net/core/rtnetlink.c

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