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         int hdrlen;
1048
1049         s_h = cb->args[0];
1050         s_idx = cb->args[1];
1051
1052         rcu_read_lock();
1053         cb->seq = net->dev_base_seq;
1054
1055         /* A hack to preserve kernel<->userspace interface.
1056          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1057          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1058          * what iproute2 < v3.9.0 used.
1059          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1060          * attribute, its netlink message is shorter than struct ifinfomsg.
1061          */
1062         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1063                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1064
1065         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1066
1067                 if (tb[IFLA_EXT_MASK])
1068                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1069         }
1070
1071         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1072                 idx = 0;
1073                 head = &net->dev_index_head[h];
1074                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1075                         if (idx < s_idx)
1076                                 goto cont;
1077                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1078                                                NETLINK_CB(cb->skb).portid,
1079                                                cb->nlh->nlmsg_seq, 0,
1080                                                NLM_F_MULTI,
1081                                                ext_filter_mask);
1082                         /* If we ran out of room on the first message,
1083                          * we're in trouble
1084                          */
1085                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1086
1087                         if (err <= 0)
1088                                 goto out;
1089
1090                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1091 cont:
1092                         idx++;
1093                 }
1094         }
1095 out:
1096         rcu_read_unlock();
1097         cb->args[1] = idx;
1098         cb->args[0] = h;
1099
1100         return skb->len;
1101 }
1102
1103 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1104         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1105         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1107         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1108         [IFLA_MTU]              = { .type = NLA_U32 },
1109         [IFLA_LINK]             = { .type = NLA_U32 },
1110         [IFLA_MASTER]           = { .type = NLA_U32 },
1111         [IFLA_CARRIER]          = { .type = NLA_U8 },
1112         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1113         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1114         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1115         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1116         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1117         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1118         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1119         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1120         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1121         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1122         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1123         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1124         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1125         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1126         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1127         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1128 };
1129 EXPORT_SYMBOL(ifla_policy);
1130
1131 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1132         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1133         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1134 };
1135
1136 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1137         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1138 };
1139
1140 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1141         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1142                                     .len = sizeof(struct ifla_vf_mac) },
1143         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1144                                     .len = sizeof(struct ifla_vf_vlan) },
1145         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1146                                     .len = sizeof(struct ifla_vf_tx_rate) },
1147         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1148                                     .len = sizeof(struct ifla_vf_spoofchk) },
1149 };
1150
1151 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1152         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1153         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1154                                     .len = PORT_PROFILE_MAX },
1155         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1156                                     .len = sizeof(struct ifla_port_vsi)},
1157         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1158                                       .len = PORT_UUID_MAX },
1159         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1160                                     .len = PORT_UUID_MAX },
1161         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1162         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1163 };
1164
1165 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1166 {
1167         struct net *net;
1168         /* Examine the link attributes and figure out which
1169          * network namespace we are talking about.
1170          */
1171         if (tb[IFLA_NET_NS_PID])
1172                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1173         else if (tb[IFLA_NET_NS_FD])
1174                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1175         else
1176                 net = get_net(src_net);
1177         return net;
1178 }
1179 EXPORT_SYMBOL(rtnl_link_get_net);
1180
1181 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1182 {
1183         if (dev) {
1184                 if (tb[IFLA_ADDRESS] &&
1185                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1186                         return -EINVAL;
1187
1188                 if (tb[IFLA_BROADCAST] &&
1189                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1190                         return -EINVAL;
1191         }
1192
1193         if (tb[IFLA_AF_SPEC]) {
1194                 struct nlattr *af;
1195                 int rem, err;
1196
1197                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1198                         const struct rtnl_af_ops *af_ops;
1199
1200                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1201                                 return -EAFNOSUPPORT;
1202
1203                         if (!af_ops->set_link_af)
1204                                 return -EOPNOTSUPP;
1205
1206                         if (af_ops->validate_link_af) {
1207                                 err = af_ops->validate_link_af(dev, af);
1208                                 if (err < 0)
1209                                         return err;
1210                         }
1211                 }
1212         }
1213
1214         return 0;
1215 }
1216
1217 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1218 {
1219         int rem, err = -EINVAL;
1220         struct nlattr *vf;
1221         const struct net_device_ops *ops = dev->netdev_ops;
1222
1223         nla_for_each_nested(vf, attr, rem) {
1224                 switch (nla_type(vf)) {
1225                 case IFLA_VF_MAC: {
1226                         struct ifla_vf_mac *ivm;
1227                         ivm = nla_data(vf);
1228                         err = -EOPNOTSUPP;
1229                         if (ops->ndo_set_vf_mac)
1230                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1231                                                           ivm->mac);
1232                         break;
1233                 }
1234                 case IFLA_VF_VLAN: {
1235                         struct ifla_vf_vlan *ivv;
1236                         ivv = nla_data(vf);
1237                         err = -EOPNOTSUPP;
1238                         if (ops->ndo_set_vf_vlan)
1239                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1240                                                            ivv->vlan,
1241                                                            ivv->qos);
1242                         break;
1243                 }
1244                 case IFLA_VF_TX_RATE: {
1245                         struct ifla_vf_tx_rate *ivt;
1246                         ivt = nla_data(vf);
1247                         err = -EOPNOTSUPP;
1248                         if (ops->ndo_set_vf_tx_rate)
1249                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1250                                                               ivt->rate);
1251                         break;
1252                 }
1253                 case IFLA_VF_SPOOFCHK: {
1254                         struct ifla_vf_spoofchk *ivs;
1255                         ivs = nla_data(vf);
1256                         err = -EOPNOTSUPP;
1257                         if (ops->ndo_set_vf_spoofchk)
1258                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1259                                                                ivs->setting);
1260                         break;
1261                 }
1262                 default:
1263                         err = -EINVAL;
1264                         break;
1265                 }
1266                 if (err)
1267                         break;
1268         }
1269         return err;
1270 }
1271
1272 static int do_set_master(struct net_device *dev, int ifindex)
1273 {
1274         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1275         const struct net_device_ops *ops;
1276         int err;
1277
1278         if (upper_dev) {
1279                 if (upper_dev->ifindex == ifindex)
1280                         return 0;
1281                 ops = upper_dev->netdev_ops;
1282                 if (ops->ndo_del_slave) {
1283                         err = ops->ndo_del_slave(upper_dev, dev);
1284                         if (err)
1285                                 return err;
1286                 } else {
1287                         return -EOPNOTSUPP;
1288                 }
1289         }
1290
1291         if (ifindex) {
1292                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1293                 if (!upper_dev)
1294                         return -EINVAL;
1295                 ops = upper_dev->netdev_ops;
1296                 if (ops->ndo_add_slave) {
1297                         err = ops->ndo_add_slave(upper_dev, dev);
1298                         if (err)
1299                                 return err;
1300                 } else {
1301                         return -EOPNOTSUPP;
1302                 }
1303         }
1304         return 0;
1305 }
1306
1307 static int do_setlink(const struct sk_buff *skb,
1308                       struct net_device *dev, struct ifinfomsg *ifm,
1309                       struct nlattr **tb, char *ifname, int modified)
1310 {
1311         const struct net_device_ops *ops = dev->netdev_ops;
1312         int err;
1313
1314         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1315                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1316                 if (IS_ERR(net)) {
1317                         err = PTR_ERR(net);
1318                         goto errout;
1319                 }
1320                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1321                         err = -EPERM;
1322                         goto errout;
1323                 }
1324                 err = dev_change_net_namespace(dev, net, ifname);
1325                 put_net(net);
1326                 if (err)
1327                         goto errout;
1328                 modified = 1;
1329         }
1330
1331         if (tb[IFLA_MAP]) {
1332                 struct rtnl_link_ifmap *u_map;
1333                 struct ifmap k_map;
1334
1335                 if (!ops->ndo_set_config) {
1336                         err = -EOPNOTSUPP;
1337                         goto errout;
1338                 }
1339
1340                 if (!netif_device_present(dev)) {
1341                         err = -ENODEV;
1342                         goto errout;
1343                 }
1344
1345                 u_map = nla_data(tb[IFLA_MAP]);
1346                 k_map.mem_start = (unsigned long) u_map->mem_start;
1347                 k_map.mem_end = (unsigned long) u_map->mem_end;
1348                 k_map.base_addr = (unsigned short) u_map->base_addr;
1349                 k_map.irq = (unsigned char) u_map->irq;
1350                 k_map.dma = (unsigned char) u_map->dma;
1351                 k_map.port = (unsigned char) u_map->port;
1352
1353                 err = ops->ndo_set_config(dev, &k_map);
1354                 if (err < 0)
1355                         goto errout;
1356
1357                 modified = 1;
1358         }
1359
1360         if (tb[IFLA_ADDRESS]) {
1361                 struct sockaddr *sa;
1362                 int len;
1363
1364                 len = sizeof(sa_family_t) + dev->addr_len;
1365                 sa = kmalloc(len, GFP_KERNEL);
1366                 if (!sa) {
1367                         err = -ENOMEM;
1368                         goto errout;
1369                 }
1370                 sa->sa_family = dev->type;
1371                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1372                        dev->addr_len);
1373                 err = dev_set_mac_address(dev, sa);
1374                 kfree(sa);
1375                 if (err)
1376                         goto errout;
1377                 modified = 1;
1378         }
1379
1380         if (tb[IFLA_MTU]) {
1381                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1382                 if (err < 0)
1383                         goto errout;
1384                 modified = 1;
1385         }
1386
1387         if (tb[IFLA_GROUP]) {
1388                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1389                 modified = 1;
1390         }
1391
1392         /*
1393          * Interface selected by interface index but interface
1394          * name provided implies that a name change has been
1395          * requested.
1396          */
1397         if (ifm->ifi_index > 0 && ifname[0]) {
1398                 err = dev_change_name(dev, ifname);
1399                 if (err < 0)
1400                         goto errout;
1401                 modified = 1;
1402         }
1403
1404         if (tb[IFLA_IFALIAS]) {
1405                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1406                                     nla_len(tb[IFLA_IFALIAS]));
1407                 if (err < 0)
1408                         goto errout;
1409                 modified = 1;
1410         }
1411
1412         if (tb[IFLA_BROADCAST]) {
1413                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1414                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1415         }
1416
1417         if (ifm->ifi_flags || ifm->ifi_change) {
1418                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1419                 if (err < 0)
1420                         goto errout;
1421         }
1422
1423         if (tb[IFLA_MASTER]) {
1424                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1425                 if (err)
1426                         goto errout;
1427                 modified = 1;
1428         }
1429
1430         if (tb[IFLA_CARRIER]) {
1431                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1432                 if (err)
1433                         goto errout;
1434                 modified = 1;
1435         }
1436
1437         if (tb[IFLA_TXQLEN])
1438                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1439
1440         if (tb[IFLA_OPERSTATE])
1441                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1442
1443         if (tb[IFLA_LINKMODE]) {
1444                 write_lock_bh(&dev_base_lock);
1445                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1446                 write_unlock_bh(&dev_base_lock);
1447         }
1448
1449         if (tb[IFLA_VFINFO_LIST]) {
1450                 struct nlattr *attr;
1451                 int rem;
1452                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1453                         if (nla_type(attr) != IFLA_VF_INFO) {
1454                                 err = -EINVAL;
1455                                 goto errout;
1456                         }
1457                         err = do_setvfinfo(dev, attr);
1458                         if (err < 0)
1459                                 goto errout;
1460                         modified = 1;
1461                 }
1462         }
1463         err = 0;
1464
1465         if (tb[IFLA_VF_PORTS]) {
1466                 struct nlattr *port[IFLA_PORT_MAX+1];
1467                 struct nlattr *attr;
1468                 int vf;
1469                 int rem;
1470
1471                 err = -EOPNOTSUPP;
1472                 if (!ops->ndo_set_vf_port)
1473                         goto errout;
1474
1475                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1476                         if (nla_type(attr) != IFLA_VF_PORT)
1477                                 continue;
1478                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1479                                 attr, ifla_port_policy);
1480                         if (err < 0)
1481                                 goto errout;
1482                         if (!port[IFLA_PORT_VF]) {
1483                                 err = -EOPNOTSUPP;
1484                                 goto errout;
1485                         }
1486                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1487                         err = ops->ndo_set_vf_port(dev, vf, port);
1488                         if (err < 0)
1489                                 goto errout;
1490                         modified = 1;
1491                 }
1492         }
1493         err = 0;
1494
1495         if (tb[IFLA_PORT_SELF]) {
1496                 struct nlattr *port[IFLA_PORT_MAX+1];
1497
1498                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1499                         tb[IFLA_PORT_SELF], ifla_port_policy);
1500                 if (err < 0)
1501                         goto errout;
1502
1503                 err = -EOPNOTSUPP;
1504                 if (ops->ndo_set_vf_port)
1505                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1506                 if (err < 0)
1507                         goto errout;
1508                 modified = 1;
1509         }
1510
1511         if (tb[IFLA_AF_SPEC]) {
1512                 struct nlattr *af;
1513                 int rem;
1514
1515                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1516                         const struct rtnl_af_ops *af_ops;
1517
1518                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1519                                 BUG();
1520
1521                         err = af_ops->set_link_af(dev, af);
1522                         if (err < 0)
1523                                 goto errout;
1524
1525                         modified = 1;
1526                 }
1527         }
1528         err = 0;
1529
1530 errout:
1531         if (err < 0 && modified)
1532                 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",
1533                                      dev->name);
1534
1535         return err;
1536 }
1537
1538 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1539 {
1540         struct net *net = sock_net(skb->sk);
1541         struct ifinfomsg *ifm;
1542         struct net_device *dev;
1543         int err;
1544         struct nlattr *tb[IFLA_MAX+1];
1545         char ifname[IFNAMSIZ];
1546
1547         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1548         if (err < 0)
1549                 goto errout;
1550
1551         if (tb[IFLA_IFNAME])
1552                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1553         else
1554                 ifname[0] = '\0';
1555
1556         err = -EINVAL;
1557         ifm = nlmsg_data(nlh);
1558         if (ifm->ifi_index > 0)
1559                 dev = __dev_get_by_index(net, ifm->ifi_index);
1560         else if (tb[IFLA_IFNAME])
1561                 dev = __dev_get_by_name(net, ifname);
1562         else
1563                 goto errout;
1564
1565         if (dev == NULL) {
1566                 err = -ENODEV;
1567                 goto errout;
1568         }
1569
1570         err = validate_linkmsg(dev, tb);
1571         if (err < 0)
1572                 goto errout;
1573
1574         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1575 errout:
1576         return err;
1577 }
1578
1579 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1580 {
1581         struct net *net = sock_net(skb->sk);
1582         const struct rtnl_link_ops *ops;
1583         struct net_device *dev;
1584         struct ifinfomsg *ifm;
1585         char ifname[IFNAMSIZ];
1586         struct nlattr *tb[IFLA_MAX+1];
1587         int err;
1588         LIST_HEAD(list_kill);
1589
1590         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1591         if (err < 0)
1592                 return err;
1593
1594         if (tb[IFLA_IFNAME])
1595                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1596
1597         ifm = nlmsg_data(nlh);
1598         if (ifm->ifi_index > 0)
1599                 dev = __dev_get_by_index(net, ifm->ifi_index);
1600         else if (tb[IFLA_IFNAME])
1601                 dev = __dev_get_by_name(net, ifname);
1602         else
1603                 return -EINVAL;
1604
1605         if (!dev)
1606                 return -ENODEV;
1607
1608         ops = dev->rtnl_link_ops;
1609         if (!ops)
1610                 return -EOPNOTSUPP;
1611
1612         ops->dellink(dev, &list_kill);
1613         unregister_netdevice_many(&list_kill);
1614         return 0;
1615 }
1616
1617 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1618 {
1619         unsigned int old_flags;
1620         int err;
1621
1622         old_flags = dev->flags;
1623         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1624                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1625                 if (err < 0)
1626                         return err;
1627         }
1628
1629         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1630         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1631
1632         __dev_notify_flags(dev, old_flags);
1633         return 0;
1634 }
1635 EXPORT_SYMBOL(rtnl_configure_link);
1636
1637 struct net_device *rtnl_create_link(struct net *net,
1638         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1639 {
1640         int err;
1641         struct net_device *dev;
1642         unsigned int num_tx_queues = 1;
1643         unsigned int num_rx_queues = 1;
1644
1645         if (tb[IFLA_NUM_TX_QUEUES])
1646                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1647         else if (ops->get_num_tx_queues)
1648                 num_tx_queues = ops->get_num_tx_queues();
1649
1650         if (tb[IFLA_NUM_RX_QUEUES])
1651                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1652         else if (ops->get_num_rx_queues)
1653                 num_rx_queues = ops->get_num_rx_queues();
1654
1655         err = -ENOMEM;
1656         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1657                                num_tx_queues, num_rx_queues);
1658         if (!dev)
1659                 goto err;
1660
1661         dev_net_set(dev, net);
1662         dev->rtnl_link_ops = ops;
1663         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1664
1665         if (tb[IFLA_MTU])
1666                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1667         if (tb[IFLA_ADDRESS]) {
1668                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1669                                 nla_len(tb[IFLA_ADDRESS]));
1670                 dev->addr_assign_type = NET_ADDR_SET;
1671         }
1672         if (tb[IFLA_BROADCAST])
1673                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1674                                 nla_len(tb[IFLA_BROADCAST]));
1675         if (tb[IFLA_TXQLEN])
1676                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1677         if (tb[IFLA_OPERSTATE])
1678                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1679         if (tb[IFLA_LINKMODE])
1680                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1681         if (tb[IFLA_GROUP])
1682                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1683
1684         return dev;
1685
1686 err:
1687         return ERR_PTR(err);
1688 }
1689 EXPORT_SYMBOL(rtnl_create_link);
1690
1691 static int rtnl_group_changelink(const struct sk_buff *skb,
1692                 struct net *net, int group,
1693                 struct ifinfomsg *ifm,
1694                 struct nlattr **tb)
1695 {
1696         struct net_device *dev;
1697         int err;
1698
1699         for_each_netdev(net, dev) {
1700                 if (dev->group == group) {
1701                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1702                         if (err < 0)
1703                                 return err;
1704                 }
1705         }
1706
1707         return 0;
1708 }
1709
1710 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1711 {
1712         struct net *net = sock_net(skb->sk);
1713         const struct rtnl_link_ops *ops;
1714         struct net_device *dev;
1715         struct ifinfomsg *ifm;
1716         char kind[MODULE_NAME_LEN];
1717         char ifname[IFNAMSIZ];
1718         struct nlattr *tb[IFLA_MAX+1];
1719         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1720         int err;
1721
1722 #ifdef CONFIG_MODULES
1723 replay:
1724 #endif
1725         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1726         if (err < 0)
1727                 return err;
1728
1729         if (tb[IFLA_IFNAME])
1730                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1731         else
1732                 ifname[0] = '\0';
1733
1734         ifm = nlmsg_data(nlh);
1735         if (ifm->ifi_index > 0)
1736                 dev = __dev_get_by_index(net, ifm->ifi_index);
1737         else {
1738                 if (ifname[0])
1739                         dev = __dev_get_by_name(net, ifname);
1740                 else
1741                         dev = NULL;
1742         }
1743
1744         err = validate_linkmsg(dev, tb);
1745         if (err < 0)
1746                 return err;
1747
1748         if (tb[IFLA_LINKINFO]) {
1749                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1750                                        tb[IFLA_LINKINFO], ifla_info_policy);
1751                 if (err < 0)
1752                         return err;
1753         } else
1754                 memset(linkinfo, 0, sizeof(linkinfo));
1755
1756         if (linkinfo[IFLA_INFO_KIND]) {
1757                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1758                 ops = rtnl_link_ops_get(kind);
1759         } else {
1760                 kind[0] = '\0';
1761                 ops = NULL;
1762         }
1763
1764         if (1) {
1765                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1766                 struct net *dest_net;
1767
1768                 if (ops) {
1769                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1770                                 err = nla_parse_nested(attr, ops->maxtype,
1771                                                        linkinfo[IFLA_INFO_DATA],
1772                                                        ops->policy);
1773                                 if (err < 0)
1774                                         return err;
1775                                 data = attr;
1776                         }
1777                         if (ops->validate) {
1778                                 err = ops->validate(tb, data);
1779                                 if (err < 0)
1780                                         return err;
1781                         }
1782                 }
1783
1784                 if (dev) {
1785                         int modified = 0;
1786
1787                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1788                                 return -EEXIST;
1789                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1790                                 return -EOPNOTSUPP;
1791
1792                         if (linkinfo[IFLA_INFO_DATA]) {
1793                                 if (!ops || ops != dev->rtnl_link_ops ||
1794                                     !ops->changelink)
1795                                         return -EOPNOTSUPP;
1796
1797                                 err = ops->changelink(dev, tb, data);
1798                                 if (err < 0)
1799                                         return err;
1800                                 modified = 1;
1801                         }
1802
1803                         return do_setlink(skb, dev, ifm, tb, ifname, modified);
1804                 }
1805
1806                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1807                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1808                                 return rtnl_group_changelink(skb, net,
1809                                                 nla_get_u32(tb[IFLA_GROUP]),
1810                                                 ifm, tb);
1811                         return -ENODEV;
1812                 }
1813
1814                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1815                         return -EOPNOTSUPP;
1816
1817                 if (!ops) {
1818 #ifdef CONFIG_MODULES
1819                         if (kind[0]) {
1820                                 __rtnl_unlock();
1821                                 request_module("rtnl-link-%s", kind);
1822                                 rtnl_lock();
1823                                 ops = rtnl_link_ops_get(kind);
1824                                 if (ops)
1825                                         goto replay;
1826                         }
1827 #endif
1828                         return -EOPNOTSUPP;
1829                 }
1830
1831                 if (!ifname[0])
1832                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1833
1834                 dest_net = rtnl_link_get_net(net, tb);
1835                 if (IS_ERR(dest_net))
1836                         return PTR_ERR(dest_net);
1837
1838                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1839                 if (IS_ERR(dev)) {
1840                         err = PTR_ERR(dev);
1841                         goto out;
1842                 }
1843
1844                 dev->ifindex = ifm->ifi_index;
1845
1846                 if (ops->newlink)
1847                         err = ops->newlink(net, dev, tb, data);
1848                 else
1849                         err = register_netdevice(dev);
1850
1851                 if (err < 0 && !IS_ERR(dev))
1852                         free_netdev(dev);
1853                 if (err < 0)
1854                         goto out;
1855
1856                 err = rtnl_configure_link(dev, ifm);
1857                 if (err < 0)
1858                         unregister_netdevice(dev);
1859 out:
1860                 put_net(dest_net);
1861                 return err;
1862         }
1863 }
1864
1865 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1866 {
1867         struct net *net = sock_net(skb->sk);
1868         struct ifinfomsg *ifm;
1869         char ifname[IFNAMSIZ];
1870         struct nlattr *tb[IFLA_MAX+1];
1871         struct net_device *dev = NULL;
1872         struct sk_buff *nskb;
1873         int err;
1874         u32 ext_filter_mask = 0;
1875
1876         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1877         if (err < 0)
1878                 return err;
1879
1880         if (tb[IFLA_IFNAME])
1881                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1882
1883         if (tb[IFLA_EXT_MASK])
1884                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1885
1886         ifm = nlmsg_data(nlh);
1887         if (ifm->ifi_index > 0)
1888                 dev = __dev_get_by_index(net, ifm->ifi_index);
1889         else if (tb[IFLA_IFNAME])
1890                 dev = __dev_get_by_name(net, ifname);
1891         else
1892                 return -EINVAL;
1893
1894         if (dev == NULL)
1895                 return -ENODEV;
1896
1897         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1898         if (nskb == NULL)
1899                 return -ENOBUFS;
1900
1901         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1902                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1903         if (err < 0) {
1904                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1905                 WARN_ON(err == -EMSGSIZE);
1906                 kfree_skb(nskb);
1907         } else
1908                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1909
1910         return err;
1911 }
1912
1913 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1914 {
1915         struct net *net = sock_net(skb->sk);
1916         struct net_device *dev;
1917         struct nlattr *tb[IFLA_MAX+1];
1918         u32 ext_filter_mask = 0;
1919         u16 min_ifinfo_dump_size = 0;
1920         int hdrlen;
1921
1922         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
1923         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1924                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1925
1926         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1927                 if (tb[IFLA_EXT_MASK])
1928                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1929         }
1930
1931         if (!ext_filter_mask)
1932                 return NLMSG_GOODSIZE;
1933         /*
1934          * traverse the list of net devices and compute the minimum
1935          * buffer size based upon the filter mask.
1936          */
1937         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1938                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1939                                              if_nlmsg_size(dev,
1940                                                            ext_filter_mask));
1941         }
1942
1943         return min_ifinfo_dump_size;
1944 }
1945
1946 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1947 {
1948         int idx;
1949         int s_idx = cb->family;
1950
1951         if (s_idx == 0)
1952                 s_idx = 1;
1953         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1954                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1955                 if (idx < s_idx || idx == PF_PACKET)
1956                         continue;
1957                 if (rtnl_msg_handlers[idx] == NULL ||
1958                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1959                         continue;
1960                 if (idx > s_idx) {
1961                         memset(&cb->args[0], 0, sizeof(cb->args));
1962                         cb->prev_seq = 0;
1963                         cb->seq = 0;
1964                 }
1965                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1966                         break;
1967         }
1968         cb->family = idx;
1969
1970         return skb->len;
1971 }
1972
1973 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1974 {
1975         struct net *net = dev_net(dev);
1976         struct sk_buff *skb;
1977         int err = -ENOBUFS;
1978         size_t if_info_size;
1979
1980         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1981         if (skb == NULL)
1982                 goto errout;
1983
1984         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1985         if (err < 0) {
1986                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1987                 WARN_ON(err == -EMSGSIZE);
1988                 kfree_skb(skb);
1989                 goto errout;
1990         }
1991         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1992         return;
1993 errout:
1994         if (err < 0)
1995                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1996 }
1997 EXPORT_SYMBOL(rtmsg_ifinfo);
1998
1999 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2000                                    struct net_device *dev,
2001                                    u8 *addr, u32 pid, u32 seq,
2002                                    int type, unsigned int flags,
2003                                    int nlflags)
2004 {
2005         struct nlmsghdr *nlh;
2006         struct ndmsg *ndm;
2007
2008         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2009         if (!nlh)
2010                 return -EMSGSIZE;
2011
2012         ndm = nlmsg_data(nlh);
2013         ndm->ndm_family  = AF_BRIDGE;
2014         ndm->ndm_pad1    = 0;
2015         ndm->ndm_pad2    = 0;
2016         ndm->ndm_flags   = flags;
2017         ndm->ndm_type    = 0;
2018         ndm->ndm_ifindex = dev->ifindex;
2019         ndm->ndm_state   = NUD_PERMANENT;
2020
2021         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2022                 goto nla_put_failure;
2023
2024         return nlmsg_end(skb, nlh);
2025
2026 nla_put_failure:
2027         nlmsg_cancel(skb, nlh);
2028         return -EMSGSIZE;
2029 }
2030
2031 static inline size_t rtnl_fdb_nlmsg_size(void)
2032 {
2033         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2034 }
2035
2036 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2037 {
2038         struct net *net = dev_net(dev);
2039         struct sk_buff *skb;
2040         int err = -ENOBUFS;
2041
2042         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2043         if (!skb)
2044                 goto errout;
2045
2046         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2047         if (err < 0) {
2048                 kfree_skb(skb);
2049                 goto errout;
2050         }
2051
2052         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2053         return;
2054 errout:
2055         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2056 }
2057
2058 /**
2059  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2060  */
2061 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2062                      struct nlattr *tb[],
2063                      struct net_device *dev,
2064                      const unsigned char *addr,
2065                      u16 flags)
2066 {
2067         int err = -EINVAL;
2068
2069         /* If aging addresses are supported device will need to
2070          * implement its own handler for this.
2071          */
2072         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2073                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2074                 return err;
2075         }
2076
2077         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2078                 err = dev_uc_add_excl(dev, addr);
2079         else if (is_multicast_ether_addr(addr))
2080                 err = dev_mc_add_excl(dev, addr);
2081
2082         /* Only return duplicate errors if NLM_F_EXCL is set */
2083         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2084                 err = 0;
2085
2086         return err;
2087 }
2088 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2089
2090 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2091 {
2092         struct net *net = sock_net(skb->sk);
2093         struct ndmsg *ndm;
2094         struct nlattr *tb[NDA_MAX+1];
2095         struct net_device *dev;
2096         u8 *addr;
2097         int err;
2098
2099         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2100         if (err < 0)
2101                 return err;
2102
2103         ndm = nlmsg_data(nlh);
2104         if (ndm->ndm_ifindex == 0) {
2105                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2106                 return -EINVAL;
2107         }
2108
2109         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2110         if (dev == NULL) {
2111                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2112                 return -ENODEV;
2113         }
2114
2115         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2116                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2117                 return -EINVAL;
2118         }
2119
2120         addr = nla_data(tb[NDA_LLADDR]);
2121         if (is_zero_ether_addr(addr)) {
2122                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2123                 return -EINVAL;
2124         }
2125
2126         err = -EOPNOTSUPP;
2127
2128         /* Support fdb on master device the net/bridge default case */
2129         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2130             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2131                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2132                 const struct net_device_ops *ops = br_dev->netdev_ops;
2133
2134                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2135                 if (err)
2136                         goto out;
2137                 else
2138                         ndm->ndm_flags &= ~NTF_MASTER;
2139         }
2140
2141         /* Embedded bridge, macvlan, and any other device support */
2142         if ((ndm->ndm_flags & NTF_SELF)) {
2143                 if (dev->netdev_ops->ndo_fdb_add)
2144                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2145                                                            nlh->nlmsg_flags);
2146                 else
2147                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2148                                                nlh->nlmsg_flags);
2149
2150                 if (!err) {
2151                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2152                         ndm->ndm_flags &= ~NTF_SELF;
2153                 }
2154         }
2155 out:
2156         return err;
2157 }
2158
2159 /**
2160  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2161  */
2162 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2163                      struct nlattr *tb[],
2164                      struct net_device *dev,
2165                      const unsigned char *addr)
2166 {
2167         int err = -EOPNOTSUPP;
2168
2169         /* If aging addresses are supported device will need to
2170          * implement its own handler for this.
2171          */
2172         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2173                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2174                 return -EINVAL;
2175         }
2176
2177         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2178                 err = dev_uc_del(dev, addr);
2179         else if (is_multicast_ether_addr(addr))
2180                 err = dev_mc_del(dev, addr);
2181         else
2182                 err = -EINVAL;
2183
2184         return err;
2185 }
2186 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2187
2188 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2189 {
2190         struct net *net = sock_net(skb->sk);
2191         struct ndmsg *ndm;
2192         struct nlattr *tb[NDA_MAX+1];
2193         struct net_device *dev;
2194         int err = -EINVAL;
2195         __u8 *addr;
2196
2197         if (!netlink_capable(skb, CAP_NET_ADMIN))
2198                 return -EPERM;
2199
2200         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2201         if (err < 0)
2202                 return err;
2203
2204         ndm = nlmsg_data(nlh);
2205         if (ndm->ndm_ifindex == 0) {
2206                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2207                 return -EINVAL;
2208         }
2209
2210         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2211         if (dev == NULL) {
2212                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2213                 return -ENODEV;
2214         }
2215
2216         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2217                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2218                 return -EINVAL;
2219         }
2220
2221         addr = nla_data(tb[NDA_LLADDR]);
2222         if (is_zero_ether_addr(addr)) {
2223                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2224                 return -EINVAL;
2225         }
2226
2227         err = -EOPNOTSUPP;
2228
2229         /* Support fdb on master device the net/bridge default case */
2230         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2231             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2232                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2233                 const struct net_device_ops *ops = br_dev->netdev_ops;
2234
2235                 if (ops->ndo_fdb_del)
2236                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2237
2238                 if (err)
2239                         goto out;
2240                 else
2241                         ndm->ndm_flags &= ~NTF_MASTER;
2242         }
2243
2244         /* Embedded bridge, macvlan, and any other device support */
2245         if (ndm->ndm_flags & NTF_SELF) {
2246                 if (dev->netdev_ops->ndo_fdb_del)
2247                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2248                 else
2249                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2250
2251                 if (!err) {
2252                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2253                         ndm->ndm_flags &= ~NTF_SELF;
2254                 }
2255         }
2256 out:
2257         return err;
2258 }
2259
2260 static int nlmsg_populate_fdb(struct sk_buff *skb,
2261                               struct netlink_callback *cb,
2262                               struct net_device *dev,
2263                               int *idx,
2264                               struct netdev_hw_addr_list *list)
2265 {
2266         struct netdev_hw_addr *ha;
2267         int err;
2268         u32 portid, seq;
2269
2270         portid = NETLINK_CB(cb->skb).portid;
2271         seq = cb->nlh->nlmsg_seq;
2272
2273         list_for_each_entry(ha, &list->list, list) {
2274                 if (*idx < cb->args[0])
2275                         goto skip;
2276
2277                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2278                                               portid, seq,
2279                                               RTM_NEWNEIGH, NTF_SELF,
2280                                               NLM_F_MULTI);
2281                 if (err < 0)
2282                         return err;
2283 skip:
2284                 *idx += 1;
2285         }
2286         return 0;
2287 }
2288
2289 /**
2290  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2291  * @nlh: netlink message header
2292  * @dev: netdevice
2293  *
2294  * Default netdevice operation to dump the existing unicast address list.
2295  * Returns number of addresses from list put in skb.
2296  */
2297 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2298                       struct netlink_callback *cb,
2299                       struct net_device *dev,
2300                       int idx)
2301 {
2302         int err;
2303
2304         netif_addr_lock_bh(dev);
2305         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2306         if (err)
2307                 goto out;
2308         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2309 out:
2310         netif_addr_unlock_bh(dev);
2311         return idx;
2312 }
2313 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2314
2315 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2316 {
2317         int idx = 0;
2318         struct net *net = sock_net(skb->sk);
2319         struct net_device *dev;
2320
2321         rcu_read_lock();
2322         for_each_netdev_rcu(net, dev) {
2323                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2324                         struct net_device *br_dev;
2325                         const struct net_device_ops *ops;
2326
2327                         br_dev = netdev_master_upper_dev_get(dev);
2328                         ops = br_dev->netdev_ops;
2329                         if (ops->ndo_fdb_dump)
2330                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2331                 }
2332
2333                 if (dev->netdev_ops->ndo_fdb_dump)
2334                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2335                 else
2336                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2337         }
2338         rcu_read_unlock();
2339
2340         cb->args[0] = idx;
2341         return skb->len;
2342 }
2343
2344 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2345                             struct net_device *dev, u16 mode)
2346 {
2347         struct nlmsghdr *nlh;
2348         struct ifinfomsg *ifm;
2349         struct nlattr *br_afspec;
2350         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2351         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2352
2353         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2354         if (nlh == NULL)
2355                 return -EMSGSIZE;
2356
2357         ifm = nlmsg_data(nlh);
2358         ifm->ifi_family = AF_BRIDGE;
2359         ifm->__ifi_pad = 0;
2360         ifm->ifi_type = dev->type;
2361         ifm->ifi_index = dev->ifindex;
2362         ifm->ifi_flags = dev_get_flags(dev);
2363         ifm->ifi_change = 0;
2364
2365
2366         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2367             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2368             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2369             (br_dev &&
2370              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2371             (dev->addr_len &&
2372              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2373             (dev->ifindex != dev->iflink &&
2374              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2375                 goto nla_put_failure;
2376
2377         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2378         if (!br_afspec)
2379                 goto nla_put_failure;
2380
2381         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2382             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2383                 nla_nest_cancel(skb, br_afspec);
2384                 goto nla_put_failure;
2385         }
2386         nla_nest_end(skb, br_afspec);
2387
2388         return nlmsg_end(skb, nlh);
2389 nla_put_failure:
2390         nlmsg_cancel(skb, nlh);
2391         return -EMSGSIZE;
2392 }
2393 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2394
2395 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2396 {
2397         struct net *net = sock_net(skb->sk);
2398         struct net_device *dev;
2399         int idx = 0;
2400         u32 portid = NETLINK_CB(cb->skb).portid;
2401         u32 seq = cb->nlh->nlmsg_seq;
2402         struct nlattr *extfilt;
2403         u32 filter_mask = 0;
2404
2405         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2406                                   IFLA_EXT_MASK);
2407         if (extfilt)
2408                 filter_mask = nla_get_u32(extfilt);
2409
2410         rcu_read_lock();
2411         for_each_netdev_rcu(net, dev) {
2412                 const struct net_device_ops *ops = dev->netdev_ops;
2413                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2414
2415                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2416                         if (idx >= cb->args[0] &&
2417                             br_dev->netdev_ops->ndo_bridge_getlink(
2418                                     skb, portid, seq, dev, filter_mask) < 0)
2419                                 break;
2420                         idx++;
2421                 }
2422
2423                 if (ops->ndo_bridge_getlink) {
2424                         if (idx >= cb->args[0] &&
2425                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2426                                                     filter_mask) < 0)
2427                                 break;
2428                         idx++;
2429                 }
2430         }
2431         rcu_read_unlock();
2432         cb->args[0] = idx;
2433
2434         return skb->len;
2435 }
2436
2437 static inline size_t bridge_nlmsg_size(void)
2438 {
2439         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2440                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2441                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2442                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2443                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2444                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2445                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2446                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2447                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2448                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2449                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2450 }
2451
2452 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2453 {
2454         struct net *net = dev_net(dev);
2455         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2456         struct sk_buff *skb;
2457         int err = -EOPNOTSUPP;
2458
2459         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2460         if (!skb) {
2461                 err = -ENOMEM;
2462                 goto errout;
2463         }
2464
2465         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2466             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2467                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2468                 if (err < 0)
2469                         goto errout;
2470         }
2471
2472         if ((flags & BRIDGE_FLAGS_SELF) &&
2473             dev->netdev_ops->ndo_bridge_getlink) {
2474                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2475                 if (err < 0)
2476                         goto errout;
2477         }
2478
2479         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2480         return 0;
2481 errout:
2482         WARN_ON(err == -EMSGSIZE);
2483         kfree_skb(skb);
2484         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2485         return err;
2486 }
2487
2488 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2489 {
2490         struct net *net = sock_net(skb->sk);
2491         struct ifinfomsg *ifm;
2492         struct net_device *dev;
2493         struct nlattr *br_spec, *attr = NULL;
2494         int rem, err = -EOPNOTSUPP;
2495         u16 oflags, flags = 0;
2496         bool have_flags = false;
2497
2498         if (nlmsg_len(nlh) < sizeof(*ifm))
2499                 return -EINVAL;
2500
2501         ifm = nlmsg_data(nlh);
2502         if (ifm->ifi_family != AF_BRIDGE)
2503                 return -EPFNOSUPPORT;
2504
2505         dev = __dev_get_by_index(net, ifm->ifi_index);
2506         if (!dev) {
2507                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2508                 return -ENODEV;
2509         }
2510
2511         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2512         if (br_spec) {
2513                 nla_for_each_nested(attr, br_spec, rem) {
2514                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2515                                 have_flags = true;
2516                                 flags = nla_get_u16(attr);
2517                                 break;
2518                         }
2519                 }
2520         }
2521
2522         oflags = flags;
2523
2524         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2525                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2526
2527                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2528                         err = -EOPNOTSUPP;
2529                         goto out;
2530                 }
2531
2532                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2533                 if (err)
2534                         goto out;
2535
2536                 flags &= ~BRIDGE_FLAGS_MASTER;
2537         }
2538
2539         if ((flags & BRIDGE_FLAGS_SELF)) {
2540                 if (!dev->netdev_ops->ndo_bridge_setlink)
2541                         err = -EOPNOTSUPP;
2542                 else
2543                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2544
2545                 if (!err)
2546                         flags &= ~BRIDGE_FLAGS_SELF;
2547         }
2548
2549         if (have_flags)
2550                 memcpy(nla_data(attr), &flags, sizeof(flags));
2551         /* Generate event to notify upper layer of bridge change */
2552         if (!err)
2553                 err = rtnl_bridge_notify(dev, oflags);
2554 out:
2555         return err;
2556 }
2557
2558 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2559 {
2560         struct net *net = sock_net(skb->sk);
2561         struct ifinfomsg *ifm;
2562         struct net_device *dev;
2563         struct nlattr *br_spec, *attr = NULL;
2564         int rem, err = -EOPNOTSUPP;
2565         u16 oflags, flags = 0;
2566         bool have_flags = false;
2567
2568         if (nlmsg_len(nlh) < sizeof(*ifm))
2569                 return -EINVAL;
2570
2571         ifm = nlmsg_data(nlh);
2572         if (ifm->ifi_family != AF_BRIDGE)
2573                 return -EPFNOSUPPORT;
2574
2575         dev = __dev_get_by_index(net, ifm->ifi_index);
2576         if (!dev) {
2577                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2578                 return -ENODEV;
2579         }
2580
2581         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2582         if (br_spec) {
2583                 nla_for_each_nested(attr, br_spec, rem) {
2584                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2585                                 have_flags = true;
2586                                 flags = nla_get_u16(attr);
2587                                 break;
2588                         }
2589                 }
2590         }
2591
2592         oflags = flags;
2593
2594         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2595                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2596
2597                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2598                         err = -EOPNOTSUPP;
2599                         goto out;
2600                 }
2601
2602                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2603                 if (err)
2604                         goto out;
2605
2606                 flags &= ~BRIDGE_FLAGS_MASTER;
2607         }
2608
2609         if ((flags & BRIDGE_FLAGS_SELF)) {
2610                 if (!dev->netdev_ops->ndo_bridge_dellink)
2611                         err = -EOPNOTSUPP;
2612                 else
2613                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2614
2615                 if (!err)
2616                         flags &= ~BRIDGE_FLAGS_SELF;
2617         }
2618
2619         if (have_flags)
2620                 memcpy(nla_data(attr), &flags, sizeof(flags));
2621         /* Generate event to notify upper layer of bridge change */
2622         if (!err)
2623                 err = rtnl_bridge_notify(dev, oflags);
2624 out:
2625         return err;
2626 }
2627
2628 /* Process one rtnetlink message. */
2629
2630 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2631 {
2632         struct net *net = sock_net(skb->sk);
2633         rtnl_doit_func doit;
2634         int sz_idx, kind;
2635         int family;
2636         int type;
2637         int err;
2638
2639         type = nlh->nlmsg_type;
2640         if (type > RTM_MAX)
2641                 return -EOPNOTSUPP;
2642
2643         type -= RTM_BASE;
2644
2645         /* All the messages must have at least 1 byte length */
2646         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2647                 return 0;
2648
2649         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2650         sz_idx = type>>2;
2651         kind = type&3;
2652
2653         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2654                 return -EPERM;
2655
2656         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2657                 struct sock *rtnl;
2658                 rtnl_dumpit_func dumpit;
2659                 rtnl_calcit_func calcit;
2660                 u16 min_dump_alloc = 0;
2661
2662                 dumpit = rtnl_get_dumpit(family, type);
2663                 if (dumpit == NULL)
2664                         return -EOPNOTSUPP;
2665                 calcit = rtnl_get_calcit(family, type);
2666                 if (calcit)
2667                         min_dump_alloc = calcit(skb, nlh);
2668
2669                 __rtnl_unlock();
2670                 rtnl = net->rtnl;
2671                 {
2672                         struct netlink_dump_control c = {
2673                                 .dump           = dumpit,
2674                                 .min_dump_alloc = min_dump_alloc,
2675                         };
2676                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2677                 }
2678                 rtnl_lock();
2679                 return err;
2680         }
2681
2682         doit = rtnl_get_doit(family, type);
2683         if (doit == NULL)
2684                 return -EOPNOTSUPP;
2685
2686         return doit(skb, nlh);
2687 }
2688
2689 static void rtnetlink_rcv(struct sk_buff *skb)
2690 {
2691         rtnl_lock();
2692         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2693         rtnl_unlock();
2694 }
2695
2696 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2697 {
2698         struct net_device *dev = ptr;
2699
2700         switch (event) {
2701         case NETDEV_UP:
2702         case NETDEV_DOWN:
2703         case NETDEV_PRE_UP:
2704         case NETDEV_POST_INIT:
2705         case NETDEV_REGISTER:
2706         case NETDEV_CHANGE:
2707         case NETDEV_PRE_TYPE_CHANGE:
2708         case NETDEV_GOING_DOWN:
2709         case NETDEV_UNREGISTER:
2710         case NETDEV_UNREGISTER_FINAL:
2711         case NETDEV_RELEASE:
2712         case NETDEV_JOIN:
2713                 break;
2714         default:
2715                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2716                 break;
2717         }
2718         return NOTIFY_DONE;
2719 }
2720
2721 static struct notifier_block rtnetlink_dev_notifier = {
2722         .notifier_call  = rtnetlink_event,
2723 };
2724
2725
2726 static int __net_init rtnetlink_net_init(struct net *net)
2727 {
2728         struct sock *sk;
2729         struct netlink_kernel_cfg cfg = {
2730                 .groups         = RTNLGRP_MAX,
2731                 .input          = rtnetlink_rcv,
2732                 .cb_mutex       = &rtnl_mutex,
2733                 .flags          = NL_CFG_F_NONROOT_RECV,
2734         };
2735
2736         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2737         if (!sk)
2738                 return -ENOMEM;
2739         net->rtnl = sk;
2740         return 0;
2741 }
2742
2743 static void __net_exit rtnetlink_net_exit(struct net *net)
2744 {
2745         netlink_kernel_release(net->rtnl);
2746         net->rtnl = NULL;
2747 }
2748
2749 static struct pernet_operations rtnetlink_net_ops = {
2750         .init = rtnetlink_net_init,
2751         .exit = rtnetlink_net_exit,
2752 };
2753
2754 void __init rtnetlink_init(void)
2755 {
2756         if (register_pernet_subsys(&rtnetlink_net_ops))
2757                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2758
2759         register_netdevice_notifier(&rtnetlink_dev_notifier);
2760
2761         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2762                       rtnl_dump_ifinfo, rtnl_calcit);
2763         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2764         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2765         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2766
2767         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2768         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2769
2770         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2771         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2772         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2773
2774         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2775         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2776         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2777 }
2778