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