net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 void __rtnl_unlock(void)
  75 {
  76         mutex_unlock(&rtnl_mutex);
  77 }
  78
  79 void rtnl_unlock(void)
  80 {
  81         /* This fellow will unlock it for us. */
  82         netdev_run_todo();
  83 }
  84 EXPORT_SYMBOL(rtnl_unlock);
  85
  86 int rtnl_trylock(void)
  87 {
  88         return mutex_trylock(&rtnl_mutex);
  89 }
  90 EXPORT_SYMBOL(rtnl_trylock);
  91
  92 int rtnl_is_locked(void)
  93 {
  94         return mutex_is_locked(&rtnl_mutex);
  95 }
  96 EXPORT_SYMBOL(rtnl_is_locked);
  97
  98 #ifdef CONFIG_PROVE_LOCKING
  99 bool lockdep_rtnl_is_held(void)
 100 {
 101         return lockdep_is_held(&rtnl_mutex);
 102 }
 103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 105
 106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 107
 108 static inline int rtm_msgindex(int msgtype)
 109 {
 110         int msgindex = msgtype - RTM_BASE;
 111
 112         /*
 113          * msgindex < 0 implies someone tried to register a netlink
 114          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 115          * the message type has not been added to linux/rtnetlink.h
 116          */
 117         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 118
 119         return msgindex;
 120 }
 121
 122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 123 {
 124         struct rtnl_link *tab;
 125
 126         if (protocol <= RTNL_FAMILY_MAX)
 127                 tab = rtnl_msg_handlers[protocol];
 128         else
 129                 tab = NULL;
 130
 131         if (tab == NULL || tab[msgindex].doit == NULL)
 132                 tab = rtnl_msg_handlers[PF_UNSPEC];
 133
 134         return tab[msgindex].doit;
 135 }
 136
 137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 138 {
 139         struct rtnl_link *tab;
 140
 141         if (protocol <= RTNL_FAMILY_MAX)
 142                 tab = rtnl_msg_handlers[protocol];
 143         else
 144                 tab = NULL;
 145
 146         if (tab == NULL || tab[msgindex].dumpit == NULL)
 147                 tab = rtnl_msg_handlers[PF_UNSPEC];
 148
 149         return tab[msgindex].dumpit;
 150 }
 151
 152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 153 {
 154         struct rtnl_link *tab;
 155
 156         if (protocol <= RTNL_FAMILY_MAX)
 157                 tab = rtnl_msg_handlers[protocol];
 158         else
 159                 tab = NULL;
 160
 161         if (tab == NULL || tab[msgindex].calcit == NULL)
 162                 tab = rtnl_msg_handlers[PF_UNSPEC];
 163
 164         return tab[msgindex].calcit;
 165 }
 166
 167 /**
 168  * __rtnl_register - Register a rtnetlink message type
 169  * @protocol: Protocol family or PF_UNSPEC
 170  * @msgtype: rtnetlink message type
 171  * @doit: Function pointer called for each request message
 172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 173  * @calcit: Function pointer to calc size of dump message
 174  *
 175  * Registers the specified function pointers (at least one of them has
 176  * to be non-NULL) to be called whenever a request message for the
 177  * specified protocol family and message type is received.
 178  *
 179  * The special protocol family PF_UNSPEC may be used to define fallback
 180  * function pointers for the case when no entry for the specific protocol
 181  * family exists.
 182  *
 183  * Returns 0 on success or a negative error code.
 184  */
 185 int __rtnl_register(int protocol, int msgtype,
 186                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 187                     rtnl_calcit_func calcit)
 188 {
 189         struct rtnl_link *tab;
 190         int msgindex;
 191
 192         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 193         msgindex = rtm_msgindex(msgtype);
 194
 195         tab = rtnl_msg_handlers[protocol];
 196         if (tab == NULL) {
 197                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 198                 if (tab == NULL)
 199                         return -ENOBUFS;
 200
 201                 rtnl_msg_handlers[protocol] = tab;
 202         }
 203
 204         if (doit)
 205                 tab[msgindex].doit = doit;
 206
 207         if (dumpit)
 208                 tab[msgindex].dumpit = dumpit;
 209
 210         if (calcit)
 211                 tab[msgindex].calcit = calcit;
 212
 213         return 0;
 214 }
 215 EXPORT_SYMBOL_GPL(__rtnl_register);
 216
 217 /**
 218  * rtnl_register - Register a rtnetlink message type
 219  *
 220  * Identical to __rtnl_register() but panics on failure. This is useful
 221  * as failure of this function is very unlikely, it can only happen due
 222  * to lack of memory when allocating the chain to store all message
 223  * handlers for a protocol. Meant for use in init functions where lack
 224  * of memory implies no sense in continuing.
 225  */
 226 void rtnl_register(int protocol, int msgtype,
 227                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 228                    rtnl_calcit_func calcit)
 229 {
 230         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 231                 panic("Unable to register rtnetlink message handler, "
 232                       "protocol = %d, message type = %d\n",
 233                       protocol, msgtype);
 234 }
 235 EXPORT_SYMBOL_GPL(rtnl_register);
 236
 237 /**
 238  * rtnl_unregister - Unregister a rtnetlink message type
 239  * @protocol: Protocol family or PF_UNSPEC
 240  * @msgtype: rtnetlink message type
 241  *
 242  * Returns 0 on success or a negative error code.
 243  */
 244 int rtnl_unregister(int protocol, int msgtype)
 245 {
 246         int msgindex;
 247
 248         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 249         msgindex = rtm_msgindex(msgtype);
 250
 251         if (rtnl_msg_handlers[protocol] == NULL)
 252                 return -ENOENT;
 253
 254         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 255         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 256         rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
 257
 258         return 0;
 259 }
 260 EXPORT_SYMBOL_GPL(rtnl_unregister);
 261
 262 /**
 263  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 264  * @protocol : Protocol family or PF_UNSPEC
 265  *
 266  * Identical to calling rtnl_unregster() for all registered message types
 267  * of a certain protocol family.
 268  */
 269 void rtnl_unregister_all(int protocol)
 270 {
 271         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 272
 273         kfree(rtnl_msg_handlers[protocol]);
 274         rtnl_msg_handlers[protocol] = NULL;
 275 }
 276 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 277
 278 static LIST_HEAD(link_ops);
 279
 280 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 281 {
 282         const struct rtnl_link_ops *ops;
 283
 284         list_for_each_entry(ops, &link_ops, list) {
 285                 if (!strcmp(ops->kind, kind))
 286                         return ops;
 287         }
 288         return NULL;
 289 }
 290
 291 /**
 292  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 293  * @ops: struct rtnl_link_ops * to register
 294  *
 295  * The caller must hold the rtnl_mutex. This function should be used
 296  * by drivers that create devices during module initialization. It
 297  * must be called before registering the devices.
 298  *
 299  * Returns 0 on success or a negative error code.
 300  */
 301 int __rtnl_link_register(struct rtnl_link_ops *ops)
 302 {
 303         if (rtnl_link_ops_get(ops->kind))
 304                 return -EEXIST;
 305
 306         /* The check for setup is here because if ops
 307          * does not have that filled up, it is not possible
 308          * to use the ops for creating device. So do not
 309          * fill up dellink as well. That disables rtnl_dellink.
 310          */
 311         if (ops->setup && !ops->dellink)
 312                 ops->dellink = unregister_netdevice_queue;
 313
 314         list_add_tail(&ops->list, &link_ops);
 315         return 0;
 316 }
 317 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 318
 319 /**
 320  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 321  * @ops: struct rtnl_link_ops * to register
 322  *
 323  * Returns 0 on success or a negative error code.
 324  */
 325 int rtnl_link_register(struct rtnl_link_ops *ops)
 326 {
 327         int err;
 328
 329         rtnl_lock();
 330         err = __rtnl_link_register(ops);
 331         rtnl_unlock();
 332         return err;
 333 }
 334 EXPORT_SYMBOL_GPL(rtnl_link_register);
 335
 336 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 337 {
 338         struct net_device *dev;
 339         LIST_HEAD(list_kill);
 340
 341         for_each_netdev(net, dev) {
 342                 if (dev->rtnl_link_ops == ops)
 343                         ops->dellink(dev, &list_kill);
 344         }
 345         unregister_netdevice_many(&list_kill);
 346 }
 347
 348 /**
 349  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 350  * @ops: struct rtnl_link_ops * to unregister
 351  *
 352  * The caller must hold the rtnl_mutex.
 353  */
 354 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 355 {
 356         struct net *net;
 357
 358         for_each_net(net) {
 359                 __rtnl_kill_links(net, ops);
 360         }
 361         list_del(&ops->list);
 362 }
 363 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 364
 365 /* Return with the rtnl_lock held when there are no network
 366  * devices unregistering in any network namespace.
 367  */
 368 static void rtnl_lock_unregistering_all(void)
 369 {
 370         struct net *net;
 371         bool unregistering;
 372         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 373
 374         add_wait_queue(&netdev_unregistering_wq, &wait);
 375         for (;;) {
 376                 unregistering = false;
 377                 rtnl_lock();
 378                 for_each_net(net) {
 379                         if (net->dev_unreg_count > 0) {
 380                                 unregistering = true;
 381                                 break;
 382                         }
 383                 }
 384                 if (!unregistering)
 385                         break;
 386                 __rtnl_unlock();
 387
 388                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 389         }
 390         remove_wait_queue(&netdev_unregistering_wq, &wait);
 391 }
 392
 393 /**
 394  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 395  * @ops: struct rtnl_link_ops * to unregister
 396  */
 397 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 398 {
 399         /* Close the race with cleanup_net() */
 400         mutex_lock(&net_mutex);
 401         rtnl_lock_unregistering_all();
 402         __rtnl_link_unregister(ops);
 403         rtnl_unlock();
 404         mutex_unlock(&net_mutex);
 405 }
 406 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 407
 408 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 409 {
 410         struct net_device *master_dev;
 411         const struct rtnl_link_ops *ops;
 412
 413         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 414         if (!master_dev)
 415                 return 0;
 416         ops = master_dev->rtnl_link_ops;
 417         if (!ops || !ops->get_slave_size)
 418                 return 0;
 419         /* IFLA_INFO_SLAVE_DATA + nested data */
 420         return nla_total_size(sizeof(struct nlattr)) +
 421                ops->get_slave_size(master_dev, dev);
 422 }
 423
 424 static size_t rtnl_link_get_size(const struct net_device *dev)
 425 {
 426         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 427         size_t size;
 428
 429         if (!ops)
 430                 return 0;
 431
 432         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 433                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 434
 435         if (ops->get_size)
 436                 /* IFLA_INFO_DATA + nested data */
 437                 size += nla_total_size(sizeof(struct nlattr)) +
 438                         ops->get_size(dev);
 439
 440         if (ops->get_xstats_size)
 441                 /* IFLA_INFO_XSTATS */
 442                 size += nla_total_size(ops->get_xstats_size(dev));
 443
 444         size += rtnl_link_get_slave_info_data_size(dev);
 445
 446         return size;
 447 }
 448
 449 static LIST_HEAD(rtnl_af_ops);
 450
 451 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 452 {
 453         const struct rtnl_af_ops *ops;
 454
 455         list_for_each_entry(ops, &rtnl_af_ops, list) {
 456                 if (ops->family == family)
 457                         return ops;
 458         }
 459
 460         return NULL;
 461 }
 462
 463 /**
 464  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 465  * @ops: struct rtnl_af_ops * to register
 466  *
 467  * Returns 0 on success or a negative error code.
 468  */
 469 void rtnl_af_register(struct rtnl_af_ops *ops)
 470 {
 471         rtnl_lock();
 472         list_add_tail(&ops->list, &rtnl_af_ops);
 473         rtnl_unlock();
 474 }
 475 EXPORT_SYMBOL_GPL(rtnl_af_register);
 476
 477 /**
 478  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 479  * @ops: struct rtnl_af_ops * to unregister
 480  *
 481  * The caller must hold the rtnl_mutex.
 482  */
 483 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 484 {
 485         list_del(&ops->list);
 486 }
 487 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 488
 489 /**
 490  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 491  * @ops: struct rtnl_af_ops * to unregister
 492  */
 493 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 494 {
 495         rtnl_lock();
 496         __rtnl_af_unregister(ops);
 497         rtnl_unlock();
 498 }
 499 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 500
 501 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 502                                     u32 ext_filter_mask)
 503 {
 504         struct rtnl_af_ops *af_ops;
 505         size_t size;
 506
 507         /* IFLA_AF_SPEC */
 508         size = nla_total_size(sizeof(struct nlattr));
 509
 510         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 511                 if (af_ops->get_link_af_size) {
 512                         /* AF_* + nested data */
 513                         size += nla_total_size(sizeof(struct nlattr)) +
 514                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 515                 }
 516         }
 517
 518         return size;
 519 }
 520
 521 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 522 {
 523         struct net_device *master_dev;
 524
 525         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 526         if (master_dev && master_dev->rtnl_link_ops)
 527                 return true;
 528         return false;
 529 }
 530
 531 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 532                                      const struct net_device *dev)
 533 {
 534         struct net_device *master_dev;
 535         const struct rtnl_link_ops *ops;
 536         struct nlattr *slave_data;
 537         int err;
 538
 539         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 540         if (!master_dev)
 541                 return 0;
 542         ops = master_dev->rtnl_link_ops;
 543         if (!ops)
 544                 return 0;
 545         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 546                 return -EMSGSIZE;
 547         if (ops->fill_slave_info) {
 548                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 549                 if (!slave_data)
 550                         return -EMSGSIZE;
 551                 err = ops->fill_slave_info(skb, master_dev, dev);
 552                 if (err < 0)
 553                         goto err_cancel_slave_data;
 554                 nla_nest_end(skb, slave_data);
 555         }
 556         return 0;
 557
 558 err_cancel_slave_data:
 559         nla_nest_cancel(skb, slave_data);
 560         return err;
 561 }
 562
 563 static int rtnl_link_info_fill(struct sk_buff *skb,
 564                                const struct net_device *dev)
 565 {
 566         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 567         struct nlattr *data;
 568         int err;
 569
 570         if (!ops)
 571                 return 0;
 572         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 573                 return -EMSGSIZE;
 574         if (ops->fill_xstats) {
 575                 err = ops->fill_xstats(skb, dev);
 576                 if (err < 0)
 577                         return err;
 578         }
 579         if (ops->fill_info) {
 580                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 581                 if (data == NULL)
 582                         return -EMSGSIZE;
 583                 err = ops->fill_info(skb, dev);
 584                 if (err < 0)
 585                         goto err_cancel_data;
 586                 nla_nest_end(skb, data);
 587         }
 588         return 0;
 589
 590 err_cancel_data:
 591         nla_nest_cancel(skb, data);
 592         return err;
 593 }
 594
 595 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 596 {
 597         struct nlattr *linkinfo;
 598         int err = -EMSGSIZE;
 599
 600         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 601         if (linkinfo == NULL)
 602                 goto out;
 603
 604         err = rtnl_link_info_fill(skb, dev);
 605         if (err < 0)
 606                 goto err_cancel_link;
 607
 608         err = rtnl_link_slave_info_fill(skb, dev);
 609         if (err < 0)
 610                 goto err_cancel_link;
 611
 612         nla_nest_end(skb, linkinfo);
 613         return 0;
 614
 615 err_cancel_link:
 616         nla_nest_cancel(skb, linkinfo);
 617 out:
 618         return err;
 619 }
 620
 621 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 622 {
 623         struct sock *rtnl = net->rtnl;
 624         int err = 0;
 625
 626         NETLINK_CB(skb).dst_group = group;
 627         if (echo)
 628                 atomic_inc(&skb->users);
 629         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 630         if (echo)
 631                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 632         return err;
 633 }
 634
 635 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 636 {
 637         struct sock *rtnl = net->rtnl;
 638
 639         return nlmsg_unicast(rtnl, skb, pid);
 640 }
 641 EXPORT_SYMBOL(rtnl_unicast);
 642
 643 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 644                  struct nlmsghdr *nlh, gfp_t flags)
 645 {
 646         struct sock *rtnl = net->rtnl;
 647         int report = 0;
 648
 649         if (nlh)
 650                 report = nlmsg_report(nlh);
 651
 652         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 653 }
 654 EXPORT_SYMBOL(rtnl_notify);
 655
 656 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 657 {
 658         struct sock *rtnl = net->rtnl;
 659
 660         netlink_set_err(rtnl, 0, group, error);
 661 }
 662 EXPORT_SYMBOL(rtnl_set_sk_err);
 663
 664 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 665 {
 666         struct nlattr *mx;
 667         int i, valid = 0;
 668
 669         mx = nla_nest_start(skb, RTA_METRICS);
 670         if (mx == NULL)
 671                 return -ENOBUFS;
 672
 673         for (i = 0; i < RTAX_MAX; i++) {
 674                 if (metrics[i]) {
 675                         if (i == RTAX_CC_ALGO - 1) {
 676                                 char tmp[TCP_CA_NAME_MAX], *name;
 677
 678                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 679                                 if (!name)
 680                                         continue;
 681                                 if (nla_put_string(skb, i + 1, name))
 682                                         goto nla_put_failure;
 683                         } else if (i == RTAX_FEATURES - 1) {
 684                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 685
 686                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 687                                 if (nla_put_u32(skb, i + 1, user_features))
 688                                         goto nla_put_failure;
 689                         } else {
 690                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 691                                         goto nla_put_failure;
 692                         }
 693                         valid++;
 694                 }
 695         }
 696
 697         if (!valid) {
 698                 nla_nest_cancel(skb, mx);
 699                 return 0;
 700         }
 701
 702         return nla_nest_end(skb, mx);
 703
 704 nla_put_failure:
 705         nla_nest_cancel(skb, mx);
 706         return -EMSGSIZE;
 707 }
 708 EXPORT_SYMBOL(rtnetlink_put_metrics);
 709
 710 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 711                        long expires, u32 error)
 712 {
 713         struct rta_cacheinfo ci = {
 714                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 715                 .rta_used = dst->__use,
 716                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 717                 .rta_error = error,
 718                 .rta_id =  id,
 719         };
 720
 721         if (expires) {
 722                 unsigned long clock;
 723
 724                 clock = jiffies_to_clock_t(abs(expires));
 725                 clock = min_t(unsigned long, clock, INT_MAX);
 726                 ci.rta_expires = (expires > 0) ? clock : -clock;
 727         }
 728         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 729 }
 730 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 731
 732 static void set_operstate(struct net_device *dev, unsigned char transition)
 733 {
 734         unsigned char operstate = dev->operstate;
 735
 736         switch (transition) {
 737         case IF_OPER_UP:
 738                 if ((operstate == IF_OPER_DORMANT ||
 739                      operstate == IF_OPER_UNKNOWN) &&
 740                     !netif_dormant(dev))
 741                         operstate = IF_OPER_UP;
 742                 break;
 743
 744         case IF_OPER_DORMANT:
 745                 if (operstate == IF_OPER_UP ||
 746                     operstate == IF_OPER_UNKNOWN)
 747                         operstate = IF_OPER_DORMANT;
 748                 break;
 749         }
 750
 751         if (dev->operstate != operstate) {
 752                 write_lock_bh(&dev_base_lock);
 753                 dev->operstate = operstate;
 754                 write_unlock_bh(&dev_base_lock);
 755                 netdev_state_change(dev);
 756         }
 757 }
 758
 759 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 760 {
 761         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 762                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 763 }
 764
 765 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 766                                            const struct ifinfomsg *ifm)
 767 {
 768         unsigned int flags = ifm->ifi_flags;
 769
 770         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 771         if (ifm->ifi_change)
 772                 flags = (flags & ifm->ifi_change) |
 773                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 774
 775         return flags;
 776 }
 777
 778 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 779                                  const struct rtnl_link_stats64 *b)
 780 {
 781         a->rx_packets = b->rx_packets;
 782         a->tx_packets = b->tx_packets;
 783         a->rx_bytes = b->rx_bytes;
 784         a->tx_bytes = b->tx_bytes;
 785         a->rx_errors = b->rx_errors;
 786         a->tx_errors = b->tx_errors;
 787         a->rx_dropped = b->rx_dropped;
 788         a->tx_dropped = b->tx_dropped;
 789
 790         a->multicast = b->multicast;
 791         a->collisions = b->collisions;
 792
 793         a->rx_length_errors = b->rx_length_errors;
 794         a->rx_over_errors = b->rx_over_errors;
 795         a->rx_crc_errors = b->rx_crc_errors;
 796         a->rx_frame_errors = b->rx_frame_errors;
 797         a->rx_fifo_errors = b->rx_fifo_errors;
 798         a->rx_missed_errors = b->rx_missed_errors;
 799
 800         a->tx_aborted_errors = b->tx_aborted_errors;
 801         a->tx_carrier_errors = b->tx_carrier_errors;
 802         a->tx_fifo_errors = b->tx_fifo_errors;
 803         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 804         a->tx_window_errors = b->tx_window_errors;
 805
 806         a->rx_compressed = b->rx_compressed;
 807         a->tx_compressed = b->tx_compressed;
 808 }
 809
 810 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 811 {
 812         memcpy(v, b, sizeof(*b));
 813 }
 814
 815 /* All VF info */
 816 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 817                                    u32 ext_filter_mask)
 818 {
 819         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 820             (ext_filter_mask & RTEXT_FILTER_VF)) {
 821                 int num_vfs = dev_num_vf(dev->dev.parent);
 822                 size_t size = nla_total_size(sizeof(struct nlattr));
 823                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 824                 size += num_vfs *
 825                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 826                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 827                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 828                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 829                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 830                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 831                          /* IFLA_VF_STATS_RX_PACKETS */
 832                          nla_total_size(sizeof(__u64)) +
 833                          /* IFLA_VF_STATS_TX_PACKETS */
 834                          nla_total_size(sizeof(__u64)) +
 835                          /* IFLA_VF_STATS_RX_BYTES */
 836                          nla_total_size(sizeof(__u64)) +
 837                          /* IFLA_VF_STATS_TX_BYTES */
 838                          nla_total_size(sizeof(__u64)) +
 839                          /* IFLA_VF_STATS_BROADCAST */
 840                          nla_total_size(sizeof(__u64)) +
 841                          /* IFLA_VF_STATS_MULTICAST */
 842                          nla_total_size(sizeof(__u64)) +
 843                          nla_total_size(sizeof(struct ifla_vf_trust)));
 844                 return size;
 845         } else
 846                 return 0;
 847 }
 848
 849 static size_t rtnl_port_size(const struct net_device *dev,
 850                              u32 ext_filter_mask)
 851 {
 852         size_t port_size = nla_total_size(4)            /* PORT_VF */
 853                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 854                 + nla_total_size(sizeof(struct ifla_port_vsi))
 855                                                         /* PORT_VSI_TYPE */
 856                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 857                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 858                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 859                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 860         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 861         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 862                 + port_size;
 863         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 864                 + port_size;
 865
 866         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 867             !(ext_filter_mask & RTEXT_FILTER_VF))
 868                 return 0;
 869         if (dev_num_vf(dev->dev.parent))
 870                 return port_self_size + vf_ports_size +
 871                         vf_port_size * dev_num_vf(dev->dev.parent);
 872         else
 873                 return port_self_size;
 874 }
 875
 876 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 877                                      u32 ext_filter_mask)
 878 {
 879         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 880                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 881                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 882                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 883                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 884                + nla_total_size(sizeof(struct rtnl_link_stats))
 885                + nla_total_size(sizeof(struct rtnl_link_stats64))
 886                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 887                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 888                + nla_total_size(4) /* IFLA_TXQLEN */
 889                + nla_total_size(4) /* IFLA_WEIGHT */
 890                + nla_total_size(4) /* IFLA_MTU */
 891                + nla_total_size(4) /* IFLA_LINK */
 892                + nla_total_size(4) /* IFLA_MASTER */
 893                + nla_total_size(1) /* IFLA_CARRIER */
 894                + nla_total_size(4) /* IFLA_PROMISCUITY */
 895                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 896                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 897                + nla_total_size(1) /* IFLA_OPERSTATE */
 898                + nla_total_size(1) /* IFLA_LINKMODE */
 899                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 900                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 901                + nla_total_size(4) /* IFLA_GROUP */
 902                + nla_total_size(ext_filter_mask
 903                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 904                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 905                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 906                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 907                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 908                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 909                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 910                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 911                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 912
 913 }
 914
 915 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 916 {
 917         struct nlattr *vf_ports;
 918         struct nlattr *vf_port;
 919         int vf;
 920         int err;
 921
 922         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 923         if (!vf_ports)
 924                 return -EMSGSIZE;
 925
 926         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 927                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 928                 if (!vf_port)
 929                         goto nla_put_failure;
 930                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 931                         goto nla_put_failure;
 932                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 933                 if (err == -EMSGSIZE)
 934                         goto nla_put_failure;
 935                 if (err) {
 936                         nla_nest_cancel(skb, vf_port);
 937                         continue;
 938                 }
 939                 nla_nest_end(skb, vf_port);
 940         }
 941
 942         nla_nest_end(skb, vf_ports);
 943
 944         return 0;
 945
 946 nla_put_failure:
 947         nla_nest_cancel(skb, vf_ports);
 948         return -EMSGSIZE;
 949 }
 950
 951 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 952 {
 953         struct nlattr *port_self;
 954         int err;
 955
 956         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 957         if (!port_self)
 958                 return -EMSGSIZE;
 959
 960         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 961         if (err) {
 962                 nla_nest_cancel(skb, port_self);
 963                 return (err == -EMSGSIZE) ? err : 0;
 964         }
 965
 966         nla_nest_end(skb, port_self);
 967
 968         return 0;
 969 }
 970
 971 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 972                           u32 ext_filter_mask)
 973 {
 974         int err;
 975
 976         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 977             !(ext_filter_mask & RTEXT_FILTER_VF))
 978                 return 0;
 979
 980         err = rtnl_port_self_fill(skb, dev);
 981         if (err)
 982                 return err;
 983
 984         if (dev_num_vf(dev->dev.parent)) {
 985                 err = rtnl_vf_ports_fill(skb, dev);
 986                 if (err)
 987                         return err;
 988         }
 989
 990         return 0;
 991 }
 992
 993 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 994 {
 995         int err;
 996         struct netdev_phys_item_id ppid;
 997
 998         err = dev_get_phys_port_id(dev, &ppid);
 999         if (err) {
1000                 if (err == -EOPNOTSUPP)
1001                         return 0;
1002                 return err;
1003         }
1004
1005         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1006                 return -EMSGSIZE;
1007
1008         return 0;
1009 }
1010
1011 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1012 {
1013         char name[IFNAMSIZ];
1014         int err;
1015
1016         err = dev_get_phys_port_name(dev, name, sizeof(name));
1017         if (err) {
1018                 if (err == -EOPNOTSUPP)
1019                         return 0;
1020                 return err;
1021         }
1022
1023         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1024                 return -EMSGSIZE;
1025
1026         return 0;
1027 }
1028
1029 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1030 {
1031         int err;
1032         struct switchdev_attr attr = {
1033                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1034                 .flags = SWITCHDEV_F_NO_RECURSE,
1035         };
1036
1037         err = switchdev_port_attr_get(dev, &attr);
1038         if (err) {
1039                 if (err == -EOPNOTSUPP)
1040                         return 0;
1041                 return err;
1042         }
1043
1044         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1045                     attr.u.ppid.id))
1046                 return -EMSGSIZE;
1047
1048         return 0;
1049 }
1050
1051 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1052                                               struct net_device *dev)
1053 {
1054         const struct rtnl_link_stats64 *stats;
1055         struct rtnl_link_stats64 temp;
1056         struct nlattr *attr;
1057
1058         stats = dev_get_stats(dev, &temp);
1059
1060         attr = nla_reserve(skb, IFLA_STATS,
1061                            sizeof(struct rtnl_link_stats));
1062         if (!attr)
1063                 return -EMSGSIZE;
1064
1065         copy_rtnl_link_stats(nla_data(attr), stats);
1066
1067         attr = nla_reserve(skb, IFLA_STATS64,
1068                            sizeof(struct rtnl_link_stats64));
1069         if (!attr)
1070                 return -EMSGSIZE;
1071
1072         copy_rtnl_link_stats64(nla_data(attr), stats);
1073
1074         return 0;
1075 }
1076
1077 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1078                                                struct net_device *dev,
1079                                                int vfs_num,
1080                                                struct nlattr *vfinfo)
1081 {
1082         struct ifla_vf_rss_query_en vf_rss_query_en;
1083         struct ifla_vf_link_state vf_linkstate;
1084         struct ifla_vf_spoofchk vf_spoofchk;
1085         struct ifla_vf_tx_rate vf_tx_rate;
1086         struct ifla_vf_stats vf_stats;
1087         struct ifla_vf_trust vf_trust;
1088         struct ifla_vf_vlan vf_vlan;
1089         struct ifla_vf_rate vf_rate;
1090         struct nlattr *vf, *vfstats;
1091         struct ifla_vf_mac vf_mac;
1092         struct ifla_vf_info ivi;
1093
1094         memset(&ivi, 0, sizeof(ivi));
1095
1096         /* Not all SR-IOV capable drivers support the
1097          * spoofcheck and "RSS query enable" query.  Preset to
1098          * -1 so the user space tool can detect that the driver
1099          * didn't report anything.
1100          */
1101         ivi.spoofchk = -1;
1102         ivi.rss_query_en = -1;
1103         ivi.trusted = -1;
1104         /* The default value for VF link state is "auto"
1105          * IFLA_VF_LINK_STATE_AUTO which equals zero
1106          */
1107         ivi.linkstate = 0;
1108         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1109                 return 0;
1110
1111         vf_mac.vf =
1112                 vf_vlan.vf =
1113                 vf_rate.vf =
1114                 vf_tx_rate.vf =
1115                 vf_spoofchk.vf =
1116                 vf_linkstate.vf =
1117                 vf_rss_query_en.vf =
1118                 vf_trust.vf = ivi.vf;
1119
1120         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1121         vf_vlan.vlan = ivi.vlan;
1122         vf_vlan.qos = ivi.qos;
1123         vf_tx_rate.rate = ivi.max_tx_rate;
1124         vf_rate.min_tx_rate = ivi.min_tx_rate;
1125         vf_rate.max_tx_rate = ivi.max_tx_rate;
1126         vf_spoofchk.setting = ivi.spoofchk;
1127         vf_linkstate.link_state = ivi.linkstate;
1128         vf_rss_query_en.setting = ivi.rss_query_en;
1129         vf_trust.setting = ivi.trusted;
1130         vf = nla_nest_start(skb, IFLA_VF_INFO);
1131         if (!vf) {
1132                 nla_nest_cancel(skb, vfinfo);
1133                 return -EMSGSIZE;
1134         }
1135         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1136             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1137             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1138                     &vf_rate) ||
1139             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1140                     &vf_tx_rate) ||
1141             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1142                     &vf_spoofchk) ||
1143             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1144                     &vf_linkstate) ||
1145             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1146                     sizeof(vf_rss_query_en),
1147                     &vf_rss_query_en) ||
1148             nla_put(skb, IFLA_VF_TRUST,
1149                     sizeof(vf_trust), &vf_trust))
1150                 return -EMSGSIZE;
1151         memset(&vf_stats, 0, sizeof(vf_stats));
1152         if (dev->netdev_ops->ndo_get_vf_stats)
1153                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1154                                                 &vf_stats);
1155         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1156         if (!vfstats) {
1157                 nla_nest_cancel(skb, vf);
1158                 nla_nest_cancel(skb, vfinfo);
1159                 return -EMSGSIZE;
1160         }
1161         if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1162                         vf_stats.rx_packets) ||
1163             nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1164                         vf_stats.tx_packets) ||
1165             nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1166                         vf_stats.rx_bytes) ||
1167             nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1168                         vf_stats.tx_bytes) ||
1169             nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1170                         vf_stats.broadcast) ||
1171             nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1172                         vf_stats.multicast))
1173                 return -EMSGSIZE;
1174         nla_nest_end(skb, vfstats);
1175         nla_nest_end(skb, vf);
1176         return 0;
1177 }
1178
1179 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1180 {
1181         struct rtnl_link_ifmap map;
1182
1183         memset(&map, 0, sizeof(map));
1184         map.mem_start   = dev->mem_start;
1185         map.mem_end     = dev->mem_end;
1186         map.base_addr   = dev->base_addr;
1187         map.irq         = dev->irq;
1188         map.dma         = dev->dma;
1189         map.port        = dev->if_port;
1190
1191         if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1192                 return -EMSGSIZE;
1193
1194         return 0;
1195 }
1196
1197 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1198                             int type, u32 pid, u32 seq, u32 change,
1199                             unsigned int flags, u32 ext_filter_mask)
1200 {
1201         struct ifinfomsg *ifm;
1202         struct nlmsghdr *nlh;
1203         struct nlattr *af_spec;
1204         struct rtnl_af_ops *af_ops;
1205         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1206
1207         ASSERT_RTNL();
1208         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1209         if (nlh == NULL)
1210                 return -EMSGSIZE;
1211
1212         ifm = nlmsg_data(nlh);
1213         ifm->ifi_family = AF_UNSPEC;
1214         ifm->__ifi_pad = 0;
1215         ifm->ifi_type = dev->type;
1216         ifm->ifi_index = dev->ifindex;
1217         ifm->ifi_flags = dev_get_flags(dev);
1218         ifm->ifi_change = change;
1219
1220         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1221             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1222             nla_put_u8(skb, IFLA_OPERSTATE,
1223                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1224             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1225             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1226             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1227             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1228             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1229 #ifdef CONFIG_RPS
1230             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1231 #endif
1232             (dev->ifindex != dev_get_iflink(dev) &&
1233              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1234             (upper_dev &&
1235              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1236             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1237             (dev->qdisc &&
1238              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1239             (dev->ifalias &&
1240              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1241             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1242                         atomic_read(&dev->carrier_changes)) ||
1243             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1244                 goto nla_put_failure;
1245
1246         if (rtnl_fill_link_ifmap(skb, dev))
1247                 goto nla_put_failure;
1248
1249         if (dev->addr_len) {
1250                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1251                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1252                         goto nla_put_failure;
1253         }
1254
1255         if (rtnl_phys_port_id_fill(skb, dev))
1256                 goto nla_put_failure;
1257
1258         if (rtnl_phys_port_name_fill(skb, dev))
1259                 goto nla_put_failure;
1260
1261         if (rtnl_phys_switch_id_fill(skb, dev))
1262                 goto nla_put_failure;
1263
1264         if (rtnl_fill_stats(skb, dev))
1265                 goto nla_put_failure;
1266
1267         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1268             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1269                 goto nla_put_failure;
1270
1271         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1272             ext_filter_mask & RTEXT_FILTER_VF) {
1273                 int i;
1274                 struct nlattr *vfinfo;
1275                 int num_vfs = dev_num_vf(dev->dev.parent);
1276
1277                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1278                 if (!vfinfo)
1279                         goto nla_put_failure;
1280                 for (i = 0; i < num_vfs; i++) {
1281                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1282                                 goto nla_put_failure;
1283                 }
1284
1285                 nla_nest_end(skb, vfinfo);
1286         }
1287
1288         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1289                 goto nla_put_failure;
1290
1291         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1292                 if (rtnl_link_fill(skb, dev) < 0)
1293                         goto nla_put_failure;
1294         }
1295
1296         if (dev->rtnl_link_ops &&
1297             dev->rtnl_link_ops->get_link_net) {
1298                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1299
1300                 if (!net_eq(dev_net(dev), link_net)) {
1301                         int id = peernet2id_alloc(dev_net(dev), link_net);
1302
1303                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1304                                 goto nla_put_failure;
1305                 }
1306         }
1307
1308         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1309                 goto nla_put_failure;
1310
1311         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1312                 if (af_ops->fill_link_af) {
1313                         struct nlattr *af;
1314                         int err;
1315
1316                         if (!(af = nla_nest_start(skb, af_ops->family)))
1317                                 goto nla_put_failure;
1318
1319                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1320
1321                         /*
1322                          * Caller may return ENODATA to indicate that there
1323                          * was no data to be dumped. This is not an error, it
1324                          * means we should trim the attribute header and
1325                          * continue.
1326                          */
1327                         if (err == -ENODATA)
1328                                 nla_nest_cancel(skb, af);
1329                         else if (err < 0)
1330                                 goto nla_put_failure;
1331
1332                         nla_nest_end(skb, af);
1333                 }
1334         }
1335
1336         nla_nest_end(skb, af_spec);
1337
1338         nlmsg_end(skb, nlh);
1339         return 0;
1340
1341 nla_put_failure:
1342         nlmsg_cancel(skb, nlh);
1343         return -EMSGSIZE;
1344 }
1345
1346 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1347         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1348         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1349         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1350         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1351         [IFLA_MTU]              = { .type = NLA_U32 },
1352         [IFLA_LINK]             = { .type = NLA_U32 },
1353         [IFLA_MASTER]           = { .type = NLA_U32 },
1354         [IFLA_CARRIER]          = { .type = NLA_U8 },
1355         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1356         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1357         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1358         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1359         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1360         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1361         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1362         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1363         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1364         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1365         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1366         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1367         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1368         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1369         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1370         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1371         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1372         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1373         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1374         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1375         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1376         [IFLA_GROUP]            = { .type = NLA_U32 },
1377 };
1378
1379 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1380         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1381         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1382         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1383         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1384 };
1385
1386 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1387         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1388         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1389         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1390         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1391         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1392         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1393         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1394         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1395         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1396 };
1397
1398 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1399         [IFLA_VF_STATS_RX_PACKETS]      = { .type = NLA_U64 },
1400         [IFLA_VF_STATS_TX_PACKETS]      = { .type = NLA_U64 },
1401         [IFLA_VF_STATS_RX_BYTES]        = { .type = NLA_U64 },
1402         [IFLA_VF_STATS_TX_BYTES]        = { .type = NLA_U64 },
1403         [IFLA_VF_STATS_BROADCAST]       = { .type = NLA_U64 },
1404         [IFLA_VF_STATS_MULTICAST]       = { .type = NLA_U64 },
1405 };
1406
1407 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1408         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1409         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1410                                     .len = PORT_PROFILE_MAX },
1411         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1412                                     .len = sizeof(struct ifla_port_vsi)},
1413         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1414                                       .len = PORT_UUID_MAX },
1415         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1416                                     .len = PORT_UUID_MAX },
1417         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1418         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1419 };
1420
1421 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1422 {
1423         struct net *net = sock_net(skb->sk);
1424         int h, s_h;
1425         int idx = 0, s_idx;
1426         struct net_device *dev;
1427         struct hlist_head *head;
1428         struct nlattr *tb[IFLA_MAX+1];
1429         u32 ext_filter_mask = 0;
1430         int err;
1431         int hdrlen;
1432
1433         s_h = cb->args[0];
1434         s_idx = cb->args[1];
1435
1436         cb->seq = net->dev_base_seq;
1437
1438         /* A hack to preserve kernel<->userspace interface.
1439          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1440          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1441          * what iproute2 < v3.9.0 used.
1442          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1443          * attribute, its netlink message is shorter than struct ifinfomsg.
1444          */
1445         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1446                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1447
1448         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1449
1450                 if (tb[IFLA_EXT_MASK])
1451                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1452         }
1453
1454         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1455                 idx = 0;
1456                 head = &net->dev_index_head[h];
1457                 hlist_for_each_entry(dev, head, index_hlist) {
1458                         if (idx < s_idx)
1459                                 goto cont;
1460                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1461                                                NETLINK_CB(cb->skb).portid,
1462                                                cb->nlh->nlmsg_seq, 0,
1463                                                NLM_F_MULTI,
1464                                                ext_filter_mask);
1465
1466                         if (err < 0) {
1467                                 if (likely(skb->len))
1468                                         goto out;
1469
1470                                 goto out_err;
1471                         }
1472
1473                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1474 cont:
1475                         idx++;
1476                 }
1477         }
1478 out:
1479         err = skb->len;
1480 out_err:
1481         cb->args[1] = idx;
1482         cb->args[0] = h;
1483
1484         return err;
1485 }
1486
1487 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1488 {
1489         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1490 }
1491 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1492
1493 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1494 {
1495         struct net *net;
1496         /* Examine the link attributes and figure out which
1497          * network namespace we are talking about.
1498          */
1499         if (tb[IFLA_NET_NS_PID])
1500                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1501         else if (tb[IFLA_NET_NS_FD])
1502                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1503         else
1504                 net = get_net(src_net);
1505         return net;
1506 }
1507 EXPORT_SYMBOL(rtnl_link_get_net);
1508
1509 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1510 {
1511         if (dev) {
1512                 if (tb[IFLA_ADDRESS] &&
1513                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1514                         return -EINVAL;
1515
1516                 if (tb[IFLA_BROADCAST] &&
1517                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1518                         return -EINVAL;
1519         }
1520
1521         if (tb[IFLA_AF_SPEC]) {
1522                 struct nlattr *af;
1523                 int rem, err;
1524
1525                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1526                         const struct rtnl_af_ops *af_ops;
1527
1528                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1529                                 return -EAFNOSUPPORT;
1530
1531                         if (!af_ops->set_link_af)
1532                                 return -EOPNOTSUPP;
1533
1534                         if (af_ops->validate_link_af) {
1535                                 err = af_ops->validate_link_af(dev, af);
1536                                 if (err < 0)
1537                                         return err;
1538                         }
1539                 }
1540         }
1541
1542         return 0;
1543 }
1544
1545 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1546 {
1547         const struct net_device_ops *ops = dev->netdev_ops;
1548         int err = -EINVAL;
1549
1550         if (tb[IFLA_VF_MAC]) {
1551                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1552
1553                 if (ivm->vf >= INT_MAX)
1554                         return -EINVAL;
1555                 err = -EOPNOTSUPP;
1556                 if (ops->ndo_set_vf_mac)
1557                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1558                                                   ivm->mac);
1559                 if (err < 0)
1560                         return err;
1561         }
1562
1563         if (tb[IFLA_VF_VLAN]) {
1564                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1565
1566                 if (ivv->vf >= INT_MAX)
1567                         return -EINVAL;
1568                 err = -EOPNOTSUPP;
1569                 if (ops->ndo_set_vf_vlan)
1570                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1571                                                    ivv->qos);
1572                 if (err < 0)
1573                         return err;
1574         }
1575
1576         if (tb[IFLA_VF_TX_RATE]) {
1577                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1578                 struct ifla_vf_info ivf;
1579
1580                 if (ivt->vf >= INT_MAX)
1581                         return -EINVAL;
1582                 err = -EOPNOTSUPP;
1583                 if (ops->ndo_get_vf_config)
1584                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1585                 if (err < 0)
1586                         return err;
1587
1588                 err = -EOPNOTSUPP;
1589                 if (ops->ndo_set_vf_rate)
1590                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1591                                                    ivf.min_tx_rate,
1592                                                    ivt->rate);
1593                 if (err < 0)
1594                         return err;
1595         }
1596
1597         if (tb[IFLA_VF_RATE]) {
1598                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1599
1600                 if (ivt->vf >= INT_MAX)
1601                         return -EINVAL;
1602                 err = -EOPNOTSUPP;
1603                 if (ops->ndo_set_vf_rate)
1604                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1605                                                    ivt->min_tx_rate,
1606                                                    ivt->max_tx_rate);
1607                 if (err < 0)
1608                         return err;
1609         }
1610
1611         if (tb[IFLA_VF_SPOOFCHK]) {
1612                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1613
1614                 if (ivs->vf >= INT_MAX)
1615                         return -EINVAL;
1616                 err = -EOPNOTSUPP;
1617                 if (ops->ndo_set_vf_spoofchk)
1618                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1619                                                        ivs->setting);
1620                 if (err < 0)
1621                         return err;
1622         }
1623
1624         if (tb[IFLA_VF_LINK_STATE]) {
1625                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1626
1627                 if (ivl->vf >= INT_MAX)
1628                         return -EINVAL;
1629                 err = -EOPNOTSUPP;
1630                 if (ops->ndo_set_vf_link_state)
1631                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1632                                                          ivl->link_state);
1633                 if (err < 0)
1634                         return err;
1635         }
1636
1637         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1638                 struct ifla_vf_rss_query_en *ivrssq_en;
1639
1640                 err = -EOPNOTSUPP;
1641                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1642                 if (ivrssq_en->vf >= INT_MAX)
1643                         return -EINVAL;
1644                 if (ops->ndo_set_vf_rss_query_en)
1645                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1646                                                            ivrssq_en->setting);
1647                 if (err < 0)
1648                         return err;
1649         }
1650
1651         if (tb[IFLA_VF_TRUST]) {
1652                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1653
1654                 if (ivt->vf >= INT_MAX)
1655                         return -EINVAL;
1656                 err = -EOPNOTSUPP;
1657                 if (ops->ndo_set_vf_trust)
1658                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1659                 if (err < 0)
1660                         return err;
1661         }
1662
1663         return err;
1664 }
1665
1666 static int do_set_master(struct net_device *dev, int ifindex)
1667 {
1668         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1669         const struct net_device_ops *ops;
1670         int err;
1671
1672         if (upper_dev) {
1673                 if (upper_dev->ifindex == ifindex)
1674                         return 0;
1675                 ops = upper_dev->netdev_ops;
1676                 if (ops->ndo_del_slave) {
1677                         err = ops->ndo_del_slave(upper_dev, dev);
1678                         if (err)
1679                                 return err;
1680                 } else {
1681                         return -EOPNOTSUPP;
1682                 }
1683         }
1684
1685         if (ifindex) {
1686                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1687                 if (!upper_dev)
1688                         return -EINVAL;
1689                 ops = upper_dev->netdev_ops;
1690                 if (ops->ndo_add_slave) {
1691                         err = ops->ndo_add_slave(upper_dev, dev);
1692                         if (err)
1693                                 return err;
1694                 } else {
1695                         return -EOPNOTSUPP;
1696                 }
1697         }
1698         return 0;
1699 }
1700
1701 #define DO_SETLINK_MODIFIED     0x01
1702 /* notify flag means notify + modified. */
1703 #define DO_SETLINK_NOTIFY       0x03
1704 static int do_setlink(const struct sk_buff *skb,
1705                       struct net_device *dev, struct ifinfomsg *ifm,
1706                       struct nlattr **tb, char *ifname, int status)
1707 {
1708         const struct net_device_ops *ops = dev->netdev_ops;
1709         int err;
1710
1711         err = validate_linkmsg(dev, tb);
1712         if (err < 0)
1713                 return err;
1714
1715         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1716                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1717                 if (IS_ERR(net)) {
1718                         err = PTR_ERR(net);
1719                         goto errout;
1720                 }
1721                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1722                         put_net(net);
1723                         err = -EPERM;
1724                         goto errout;
1725                 }
1726                 err = dev_change_net_namespace(dev, net, ifname);
1727                 put_net(net);
1728                 if (err)
1729                         goto errout;
1730                 status |= DO_SETLINK_MODIFIED;
1731         }
1732
1733         if (tb[IFLA_MAP]) {
1734                 struct rtnl_link_ifmap *u_map;
1735                 struct ifmap k_map;
1736
1737                 if (!ops->ndo_set_config) {
1738                         err = -EOPNOTSUPP;
1739                         goto errout;
1740                 }
1741
1742                 if (!netif_device_present(dev)) {
1743                         err = -ENODEV;
1744                         goto errout;
1745                 }
1746
1747                 u_map = nla_data(tb[IFLA_MAP]);
1748                 k_map.mem_start = (unsigned long) u_map->mem_start;
1749                 k_map.mem_end = (unsigned long) u_map->mem_end;
1750                 k_map.base_addr = (unsigned short) u_map->base_addr;
1751                 k_map.irq = (unsigned char) u_map->irq;
1752                 k_map.dma = (unsigned char) u_map->dma;
1753                 k_map.port = (unsigned char) u_map->port;
1754
1755                 err = ops->ndo_set_config(dev, &k_map);
1756                 if (err < 0)
1757                         goto errout;
1758
1759                 status |= DO_SETLINK_NOTIFY;
1760         }
1761
1762         if (tb[IFLA_ADDRESS]) {
1763                 struct sockaddr *sa;
1764                 int len;
1765
1766                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
1767                                                   sizeof(*sa));
1768                 sa = kmalloc(len, GFP_KERNEL);
1769                 if (!sa) {
1770                         err = -ENOMEM;
1771                         goto errout;
1772                 }
1773                 sa->sa_family = dev->type;
1774                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1775                        dev->addr_len);
1776                 err = dev_set_mac_address(dev, sa);
1777                 kfree(sa);
1778                 if (err)
1779                         goto errout;
1780                 status |= DO_SETLINK_MODIFIED;
1781         }
1782
1783         if (tb[IFLA_MTU]) {
1784                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1785                 if (err < 0)
1786                         goto errout;
1787                 status |= DO_SETLINK_MODIFIED;
1788         }
1789
1790         if (tb[IFLA_GROUP]) {
1791                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1792                 status |= DO_SETLINK_NOTIFY;
1793         }
1794
1795         /*
1796          * Interface selected by interface index but interface
1797          * name provided implies that a name change has been
1798          * requested.
1799          */
1800         if (ifm->ifi_index > 0 && ifname[0]) {
1801                 err = dev_change_name(dev, ifname);
1802                 if (err < 0)
1803                         goto errout;
1804                 status |= DO_SETLINK_MODIFIED;
1805         }
1806
1807         if (tb[IFLA_IFALIAS]) {
1808                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1809                                     nla_len(tb[IFLA_IFALIAS]));
1810                 if (err < 0)
1811                         goto errout;
1812                 status |= DO_SETLINK_NOTIFY;
1813         }
1814
1815         if (tb[IFLA_BROADCAST]) {
1816                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1817                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1818         }
1819
1820         if (ifm->ifi_flags || ifm->ifi_change) {
1821                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1822                 if (err < 0)
1823                         goto errout;
1824         }
1825
1826         if (tb[IFLA_MASTER]) {
1827                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1828                 if (err)
1829                         goto errout;
1830                 status |= DO_SETLINK_MODIFIED;
1831         }
1832
1833         if (tb[IFLA_CARRIER]) {
1834                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1835                 if (err)
1836                         goto errout;
1837                 status |= DO_SETLINK_MODIFIED;
1838         }
1839
1840         if (tb[IFLA_TXQLEN]) {
1841                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1842
1843                 if (dev->tx_queue_len ^ value)
1844                         status |= DO_SETLINK_NOTIFY;
1845
1846                 dev->tx_queue_len = value;
1847         }
1848
1849         if (tb[IFLA_OPERSTATE])
1850                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1851
1852         if (tb[IFLA_LINKMODE]) {
1853                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1854
1855                 write_lock_bh(&dev_base_lock);
1856                 if (dev->link_mode ^ value)
1857                         status |= DO_SETLINK_NOTIFY;
1858                 dev->link_mode = value;
1859                 write_unlock_bh(&dev_base_lock);
1860         }
1861
1862         if (tb[IFLA_VFINFO_LIST]) {
1863                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1864                 struct nlattr *attr;
1865                 int rem;
1866
1867                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1868                         if (nla_type(attr) != IFLA_VF_INFO ||
1869                             nla_len(attr) < NLA_HDRLEN) {
1870                                 err = -EINVAL;
1871                                 goto errout;
1872                         }
1873                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1874                                                ifla_vf_policy);
1875                         if (err < 0)
1876                                 goto errout;
1877                         err = do_setvfinfo(dev, vfinfo);
1878                         if (err < 0)
1879                                 goto errout;
1880                         status |= DO_SETLINK_NOTIFY;
1881                 }
1882         }
1883         err = 0;
1884
1885         if (tb[IFLA_VF_PORTS]) {
1886                 struct nlattr *port[IFLA_PORT_MAX+1];
1887                 struct nlattr *attr;
1888                 int vf;
1889                 int rem;
1890
1891                 err = -EOPNOTSUPP;
1892                 if (!ops->ndo_set_vf_port)
1893                         goto errout;
1894
1895                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1896                         if (nla_type(attr) != IFLA_VF_PORT ||
1897                             nla_len(attr) < NLA_HDRLEN) {
1898                                 err = -EINVAL;
1899                                 goto errout;
1900                         }
1901                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1902                                                ifla_port_policy);
1903                         if (err < 0)
1904                                 goto errout;
1905                         if (!port[IFLA_PORT_VF]) {
1906                                 err = -EOPNOTSUPP;
1907                                 goto errout;
1908                         }
1909                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1910                         err = ops->ndo_set_vf_port(dev, vf, port);
1911                         if (err < 0)
1912                                 goto errout;
1913                         status |= DO_SETLINK_NOTIFY;
1914                 }
1915         }
1916         err = 0;
1917
1918         if (tb[IFLA_PORT_SELF]) {
1919                 struct nlattr *port[IFLA_PORT_MAX+1];
1920
1921                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1922                         tb[IFLA_PORT_SELF], ifla_port_policy);
1923                 if (err < 0)
1924                         goto errout;
1925
1926                 err = -EOPNOTSUPP;
1927                 if (ops->ndo_set_vf_port)
1928                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1929                 if (err < 0)
1930                         goto errout;
1931                 status |= DO_SETLINK_NOTIFY;
1932         }
1933
1934         if (tb[IFLA_AF_SPEC]) {
1935                 struct nlattr *af;
1936                 int rem;
1937
1938                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1939                         const struct rtnl_af_ops *af_ops;
1940
1941                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1942                                 BUG();
1943
1944                         err = af_ops->set_link_af(dev, af);
1945                         if (err < 0)
1946                                 goto errout;
1947
1948                         status |= DO_SETLINK_NOTIFY;
1949                 }
1950         }
1951         err = 0;
1952
1953         if (tb[IFLA_PROTO_DOWN]) {
1954                 err = dev_change_proto_down(dev,
1955                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
1956                 if (err)
1957                         goto errout;
1958                 status |= DO_SETLINK_NOTIFY;
1959         }
1960
1961 errout:
1962         if (status & DO_SETLINK_MODIFIED) {
1963                 if (status & DO_SETLINK_NOTIFY)
1964                         netdev_state_change(dev);
1965
1966                 if (err < 0)
1967                         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",
1968                                              dev->name);
1969         }
1970
1971         return err;
1972 }
1973
1974 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1975 {
1976         struct net *net = sock_net(skb->sk);
1977         struct ifinfomsg *ifm;
1978         struct net_device *dev;
1979         int err;
1980         struct nlattr *tb[IFLA_MAX+1];
1981         char ifname[IFNAMSIZ];
1982
1983         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1984         if (err < 0)
1985                 goto errout;
1986
1987         if (tb[IFLA_IFNAME])
1988                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1989         else
1990                 ifname[0] = '\0';
1991
1992         err = -EINVAL;
1993         ifm = nlmsg_data(nlh);
1994         if (ifm->ifi_index > 0)
1995                 dev = __dev_get_by_index(net, ifm->ifi_index);
1996         else if (tb[IFLA_IFNAME])
1997                 dev = __dev_get_by_name(net, ifname);
1998         else
1999                 goto errout;
2000
2001         if (dev == NULL) {
2002                 err = -ENODEV;
2003                 goto errout;
2004         }
2005
2006         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2007 errout:
2008         return err;
2009 }
2010
2011 static int rtnl_group_dellink(const struct net *net, int group)
2012 {
2013         struct net_device *dev, *aux;
2014         LIST_HEAD(list_kill);
2015         bool found = false;
2016
2017         if (!group)
2018                 return -EPERM;
2019
2020         for_each_netdev(net, dev) {
2021                 if (dev->group == group) {
2022                         const struct rtnl_link_ops *ops;
2023
2024                         found = true;
2025                         ops = dev->rtnl_link_ops;
2026                         if (!ops || !ops->dellink)
2027                                 return -EOPNOTSUPP;
2028                 }
2029         }
2030
2031         if (!found)
2032                 return -ENODEV;
2033
2034         for_each_netdev_safe(net, dev, aux) {
2035                 if (dev->group == group) {
2036                         const struct rtnl_link_ops *ops;
2037
2038                         ops = dev->rtnl_link_ops;
2039                         ops->dellink(dev, &list_kill);
2040                 }
2041         }
2042         unregister_netdevice_many(&list_kill);
2043
2044         return 0;
2045 }
2046
2047 int rtnl_delete_link(struct net_device *dev)
2048 {
2049         const struct rtnl_link_ops *ops;
2050         LIST_HEAD(list_kill);
2051
2052         ops = dev->rtnl_link_ops;
2053         if (!ops || !ops->dellink)
2054                 return -EOPNOTSUPP;
2055
2056         ops->dellink(dev, &list_kill);
2057         unregister_netdevice_many(&list_kill);
2058
2059         return 0;
2060 }
2061 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2062
2063 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2064 {
2065         struct net *net = sock_net(skb->sk);
2066         struct net_device *dev;
2067         struct ifinfomsg *ifm;
2068         char ifname[IFNAMSIZ];
2069         struct nlattr *tb[IFLA_MAX+1];
2070         int err;
2071
2072         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2073         if (err < 0)
2074                 return err;
2075
2076         if (tb[IFLA_IFNAME])
2077                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2078
2079         ifm = nlmsg_data(nlh);
2080         if (ifm->ifi_index > 0)
2081                 dev = __dev_get_by_index(net, ifm->ifi_index);
2082         else if (tb[IFLA_IFNAME])
2083                 dev = __dev_get_by_name(net, ifname);
2084         else if (tb[IFLA_GROUP])
2085                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2086         else
2087                 return -EINVAL;
2088
2089         if (!dev)
2090                 return -ENODEV;
2091
2092         return rtnl_delete_link(dev);
2093 }
2094
2095 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2096 {
2097         unsigned int old_flags;
2098         int err;
2099
2100         old_flags = dev->flags;
2101         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2102                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2103                 if (err < 0)
2104                         return err;
2105         }
2106
2107         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2108                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2109         } else {
2110                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2111                 __dev_notify_flags(dev, old_flags, ~0U);
2112         }
2113         return 0;
2114 }
2115 EXPORT_SYMBOL(rtnl_configure_link);
2116
2117 struct net_device *rtnl_create_link(struct net *net,
2118         const char *ifname, unsigned char name_assign_type,
2119         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2120 {
2121         int err;
2122         struct net_device *dev;
2123         unsigned int num_tx_queues = 1;
2124         unsigned int num_rx_queues = 1;
2125
2126         if (tb[IFLA_NUM_TX_QUEUES])
2127                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2128         else if (ops->get_num_tx_queues)
2129                 num_tx_queues = ops->get_num_tx_queues();
2130
2131         if (tb[IFLA_NUM_RX_QUEUES])
2132                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2133         else if (ops->get_num_rx_queues)
2134                 num_rx_queues = ops->get_num_rx_queues();
2135
2136         if (num_tx_queues < 1 || num_tx_queues > 4096)
2137                 return ERR_PTR(-EINVAL);
2138
2139         if (num_rx_queues < 1 || num_rx_queues > 4096)
2140                 return ERR_PTR(-EINVAL);
2141
2142         err = -ENOMEM;
2143         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2144                                ops->setup, num_tx_queues, num_rx_queues);
2145         if (!dev)
2146                 goto err;
2147
2148         dev_net_set(dev, net);
2149         dev->rtnl_link_ops = ops;
2150         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2151
2152         if (tb[IFLA_MTU])
2153                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2154         if (tb[IFLA_ADDRESS]) {
2155                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2156                                 nla_len(tb[IFLA_ADDRESS]));
2157                 dev->addr_assign_type = NET_ADDR_SET;
2158         }
2159         if (tb[IFLA_BROADCAST])
2160                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2161                                 nla_len(tb[IFLA_BROADCAST]));
2162         if (tb[IFLA_TXQLEN])
2163                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2164         if (tb[IFLA_OPERSTATE])
2165                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2166         if (tb[IFLA_LINKMODE])
2167                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2168         if (tb[IFLA_GROUP])
2169                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2170
2171         return dev;
2172
2173 err:
2174         return ERR_PTR(err);
2175 }
2176 EXPORT_SYMBOL(rtnl_create_link);
2177
2178 static int rtnl_group_changelink(const struct sk_buff *skb,
2179                 struct net *net, int group,
2180                 struct ifinfomsg *ifm,
2181                 struct nlattr **tb)
2182 {
2183         struct net_device *dev, *aux;
2184         int err;
2185
2186         for_each_netdev_safe(net, dev, aux) {
2187                 if (dev->group == group) {
2188                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2189                         if (err < 0)
2190                                 return err;
2191                 }
2192         }
2193
2194         return 0;
2195 }
2196
2197 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2198 {
2199         struct net *net = sock_net(skb->sk);
2200         const struct rtnl_link_ops *ops;
2201         const struct rtnl_link_ops *m_ops = NULL;
2202         struct net_device *dev;
2203         struct net_device *master_dev = NULL;
2204         struct ifinfomsg *ifm;
2205         char kind[MODULE_NAME_LEN];
2206         char ifname[IFNAMSIZ];
2207         struct nlattr *tb[IFLA_MAX+1];
2208         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2209         unsigned char name_assign_type = NET_NAME_USER;
2210         int err;
2211
2212 #ifdef CONFIG_MODULES
2213 replay:
2214 #endif
2215         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2216         if (err < 0)
2217                 return err;
2218
2219         if (tb[IFLA_IFNAME])
2220                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2221         else
2222                 ifname[0] = '\0';
2223
2224         ifm = nlmsg_data(nlh);
2225         if (ifm->ifi_index > 0)
2226                 dev = __dev_get_by_index(net, ifm->ifi_index);
2227         else {
2228                 if (ifname[0])
2229                         dev = __dev_get_by_name(net, ifname);
2230                 else
2231                         dev = NULL;
2232         }
2233
2234         if (dev) {
2235                 master_dev = netdev_master_upper_dev_get(dev);
2236                 if (master_dev)
2237                         m_ops = master_dev->rtnl_link_ops;
2238         }
2239
2240         err = validate_linkmsg(dev, tb);
2241         if (err < 0)
2242                 return err;
2243
2244         if (tb[IFLA_LINKINFO]) {
2245                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2246                                        tb[IFLA_LINKINFO], ifla_info_policy);
2247                 if (err < 0)
2248                         return err;
2249         } else
2250                 memset(linkinfo, 0, sizeof(linkinfo));
2251
2252         if (linkinfo[IFLA_INFO_KIND]) {
2253                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2254                 ops = rtnl_link_ops_get(kind);
2255         } else {
2256                 kind[0] = '\0';
2257                 ops = NULL;
2258         }
2259
2260         if (1) {
2261                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2262                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2263                 struct nlattr **data = NULL;
2264                 struct nlattr **slave_data = NULL;
2265                 struct net *dest_net, *link_net = NULL;
2266
2267                 if (ops) {
2268                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2269                                 err = nla_parse_nested(attr, ops->maxtype,
2270                                                        linkinfo[IFLA_INFO_DATA],
2271                                                        ops->policy);
2272                                 if (err < 0)
2273                                         return err;
2274                                 data = attr;
2275                         }
2276                         if (ops->validate) {
2277                                 err = ops->validate(tb, data);
2278                                 if (err < 0)
2279                                         return err;
2280                         }
2281                 }
2282
2283                 if (m_ops) {
2284                         if (m_ops->slave_maxtype &&
2285                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2286                                 err = nla_parse_nested(slave_attr,
2287                                                        m_ops->slave_maxtype,
2288                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2289                                                        m_ops->slave_policy);
2290                                 if (err < 0)
2291                                         return err;
2292                                 slave_data = slave_attr;
2293                         }
2294                         if (m_ops->slave_validate) {
2295                                 err = m_ops->slave_validate(tb, slave_data);
2296                                 if (err < 0)
2297                                         return err;
2298                         }
2299                 }
2300
2301                 if (dev) {
2302                         int status = 0;
2303
2304                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2305                                 return -EEXIST;
2306                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2307                                 return -EOPNOTSUPP;
2308
2309                         if (linkinfo[IFLA_INFO_DATA]) {
2310                                 if (!ops || ops != dev->rtnl_link_ops ||
2311                                     !ops->changelink)
2312                                         return -EOPNOTSUPP;
2313
2314                                 err = ops->changelink(dev, tb, data);
2315                                 if (err < 0)
2316                                         return err;
2317                                 status |= DO_SETLINK_NOTIFY;
2318                         }
2319
2320                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2321                                 if (!m_ops || !m_ops->slave_changelink)
2322                                         return -EOPNOTSUPP;
2323
2324                                 err = m_ops->slave_changelink(master_dev, dev,
2325                                                               tb, slave_data);
2326                                 if (err < 0)
2327                                         return err;
2328                                 status |= DO_SETLINK_NOTIFY;
2329                         }
2330
2331                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2332                 }
2333
2334                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2335                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2336                                 return rtnl_group_changelink(skb, net,
2337                                                 nla_get_u32(tb[IFLA_GROUP]),
2338                                                 ifm, tb);
2339                         return -ENODEV;
2340                 }
2341
2342                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2343                         return -EOPNOTSUPP;
2344
2345                 if (!ops) {
2346 #ifdef CONFIG_MODULES
2347                         if (kind[0]) {
2348                                 __rtnl_unlock();
2349                                 request_module("rtnl-link-%s", kind);
2350                                 rtnl_lock();
2351                                 ops = rtnl_link_ops_get(kind);
2352                                 if (ops)
2353                                         goto replay;
2354                         }
2355 #endif
2356                         return -EOPNOTSUPP;
2357                 }
2358
2359                 if (!ops->setup)
2360                         return -EOPNOTSUPP;
2361
2362                 if (!ifname[0]) {
2363                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2364                         name_assign_type = NET_NAME_ENUM;
2365                 }
2366
2367                 dest_net = rtnl_link_get_net(net, tb);
2368                 if (IS_ERR(dest_net))
2369                         return PTR_ERR(dest_net);
2370
2371                 err = -EPERM;
2372                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2373                         goto out;
2374
2375                 if (tb[IFLA_LINK_NETNSID]) {
2376                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2377
2378                         link_net = get_net_ns_by_id(dest_net, id);
2379                         if (!link_net) {
2380                                 err =  -EINVAL;
2381                                 goto out;
2382                         }
2383                         err = -EPERM;
2384                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2385                                 goto out;
2386                 }
2387
2388                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2389                                        name_assign_type, ops, tb);
2390                 if (IS_ERR(dev)) {
2391                         err = PTR_ERR(dev);
2392                         goto out;
2393                 }
2394
2395                 dev->ifindex = ifm->ifi_index;
2396
2397                 if (ops->newlink) {
2398                         err = ops->newlink(link_net ? : net, dev, tb, data);
2399                         /* Drivers should call free_netdev() in ->destructor
2400                          * and unregister it on failure after registration
2401                          * so that device could be finally freed in rtnl_unlock.
2402                          */
2403                         if (err < 0) {
2404                                 /* If device is not registered at all, free it now */
2405                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2406                                         free_netdev(dev);
2407                                 goto out;
2408                         }
2409                 } else {
2410                         err = register_netdevice(dev);
2411                         if (err < 0) {
2412                                 free_netdev(dev);
2413                                 goto out;
2414                         }
2415                 }
2416                 err = rtnl_configure_link(dev, ifm);
2417                 if (err < 0)
2418                         goto out_unregister;
2419                 if (link_net) {
2420                         err = dev_change_net_namespace(dev, dest_net, ifname);
2421                         if (err < 0)
2422                                 goto out_unregister;
2423                 }
2424 out:
2425                 if (link_net)
2426                         put_net(link_net);
2427                 put_net(dest_net);
2428                 return err;
2429 out_unregister:
2430                 if (ops->newlink) {
2431                         LIST_HEAD(list_kill);
2432
2433                         ops->dellink(dev, &list_kill);
2434                         unregister_netdevice_many(&list_kill);
2435                 } else {
2436                         unregister_netdevice(dev);
2437                 }
2438                 goto out;
2439         }
2440 }
2441
2442 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2443 {
2444         struct net *net = sock_net(skb->sk);
2445         struct ifinfomsg *ifm;
2446         char ifname[IFNAMSIZ];
2447         struct nlattr *tb[IFLA_MAX+1];
2448         struct net_device *dev = NULL;
2449         struct sk_buff *nskb;
2450         int err;
2451         u32 ext_filter_mask = 0;
2452
2453         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2454         if (err < 0)
2455                 return err;
2456
2457         if (tb[IFLA_IFNAME])
2458                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2459
2460         if (tb[IFLA_EXT_MASK])
2461                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2462
2463         ifm = nlmsg_data(nlh);
2464         if (ifm->ifi_index > 0)
2465                 dev = __dev_get_by_index(net, ifm->ifi_index);
2466         else if (tb[IFLA_IFNAME])
2467                 dev = __dev_get_by_name(net, ifname);
2468         else
2469                 return -EINVAL;
2470
2471         if (dev == NULL)
2472                 return -ENODEV;
2473
2474         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2475         if (nskb == NULL)
2476                 return -ENOBUFS;
2477
2478         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2479                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2480         if (err < 0) {
2481                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2482                 WARN_ON(err == -EMSGSIZE);
2483                 kfree_skb(nskb);
2484         } else
2485                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2486
2487         return err;
2488 }
2489
2490 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2491 {
2492         struct net *net = sock_net(skb->sk);
2493         struct net_device *dev;
2494         struct nlattr *tb[IFLA_MAX+1];
2495         u32 ext_filter_mask = 0;
2496         u16 min_ifinfo_dump_size = 0;
2497         int hdrlen;
2498
2499         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2500         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2501                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2502
2503         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2504                 if (tb[IFLA_EXT_MASK])
2505                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2506         }
2507
2508         if (!ext_filter_mask)
2509                 return NLMSG_GOODSIZE;
2510         /*
2511          * traverse the list of net devices and compute the minimum
2512          * buffer size based upon the filter mask.
2513          */
2514         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2515                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2516                                              if_nlmsg_size(dev,
2517                                                            ext_filter_mask));
2518         }
2519
2520         return min_ifinfo_dump_size;
2521 }
2522
2523 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2524 {
2525         int idx;
2526         int s_idx = cb->family;
2527
2528         if (s_idx == 0)
2529                 s_idx = 1;
2530         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2531                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2532                 if (idx < s_idx || idx == PF_PACKET)
2533                         continue;
2534                 if (rtnl_msg_handlers[idx] == NULL ||
2535                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2536                         continue;
2537                 if (idx > s_idx) {
2538                         memset(&cb->args[0], 0, sizeof(cb->args));
2539                         cb->prev_seq = 0;
2540                         cb->seq = 0;
2541                 }
2542                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2543                         break;
2544         }
2545         cb->family = idx;
2546
2547         return skb->len;
2548 }
2549
2550 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2551                                        unsigned int change, gfp_t flags)
2552 {
2553         struct net *net = dev_net(dev);
2554         struct sk_buff *skb;
2555         int err = -ENOBUFS;
2556         size_t if_info_size;
2557
2558         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2559         if (skb == NULL)
2560                 goto errout;
2561
2562         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2563         if (err < 0) {
2564                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2565                 WARN_ON(err == -EMSGSIZE);
2566                 kfree_skb(skb);
2567                 goto errout;
2568         }
2569         return skb;
2570 errout:
2571         if (err < 0)
2572                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2573         return NULL;
2574 }
2575
2576 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2577 {
2578         struct net *net = dev_net(dev);
2579
2580         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2581 }
2582
2583 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2584                   gfp_t flags)
2585 {
2586         struct sk_buff *skb;
2587
2588         if (dev->reg_state != NETREG_REGISTERED)
2589                 return;
2590
2591         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2592         if (skb)
2593                 rtmsg_ifinfo_send(skb, dev, flags);
2594 }
2595 EXPORT_SYMBOL(rtmsg_ifinfo);
2596
2597 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2598                                    struct net_device *dev,
2599                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2600                                    int type, unsigned int flags,
2601                                    int nlflags)
2602 {
2603         struct nlmsghdr *nlh;
2604         struct ndmsg *ndm;
2605
2606         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2607         if (!nlh)
2608                 return -EMSGSIZE;
2609
2610         ndm = nlmsg_data(nlh);
2611         ndm->ndm_family  = AF_BRIDGE;
2612         ndm->ndm_pad1    = 0;
2613         ndm->ndm_pad2    = 0;
2614         ndm->ndm_flags   = flags;
2615         ndm->ndm_type    = 0;
2616         ndm->ndm_ifindex = dev->ifindex;
2617         ndm->ndm_state   = NUD_PERMANENT;
2618
2619         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2620                 goto nla_put_failure;
2621         if (vid)
2622                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2623                         goto nla_put_failure;
2624
2625         nlmsg_end(skb, nlh);
2626         return 0;
2627
2628 nla_put_failure:
2629         nlmsg_cancel(skb, nlh);
2630         return -EMSGSIZE;
2631 }
2632
2633 static inline size_t rtnl_fdb_nlmsg_size(void)
2634 {
2635         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2636                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2637                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2638                0;
2639 }
2640
2641 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2642 {
2643         struct net *net = dev_net(dev);
2644         struct sk_buff *skb;
2645         int err = -ENOBUFS;
2646
2647         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2648         if (!skb)
2649                 goto errout;
2650
2651         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2652                                       0, 0, type, NTF_SELF, 0);
2653         if (err < 0) {
2654                 kfree_skb(skb);
2655                 goto errout;
2656         }
2657
2658         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2659         return;
2660 errout:
2661         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2662 }
2663
2664 /**
2665  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2666  */
2667 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2668                      struct nlattr *tb[],
2669                      struct net_device *dev,
2670                      const unsigned char *addr, u16 vid,
2671                      u16 flags)
2672 {
2673         int err = -EINVAL;
2674
2675         /* If aging addresses are supported device will need to
2676          * implement its own handler for this.
2677          */
2678         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2679                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2680                 return err;
2681         }
2682
2683         if (vid) {
2684                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2685                 return err;
2686         }
2687
2688         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2689                 err = dev_uc_add_excl(dev, addr);
2690         else if (is_multicast_ether_addr(addr))
2691                 err = dev_mc_add_excl(dev, addr);
2692
2693         /* Only return duplicate errors if NLM_F_EXCL is set */
2694         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2695                 err = 0;
2696
2697         return err;
2698 }
2699 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2700
2701 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2702 {
2703         u16 vid = 0;
2704
2705         if (vlan_attr) {
2706                 if (nla_len(vlan_attr) != sizeof(u16)) {
2707                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2708                         return -EINVAL;
2709                 }
2710
2711                 vid = nla_get_u16(vlan_attr);
2712
2713                 if (!vid || vid >= VLAN_VID_MASK) {
2714                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2715                                 vid);
2716                         return -EINVAL;
2717                 }
2718         }
2719         *p_vid = vid;
2720         return 0;
2721 }
2722
2723 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2724 {
2725         struct net *net = sock_net(skb->sk);
2726         struct ndmsg *ndm;
2727         struct nlattr *tb[NDA_MAX+1];
2728         struct net_device *dev;
2729         u8 *addr;
2730         u16 vid;
2731         int err;
2732
2733         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2734         if (err < 0)
2735                 return err;
2736
2737         ndm = nlmsg_data(nlh);
2738         if (ndm->ndm_ifindex == 0) {
2739                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2740                 return -EINVAL;
2741         }
2742
2743         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2744         if (dev == NULL) {
2745                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2746                 return -ENODEV;
2747         }
2748
2749         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2750                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2751                 return -EINVAL;
2752         }
2753
2754         if (dev->type != ARPHRD_ETHER) {
2755                 pr_info("PF_BRIDGE: FDB add only supported for Ethernet devices");
2756                 return -EINVAL;
2757         }
2758
2759         addr = nla_data(tb[NDA_LLADDR]);
2760
2761         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2762         if (err)
2763                 return err;
2764
2765         err = -EOPNOTSUPP;
2766
2767         /* Support fdb on master device the net/bridge default case */
2768         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2769             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2770                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2771                 const struct net_device_ops *ops = br_dev->netdev_ops;
2772
2773                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2774                                        nlh->nlmsg_flags);
2775                 if (err)
2776                         goto out;
2777                 else
2778                         ndm->ndm_flags &= ~NTF_MASTER;
2779         }
2780
2781         /* Embedded bridge, macvlan, and any other device support */
2782         if ((ndm->ndm_flags & NTF_SELF)) {
2783                 if (dev->netdev_ops->ndo_fdb_add)
2784                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2785                                                            vid,
2786                                                            nlh->nlmsg_flags);
2787                 else
2788                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2789                                                nlh->nlmsg_flags);
2790
2791                 if (!err) {
2792                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2793                         ndm->ndm_flags &= ~NTF_SELF;
2794                 }
2795         }
2796 out:
2797         return err;
2798 }
2799
2800 /**
2801  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2802  */
2803 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2804                      struct nlattr *tb[],
2805                      struct net_device *dev,
2806                      const unsigned char *addr, u16 vid)
2807 {
2808         int err = -EINVAL;
2809
2810         /* If aging addresses are supported device will need to
2811          * implement its own handler for this.
2812          */
2813         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2814                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2815                 return err;
2816         }
2817
2818         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2819                 err = dev_uc_del(dev, addr);
2820         else if (is_multicast_ether_addr(addr))
2821                 err = dev_mc_del(dev, addr);
2822
2823         return err;
2824 }
2825 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2826
2827 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2828 {
2829         struct net *net = sock_net(skb->sk);
2830         struct ndmsg *ndm;
2831         struct nlattr *tb[NDA_MAX+1];
2832         struct net_device *dev;
2833         int err = -EINVAL;
2834         __u8 *addr;
2835         u16 vid;
2836
2837         if (!netlink_capable(skb, CAP_NET_ADMIN))
2838                 return -EPERM;
2839
2840         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2841         if (err < 0)
2842                 return err;
2843
2844         ndm = nlmsg_data(nlh);
2845         if (ndm->ndm_ifindex == 0) {
2846                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2847                 return -EINVAL;
2848         }
2849
2850         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2851         if (dev == NULL) {
2852                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2853                 return -ENODEV;
2854         }
2855
2856         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2857                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2858                 return -EINVAL;
2859         }
2860
2861         if (dev->type != ARPHRD_ETHER) {
2862                 pr_info("PF_BRIDGE: FDB delete only supported for Ethernet devices");
2863                 return -EINVAL;
2864         }
2865
2866         addr = nla_data(tb[NDA_LLADDR]);
2867
2868         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2869         if (err)
2870                 return err;
2871
2872         err = -EOPNOTSUPP;
2873
2874         /* Support fdb on master device the net/bridge default case */
2875         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2876             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2877                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2878                 const struct net_device_ops *ops = br_dev->netdev_ops;
2879
2880                 if (ops->ndo_fdb_del)
2881                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2882
2883                 if (err)
2884                         goto out;
2885                 else
2886                         ndm->ndm_flags &= ~NTF_MASTER;
2887         }
2888
2889         /* Embedded bridge, macvlan, and any other device support */
2890         if (ndm->ndm_flags & NTF_SELF) {
2891                 if (dev->netdev_ops->ndo_fdb_del)
2892                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2893                                                            vid);
2894                 else
2895                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2896
2897                 if (!err) {
2898                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2899                         ndm->ndm_flags &= ~NTF_SELF;
2900                 }
2901         }
2902 out:
2903         return err;
2904 }
2905
2906 static int nlmsg_populate_fdb(struct sk_buff *skb,
2907                               struct netlink_callback *cb,
2908                               struct net_device *dev,
2909                               int *idx,
2910                               struct netdev_hw_addr_list *list)
2911 {
2912         struct netdev_hw_addr *ha;
2913         int err;
2914         u32 portid, seq;
2915
2916         portid = NETLINK_CB(cb->skb).portid;
2917         seq = cb->nlh->nlmsg_seq;
2918
2919         list_for_each_entry(ha, &list->list, list) {
2920                 if (*idx < cb->args[0])
2921                         goto skip;
2922
2923                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2924                                               portid, seq,
2925                                               RTM_NEWNEIGH, NTF_SELF,
2926                                               NLM_F_MULTI);
2927                 if (err < 0)
2928                         return err;
2929 skip:
2930                 *idx += 1;
2931         }
2932         return 0;
2933 }
2934
2935 /**
2936  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2937  * @nlh: netlink message header
2938  * @dev: netdevice
2939  *
2940  * Default netdevice operation to dump the existing unicast address list.
2941  * Returns number of addresses from list put in skb.
2942  */
2943 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2944                       struct netlink_callback *cb,
2945                       struct net_device *dev,
2946                       struct net_device *filter_dev,
2947                       int idx)
2948 {
2949         int err;
2950
2951         if (dev->type != ARPHRD_ETHER)
2952                 return -EINVAL;
2953
2954         netif_addr_lock_bh(dev);
2955         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2956         if (err)
2957                 goto out;
2958         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2959 out:
2960         netif_addr_unlock_bh(dev);
2961         return idx;
2962 }
2963 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2964
2965 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2966 {
2967         struct net_device *dev;
2968         struct nlattr *tb[IFLA_MAX+1];
2969         struct net_device *br_dev = NULL;
2970         const struct net_device_ops *ops = NULL;
2971         const struct net_device_ops *cops = NULL;
2972         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2973         struct net *net = sock_net(skb->sk);
2974         int brport_idx = 0;
2975         int br_idx = 0;
2976         int idx = 0;
2977
2978         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2979                         ifla_policy) == 0) {
2980                 if (tb[IFLA_MASTER])
2981                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2982         }
2983
2984         brport_idx = ifm->ifi_index;
2985
2986         if (br_idx) {
2987                 br_dev = __dev_get_by_index(net, br_idx);
2988                 if (!br_dev)
2989                         return -ENODEV;
2990
2991                 ops = br_dev->netdev_ops;
2992         }
2993
2994         for_each_netdev(net, dev) {
2995                 if (brport_idx && (dev->ifindex != brport_idx))
2996                         continue;
2997
2998                 if (!br_idx) { /* user did not specify a specific bridge */
2999                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3000                                 br_dev = netdev_master_upper_dev_get(dev);
3001                                 cops = br_dev->netdev_ops;
3002                         }
3003
3004                 } else {
3005                         if (dev != br_dev &&
3006                             !(dev->priv_flags & IFF_BRIDGE_PORT))
3007                                 continue;
3008
3009                         if (br_dev != netdev_master_upper_dev_get(dev) &&
3010                             !(dev->priv_flags & IFF_EBRIDGE))
3011                                 continue;
3012
3013                         cops = ops;
3014                 }
3015
3016                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3017                         if (cops && cops->ndo_fdb_dump)
3018                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3019                                                          idx);
3020                 }
3021
3022                 if (dev->netdev_ops->ndo_fdb_dump)
3023                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3024                                                             idx);
3025                 else
3026                         idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3027
3028                 cops = NULL;
3029         }
3030
3031         cb->args[0] = idx;
3032         return skb->len;
3033 }
3034
3035 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3036                                unsigned int attrnum, unsigned int flag)
3037 {
3038         if (mask & flag)
3039                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3040         return 0;
3041 }
3042
3043 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3044                             struct net_device *dev, u16 mode,
3045                             u32 flags, u32 mask, int nlflags,
3046                             u32 filter_mask,
3047                             int (*vlan_fill)(struct sk_buff *skb,
3048                                              struct net_device *dev,
3049                                              u32 filter_mask))
3050 {
3051         struct nlmsghdr *nlh;
3052         struct ifinfomsg *ifm;
3053         struct nlattr *br_afspec;
3054         struct nlattr *protinfo;
3055         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3056         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3057         int err = 0;
3058
3059         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3060         if (nlh == NULL)
3061                 return -EMSGSIZE;
3062
3063         ifm = nlmsg_data(nlh);
3064         ifm->ifi_family = AF_BRIDGE;
3065         ifm->__ifi_pad = 0;
3066         ifm->ifi_type = dev->type;
3067         ifm->ifi_index = dev->ifindex;
3068         ifm->ifi_flags = dev_get_flags(dev);
3069         ifm->ifi_change = 0;
3070
3071
3072         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3073             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3074             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3075             (br_dev &&
3076              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3077             (dev->addr_len &&
3078              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3079             (dev->ifindex != dev_get_iflink(dev) &&
3080              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3081                 goto nla_put_failure;
3082
3083         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3084         if (!br_afspec)
3085                 goto nla_put_failure;
3086
3087         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3088                 nla_nest_cancel(skb, br_afspec);
3089                 goto nla_put_failure;
3090         }
3091
3092         if (mode != BRIDGE_MODE_UNDEF) {
3093                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3094                         nla_nest_cancel(skb, br_afspec);
3095                         goto nla_put_failure;
3096                 }
3097         }
3098         if (vlan_fill) {
3099                 err = vlan_fill(skb, dev, filter_mask);
3100                 if (err) {
3101                         nla_nest_cancel(skb, br_afspec);
3102                         goto nla_put_failure;
3103                 }
3104         }
3105         nla_nest_end(skb, br_afspec);
3106
3107         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3108         if (!protinfo)
3109                 goto nla_put_failure;
3110
3111         if (brport_nla_put_flag(skb, flags, mask,
3112                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3113             brport_nla_put_flag(skb, flags, mask,
3114                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3115             brport_nla_put_flag(skb, flags, mask,
3116                                 IFLA_BRPORT_FAST_LEAVE,
3117                                 BR_MULTICAST_FAST_LEAVE) ||
3118             brport_nla_put_flag(skb, flags, mask,
3119                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3120             brport_nla_put_flag(skb, flags, mask,
3121                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3122             brport_nla_put_flag(skb, flags, mask,
3123                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3124             brport_nla_put_flag(skb, flags, mask,
3125                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3126             brport_nla_put_flag(skb, flags, mask,
3127                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3128                 nla_nest_cancel(skb, protinfo);
3129                 goto nla_put_failure;
3130         }
3131
3132         nla_nest_end(skb, protinfo);
3133
3134         nlmsg_end(skb, nlh);
3135         return 0;
3136 nla_put_failure:
3137         nlmsg_cancel(skb, nlh);
3138         return err ? err : -EMSGSIZE;
3139 }
3140 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3141
3142 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3143 {
3144         struct net *net = sock_net(skb->sk);
3145         struct net_device *dev;
3146         int idx = 0;
3147         u32 portid = NETLINK_CB(cb->skb).portid;
3148         u32 seq = cb->nlh->nlmsg_seq;
3149         u32 filter_mask = 0;
3150         int err;
3151
3152         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3153                 struct nlattr *extfilt;
3154
3155                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3156                                           IFLA_EXT_MASK);
3157                 if (extfilt) {
3158                         if (nla_len(extfilt) < sizeof(filter_mask))
3159                                 return -EINVAL;
3160
3161                         filter_mask = nla_get_u32(extfilt);
3162                 }
3163         }
3164
3165         rcu_read_lock();
3166         for_each_netdev_rcu(net, dev) {
3167                 const struct net_device_ops *ops = dev->netdev_ops;
3168                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3169
3170                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3171                         if (idx >= cb->args[0]) {
3172                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3173                                                 skb, portid, seq, dev,
3174                                                 filter_mask, NLM_F_MULTI);
3175                                 if (err < 0 && err != -EOPNOTSUPP) {
3176                                         if (likely(skb->len))
3177                                                 break;
3178
3179                                         goto out_err;
3180                                 }
3181                         }
3182                         idx++;
3183                 }
3184
3185                 if (ops->ndo_bridge_getlink) {
3186                         if (idx >= cb->args[0]) {
3187                                 err = ops->ndo_bridge_getlink(skb, portid,
3188                                                               seq, dev,
3189                                                               filter_mask,
3190                                                               NLM_F_MULTI);
3191                                 if (err < 0 && err != -EOPNOTSUPP) {
3192                                         if (likely(skb->len))
3193                                                 break;
3194
3195                                         goto out_err;
3196                                 }
3197                         }
3198                         idx++;
3199                 }
3200         }
3201         err = skb->len;
3202 out_err:
3203         rcu_read_unlock();
3204         cb->args[0] = idx;
3205
3206         return err;
3207 }
3208
3209 static inline size_t bridge_nlmsg_size(void)
3210 {
3211         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3212                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3213                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3214                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3215                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3216                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3217                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3218                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3219                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3220                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3221                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3222 }
3223
3224 static int rtnl_bridge_notify(struct net_device *dev)
3225 {
3226         struct net *net = dev_net(dev);
3227         struct sk_buff *skb;
3228         int err = -EOPNOTSUPP;
3229
3230         if (!dev->netdev_ops->ndo_bridge_getlink)
3231                 return 0;
3232
3233         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3234         if (!skb) {
3235                 err = -ENOMEM;
3236                 goto errout;
3237         }
3238
3239         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3240         if (err < 0)
3241                 goto errout;
3242
3243         /* Notification info is only filled for bridge ports, not the bridge
3244          * device itself. Therefore, a zero notification length is valid and
3245          * should not result in an error.
3246          */
3247         if (!skb->len)
3248                 goto errout;
3249
3250         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3251         return 0;
3252 errout:
3253         WARN_ON(err == -EMSGSIZE);
3254         kfree_skb(skb);
3255         if (err)
3256                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3257         return err;
3258 }
3259
3260 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3261 {
3262         struct net *net = sock_net(skb->sk);
3263         struct ifinfomsg *ifm;
3264         struct net_device *dev;
3265         struct nlattr *br_spec, *attr = NULL;
3266         int rem, err = -EOPNOTSUPP;
3267         u16 flags = 0;
3268         bool have_flags = false;
3269
3270         if (nlmsg_len(nlh) < sizeof(*ifm))
3271                 return -EINVAL;
3272
3273         ifm = nlmsg_data(nlh);
3274         if (ifm->ifi_family != AF_BRIDGE)
3275                 return -EPFNOSUPPORT;
3276
3277         dev = __dev_get_by_index(net, ifm->ifi_index);
3278         if (!dev) {
3279                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3280                 return -ENODEV;
3281         }
3282
3283         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3284         if (br_spec) {
3285                 nla_for_each_nested(attr, br_spec, rem) {
3286                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3287                                 if (nla_len(attr) < sizeof(flags))
3288                                         return -EINVAL;
3289
3290                                 have_flags = true;
3291                                 flags = nla_get_u16(attr);
3292                                 break;
3293                         }
3294                 }
3295         }
3296
3297         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3298                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3299
3300                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3301                         err = -EOPNOTSUPP;
3302                         goto out;
3303                 }
3304
3305                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3306                 if (err)
3307                         goto out;
3308
3309                 flags &= ~BRIDGE_FLAGS_MASTER;
3310         }
3311
3312         if ((flags & BRIDGE_FLAGS_SELF)) {
3313                 if (!dev->netdev_ops->ndo_bridge_setlink)
3314                         err = -EOPNOTSUPP;
3315                 else
3316                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3317                                                                   flags);
3318                 if (!err) {
3319                         flags &= ~BRIDGE_FLAGS_SELF;
3320
3321                         /* Generate event to notify upper layer of bridge
3322                          * change
3323                          */
3324                         err = rtnl_bridge_notify(dev);
3325                 }
3326         }
3327
3328         if (have_flags)
3329                 memcpy(nla_data(attr), &flags, sizeof(flags));
3330 out:
3331         return err;
3332 }
3333
3334 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3335 {
3336         struct net *net = sock_net(skb->sk);
3337         struct ifinfomsg *ifm;
3338         struct net_device *dev;
3339         struct nlattr *br_spec, *attr = NULL;
3340         int rem, err = -EOPNOTSUPP;
3341         u16 flags = 0;
3342         bool have_flags = false;
3343
3344         if (nlmsg_len(nlh) < sizeof(*ifm))
3345                 return -EINVAL;
3346
3347         ifm = nlmsg_data(nlh);
3348         if (ifm->ifi_family != AF_BRIDGE)
3349                 return -EPFNOSUPPORT;
3350
3351         dev = __dev_get_by_index(net, ifm->ifi_index);
3352         if (!dev) {
3353                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3354                 return -ENODEV;
3355         }
3356
3357         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3358         if (br_spec) {
3359                 nla_for_each_nested(attr, br_spec, rem) {
3360                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3361                                 if (nla_len(attr) < sizeof(flags))
3362                                         return -EINVAL;
3363
3364                                 have_flags = true;
3365                                 flags = nla_get_u16(attr);
3366                                 break;
3367                         }
3368                 }
3369         }
3370
3371         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3372                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3373
3374                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3375                         err = -EOPNOTSUPP;
3376                         goto out;
3377                 }
3378
3379                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3380                 if (err)
3381                         goto out;
3382
3383                 flags &= ~BRIDGE_FLAGS_MASTER;
3384         }
3385
3386         if ((flags & BRIDGE_FLAGS_SELF)) {
3387                 if (!dev->netdev_ops->ndo_bridge_dellink)
3388                         err = -EOPNOTSUPP;
3389                 else
3390                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3391                                                                   flags);
3392
3393                 if (!err) {
3394                         flags &= ~BRIDGE_FLAGS_SELF;
3395
3396                         /* Generate event to notify upper layer of bridge
3397                          * change
3398                          */
3399                         err = rtnl_bridge_notify(dev);
3400                 }
3401         }
3402
3403         if (have_flags)
3404                 memcpy(nla_data(attr), &flags, sizeof(flags));
3405 out:
3406         return err;
3407 }
3408
3409 /* Process one rtnetlink message. */
3410
3411 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3412 {
3413         struct net *net = sock_net(skb->sk);
3414         rtnl_doit_func doit;
3415         int sz_idx, kind;
3416         int family;
3417         int type;
3418         int err;
3419
3420         type = nlh->nlmsg_type;
3421         if (type > RTM_MAX)
3422                 return -EOPNOTSUPP;
3423
3424         type -= RTM_BASE;
3425
3426         /* All the messages must have at least 1 byte length */
3427         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3428                 return 0;
3429
3430         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3431         sz_idx = type>>2;
3432         kind = type&3;
3433
3434         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3435                 return -EPERM;
3436
3437         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3438                 struct sock *rtnl;
3439                 rtnl_dumpit_func dumpit;
3440                 rtnl_calcit_func calcit;
3441                 u16 min_dump_alloc = 0;
3442
3443                 dumpit = rtnl_get_dumpit(family, type);
3444                 if (dumpit == NULL)
3445                         return -EOPNOTSUPP;
3446                 calcit = rtnl_get_calcit(family, type);
3447                 if (calcit)
3448                         min_dump_alloc = calcit(skb, nlh);
3449
3450                 __rtnl_unlock();
3451                 rtnl = net->rtnl;
3452                 {
3453                         struct netlink_dump_control c = {
3454                                 .dump           = dumpit,
3455                                 .min_dump_alloc = min_dump_alloc,
3456                         };
3457                         err = netlink_dump_start(rtnl, skb, nlh, &c);
3458                 }
3459                 rtnl_lock();
3460                 return err;
3461         }
3462
3463         doit = rtnl_get_doit(family, type);
3464         if (doit == NULL)
3465                 return -EOPNOTSUPP;
3466
3467         return doit(skb, nlh);
3468 }
3469
3470 static void rtnetlink_rcv(struct sk_buff *skb)
3471 {
3472         rtnl_lock();
3473         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3474         rtnl_unlock();
3475 }
3476
3477 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3478 {
3479         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3480
3481         switch (event) {
3482         case NETDEV_UP:
3483         case NETDEV_DOWN:
3484         case NETDEV_PRE_UP:
3485         case NETDEV_POST_INIT:
3486         case NETDEV_REGISTER:
3487         case NETDEV_CHANGE:
3488         case NETDEV_PRE_TYPE_CHANGE:
3489         case NETDEV_GOING_DOWN:
3490         case NETDEV_UNREGISTER:
3491         case NETDEV_UNREGISTER_FINAL:
3492         case NETDEV_RELEASE:
3493         case NETDEV_JOIN:
3494         case NETDEV_BONDING_INFO:
3495                 break;
3496         default:
3497                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3498                 break;
3499         }
3500         return NOTIFY_DONE;
3501 }
3502
3503 static struct notifier_block rtnetlink_dev_notifier = {
3504         .notifier_call  = rtnetlink_event,
3505 };
3506
3507
3508 static int __net_init rtnetlink_net_init(struct net *net)
3509 {
3510         struct sock *sk;
3511         struct netlink_kernel_cfg cfg = {
3512                 .groups         = RTNLGRP_MAX,
3513                 .input          = rtnetlink_rcv,
3514                 .cb_mutex       = &rtnl_mutex,
3515                 .flags          = NL_CFG_F_NONROOT_RECV,
3516         };
3517
3518         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3519         if (!sk)
3520                 return -ENOMEM;
3521         net->rtnl = sk;
3522         return 0;
3523 }
3524
3525 static void __net_exit rtnetlink_net_exit(struct net *net)
3526 {
3527         netlink_kernel_release(net->rtnl);
3528         net->rtnl = NULL;
3529 }
3530
3531 static struct pernet_operations rtnetlink_net_ops = {
3532         .init = rtnetlink_net_init,
3533         .exit = rtnetlink_net_exit,
3534 };
3535
3536 void __init rtnetlink_init(void)
3537 {
3538         if (register_pernet_subsys(&rtnetlink_net_ops))
3539                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3540
3541         register_netdevice_notifier(&rtnetlink_dev_notifier);
3542
3543         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3544                       rtnl_dump_ifinfo, rtnl_calcit);
3545         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3546         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3547         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3548
3549         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3550         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3551
3552         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3553         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3554         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3555
3556         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3557         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3558         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3559 }