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 ? tab[msgindex].doit : NULL;
 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 ? tab[msgindex].dumpit : NULL;
 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 ? tab[msgindex].calcit : NULL;
 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                         goto err_cancel_link;
 501                 err = ops->fill_info(skb, dev);
 502                 if (err < 0)
 503                         goto err_cancel_data;
 504                 nla_nest_end(skb, data);
 505         }
 506
 507         nla_nest_end(skb, linkinfo);
 508         return 0;
 509
 510 err_cancel_data:
 511         nla_nest_cancel(skb, data);
 512 err_cancel_link:
 513         nla_nest_cancel(skb, linkinfo);
 514 out:
 515         return err;
 516 }
 517
 518 static const int rtm_min[RTM_NR_FAMILIES] =
 519 {
 520         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 521         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 522         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 523         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 524         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 527         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 528         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 530 };
 531
 532 static const int rta_max[RTM_NR_FAMILIES] =
 533 {
 534         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 535         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 536         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 537         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 538         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 539         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 540         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 541         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 542 };
 543
 544 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 545 {
 546         struct sock *rtnl = net->rtnl;
 547         int err = 0;
 548
 549         NETLINK_CB(skb).dst_group = group;
 550         if (echo)
 551                 atomic_inc(&skb->users);
 552         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 553         if (echo)
 554                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 555         return err;
 556 }
 557
 558 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 559 {
 560         struct sock *rtnl = net->rtnl;
 561
 562         return nlmsg_unicast(rtnl, skb, pid);
 563 }
 564 EXPORT_SYMBOL(rtnl_unicast);
 565
 566 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 567                  struct nlmsghdr *nlh, gfp_t flags)
 568 {
 569         struct sock *rtnl = net->rtnl;
 570         int report = 0;
 571
 572         if (nlh)
 573                 report = nlmsg_report(nlh);
 574
 575         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 576 }
 577 EXPORT_SYMBOL(rtnl_notify);
 578
 579 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 580 {
 581         struct sock *rtnl = net->rtnl;
 582
 583         netlink_set_err(rtnl, 0, group, error);
 584 }
 585 EXPORT_SYMBOL(rtnl_set_sk_err);
 586
 587 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 588 {
 589         struct nlattr *mx;
 590         int i, valid = 0;
 591
 592         mx = nla_nest_start(skb, RTA_METRICS);
 593         if (mx == NULL)
 594                 return -ENOBUFS;
 595
 596         for (i = 0; i < RTAX_MAX; i++) {
 597                 if (metrics[i]) {
 598                         valid++;
 599                         if (nla_put_u32(skb, i+1, metrics[i]))
 600                                 goto nla_put_failure;
 601                 }
 602         }
 603
 604         if (!valid) {
 605                 nla_nest_cancel(skb, mx);
 606                 return 0;
 607         }
 608
 609         return nla_nest_end(skb, mx);
 610
 611 nla_put_failure:
 612         nla_nest_cancel(skb, mx);
 613         return -EMSGSIZE;
 614 }
 615 EXPORT_SYMBOL(rtnetlink_put_metrics);
 616
 617 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 618                        long expires, u32 error)
 619 {
 620         struct rta_cacheinfo ci = {
 621                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 622                 .rta_used = dst->__use,
 623                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 624                 .rta_error = error,
 625                 .rta_id =  id,
 626         };
 627
 628         if (expires)
 629                 ci.rta_expires = jiffies_to_clock_t(expires);
 630
 631         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 632 }
 633 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 634
 635 static void set_operstate(struct net_device *dev, unsigned char transition)
 636 {
 637         unsigned char operstate = dev->operstate;
 638
 639         switch (transition) {
 640         case IF_OPER_UP:
 641                 if ((operstate == IF_OPER_DORMANT ||
 642                      operstate == IF_OPER_UNKNOWN) &&
 643                     !netif_dormant(dev))
 644                         operstate = IF_OPER_UP;
 645                 break;
 646
 647         case IF_OPER_DORMANT:
 648                 if (operstate == IF_OPER_UP ||
 649                     operstate == IF_OPER_UNKNOWN)
 650                         operstate = IF_OPER_DORMANT;
 651                 break;
 652         }
 653
 654         if (dev->operstate != operstate) {
 655                 write_lock_bh(&dev_base_lock);
 656                 dev->operstate = operstate;
 657                 write_unlock_bh(&dev_base_lock);
 658                 netdev_state_change(dev);
 659         }
 660 }
 661
 662 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 663                                            const struct ifinfomsg *ifm)
 664 {
 665         unsigned int flags = ifm->ifi_flags;
 666
 667         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 668         if (ifm->ifi_change)
 669                 flags = (flags & ifm->ifi_change) |
 670                         (dev->flags & ~ifm->ifi_change);
 671
 672         return flags;
 673 }
 674
 675 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 676                                  const struct rtnl_link_stats64 *b)
 677 {
 678         a->rx_packets = b->rx_packets;
 679         a->tx_packets = b->tx_packets;
 680         a->rx_bytes = b->rx_bytes;
 681         a->tx_bytes = b->tx_bytes;
 682         a->rx_errors = b->rx_errors;
 683         a->tx_errors = b->tx_errors;
 684         a->rx_dropped = b->rx_dropped;
 685         a->tx_dropped = b->tx_dropped;
 686
 687         a->multicast = b->multicast;
 688         a->collisions = b->collisions;
 689
 690         a->rx_length_errors = b->rx_length_errors;
 691         a->rx_over_errors = b->rx_over_errors;
 692         a->rx_crc_errors = b->rx_crc_errors;
 693         a->rx_frame_errors = b->rx_frame_errors;
 694         a->rx_fifo_errors = b->rx_fifo_errors;
 695         a->rx_missed_errors = b->rx_missed_errors;
 696
 697         a->tx_aborted_errors = b->tx_aborted_errors;
 698         a->tx_carrier_errors = b->tx_carrier_errors;
 699         a->tx_fifo_errors = b->tx_fifo_errors;
 700         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 701         a->tx_window_errors = b->tx_window_errors;
 702
 703         a->rx_compressed = b->rx_compressed;
 704         a->tx_compressed = b->tx_compressed;
 705 }
 706
 707 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 708 {
 709         memcpy(v, b, sizeof(*b));
 710 }
 711
 712 /* All VF info */
 713 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 714                                    u32 ext_filter_mask)
 715 {
 716         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 717             (ext_filter_mask & RTEXT_FILTER_VF)) {
 718                 int num_vfs = dev_num_vf(dev->dev.parent);
 719                 size_t size = nla_total_size(sizeof(struct nlattr));
 720                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 721                 size += num_vfs *
 722                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 723                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 724                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 725                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 726                 return size;
 727         } else
 728                 return 0;
 729 }
 730
 731 static size_t rtnl_port_size(const struct net_device *dev)
 732 {
 733         size_t port_size = nla_total_size(4)            /* PORT_VF */
 734                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 735                 + nla_total_size(sizeof(struct ifla_port_vsi))
 736                                                         /* PORT_VSI_TYPE */
 737                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 738                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 739                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 740                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 741         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 742         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 743                 + port_size;
 744         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 745                 + port_size;
 746
 747         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 748                 return 0;
 749         if (dev_num_vf(dev->dev.parent))
 750                 return port_self_size + vf_ports_size +
 751                         vf_port_size * dev_num_vf(dev->dev.parent);
 752         else
 753                 return port_self_size;
 754 }
 755
 756 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 757                                      u32 ext_filter_mask)
 758 {
 759         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 760                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 761                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 762                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 763                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 764                + nla_total_size(sizeof(struct rtnl_link_stats))
 765                + nla_total_size(sizeof(struct rtnl_link_stats64))
 766                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 767                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 768                + nla_total_size(4) /* IFLA_TXQLEN */
 769                + nla_total_size(4) /* IFLA_WEIGHT */
 770                + nla_total_size(4) /* IFLA_MTU */
 771                + nla_total_size(4) /* IFLA_LINK */
 772                + nla_total_size(4) /* IFLA_MASTER */
 773                + nla_total_size(4) /* IFLA_PROMISCUITY */
 774                + nla_total_size(1) /* IFLA_OPERSTATE */
 775                + nla_total_size(1) /* IFLA_LINKMODE */
 776                + nla_total_size(ext_filter_mask
 777                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 778                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 779                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 780                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 781                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 782 }
 783
 784 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 785 {
 786         struct nlattr *vf_ports;
 787         struct nlattr *vf_port;
 788         int vf;
 789         int err;
 790
 791         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 792         if (!vf_ports)
 793                 return -EMSGSIZE;
 794
 795         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 796                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 797                 if (!vf_port)
 798                         goto nla_put_failure;
 799                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 800                         goto nla_put_failure;
 801                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 802                 if (err == -EMSGSIZE)
 803                         goto nla_put_failure;
 804                 if (err) {
 805                         nla_nest_cancel(skb, vf_port);
 806                         continue;
 807                 }
 808                 nla_nest_end(skb, vf_port);
 809         }
 810
 811         nla_nest_end(skb, vf_ports);
 812
 813         return 0;
 814
 815 nla_put_failure:
 816         nla_nest_cancel(skb, vf_ports);
 817         return -EMSGSIZE;
 818 }
 819
 820 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 821 {
 822         struct nlattr *port_self;
 823         int err;
 824
 825         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 826         if (!port_self)
 827                 return -EMSGSIZE;
 828
 829         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 830         if (err) {
 831                 nla_nest_cancel(skb, port_self);
 832                 return (err == -EMSGSIZE) ? err : 0;
 833         }
 834
 835         nla_nest_end(skb, port_self);
 836
 837         return 0;
 838 }
 839
 840 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 841 {
 842         int err;
 843
 844         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 845                 return 0;
 846
 847         err = rtnl_port_self_fill(skb, dev);
 848         if (err)
 849                 return err;
 850
 851         if (dev_num_vf(dev->dev.parent)) {
 852                 err = rtnl_vf_ports_fill(skb, dev);
 853                 if (err)
 854                         return err;
 855         }
 856
 857         return 0;
 858 }
 859
 860 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 861                             int type, u32 pid, u32 seq, u32 change,
 862                             unsigned int flags, u32 ext_filter_mask)
 863 {
 864         struct ifinfomsg *ifm;
 865         struct nlmsghdr *nlh;
 866         struct rtnl_link_stats64 temp;
 867         const struct rtnl_link_stats64 *stats;
 868         struct nlattr *attr, *af_spec;
 869         struct rtnl_af_ops *af_ops;
 870
 871         ASSERT_RTNL();
 872         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 873         if (nlh == NULL)
 874                 return -EMSGSIZE;
 875
 876         ifm = nlmsg_data(nlh);
 877         ifm->ifi_family = AF_UNSPEC;
 878         ifm->__ifi_pad = 0;
 879         ifm->ifi_type = dev->type;
 880         ifm->ifi_index = dev->ifindex;
 881         ifm->ifi_flags = dev_get_flags(dev);
 882         ifm->ifi_change = change;
 883
 884         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 885             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 886             nla_put_u8(skb, IFLA_OPERSTATE,
 887                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 888             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 889             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 890             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 891             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 892             (dev->ifindex != dev->iflink &&
 893              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 894             (dev->master &&
 895              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 896             (dev->qdisc &&
 897              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 898             (dev->ifalias &&
 899              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 900                 goto nla_put_failure;
 901
 902         if (1) {
 903                 struct rtnl_link_ifmap map = {
 904                         .mem_start   = dev->mem_start,
 905                         .mem_end     = dev->mem_end,
 906                         .base_addr   = dev->base_addr,
 907                         .irq         = dev->irq,
 908                         .dma         = dev->dma,
 909                         .port        = dev->if_port,
 910                 };
 911                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 912                         goto nla_put_failure;
 913         }
 914
 915         if (dev->addr_len) {
 916                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 917                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 918                         goto nla_put_failure;
 919         }
 920
 921         attr = nla_reserve(skb, IFLA_STATS,
 922                         sizeof(struct rtnl_link_stats));
 923         if (attr == NULL)
 924                 goto nla_put_failure;
 925
 926         stats = dev_get_stats(dev, &temp);
 927         copy_rtnl_link_stats(nla_data(attr), stats);
 928
 929         attr = nla_reserve(skb, IFLA_STATS64,
 930                         sizeof(struct rtnl_link_stats64));
 931         if (attr == NULL)
 932                 goto nla_put_failure;
 933         copy_rtnl_link_stats64(nla_data(attr), stats);
 934
 935         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 936             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 937                 goto nla_put_failure;
 938
 939         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 940             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 941                 int i;
 942
 943                 struct nlattr *vfinfo, *vf;
 944                 int num_vfs = dev_num_vf(dev->dev.parent);
 945
 946                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 947                 if (!vfinfo)
 948                         goto nla_put_failure;
 949                 for (i = 0; i < num_vfs; i++) {
 950                         struct ifla_vf_info ivi;
 951                         struct ifla_vf_mac vf_mac;
 952                         struct ifla_vf_vlan vf_vlan;
 953                         struct ifla_vf_tx_rate vf_tx_rate;
 954                         struct ifla_vf_spoofchk vf_spoofchk;
 955
 956                         /*
 957                          * Not all SR-IOV capable drivers support the
 958                          * spoofcheck query.  Preset to -1 so the user
 959                          * space tool can detect that the driver didn't
 960                          * report anything.
 961                          */
 962                         ivi.spoofchk = -1;
 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))
 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 hlist_node *node;
1045         struct nlattr *tb[IFLA_MAX+1];
1046         u32 ext_filter_mask = 0;
1047
1048         s_h = cb->args[0];
1049         s_idx = cb->args[1];
1050
1051         rcu_read_lock();
1052         cb->seq = net->dev_base_seq;
1053
1054         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1055                         ifla_policy) >= 0) {
1056
1057                 if (tb[IFLA_EXT_MASK])
1058                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1059         }
1060
1061         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1062                 idx = 0;
1063                 head = &net->dev_index_head[h];
1064                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1065                         if (idx < s_idx)
1066                                 goto cont;
1067                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1068                                              NETLINK_CB(cb->skb).pid,
1069                                              cb->nlh->nlmsg_seq, 0,
1070                                              NLM_F_MULTI,
1071                                              ext_filter_mask) <= 0)
1072                                 goto out;
1073
1074                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1075 cont:
1076                         idx++;
1077                 }
1078         }
1079 out:
1080         rcu_read_unlock();
1081         cb->args[1] = idx;
1082         cb->args[0] = h;
1083
1084         return skb->len;
1085 }
1086
1087 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1088         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1089         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1090         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1091         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1092         [IFLA_MTU]              = { .type = NLA_U32 },
1093         [IFLA_LINK]             = { .type = NLA_U32 },
1094         [IFLA_MASTER]           = { .type = NLA_U32 },
1095         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1096         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1097         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1098         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1099         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1100         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1101         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1102         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1103         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1104         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1105         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1106         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1107         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1108         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1109 };
1110 EXPORT_SYMBOL(ifla_policy);
1111
1112 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1113         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1114         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1115 };
1116
1117 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1118         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1119 };
1120
1121 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1122         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1123                                     .len = sizeof(struct ifla_vf_mac) },
1124         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1125                                     .len = sizeof(struct ifla_vf_vlan) },
1126         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1127                                     .len = sizeof(struct ifla_vf_tx_rate) },
1128         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1129                                     .len = sizeof(struct ifla_vf_spoofchk) },
1130 };
1131
1132 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1133         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1134         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1135                                     .len = PORT_PROFILE_MAX },
1136         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1137                                     .len = sizeof(struct ifla_port_vsi)},
1138         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1139                                       .len = PORT_UUID_MAX },
1140         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1141                                     .len = PORT_UUID_MAX },
1142         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1143         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1144 };
1145
1146 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1147 {
1148         struct net *net;
1149         /* Examine the link attributes and figure out which
1150          * network namespace we are talking about.
1151          */
1152         if (tb[IFLA_NET_NS_PID])
1153                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1154         else if (tb[IFLA_NET_NS_FD])
1155                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1156         else
1157                 net = get_net(src_net);
1158         return net;
1159 }
1160 EXPORT_SYMBOL(rtnl_link_get_net);
1161
1162 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1163 {
1164         if (dev) {
1165                 if (tb[IFLA_ADDRESS] &&
1166                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1167                         return -EINVAL;
1168
1169                 if (tb[IFLA_BROADCAST] &&
1170                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1171                         return -EINVAL;
1172         }
1173
1174         if (tb[IFLA_AF_SPEC]) {
1175                 struct nlattr *af;
1176                 int rem, err;
1177
1178                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1179                         const struct rtnl_af_ops *af_ops;
1180
1181                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1182                                 return -EAFNOSUPPORT;
1183
1184                         if (!af_ops->set_link_af)
1185                                 return -EOPNOTSUPP;
1186
1187                         if (af_ops->validate_link_af) {
1188                                 err = af_ops->validate_link_af(dev, af);
1189                                 if (err < 0)
1190                                         return err;
1191                         }
1192                 }
1193         }
1194
1195         return 0;
1196 }
1197
1198 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1199 {
1200         int rem, err = -EINVAL;
1201         struct nlattr *vf;
1202         const struct net_device_ops *ops = dev->netdev_ops;
1203
1204         nla_for_each_nested(vf, attr, rem) {
1205                 switch (nla_type(vf)) {
1206                 case IFLA_VF_MAC: {
1207                         struct ifla_vf_mac *ivm;
1208                         ivm = nla_data(vf);
1209                         err = -EOPNOTSUPP;
1210                         if (ops->ndo_set_vf_mac)
1211                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1212                                                           ivm->mac);
1213                         break;
1214                 }
1215                 case IFLA_VF_VLAN: {
1216                         struct ifla_vf_vlan *ivv;
1217                         ivv = nla_data(vf);
1218                         err = -EOPNOTSUPP;
1219                         if (ops->ndo_set_vf_vlan)
1220                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1221                                                            ivv->vlan,
1222                                                            ivv->qos);
1223                         break;
1224                 }
1225                 case IFLA_VF_TX_RATE: {
1226                         struct ifla_vf_tx_rate *ivt;
1227                         ivt = nla_data(vf);
1228                         err = -EOPNOTSUPP;
1229                         if (ops->ndo_set_vf_tx_rate)
1230                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1231                                                               ivt->rate);
1232                         break;
1233                 }
1234                 case IFLA_VF_SPOOFCHK: {
1235                         struct ifla_vf_spoofchk *ivs;
1236                         ivs = nla_data(vf);
1237                         err = -EOPNOTSUPP;
1238                         if (ops->ndo_set_vf_spoofchk)
1239                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1240                                                                ivs->setting);
1241                         break;
1242                 }
1243                 default:
1244                         err = -EINVAL;
1245                         break;
1246                 }
1247                 if (err)
1248                         break;
1249         }
1250         return err;
1251 }
1252
1253 static int do_set_master(struct net_device *dev, int ifindex)
1254 {
1255         struct net_device *master_dev;
1256         const struct net_device_ops *ops;
1257         int err;
1258
1259         if (dev->master) {
1260                 if (dev->master->ifindex == ifindex)
1261                         return 0;
1262                 ops = dev->master->netdev_ops;
1263                 if (ops->ndo_del_slave) {
1264                         err = ops->ndo_del_slave(dev->master, dev);
1265                         if (err)
1266                                 return err;
1267                 } else {
1268                         return -EOPNOTSUPP;
1269                 }
1270         }
1271
1272         if (ifindex) {
1273                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1274                 if (!master_dev)
1275                         return -EINVAL;
1276                 ops = master_dev->netdev_ops;
1277                 if (ops->ndo_add_slave) {
1278                         err = ops->ndo_add_slave(master_dev, dev);
1279                         if (err)
1280                                 return err;
1281                 } else {
1282                         return -EOPNOTSUPP;
1283                 }
1284         }
1285         return 0;
1286 }
1287
1288 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1289                       struct nlattr **tb, char *ifname, int modified)
1290 {
1291         const struct net_device_ops *ops = dev->netdev_ops;
1292         int send_addr_notify = 0;
1293         int err;
1294
1295         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1296                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1297                 if (IS_ERR(net)) {
1298                         err = PTR_ERR(net);
1299                         goto errout;
1300                 }
1301                 err = dev_change_net_namespace(dev, net, ifname);
1302                 put_net(net);
1303                 if (err)
1304                         goto errout;
1305                 modified = 1;
1306         }
1307
1308         if (tb[IFLA_MAP]) {
1309                 struct rtnl_link_ifmap *u_map;
1310                 struct ifmap k_map;
1311
1312                 if (!ops->ndo_set_config) {
1313                         err = -EOPNOTSUPP;
1314                         goto errout;
1315                 }
1316
1317                 if (!netif_device_present(dev)) {
1318                         err = -ENODEV;
1319                         goto errout;
1320                 }
1321
1322                 u_map = nla_data(tb[IFLA_MAP]);
1323                 k_map.mem_start = (unsigned long) u_map->mem_start;
1324                 k_map.mem_end = (unsigned long) u_map->mem_end;
1325                 k_map.base_addr = (unsigned short) u_map->base_addr;
1326                 k_map.irq = (unsigned char) u_map->irq;
1327                 k_map.dma = (unsigned char) u_map->dma;
1328                 k_map.port = (unsigned char) u_map->port;
1329
1330                 err = ops->ndo_set_config(dev, &k_map);
1331                 if (err < 0)
1332                         goto errout;
1333
1334                 modified = 1;
1335         }
1336
1337         if (tb[IFLA_ADDRESS]) {
1338                 struct sockaddr *sa;
1339                 int len;
1340
1341                 if (!ops->ndo_set_mac_address) {
1342                         err = -EOPNOTSUPP;
1343                         goto errout;
1344                 }
1345
1346                 if (!netif_device_present(dev)) {
1347                         err = -ENODEV;
1348                         goto errout;
1349                 }
1350
1351                 len = sizeof(sa_family_t) + dev->addr_len;
1352                 sa = kmalloc(len, GFP_KERNEL);
1353                 if (!sa) {
1354                         err = -ENOMEM;
1355                         goto errout;
1356                 }
1357                 sa->sa_family = dev->type;
1358                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1359                        dev->addr_len);
1360                 err = ops->ndo_set_mac_address(dev, sa);
1361                 kfree(sa);
1362                 if (err)
1363                         goto errout;
1364                 send_addr_notify = 1;
1365                 modified = 1;
1366         }
1367
1368         if (tb[IFLA_MTU]) {
1369                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1370                 if (err < 0)
1371                         goto errout;
1372                 modified = 1;
1373         }
1374
1375         if (tb[IFLA_GROUP]) {
1376                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1377                 modified = 1;
1378         }
1379
1380         /*
1381          * Interface selected by interface index but interface
1382          * name provided implies that a name change has been
1383          * requested.
1384          */
1385         if (ifm->ifi_index > 0 && ifname[0]) {
1386                 err = dev_change_name(dev, ifname);
1387                 if (err < 0)
1388                         goto errout;
1389                 modified = 1;
1390         }
1391
1392         if (tb[IFLA_IFALIAS]) {
1393                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1394                                     nla_len(tb[IFLA_IFALIAS]));
1395                 if (err < 0)
1396                         goto errout;
1397                 modified = 1;
1398         }
1399
1400         if (tb[IFLA_BROADCAST]) {
1401                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1402                 send_addr_notify = 1;
1403         }
1404
1405         if (ifm->ifi_flags || ifm->ifi_change) {
1406                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1407                 if (err < 0)
1408                         goto errout;
1409         }
1410
1411         if (tb[IFLA_MASTER]) {
1412                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1413                 if (err)
1414                         goto errout;
1415                 modified = 1;
1416         }
1417
1418         if (tb[IFLA_TXQLEN])
1419                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1420
1421         if (tb[IFLA_OPERSTATE])
1422                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1423
1424         if (tb[IFLA_LINKMODE]) {
1425                 write_lock_bh(&dev_base_lock);
1426                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1427                 write_unlock_bh(&dev_base_lock);
1428         }
1429
1430         if (tb[IFLA_VFINFO_LIST]) {
1431                 struct nlattr *attr;
1432                 int rem;
1433                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1434                         if (nla_type(attr) != IFLA_VF_INFO) {
1435                                 err = -EINVAL;
1436                                 goto errout;
1437                         }
1438                         err = do_setvfinfo(dev, attr);
1439                         if (err < 0)
1440                                 goto errout;
1441                         modified = 1;
1442                 }
1443         }
1444         err = 0;
1445
1446         if (tb[IFLA_VF_PORTS]) {
1447                 struct nlattr *port[IFLA_PORT_MAX+1];
1448                 struct nlattr *attr;
1449                 int vf;
1450                 int rem;
1451
1452                 err = -EOPNOTSUPP;
1453                 if (!ops->ndo_set_vf_port)
1454                         goto errout;
1455
1456                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1457                         if (nla_type(attr) != IFLA_VF_PORT)
1458                                 continue;
1459                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1460                                 attr, ifla_port_policy);
1461                         if (err < 0)
1462                                 goto errout;
1463                         if (!port[IFLA_PORT_VF]) {
1464                                 err = -EOPNOTSUPP;
1465                                 goto errout;
1466                         }
1467                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1468                         err = ops->ndo_set_vf_port(dev, vf, port);
1469                         if (err < 0)
1470                                 goto errout;
1471                         modified = 1;
1472                 }
1473         }
1474         err = 0;
1475
1476         if (tb[IFLA_PORT_SELF]) {
1477                 struct nlattr *port[IFLA_PORT_MAX+1];
1478
1479                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1480                         tb[IFLA_PORT_SELF], ifla_port_policy);
1481                 if (err < 0)
1482                         goto errout;
1483
1484                 err = -EOPNOTSUPP;
1485                 if (ops->ndo_set_vf_port)
1486                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1487                 if (err < 0)
1488                         goto errout;
1489                 modified = 1;
1490         }
1491
1492         if (tb[IFLA_AF_SPEC]) {
1493                 struct nlattr *af;
1494                 int rem;
1495
1496                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1497                         const struct rtnl_af_ops *af_ops;
1498
1499                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1500                                 BUG();
1501
1502                         err = af_ops->set_link_af(dev, af);
1503                         if (err < 0)
1504                                 goto errout;
1505
1506                         modified = 1;
1507                 }
1508         }
1509         err = 0;
1510
1511 errout:
1512         if (err < 0 && modified)
1513                 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",
1514                                      dev->name);
1515
1516         if (send_addr_notify)
1517                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1518
1519         return err;
1520 }
1521
1522 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1523 {
1524         struct net *net = sock_net(skb->sk);
1525         struct ifinfomsg *ifm;
1526         struct net_device *dev;
1527         int err;
1528         struct nlattr *tb[IFLA_MAX+1];
1529         char ifname[IFNAMSIZ];
1530
1531         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1532         if (err < 0)
1533                 goto errout;
1534
1535         if (tb[IFLA_IFNAME])
1536                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1537         else
1538                 ifname[0] = '\0';
1539
1540         err = -EINVAL;
1541         ifm = nlmsg_data(nlh);
1542         if (ifm->ifi_index > 0)
1543                 dev = __dev_get_by_index(net, ifm->ifi_index);
1544         else if (tb[IFLA_IFNAME])
1545                 dev = __dev_get_by_name(net, ifname);
1546         else
1547                 goto errout;
1548
1549         if (dev == NULL) {
1550                 err = -ENODEV;
1551                 goto errout;
1552         }
1553
1554         err = validate_linkmsg(dev, tb);
1555         if (err < 0)
1556                 goto errout;
1557
1558         err = do_setlink(dev, ifm, tb, ifname, 0);
1559 errout:
1560         return err;
1561 }
1562
1563 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1564 {
1565         struct net *net = sock_net(skb->sk);
1566         const struct rtnl_link_ops *ops;
1567         struct net_device *dev;
1568         struct ifinfomsg *ifm;
1569         char ifname[IFNAMSIZ];
1570         struct nlattr *tb[IFLA_MAX+1];
1571         int err;
1572         LIST_HEAD(list_kill);
1573
1574         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1575         if (err < 0)
1576                 return err;
1577
1578         if (tb[IFLA_IFNAME])
1579                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1580
1581         ifm = nlmsg_data(nlh);
1582         if (ifm->ifi_index > 0)
1583                 dev = __dev_get_by_index(net, ifm->ifi_index);
1584         else if (tb[IFLA_IFNAME])
1585                 dev = __dev_get_by_name(net, ifname);
1586         else
1587                 return -EINVAL;
1588
1589         if (!dev)
1590                 return -ENODEV;
1591
1592         ops = dev->rtnl_link_ops;
1593         if (!ops)
1594                 return -EOPNOTSUPP;
1595
1596         ops->dellink(dev, &list_kill);
1597         unregister_netdevice_many(&list_kill);
1598         list_del(&list_kill);
1599         return 0;
1600 }
1601
1602 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1603 {
1604         unsigned int old_flags;
1605         int err;
1606
1607         old_flags = dev->flags;
1608         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1609                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1610                 if (err < 0)
1611                         return err;
1612         }
1613
1614         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1615         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1616
1617         __dev_notify_flags(dev, old_flags);
1618         return 0;
1619 }
1620 EXPORT_SYMBOL(rtnl_configure_link);
1621
1622 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1623         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1624 {
1625         int err;
1626         struct net_device *dev;
1627         unsigned int num_queues = 1;
1628
1629         if (ops->get_tx_queues) {
1630                 err = ops->get_tx_queues(src_net, tb);
1631                 if (err < 0)
1632                         goto err;
1633                 num_queues = err;
1634         }
1635
1636         err = -ENOMEM;
1637         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1638         if (!dev)
1639                 goto err;
1640
1641         dev_net_set(dev, net);
1642         dev->rtnl_link_ops = ops;
1643         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1644
1645         if (tb[IFLA_MTU])
1646                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1647         if (tb[IFLA_ADDRESS])
1648                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1649                                 nla_len(tb[IFLA_ADDRESS]));
1650         if (tb[IFLA_BROADCAST])
1651                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1652                                 nla_len(tb[IFLA_BROADCAST]));
1653         if (tb[IFLA_TXQLEN])
1654                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1655         if (tb[IFLA_OPERSTATE])
1656                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1657         if (tb[IFLA_LINKMODE])
1658                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1659         if (tb[IFLA_GROUP])
1660                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1661
1662         return dev;
1663
1664 err:
1665         return ERR_PTR(err);
1666 }
1667 EXPORT_SYMBOL(rtnl_create_link);
1668
1669 static int rtnl_group_changelink(struct net *net, int group,
1670                 struct ifinfomsg *ifm,
1671                 struct nlattr **tb)
1672 {
1673         struct net_device *dev;
1674         int err;
1675
1676         for_each_netdev(net, dev) {
1677                 if (dev->group == group) {
1678                         err = do_setlink(dev, ifm, tb, NULL, 0);
1679                         if (err < 0)
1680                                 return err;
1681                 }
1682         }
1683
1684         return 0;
1685 }
1686
1687 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1688 {
1689         struct net *net = sock_net(skb->sk);
1690         const struct rtnl_link_ops *ops;
1691         struct net_device *dev;
1692         struct ifinfomsg *ifm;
1693         char kind[MODULE_NAME_LEN];
1694         char ifname[IFNAMSIZ];
1695         struct nlattr *tb[IFLA_MAX+1];
1696         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1697         int err;
1698
1699 #ifdef CONFIG_MODULES
1700 replay:
1701 #endif
1702         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1703         if (err < 0)
1704                 return err;
1705
1706         if (tb[IFLA_IFNAME])
1707                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1708         else
1709                 ifname[0] = '\0';
1710
1711         ifm = nlmsg_data(nlh);
1712         if (ifm->ifi_index > 0)
1713                 dev = __dev_get_by_index(net, ifm->ifi_index);
1714         else {
1715                 if (ifname[0])
1716                         dev = __dev_get_by_name(net, ifname);
1717                 else
1718                         dev = NULL;
1719         }
1720
1721         err = validate_linkmsg(dev, tb);
1722         if (err < 0)
1723                 return err;
1724
1725         if (tb[IFLA_LINKINFO]) {
1726                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1727                                        tb[IFLA_LINKINFO], ifla_info_policy);
1728                 if (err < 0)
1729                         return err;
1730         } else
1731                 memset(linkinfo, 0, sizeof(linkinfo));
1732
1733         if (linkinfo[IFLA_INFO_KIND]) {
1734                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1735                 ops = rtnl_link_ops_get(kind);
1736         } else {
1737                 kind[0] = '\0';
1738                 ops = NULL;
1739         }
1740
1741         if (1) {
1742                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1743                 struct net *dest_net;
1744
1745                 if (ops) {
1746                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1747                                 err = nla_parse_nested(attr, ops->maxtype,
1748                                                        linkinfo[IFLA_INFO_DATA],
1749                                                        ops->policy);
1750                                 if (err < 0)
1751                                         return err;
1752                                 data = attr;
1753                         }
1754                         if (ops->validate) {
1755                                 err = ops->validate(tb, data);
1756                                 if (err < 0)
1757                                         return err;
1758                         }
1759                 }
1760
1761                 if (dev) {
1762                         int modified = 0;
1763
1764                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1765                                 return -EEXIST;
1766                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1767                                 return -EOPNOTSUPP;
1768
1769                         if (linkinfo[IFLA_INFO_DATA]) {
1770                                 if (!ops || ops != dev->rtnl_link_ops ||
1771                                     !ops->changelink)
1772                                         return -EOPNOTSUPP;
1773
1774                                 err = ops->changelink(dev, tb, data);
1775                                 if (err < 0)
1776                                         return err;
1777                                 modified = 1;
1778                         }
1779
1780                         return do_setlink(dev, ifm, tb, ifname, modified);
1781                 }
1782
1783                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1784                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1785                                 return rtnl_group_changelink(net,
1786                                                 nla_get_u32(tb[IFLA_GROUP]),
1787                                                 ifm, tb);
1788                         return -ENODEV;
1789                 }
1790
1791                 if (ifm->ifi_index)
1792                         return -EOPNOTSUPP;
1793                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1794                         return -EOPNOTSUPP;
1795
1796                 if (!ops) {
1797 #ifdef CONFIG_MODULES
1798                         if (kind[0]) {
1799                                 __rtnl_unlock();
1800                                 request_module("rtnl-link-%s", kind);
1801                                 rtnl_lock();
1802                                 ops = rtnl_link_ops_get(kind);
1803                                 if (ops)
1804                                         goto replay;
1805                         }
1806 #endif
1807                         return -EOPNOTSUPP;
1808                 }
1809
1810                 if (!ifname[0])
1811                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1812
1813                 dest_net = rtnl_link_get_net(net, tb);
1814                 if (IS_ERR(dest_net))
1815                         return PTR_ERR(dest_net);
1816
1817                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1818
1819                 if (IS_ERR(dev))
1820                         err = PTR_ERR(dev);
1821                 else if (ops->newlink)
1822                         err = ops->newlink(net, dev, tb, data);
1823                 else
1824                         err = register_netdevice(dev);
1825
1826                 if (err < 0 && !IS_ERR(dev))
1827                         free_netdev(dev);
1828                 if (err < 0)
1829                         goto out;
1830
1831                 err = rtnl_configure_link(dev, ifm);
1832                 if (err < 0)
1833                         unregister_netdevice(dev);
1834 out:
1835                 put_net(dest_net);
1836                 return err;
1837         }
1838 }
1839
1840 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1841 {
1842         struct net *net = sock_net(skb->sk);
1843         struct ifinfomsg *ifm;
1844         char ifname[IFNAMSIZ];
1845         struct nlattr *tb[IFLA_MAX+1];
1846         struct net_device *dev = NULL;
1847         struct sk_buff *nskb;
1848         int err;
1849         u32 ext_filter_mask = 0;
1850
1851         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1852         if (err < 0)
1853                 return err;
1854
1855         if (tb[IFLA_IFNAME])
1856                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1857
1858         if (tb[IFLA_EXT_MASK])
1859                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1860
1861         ifm = nlmsg_data(nlh);
1862         if (ifm->ifi_index > 0)
1863                 dev = __dev_get_by_index(net, ifm->ifi_index);
1864         else if (tb[IFLA_IFNAME])
1865                 dev = __dev_get_by_name(net, ifname);
1866         else
1867                 return -EINVAL;
1868
1869         if (dev == NULL)
1870                 return -ENODEV;
1871
1872         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1873         if (nskb == NULL)
1874                 return -ENOBUFS;
1875
1876         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1877                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1878         if (err < 0) {
1879                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1880                 WARN_ON(err == -EMSGSIZE);
1881                 kfree_skb(nskb);
1882         } else
1883                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1884
1885         return err;
1886 }
1887
1888 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1889 {
1890         struct net *net = sock_net(skb->sk);
1891         struct net_device *dev;
1892         struct nlattr *tb[IFLA_MAX+1];
1893         u32 ext_filter_mask = 0;
1894         u16 min_ifinfo_dump_size = 0;
1895
1896         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1897                         ifla_policy) >= 0) {
1898                 if (tb[IFLA_EXT_MASK])
1899                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1900         }
1901
1902         if (!ext_filter_mask)
1903                 return NLMSG_GOODSIZE;
1904         /*
1905          * traverse the list of net devices and compute the minimum
1906          * buffer size based upon the filter mask.
1907          */
1908         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1909                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1910                                              if_nlmsg_size(dev,
1911                                                            ext_filter_mask));
1912         }
1913
1914         return min_ifinfo_dump_size;
1915 }
1916
1917 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1918 {
1919         int idx;
1920         int s_idx = cb->family;
1921
1922         if (s_idx == 0)
1923                 s_idx = 1;
1924         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1925                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1926                 if (idx < s_idx || idx == PF_PACKET)
1927                         continue;
1928                 if (rtnl_msg_handlers[idx] == NULL ||
1929                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1930                         continue;
1931                 if (idx > s_idx)
1932                         memset(&cb->args[0], 0, sizeof(cb->args));
1933                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1934                         break;
1935         }
1936         cb->family = idx;
1937
1938         return skb->len;
1939 }
1940
1941 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1942 {
1943         struct net *net = dev_net(dev);
1944         struct sk_buff *skb;
1945         int err = -ENOBUFS;
1946         size_t if_info_size;
1947
1948         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1949         if (skb == NULL)
1950                 goto errout;
1951
1952         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1953         if (err < 0) {
1954                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1955                 WARN_ON(err == -EMSGSIZE);
1956                 kfree_skb(skb);
1957                 goto errout;
1958         }
1959         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1960         return;
1961 errout:
1962         if (err < 0)
1963                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1964 }
1965
1966 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1967                                    struct net_device *dev,
1968                                    u8 *addr, u32 pid, u32 seq,
1969                                    int type, unsigned int flags)
1970 {
1971         struct nlmsghdr *nlh;
1972         struct ndmsg *ndm;
1973
1974         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1975         if (!nlh)
1976                 return -EMSGSIZE;
1977
1978         ndm = nlmsg_data(nlh);
1979         ndm->ndm_family  = AF_BRIDGE;
1980         ndm->ndm_pad1    = 0;
1981         ndm->ndm_pad2    = 0;
1982         ndm->ndm_flags   = flags;
1983         ndm->ndm_type    = 0;
1984         ndm->ndm_ifindex = dev->ifindex;
1985         ndm->ndm_state   = NUD_PERMANENT;
1986
1987         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
1988                 goto nla_put_failure;
1989
1990         return nlmsg_end(skb, nlh);
1991
1992 nla_put_failure:
1993         nlmsg_cancel(skb, nlh);
1994         return -EMSGSIZE;
1995 }
1996
1997 static inline size_t rtnl_fdb_nlmsg_size(void)
1998 {
1999         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2000 }
2001
2002 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2003 {
2004         struct net *net = dev_net(dev);
2005         struct sk_buff *skb;
2006         int err = -ENOBUFS;
2007
2008         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2009         if (!skb)
2010                 goto errout;
2011
2012         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2013         if (err < 0) {
2014                 kfree_skb(skb);
2015                 goto errout;
2016         }
2017
2018         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2019         return;
2020 errout:
2021         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2022 }
2023
2024 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2025 {
2026         struct net *net = sock_net(skb->sk);
2027         struct net_device *master = NULL;
2028         struct ndmsg *ndm;
2029         struct nlattr *tb[NDA_MAX+1];
2030         struct net_device *dev;
2031         u8 *addr;
2032         int err;
2033
2034         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2035         if (err < 0)
2036                 return err;
2037
2038         ndm = nlmsg_data(nlh);
2039         if (ndm->ndm_ifindex == 0) {
2040                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2041                 return -EINVAL;
2042         }
2043
2044         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2045         if (dev == NULL) {
2046                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2047                 return -ENODEV;
2048         }
2049
2050         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2051                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2052                 return -EINVAL;
2053         }
2054
2055         addr = nla_data(tb[NDA_LLADDR]);
2056         if (!is_valid_ether_addr(addr)) {
2057                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2058                 return -EINVAL;
2059         }
2060
2061         err = -EOPNOTSUPP;
2062
2063         /* Support fdb on master device the net/bridge default case */
2064         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2065             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2066                 master = dev->master;
2067                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2068                                                       nlh->nlmsg_flags);
2069                 if (err)
2070                         goto out;
2071                 else
2072                         ndm->ndm_flags &= ~NTF_MASTER;
2073         }
2074
2075         /* Embedded bridge, macvlan, and any other device support */
2076         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2077                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2078                                                    nlh->nlmsg_flags);
2079
2080                 if (!err) {
2081                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2082                         ndm->ndm_flags &= ~NTF_SELF;
2083                 }
2084         }
2085 out:
2086         return err;
2087 }
2088
2089 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2090 {
2091         struct net *net = sock_net(skb->sk);
2092         struct ndmsg *ndm;
2093         struct nlattr *llattr;
2094         struct net_device *dev;
2095         int err = -EINVAL;
2096         __u8 *addr;
2097
2098         if (nlmsg_len(nlh) < sizeof(*ndm))
2099                 return -EINVAL;
2100
2101         ndm = nlmsg_data(nlh);
2102         if (ndm->ndm_ifindex == 0) {
2103                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2104                 return -EINVAL;
2105         }
2106
2107         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2108         if (dev == NULL) {
2109                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2110                 return -ENODEV;
2111         }
2112
2113         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2114         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2115                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2116                 return -EINVAL;
2117         }
2118
2119         addr = nla_data(llattr);
2120         err = -EOPNOTSUPP;
2121
2122         /* Support fdb on master device the net/bridge default case */
2123         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2124             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2125                 struct net_device *master = dev->master;
2126
2127                 if (master->netdev_ops->ndo_fdb_del)
2128                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2129
2130                 if (err)
2131                         goto out;
2132                 else
2133                         ndm->ndm_flags &= ~NTF_MASTER;
2134         }
2135
2136         /* Embedded bridge, macvlan, and any other device support */
2137         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2138                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2139
2140                 if (!err) {
2141                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2142                         ndm->ndm_flags &= ~NTF_SELF;
2143                 }
2144         }
2145 out:
2146         return err;
2147 }
2148
2149 static int nlmsg_populate_fdb(struct sk_buff *skb,
2150                               struct netlink_callback *cb,
2151                               struct net_device *dev,
2152                               int *idx,
2153                               struct netdev_hw_addr_list *list)
2154 {
2155         struct netdev_hw_addr *ha;
2156         int err;
2157         u32 pid, seq;
2158
2159         pid = NETLINK_CB(cb->skb).pid;
2160         seq = cb->nlh->nlmsg_seq;
2161
2162         list_for_each_entry(ha, &list->list, list) {
2163                 if (*idx < cb->args[0])
2164                         goto skip;
2165
2166                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2167                                               pid, seq, 0, NTF_SELF);
2168                 if (err < 0)
2169                         return err;
2170 skip:
2171                 *idx += 1;
2172         }
2173         return 0;
2174 }
2175
2176 /**
2177  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2178  * @nlh: netlink message header
2179  * @dev: netdevice
2180  *
2181  * Default netdevice operation to dump the existing unicast address list.
2182  * Returns zero on success.
2183  */
2184 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2185                       struct netlink_callback *cb,
2186                       struct net_device *dev,
2187                       int idx)
2188 {
2189         int err;
2190
2191         netif_addr_lock_bh(dev);
2192         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2193         if (err)
2194                 goto out;
2195         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2196 out:
2197         netif_addr_unlock_bh(dev);
2198         return idx;
2199 }
2200 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2201
2202 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2203 {
2204         int idx = 0;
2205         struct net *net = sock_net(skb->sk);
2206         struct net_device *dev;
2207
2208         rcu_read_lock();
2209         for_each_netdev_rcu(net, dev) {
2210                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2211                         struct net_device *master = dev->master;
2212                         const struct net_device_ops *ops = master->netdev_ops;
2213
2214                         if (ops->ndo_fdb_dump)
2215                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2216                 }
2217
2218                 if (dev->netdev_ops->ndo_fdb_dump)
2219                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2220         }
2221         rcu_read_unlock();
2222
2223         cb->args[0] = idx;
2224         return skb->len;
2225 }
2226
2227 /* Protected by RTNL sempahore.  */
2228 static struct rtattr **rta_buf;
2229 static int rtattr_max;
2230
2231 /* Process one rtnetlink message. */
2232
2233 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2234 {
2235         struct net *net = sock_net(skb->sk);
2236         rtnl_doit_func doit;
2237         int sz_idx, kind;
2238         int min_len;
2239         int family;
2240         int type;
2241         int err;
2242
2243         type = nlh->nlmsg_type;
2244         if (type > RTM_MAX)
2245                 return -EOPNOTSUPP;
2246
2247         type -= RTM_BASE;
2248
2249         /* All the messages must have at least 1 byte length */
2250         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2251                 return 0;
2252
2253         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2254         sz_idx = type>>2;
2255         kind = type&3;
2256
2257         if (kind != 2 && !capable(CAP_NET_ADMIN))
2258                 return -EPERM;
2259
2260         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2261                 struct sock *rtnl;
2262                 rtnl_dumpit_func dumpit;
2263                 rtnl_calcit_func calcit;
2264                 u16 min_dump_alloc = 0;
2265
2266                 dumpit = rtnl_get_dumpit(family, type);
2267                 if (dumpit == NULL)
2268                         return -EOPNOTSUPP;
2269                 calcit = rtnl_get_calcit(family, type);
2270                 if (calcit)
2271                         min_dump_alloc = calcit(skb, nlh);
2272
2273                 __rtnl_unlock();
2274                 rtnl = net->rtnl;
2275                 {
2276                         struct netlink_dump_control c = {
2277                                 .dump           = dumpit,
2278                                 .min_dump_alloc = min_dump_alloc,
2279                         };
2280                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2281                 }
2282                 rtnl_lock();
2283                 return err;
2284         }
2285
2286         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2287
2288         min_len = rtm_min[sz_idx];
2289         if (nlh->nlmsg_len < min_len)
2290                 return -EINVAL;
2291
2292         if (nlh->nlmsg_len > min_len) {
2293                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2294                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2295
2296                 while (RTA_OK(attr, attrlen)) {
2297                         unsigned int flavor = attr->rta_type;
2298                         if (flavor) {
2299                                 if (flavor > rta_max[sz_idx])
2300                                         return -EINVAL;
2301                                 rta_buf[flavor-1] = attr;
2302                         }
2303                         attr = RTA_NEXT(attr, attrlen);
2304                 }
2305         }
2306
2307         doit = rtnl_get_doit(family, type);
2308         if (doit == NULL)
2309                 return -EOPNOTSUPP;
2310
2311         return doit(skb, nlh, (void *)&rta_buf[0]);
2312 }
2313
2314 static void rtnetlink_rcv(struct sk_buff *skb)
2315 {
2316         rtnl_lock();
2317         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2318         rtnl_unlock();
2319 }
2320
2321 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2322 {
2323         struct net_device *dev = ptr;
2324
2325         switch (event) {
2326         case NETDEV_UP:
2327         case NETDEV_DOWN:
2328         case NETDEV_PRE_UP:
2329         case NETDEV_POST_INIT:
2330         case NETDEV_REGISTER:
2331         case NETDEV_CHANGE:
2332         case NETDEV_PRE_TYPE_CHANGE:
2333         case NETDEV_GOING_DOWN:
2334         case NETDEV_UNREGISTER:
2335         case NETDEV_UNREGISTER_BATCH:
2336         case NETDEV_RELEASE:
2337         case NETDEV_JOIN:
2338                 break;
2339         default:
2340                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2341                 break;
2342         }
2343         return NOTIFY_DONE;
2344 }
2345
2346 static struct notifier_block rtnetlink_dev_notifier = {
2347         .notifier_call  = rtnetlink_event,
2348 };
2349
2350
2351 static int __net_init rtnetlink_net_init(struct net *net)
2352 {
2353         struct sock *sk;
2354         struct netlink_kernel_cfg cfg = {
2355                 .groups         = RTNLGRP_MAX,
2356                 .input          = rtnetlink_rcv,
2357                 .cb_mutex       = &rtnl_mutex,
2358         };
2359
2360         sk = netlink_kernel_create(net, NETLINK_ROUTE, THIS_MODULE, &cfg);
2361         if (!sk)
2362                 return -ENOMEM;
2363         net->rtnl = sk;
2364         return 0;
2365 }
2366
2367 static void __net_exit rtnetlink_net_exit(struct net *net)
2368 {
2369         netlink_kernel_release(net->rtnl);
2370         net->rtnl = NULL;
2371 }
2372
2373 static struct pernet_operations rtnetlink_net_ops = {
2374         .init = rtnetlink_net_init,
2375         .exit = rtnetlink_net_exit,
2376 };
2377
2378 void __init rtnetlink_init(void)
2379 {
2380         int i;
2381
2382         rtattr_max = 0;
2383         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2384                 if (rta_max[i] > rtattr_max)
2385                         rtattr_max = rta_max[i];
2386         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2387         if (!rta_buf)
2388                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2389
2390         if (register_pernet_subsys(&rtnetlink_net_ops))
2391                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2392
2393         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2394         register_netdevice_notifier(&rtnetlink_dev_notifier);
2395
2396         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2397                       rtnl_dump_ifinfo, rtnl_calcit);
2398         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2399         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2400         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2401
2402         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2403         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2404
2405         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2406         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2407         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2408 }
2409