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