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