_put_flag(skb, flags, mask,
3023                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3024             brport_nla_put_flag(skb, flags, mask,
3025                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3026             brport_nla_put_flag(skb, flags, mask,
3027                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3028                 nla_nest_cancel(skb, protinfo);
3029                 goto nla_put_failure;
3030         }
3031
3032         nla_nest_end(skb, protinfo);
3033
3034         nlmsg_end(skb, nlh);
3035         return 0;
3036 nla_put_failure:
3037         nlmsg_cancel(skb, nlh);
3038         return err ? err : -EMSGSIZE;
3039 }
3040 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3041
3042 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3043 {
3044         struct net *net = sock_net(skb->sk);
3045         struct net_device *dev;
3046         int idx = 0;
3047         u32 portid = NETLINK_CB(cb->skb).portid;
3048         u32 seq = cb->nlh->nlmsg_seq;
3049         u32 filter_mask = 0;
3050         int err;
3051
3052         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3053                 struct nlattr *extfilt;
3054
3055                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3056                                           IFLA_EXT_MASK);
3057                 if (extfilt) {
3058                         if (nla_len(extfilt) < sizeof(filter_mask))
3059                                 return -EINVAL;
3060
3061                         filter_mask = nla_get_u32(extfilt);
3062                 }
3063         }
3064
3065         rcu_read_lock();
3066         for_each_netdev_rcu(net, dev) {
3067                 const struct net_device_ops *ops = dev->netdev_ops;
3068                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3069
3070                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3071                         if (idx >= cb->args[0]) {
3072                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3073                                                 skb, portid, seq, dev,
3074                                                 filter_mask, NLM_F_MULTI);
3075                                 if (err < 0 && err != -EOPNOTSUPP)
3076                                         break;
3077                         }
3078                         idx++;
3079                 }
3080
3081                 if (ops->ndo_bridge_getlink) {
3082                         if (idx >= cb->args[0]) {
3083                                 err = ops->ndo_bridge_getlink(skb, portid,
3084                                                               seq, dev,
3085                                                               filter_mask,
3086                                                               NLM_F_MULTI);
3087                                 if (err < 0 && err != -EOPNOTSUPP)
3088                                         break;
3089                         }
3090                         idx++;
3091                 }
3092         }
3093         rcu_read_unlock();
3094         cb->args[0] = idx;
3095
3096         return skb->len;
3097 }
3098
3099 static inline size_t bridge_nlmsg_size(void)
3100 {
3101         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3102                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3103                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3104                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3105                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3106                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3107                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3108                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3109                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3110                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3111                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3112 }
3113
3114 static int rtnl_bridge_notify(struct net_device *dev)
3115 {
3116         struct net *net = dev_net(dev);
3117         struct sk_buff *skb;
3118         int err = -EOPNOTSUPP;
3119
3120         if (!dev->netdev_ops->ndo_bridge_getlink)
3121                 return 0;
3122
3123         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3124         if (!skb) {
3125                 err = -ENOMEM;
3126                 goto errout;
3127         }
3128
3129         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3130         if (err < 0)
3131                 goto errout;
3132
3133         if (!skb->len)
3134                 goto errout;
3135
3136         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3137         return 0;
3138 errout:
3139         WARN_ON(err == -EMSGSIZE);
3140         kfree_skb(skb);
3141         if (err)
3142                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3143         return err;
3144 }
3145
3146 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3147 {
3148         struct net *net = sock_net(skb->sk);
3149         struct ifinfomsg *ifm;
3150         struct net_device *dev;
3151         struct nlattr *br_spec, *attr = NULL;
3152         int rem, err = -EOPNOTSUPP;
3153         u16 flags = 0;
3154         bool have_flags = false;
3155
3156         if (nlmsg_len(nlh) < sizeof(*ifm))
3157                 return -EINVAL;
3158
3159         ifm = nlmsg_data(nlh);
3160         if (ifm->ifi_family != AF_BRIDGE)
3161                 return -EPFNOSUPPORT;
3162
3163         dev = __dev_get_by_index(net, ifm->ifi_index);
3164         if (!dev) {
3165                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3166                 return -ENODEV;
3167         }
3168
3169         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3170         if (br_spec) {
3171                 nla_for_each_nested(attr, br_spec, rem) {
3172                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3173                                 if (nla_len(attr) < sizeof(flags))
3174                                         return -EINVAL;
3175
3176                                 have_flags = true;
3177                                 flags = nla_get_u16(attr);
3178                                 break;
3179                         }
3180                 }
3181         }
3182
3183         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3184                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3185
3186                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3187                         err = -EOPNOTSUPP;
3188                         goto out;
3189                 }
3190
3191                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3192                 if (err)
3193                         goto out;
3194
3195                 flags &= ~BRIDGE_FLAGS_MASTER;
3196         }
3197
3198         if ((flags & BRIDGE_FLAGS_SELF)) {
3199                 if (!dev->netdev_ops->ndo_bridge_setlink)
3200                         err = -EOPNOTSUPP;
3201                 else
3202                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3203                                                                   flags);
3204                 if (!err) {
3205                         flags &= ~BRIDGE_FLAGS_SELF;
3206
3207                         /* Generate event to notify upper layer of bridge
3208                          * change
3209                          */
3210                         err = rtnl_bridge_notify(dev);
3211                 }
3212         }
3213
3214         if (have_flags)
3215                 memcpy(nla_data(attr), &flags, sizeof(flags));
3216 out:
3217         return err;
3218 }
3219
3220 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3221 {
3222         struct net *net = sock_net(skb->sk);
3223         struct ifinfomsg *ifm;
3224         struct net_device *dev;
3225         struct nlattr *br_spec, *attr = NULL;
3226         int rem, err = -EOPNOTSUPP;
3227         u16 flags = 0;
3228         bool have_flags = false;
3229
3230         if (nlmsg_len(nlh) < sizeof(*ifm))
3231                 return -EINVAL;
3232
3233         ifm = nlmsg_data(nlh);
3234         if (ifm->ifi_family != AF_BRIDGE)
3235                 return -EPFNOSUPPORT;
3236
3237         dev = __dev_get_by_index(net, ifm->ifi_index);
3238         if (!dev) {
3239                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3240                 return -ENODEV;
3241         }
3242
3243         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3244         if (br_spec) {
3245                 nla_for_each_nested(attr, br_spec, rem) {
3246                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3247                                 if (nla_len(attr) < sizeof(flags))
3248                                         return -EINVAL;
3249
3250                                 have_flags = true;
3251                                 flags = nla_get_u16(attr);
3252                                 break;
3253                         }
3254                 }
3255         }
3256
3257         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3258                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3259
3260                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3261                         err = -EOPNOTSUPP;
3262                         goto out;
3263                 }
3264
3265                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3266                 if (err)
3267                         goto out;
3268
3269                 flags &= ~BRIDGE_FLAGS_MASTER;
3270         }
3271
3272         if ((flags & BRIDGE_FLAGS_SELF)) {
3273                 if (!dev->netdev_ops->ndo_bridge_dellink)
3274                         err = -EOPNOTSUPP;
3275                 else
3276                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3277                                                                   flags);
3278
3279                 if (!err) {
3280                         flags &= ~BRIDGE_FLAGS_SELF;
3281
3282                         /* Generate event to notify upper layer of bridge
3283                          * change
3284                          */
3285                         err = rtnl_bridge_notify(dev);
3286                 }
3287         }
3288
3289         if (have_flags)
3290                 memcpy(nla_data(attr), &flags, sizeof(flags));
3291 out:
3292         return err;
3293 }
3294
3295 /* Process one rtnetlink message. */
3296
3297 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3298 {
3299         struct net *net = sock_net(skb->sk);
3300         rtnl_doit_func doit;
3301         int sz_idx, kind;
3302         int family;
3303         int type;
3304         int err;
3305
3306         type = nlh->nlmsg_type;
3307         if (type > RTM_MAX)
3308                 return -EOPNOTSUPP;
3309
3310         type -= RTM_BASE;
3311
3312         /* All the messages must have at least 1 byte length */
3313         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3314                 return 0;
3315
3316         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3317         sz_idx = type>>2;
3318         kind = type&3;
3319
3320         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3321                 return -EPERM;
3322
3323         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3324                 struct sock *rtnl;
3325                 rtnl_dumpit_func dumpit;
3326                 rtnl_calcit_func calcit;
3327                 u16 min_dump_alloc = 0;
3328
3329                 dumpit = rtnl_get_dumpit(family, type);
3330                 if (dumpit == NULL)
3331                         return -EOPNOTSUPP;
3332                 calcit = rtnl_get_calcit(family, type);
3333                 if (calcit)
3334                         min_dump_alloc = calcit(skb, nlh);
3335
3336                 __rtnl_unlock();
3337                 rtnl = net->rtnl;
3338                 {
3339                         struct netlink_dump_control c = {
3340                                 .dump           = dumpit,
3341                                 .min_dump_alloc = min_dump_alloc,
3342                         };
3343                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3344                 }
3345                 rtnl_lock();
3346                 return err;
3347         }
3348
3349         doit = rtnl_get_doit(family, type);
3350         if (doit == NULL)
3351                 return -EOPNOTSUPP;
3352
3353         return doit(skb, nlh);
3354 }
3355
3356 static void rtnetlink_rcv(struct sk_buff *skb)
3357 {
3358         rtnl_lock();
3359         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3360         rtnl_unlock();
3361 }
3362
3363 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3364 {
3365         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3366
3367         switch (event) {
3368         case NETDEV_UP:
3369         case NETDEV_DOWN:
3370         case NETDEV_PRE_UP:
3371         case NETDEV_POST_INIT:
3372         case NETDEV_REGISTER:
3373         case NETDEV_CHANGE:
3374         case NETDEV_PRE_TYPE_CHANGE:
3375         case NETDEV_GOING_DOWN:
3376         case NETDEV_UNREGISTER:
3377         case NETDEV_UNREGISTER_FINAL:
3378         case NETDEV_RELEASE:
3379         case NETDEV_JOIN:
3380         case NETDEV_BONDING_INFO:
3381                 break;
3382         default:
3383                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3384                 break;
3385         }
3386         return NOTIFY_DONE;
3387 }
3388
3389 static struct notifier_block rtnetlink_dev_notifier = {
3390         .notifier_call  = rtnetlink_event,
3391 };
3392
3393
3394 static int __net_init rtnetlink_net_init(struct net *net)
3395 {
3396         struct sock *sk;
3397         struct netlink_kernel_cfg cfg = {
3398                 .groups         = RTNLGRP_MAX,
3399                 .input          = rtnetlink_rcv,
3400                 .cb_mutex       = &rtnl_mutex,
3401                 .flags          = NL_CFG_F_NONROOT_RECV,
3402         };
3403
3404         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3405         if (!sk)
3406                 return -ENOMEM;
3407         net->rtnl = sk;
3408         return 0;
3409 }
3410
3411 static void __net_exit rtnetlink_net_exit(struct net *net)
3412 {
3413         netlink_kernel_release(net->rtnl);
3414         net->rtnl = NULL;
3415 }
3416
3417 static struct pernet_operations rtnetlink_net_ops = {
3418         .init = rtnetlink_net_init,
3419         .exit = rtnetlink_net_exit,
3420 };
3421
3422 void __init rtnetlink_init(void)
3423 {
3424         if (register_pernet_subsys(&rtnetlink_net_ops))
3425                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3426
3427         register_netdevice_notifier(&rtnetlink_dev_notifier);
3428
3429         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3430                       rtnl_dump_ifinfo, rtnl_calcit);
3431         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3432         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3433         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3434
3435         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3436         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3437
3438         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3439         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3440         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3441
3442         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3443         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3444         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3445 }
3446