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(4) /* IFLA_NUM_TX_QUEUES */
 775                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 776                + nla_total_size(1) /* IFLA_OPERSTATE */
 777                + nla_total_size(1) /* IFLA_LINKMODE */
 778                + nla_total_size(ext_filter_mask
 779                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 780                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 781                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 782                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 783                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 784 }
 785
 786 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 787 {
 788         struct nlattr *vf_ports;
 789         struct nlattr *vf_port;
 790         int vf;
 791         int err;
 792
 793         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 794         if (!vf_ports)
 795                 return -EMSGSIZE;
 796
 797         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 798                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 799                 if (!vf_port)
 800                         goto nla_put_failure;
 801                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 802                         goto nla_put_failure;
 803                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 804                 if (err == -EMSGSIZE)
 805                         goto nla_put_failure;
 806                 if (err) {
 807                         nla_nest_cancel(skb, vf_port);
 808                         continue;
 809                 }
 810                 nla_nest_end(skb, vf_port);
 811         }
 812
 813         nla_nest_end(skb, vf_ports);
 814
 815         return 0;
 816
 817 nla_put_failure:
 818         nla_nest_cancel(skb, vf_ports);
 819         return -EMSGSIZE;
 820 }
 821
 822 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 823 {
 824         struct nlattr *port_self;
 825         int err;
 826
 827         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 828         if (!port_self)
 829                 return -EMSGSIZE;
 830
 831         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 832         if (err) {
 833                 nla_nest_cancel(skb, port_self);
 834                 return (err == -EMSGSIZE) ? err : 0;
 835         }
 836
 837         nla_nest_end(skb, port_self);
 838
 839         return 0;
 840 }
 841
 842 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 843 {
 844         int err;
 845
 846         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 847                 return 0;
 848
 849         err = rtnl_port_self_fill(skb, dev);
 850         if (err)
 851                 return err;
 852
 853         if (dev_num_vf(dev->dev.parent)) {
 854                 err = rtnl_vf_ports_fill(skb, dev);
 855                 if (err)
 856                         return err;
 857         }
 858
 859         return 0;
 860 }
 861
 862 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 863                             int type, u32 pid, u32 seq, u32 change,
 864                             unsigned int flags, u32 ext_filter_mask)
 865 {
 866         struct ifinfomsg *ifm;
 867         struct nlmsghdr *nlh;
 868         struct rtnl_link_stats64 temp;
 869         const struct rtnl_link_stats64 *stats;
 870         struct nlattr *attr, *af_spec;
 871         struct rtnl_af_ops *af_ops;
 872
 873         ASSERT_RTNL();
 874         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 875         if (nlh == NULL)
 876                 return -EMSGSIZE;
 877
 878         ifm = nlmsg_data(nlh);
 879         ifm->ifi_family = AF_UNSPEC;
 880         ifm->__ifi_pad = 0;
 881         ifm->ifi_type = dev->type;
 882         ifm->ifi_index = dev->ifindex;
 883         ifm->ifi_flags = dev_get_flags(dev);
 884         ifm->ifi_change = change;
 885
 886         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 887             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 888             nla_put_u8(skb, IFLA_OPERSTATE,
 889                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 890             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 891             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 892             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 893             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 894             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 895 #ifdef CONFIG_RPS
 896             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 897 #endif
 898             (dev->ifindex != dev->iflink &&
 899              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 900             (dev->master &&
 901              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 902             (dev->qdisc &&
 903              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 904             (dev->ifalias &&
 905              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 906                 goto nla_put_failure;
 907
 908         if (1) {
 909                 struct rtnl_link_ifmap map = {
 910                         .mem_start   = dev->mem_start,
 911                         .mem_end     = dev->mem_end,
 912                         .base_addr   = dev->base_addr,
 913                         .irq         = dev->irq,
 914                         .dma         = dev->dma,
 915                         .port        = dev->if_port,
 916                 };
 917                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 918                         goto nla_put_failure;
 919         }
 920
 921         if (dev->addr_len) {
 922                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 923                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 924                         goto nla_put_failure;
 925         }
 926
 927         attr = nla_reserve(skb, IFLA_STATS,
 928                         sizeof(struct rtnl_link_stats));
 929         if (attr == NULL)
 930                 goto nla_put_failure;
 931
 932         stats = dev_get_stats(dev, &temp);
 933         copy_rtnl_link_stats(nla_data(attr), stats);
 934
 935         attr = nla_reserve(skb, IFLA_STATS64,
 936                         sizeof(struct rtnl_link_stats64));
 937         if (attr == NULL)
 938                 goto nla_put_failure;
 939         copy_rtnl_link_stats64(nla_data(attr), stats);
 940
 941         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 942             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 943                 goto nla_put_failure;
 944
 945         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 946             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 947                 int i;
 948
 949                 struct nlattr *vfinfo, *vf;
 950                 int num_vfs = dev_num_vf(dev->dev.parent);
 951
 952                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 953                 if (!vfinfo)
 954                         goto nla_put_failure;
 955                 for (i = 0; i < num_vfs; i++) {
 956                         struct ifla_vf_info ivi;
 957                         struct ifla_vf_mac vf_mac;
 958                         struct ifla_vf_vlan vf_vlan;
 959                         struct ifla_vf_tx_rate vf_tx_rate;
 960                         struct ifla_vf_spoofchk vf_spoofchk;
 961
 962                         /*
 963                          * Not all SR-IOV capable drivers support the
 964                          * spoofcheck query.  Preset to -1 so the user
 965                          * space tool can detect that the driver didn't
 966                          * report anything.
 967                          */
 968                         ivi.spoofchk = -1;
 969                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 970                                 break;
 971                         vf_mac.vf =
 972                                 vf_vlan.vf =
 973                                 vf_tx_rate.vf =
 974                                 vf_spoofchk.vf = ivi.vf;
 975
 976                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 977                         vf_vlan.vlan = ivi.vlan;
 978                         vf_vlan.qos = ivi.qos;
 979                         vf_tx_rate.rate = ivi.tx_rate;
 980                         vf_spoofchk.setting = ivi.spoofchk;
 981                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 982                         if (!vf) {
 983                                 nla_nest_cancel(skb, vfinfo);
 984                                 goto nla_put_failure;
 985                         }
 986                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 987                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 988                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 989                                     &vf_tx_rate) ||
 990                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 991                                     &vf_spoofchk))
 992                                 goto nla_put_failure;
 993                         nla_nest_end(skb, vf);
 994                 }
 995                 nla_nest_end(skb, vfinfo);
 996         }
 997
 998         if (rtnl_port_fill(skb, dev))
 999                 goto nla_put_failure;
1000
1001         if (dev->rtnl_link_ops) {
1002                 if (rtnl_link_fill(skb, dev) < 0)
1003                         goto nla_put_failure;
1004         }
1005
1006         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1007                 goto nla_put_failure;
1008
1009         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1010                 if (af_ops->fill_link_af) {
1011                         struct nlattr *af;
1012                         int err;
1013
1014                         if (!(af = nla_nest_start(skb, af_ops->family)))
1015                                 goto nla_put_failure;
1016
1017                         err = af_ops->fill_link_af(skb, dev);
1018
1019                         /*
1020                          * Caller may return ENODATA to indicate that there
1021                          * was no data to be dumped. This is not an error, it
1022                          * means we should trim the attribute header and
1023                          * continue.
1024                          */
1025                         if (err == -ENODATA)
1026                                 nla_nest_cancel(skb, af);
1027                         else if (err < 0)
1028                                 goto nla_put_failure;
1029
1030                         nla_nest_end(skb, af);
1031                 }
1032         }
1033
1034         nla_nest_end(skb, af_spec);
1035
1036         return nlmsg_end(skb, nlh);
1037
1038 nla_put_failure:
1039         nlmsg_cancel(skb, nlh);
1040         return -EMSGSIZE;
1041 }
1042
1043 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1044 {
1045         struct net *net = sock_net(skb->sk);
1046         int h, s_h;
1047         int idx = 0, s_idx;
1048         struct net_device *dev;
1049         struct hlist_head *head;
1050         struct hlist_node *node;
1051         struct nlattr *tb[IFLA_MAX+1];
1052         u32 ext_filter_mask = 0;
1053
1054         s_h = cb->args[0];
1055         s_idx = cb->args[1];
1056
1057         rcu_read_lock();
1058         cb->seq = net->dev_base_seq;
1059
1060         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1061                         ifla_policy) >= 0) {
1062
1063                 if (tb[IFLA_EXT_MASK])
1064                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1065         }
1066
1067         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1068                 idx = 0;
1069                 head = &net->dev_index_head[h];
1070                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1071                         if (idx < s_idx)
1072                                 goto cont;
1073                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1074                                              NETLINK_CB(cb->skb).pid,
1075                                              cb->nlh->nlmsg_seq, 0,
1076                                              NLM_F_MULTI,
1077                                              ext_filter_mask) <= 0)
1078                                 goto out;
1079
1080                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1081 cont:
1082                         idx++;
1083                 }
1084         }
1085 out:
1086         rcu_read_unlock();
1087         cb->args[1] = idx;
1088         cb->args[0] = h;
1089
1090         return skb->len;
1091 }
1092
1093 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1094         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1095         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1096         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1097         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1098         [IFLA_MTU]              = { .type = NLA_U32 },
1099         [IFLA_LINK]             = { .type = NLA_U32 },
1100         [IFLA_MASTER]           = { .type = NLA_U32 },
1101         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1102         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1103         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1104         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1105         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1106         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1107         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1108         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1109         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1110         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1111         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1112         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1113         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1114         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1115         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1116         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1117 };
1118 EXPORT_SYMBOL(ifla_policy);
1119
1120 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1121         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1122         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1123 };
1124
1125 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1126         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1127 };
1128
1129 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1130         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1131                                     .len = sizeof(struct ifla_vf_mac) },
1132         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1133                                     .len = sizeof(struct ifla_vf_vlan) },
1134         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1135                                     .len = sizeof(struct ifla_vf_tx_rate) },
1136         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1137                                     .len = sizeof(struct ifla_vf_spoofchk) },
1138 };
1139
1140 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1141         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1142         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1143                                     .len = PORT_PROFILE_MAX },
1144         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1145                                     .len = sizeof(struct ifla_port_vsi)},
1146         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1147                                       .len = PORT_UUID_MAX },
1148         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1149                                     .len = PORT_UUID_MAX },
1150         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1151         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1152 };
1153
1154 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1155 {
1156         struct net *net;
1157         /* Examine the link attributes and figure out which
1158          * network namespace we are talking about.
1159          */
1160         if (tb[IFLA_NET_NS_PID])
1161                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1162         else if (tb[IFLA_NET_NS_FD])
1163                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1164         else
1165                 net = get_net(src_net);
1166         return net;
1167 }
1168 EXPORT_SYMBOL(rtnl_link_get_net);
1169
1170 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1171 {
1172         if (dev) {
1173                 if (tb[IFLA_ADDRESS] &&
1174                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1175                         return -EINVAL;
1176
1177                 if (tb[IFLA_BROADCAST] &&
1178                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1179                         return -EINVAL;
1180         }
1181
1182         if (tb[IFLA_AF_SPEC]) {
1183                 struct nlattr *af;
1184                 int rem, err;
1185
1186                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1187                         const struct rtnl_af_ops *af_ops;
1188
1189                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1190                                 return -EAFNOSUPPORT;
1191
1192                         if (!af_ops->set_link_af)
1193                                 return -EOPNOTSUPP;
1194
1195                         if (af_ops->validate_link_af) {
1196                                 err = af_ops->validate_link_af(dev, af);
1197                                 if (err < 0)
1198                                         return err;
1199                         }
1200                 }
1201         }
1202
1203         return 0;
1204 }
1205
1206 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1207 {
1208         int rem, err = -EINVAL;
1209         struct nlattr *vf;
1210         const struct net_device_ops *ops = dev->netdev_ops;
1211
1212         nla_for_each_nested(vf, attr, rem) {
1213                 switch (nla_type(vf)) {
1214                 case IFLA_VF_MAC: {
1215                         struct ifla_vf_mac *ivm;
1216                         ivm = nla_data(vf);
1217                         err = -EOPNOTSUPP;
1218                         if (ops->ndo_set_vf_mac)
1219                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1220                                                           ivm->mac);
1221                         break;
1222                 }
1223                 case IFLA_VF_VLAN: {
1224                         struct ifla_vf_vlan *ivv;
1225                         ivv = nla_data(vf);
1226                         err = -EOPNOTSUPP;
1227                         if (ops->ndo_set_vf_vlan)
1228                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1229                                                            ivv->vlan,
1230                                                            ivv->qos);
1231                         break;
1232                 }
1233                 case IFLA_VF_TX_RATE: {
1234                         struct ifla_vf_tx_rate *ivt;
1235                         ivt = nla_data(vf);
1236                         err = -EOPNOTSUPP;
1237                         if (ops->ndo_set_vf_tx_rate)
1238                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1239                                                               ivt->rate);
1240                         break;
1241                 }
1242                 case IFLA_VF_SPOOFCHK: {
1243                         struct ifla_vf_spoofchk *ivs;
1244                         ivs = nla_data(vf);
1245                         err = -EOPNOTSUPP;
1246                         if (ops->ndo_set_vf_spoofchk)
1247                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1248                                                                ivs->setting);
1249                         break;
1250                 }
1251                 default:
1252                         err = -EINVAL;
1253                         break;
1254                 }
1255                 if (err)
1256                         break;
1257         }
1258         return err;
1259 }
1260
1261 static int do_set_master(struct net_device *dev, int ifindex)
1262 {
1263         struct net_device *master_dev;
1264         const struct net_device_ops *ops;
1265         int err;
1266
1267         if (dev->master) {
1268                 if (dev->master->ifindex == ifindex)
1269                         return 0;
1270                 ops = dev->master->netdev_ops;
1271                 if (ops->ndo_del_slave) {
1272                         err = ops->ndo_del_slave(dev->master, dev);
1273                         if (err)
1274                                 return err;
1275                 } else {
1276                         return -EOPNOTSUPP;
1277                 }
1278         }
1279
1280         if (ifindex) {
1281                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1282                 if (!master_dev)
1283                         return -EINVAL;
1284                 ops = master_dev->netdev_ops;
1285                 if (ops->ndo_add_slave) {
1286                         err = ops->ndo_add_slave(master_dev, dev);
1287                         if (err)
1288                                 return err;
1289                 } else {
1290                         return -EOPNOTSUPP;
1291                 }
1292         }
1293         return 0;
1294 }
1295
1296 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1297                       struct nlattr **tb, char *ifname, int modified)
1298 {
1299         const struct net_device_ops *ops = dev->netdev_ops;
1300         int send_addr_notify = 0;
1301         int err;
1302
1303         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1304                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1305                 if (IS_ERR(net)) {
1306                         err = PTR_ERR(net);
1307                         goto errout;
1308                 }
1309                 err = dev_change_net_namespace(dev, net, ifname);
1310                 put_net(net);
1311                 if (err)
1312                         goto errout;
1313                 modified = 1;
1314         }
1315
1316         if (tb[IFLA_MAP]) {
1317                 struct rtnl_link_ifmap *u_map;
1318                 struct ifmap k_map;
1319
1320                 if (!ops->ndo_set_config) {
1321                         err = -EOPNOTSUPP;
1322                         goto errout;
1323                 }
1324
1325                 if (!netif_device_present(dev)) {
1326                         err = -ENODEV;
1327                         goto errout;
1328                 }
1329
1330                 u_map = nla_data(tb[IFLA_MAP]);
1331                 k_map.mem_start = (unsigned long) u_map->mem_start;
1332                 k_map.mem_end = (unsigned long) u_map->mem_end;
1333                 k_map.base_addr = (unsigned short) u_map->base_addr;
1334                 k_map.irq = (unsigned char) u_map->irq;
1335                 k_map.dma = (unsigned char) u_map->dma;
1336                 k_map.port = (unsigned char) u_map->port;
1337
1338                 err = ops->ndo_set_config(dev, &k_map);
1339                 if (err < 0)
1340                         goto errout;
1341
1342                 modified = 1;
1343         }
1344
1345         if (tb[IFLA_ADDRESS]) {
1346                 struct sockaddr *sa;
1347                 int len;
1348
1349                 if (!ops->ndo_set_mac_address) {
1350                         err = -EOPNOTSUPP;
1351                         goto errout;
1352                 }
1353
1354                 if (!netif_device_present(dev)) {
1355                         err = -ENODEV;
1356                         goto errout;
1357                 }
1358
1359                 len = sizeof(sa_family_t) + dev->addr_len;
1360                 sa = kmalloc(len, GFP_KERNEL);
1361                 if (!sa) {
1362                         err = -ENOMEM;
1363                         goto errout;
1364                 }
1365                 sa->sa_family = dev->type;
1366                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1367                        dev->addr_len);
1368                 err = ops->ndo_set_mac_address(dev, sa);
1369                 kfree(sa);
1370                 if (err)
1371                         goto errout;
1372                 send_addr_notify = 1;
1373                 modified = 1;
1374         }
1375
1376         if (tb[IFLA_MTU]) {
1377                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1378                 if (err < 0)
1379                         goto errout;
1380                 modified = 1;
1381         }
1382
1383         if (tb[IFLA_GROUP]) {
1384                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1385                 modified = 1;
1386         }
1387
1388         /*
1389          * Interface selected by interface index but interface
1390          * name provided implies that a name change has been
1391          * requested.
1392          */
1393         if (ifm->ifi_index > 0 && ifname[0]) {
1394                 err = dev_change_name(dev, ifname);
1395                 if (err < 0)
1396                         goto errout;
1397                 modified = 1;
1398         }
1399
1400         if (tb[IFLA_IFALIAS]) {
1401                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1402                                     nla_len(tb[IFLA_IFALIAS]));
1403                 if (err < 0)
1404                         goto errout;
1405                 modified = 1;
1406         }
1407
1408         if (tb[IFLA_BROADCAST]) {
1409                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1410                 send_addr_notify = 1;
1411         }
1412
1413         if (ifm->ifi_flags || ifm->ifi_change) {
1414                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1415                 if (err < 0)
1416                         goto errout;
1417         }
1418
1419         if (tb[IFLA_MASTER]) {
1420                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1421                 if (err)
1422                         goto errout;
1423                 modified = 1;
1424         }
1425
1426         if (tb[IFLA_TXQLEN])
1427                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1428
1429         if (tb[IFLA_OPERSTATE])
1430                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1431
1432         if (tb[IFLA_LINKMODE]) {
1433                 write_lock_bh(&dev_base_lock);
1434                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1435                 write_unlock_bh(&dev_base_lock);
1436         }
1437
1438         if (tb[IFLA_VFINFO_LIST]) {
1439                 struct nlattr *attr;
1440                 int rem;
1441                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1442                         if (nla_type(attr) != IFLA_VF_INFO) {
1443                                 err = -EINVAL;
1444                                 goto errout;
1445                         }
1446                         err = do_setvfinfo(dev, attr);
1447                         if (err < 0)
1448                                 goto errout;
1449                         modified = 1;
1450                 }
1451         }
1452         err = 0;
1453
1454         if (tb[IFLA_VF_PORTS]) {
1455                 struct nlattr *port[IFLA_PORT_MAX+1];
1456                 struct nlattr *attr;
1457                 int vf;
1458                 int rem;
1459
1460                 err = -EOPNOTSUPP;
1461                 if (!ops->ndo_set_vf_port)
1462                         goto errout;
1463
1464                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1465                         if (nla_type(attr) != IFLA_VF_PORT)
1466                                 continue;
1467                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1468                                 attr, ifla_port_policy);
1469                         if (err < 0)
1470                                 goto errout;
1471                         if (!port[IFLA_PORT_VF]) {
1472                                 err = -EOPNOTSUPP;
1473                                 goto errout;
1474                         }
1475                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1476                         err = ops->ndo_set_vf_port(dev, vf, port);
1477                         if (err < 0)
1478                                 goto errout;
1479                         modified = 1;
1480                 }
1481         }
1482         err = 0;
1483
1484         if (tb[IFLA_PORT_SELF]) {
1485                 struct nlattr *port[IFLA_PORT_MAX+1];
1486
1487                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1488                         tb[IFLA_PORT_SELF], ifla_port_policy);
1489                 if (err < 0)
1490                         goto errout;
1491
1492                 err = -EOPNOTSUPP;
1493                 if (ops->ndo_set_vf_port)
1494                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1495                 if (err < 0)
1496                         goto errout;
1497                 modified = 1;
1498         }
1499
1500         if (tb[IFLA_AF_SPEC]) {
1501                 struct nlattr *af;
1502                 int rem;
1503
1504                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1505                         const struct rtnl_af_ops *af_ops;
1506
1507                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1508                                 BUG();
1509
1510                         err = af_ops->set_link_af(dev, af);
1511                         if (err < 0)
1512                                 goto errout;
1513
1514                         modified = 1;
1515                 }
1516         }
1517         err = 0;
1518
1519 errout:
1520         if (err < 0 && modified)
1521                 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1522                                      dev->name);
1523
1524         if (send_addr_notify)
1525                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1526
1527         return err;
1528 }
1529
1530 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1531 {
1532         struct net *net = sock_net(skb->sk);
1533         struct ifinfomsg *ifm;
1534         struct net_device *dev;
1535         int err;
1536         struct nlattr *tb[IFLA_MAX+1];
1537         char ifname[IFNAMSIZ];
1538
1539         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1540         if (err < 0)
1541                 goto errout;
1542
1543         if (tb[IFLA_IFNAME])
1544                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1545         else
1546                 ifname[0] = '\0';
1547
1548         err = -EINVAL;
1549         ifm = nlmsg_data(nlh);
1550         if (ifm->ifi_index > 0)
1551                 dev = __dev_get_by_index(net, ifm->ifi_index);
1552         else if (tb[IFLA_IFNAME])
1553                 dev = __dev_get_by_name(net, ifname);
1554         else
1555                 goto errout;
1556
1557         if (dev == NULL) {
1558                 err = -ENODEV;
1559                 goto errout;
1560         }
1561
1562         err = validate_linkmsg(dev, tb);
1563         if (err < 0)
1564                 goto errout;
1565
1566         err = do_setlink(dev, ifm, tb, ifname, 0);
1567 errout:
1568         return err;
1569 }
1570
1571 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1572 {
1573         struct net *net = sock_net(skb->sk);
1574         const struct rtnl_link_ops *ops;
1575         struct net_device *dev;
1576         struct ifinfomsg *ifm;
1577         char ifname[IFNAMSIZ];
1578         struct nlattr *tb[IFLA_MAX+1];
1579         int err;
1580         LIST_HEAD(list_kill);
1581
1582         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1583         if (err < 0)
1584                 return err;
1585
1586         if (tb[IFLA_IFNAME])
1587                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1588
1589         ifm = nlmsg_data(nlh);
1590         if (ifm->ifi_index > 0)
1591                 dev = __dev_get_by_index(net, ifm->ifi_index);
1592         else if (tb[IFLA_IFNAME])
1593                 dev = __dev_get_by_name(net, ifname);
1594         else
1595                 return -EINVAL;
1596
1597         if (!dev)
1598                 return -ENODEV;
1599
1600         ops = dev->rtnl_link_ops;
1601         if (!ops)
1602                 return -EOPNOTSUPP;
1603
1604         ops->dellink(dev, &list_kill);
1605         unregister_netdevice_many(&list_kill);
1606         list_del(&list_kill);
1607         return 0;
1608 }
1609
1610 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1611 {
1612         unsigned int old_flags;
1613         int err;
1614
1615         old_flags = dev->flags;
1616         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1617                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1618                 if (err < 0)
1619                         return err;
1620         }
1621
1622         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1623         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1624
1625         __dev_notify_flags(dev, old_flags);
1626         return 0;
1627 }
1628 EXPORT_SYMBOL(rtnl_configure_link);
1629
1630 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1631         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1632 {
1633         int err;
1634         struct net_device *dev;
1635         unsigned int num_tx_queues = 1;
1636         unsigned int num_rx_queues = 1;
1637
1638         if (tb[IFLA_NUM_TX_QUEUES])
1639                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1640         else if (ops->get_num_tx_queues)
1641                 num_tx_queues = ops->get_num_tx_queues();
1642
1643         if (tb[IFLA_NUM_RX_QUEUES])
1644                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1645         else if (ops->get_num_rx_queues)
1646                 num_rx_queues = ops->get_num_rx_queues();
1647
1648         err = -ENOMEM;
1649         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1650                                num_tx_queues, num_rx_queues);
1651         if (!dev)
1652                 goto err;
1653
1654         dev_net_set(dev, net);
1655         dev->rtnl_link_ops = ops;
1656         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1657
1658         if (tb[IFLA_MTU])
1659                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1660         if (tb[IFLA_ADDRESS])
1661                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1662                                 nla_len(tb[IFLA_ADDRESS]));
1663         if (tb[IFLA_BROADCAST])
1664                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1665                                 nla_len(tb[IFLA_BROADCAST]));
1666         if (tb[IFLA_TXQLEN])
1667                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1668         if (tb[IFLA_OPERSTATE])
1669                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1670         if (tb[IFLA_LINKMODE])
1671                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1672         if (tb[IFLA_GROUP])
1673                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1674
1675         return dev;
1676
1677 err:
1678         return ERR_PTR(err);
1679 }
1680 EXPORT_SYMBOL(rtnl_create_link);
1681
1682 static int rtnl_group_changelink(struct net *net, int group,
1683                 struct ifinfomsg *ifm,
1684                 struct nlattr **tb)
1685 {
1686         struct net_device *dev;
1687         int err;
1688
1689         for_each_netdev(net, dev) {
1690                 if (dev->group == group) {
1691                         err = do_setlink(dev, ifm, tb, NULL, 0);
1692                         if (err < 0)
1693                                 return err;
1694                 }
1695         }
1696
1697         return 0;
1698 }
1699
1700 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1701 {
1702         struct net *net = sock_net(skb->sk);
1703         const struct rtnl_link_ops *ops;
1704         struct net_device *dev;
1705         struct ifinfomsg *ifm;
1706         char kind[MODULE_NAME_LEN];
1707         char ifname[IFNAMSIZ];
1708         struct nlattr *tb[IFLA_MAX+1];
1709         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1710         int err;
1711
1712 #ifdef CONFIG_MODULES
1713 replay:
1714 #endif
1715         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1716         if (err < 0)
1717                 return err;
1718
1719         if (tb[IFLA_IFNAME])
1720                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1721         else
1722                 ifname[0] = '\0';
1723
1724         ifm = nlmsg_data(nlh);
1725         if (ifm->ifi_index > 0)
1726                 dev = __dev_get_by_index(net, ifm->ifi_index);
1727         else {
1728                 if (ifname[0])
1729                         dev = __dev_get_by_name(net, ifname);
1730                 else
1731                         dev = NULL;
1732         }
1733
1734         err = validate_linkmsg(dev, tb);
1735         if (err < 0)
1736                 return err;
1737
1738         if (tb[IFLA_LINKINFO]) {
1739                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1740                                        tb[IFLA_LINKINFO], ifla_info_policy);
1741                 if (err < 0)
1742                         return err;
1743         } else
1744                 memset(linkinfo, 0, sizeof(linkinfo));
1745
1746         if (linkinfo[IFLA_INFO_KIND]) {
1747                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1748                 ops = rtnl_link_ops_get(kind);
1749         } else {
1750                 kind[0] = '\0';
1751                 ops = NULL;
1752         }
1753
1754         if (1) {
1755                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1756                 struct net *dest_net;
1757
1758                 if (ops) {
1759                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1760                                 err = nla_parse_nested(attr, ops->maxtype,
1761                                                        linkinfo[IFLA_INFO_DATA],
1762                                                        ops->policy);
1763                                 if (err < 0)
1764                                         return err;
1765                                 data = attr;
1766                         }
1767                         if (ops->validate) {
1768                                 err = ops->validate(tb, data);
1769                                 if (err < 0)
1770                                         return err;
1771                         }
1772                 }
1773
1774                 if (dev) {
1775                         int modified = 0;
1776
1777                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1778                                 return -EEXIST;
1779                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1780                                 return -EOPNOTSUPP;
1781
1782                         if (linkinfo[IFLA_INFO_DATA]) {
1783                                 if (!ops || ops != dev->rtnl_link_ops ||
1784                                     !ops->changelink)
1785                                         return -EOPNOTSUPP;
1786
1787                                 err = ops->changelink(dev, tb, data);
1788                                 if (err < 0)
1789                                         return err;
1790                                 modified = 1;
1791                         }
1792
1793                         return do_setlink(dev, ifm, tb, ifname, modified);
1794                 }
1795
1796                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1797                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1798                                 return rtnl_group_changelink(net,
1799                                                 nla_get_u32(tb[IFLA_GROUP]),
1800                                                 ifm, tb);
1801                         return -ENODEV;
1802                 }
1803
1804                 if (ifm->ifi_index)
1805                         return -EOPNOTSUPP;
1806                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1807                         return -EOPNOTSUPP;
1808
1809                 if (!ops) {
1810 #ifdef CONFIG_MODULES
1811                         if (kind[0]) {
1812                                 __rtnl_unlock();
1813                                 request_module("rtnl-link-%s", kind);
1814                                 rtnl_lock();
1815                                 ops = rtnl_link_ops_get(kind);
1816                                 if (ops)
1817                                         goto replay;
1818                         }
1819 #endif
1820                         return -EOPNOTSUPP;
1821                 }
1822
1823                 if (!ifname[0])
1824                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1825
1826                 dest_net = rtnl_link_get_net(net, tb);
1827                 if (IS_ERR(dest_net))
1828                         return PTR_ERR(dest_net);
1829
1830                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1831
1832                 if (IS_ERR(dev))
1833                         err = PTR_ERR(dev);
1834                 else if (ops->newlink)
1835                         err = ops->newlink(net, dev, tb, data);
1836                 else
1837                         err = register_netdevice(dev);
1838
1839                 if (err < 0 && !IS_ERR(dev))
1840                         free_netdev(dev);
1841                 if (err < 0)
1842                         goto out;
1843
1844                 err = rtnl_configure_link(dev, ifm);
1845                 if (err < 0)
1846                         unregister_netdevice(dev);
1847 out:
1848                 put_net(dest_net);
1849                 return err;
1850         }
1851 }
1852
1853 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1854 {
1855         struct net *net = sock_net(skb->sk);
1856         struct ifinfomsg *ifm;
1857         char ifname[IFNAMSIZ];
1858         struct nlattr *tb[IFLA_MAX+1];
1859         struct net_device *dev = NULL;
1860         struct sk_buff *nskb;
1861         int err;
1862         u32 ext_filter_mask = 0;
1863
1864         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1865         if (err < 0)
1866                 return err;
1867
1868         if (tb[IFLA_IFNAME])
1869                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1870
1871         if (tb[IFLA_EXT_MASK])
1872                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1873
1874         ifm = nlmsg_data(nlh);
1875         if (ifm->ifi_index > 0)
1876                 dev = __dev_get_by_index(net, ifm->ifi_index);
1877         else if (tb[IFLA_IFNAME])
1878                 dev = __dev_get_by_name(net, ifname);
1879         else
1880                 return -EINVAL;
1881
1882         if (dev == NULL)
1883                 return -ENODEV;
1884
1885         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1886         if (nskb == NULL)
1887                 return -ENOBUFS;
1888
1889         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1890                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1891         if (err < 0) {
1892                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1893                 WARN_ON(err == -EMSGSIZE);
1894                 kfree_skb(nskb);
1895         } else
1896                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1897
1898         return err;
1899 }
1900
1901 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1902 {
1903         struct net *net = sock_net(skb->sk);
1904         struct net_device *dev;
1905         struct nlattr *tb[IFLA_MAX+1];
1906         u32 ext_filter_mask = 0;
1907         u16 min_ifinfo_dump_size = 0;
1908
1909         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1910                         ifla_policy) >= 0) {
1911                 if (tb[IFLA_EXT_MASK])
1912                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1913         }
1914
1915         if (!ext_filter_mask)
1916                 return NLMSG_GOODSIZE;
1917         /*
1918          * traverse the list of net devices and compute the minimum
1919          * buffer size based upon the filter mask.
1920          */
1921         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1922                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1923                                              if_nlmsg_size(dev,
1924                                                            ext_filter_mask));
1925         }
1926
1927         return min_ifinfo_dump_size;
1928 }
1929
1930 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1931 {
1932         int idx;
1933         int s_idx = cb->family;
1934
1935         if (s_idx == 0)
1936                 s_idx = 1;
1937         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1938                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1939                 if (idx < s_idx || idx == PF_PACKET)
1940                         continue;
1941                 if (rtnl_msg_handlers[idx] == NULL ||
1942                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1943                         continue;
1944                 if (idx > s_idx)
1945                         memset(&cb->args[0], 0, sizeof(cb->args));
1946                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1947                         break;
1948         }
1949         cb->family = idx;
1950
1951         return skb->len;
1952 }
1953
1954 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1955 {
1956         struct net *net = dev_net(dev);
1957         struct sk_buff *skb;
1958         int err = -ENOBUFS;
1959         size_t if_info_size;
1960
1961         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1962         if (skb == NULL)
1963                 goto errout;
1964
1965         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1966         if (err < 0) {
1967                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1968                 WARN_ON(err == -EMSGSIZE);
1969                 kfree_skb(skb);
1970                 goto errout;
1971         }
1972         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1973         return;
1974 errout:
1975         if (err < 0)
1976                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1977 }
1978
1979 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1980                                    struct net_device *dev,
1981                                    u8 *addr, u32 pid, u32 seq,
1982                                    int type, unsigned int flags)
1983 {
1984         struct nlmsghdr *nlh;
1985         struct ndmsg *ndm;
1986
1987         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1988         if (!nlh)
1989                 return -EMSGSIZE;
1990
1991         ndm = nlmsg_data(nlh);
1992         ndm->ndm_family  = AF_BRIDGE;
1993         ndm->ndm_pad1    = 0;
1994         ndm->ndm_pad2    = 0;
1995         ndm->ndm_flags   = flags;
1996         ndm->ndm_type    = 0;
1997         ndm->ndm_ifindex = dev->ifindex;
1998         ndm->ndm_state   = NUD_PERMANENT;
1999
2000         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2001                 goto nla_put_failure;
2002
2003         return nlmsg_end(skb, nlh);
2004
2005 nla_put_failure:
2006         nlmsg_cancel(skb, nlh);
2007         return -EMSGSIZE;
2008 }
2009
2010 static inline size_t rtnl_fdb_nlmsg_size(void)
2011 {
2012         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2013 }
2014
2015 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2016 {
2017         struct net *net = dev_net(dev);
2018         struct sk_buff *skb;
2019         int err = -ENOBUFS;
2020
2021         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2022         if (!skb)
2023                 goto errout;
2024
2025         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2026         if (err < 0) {
2027                 kfree_skb(skb);
2028                 goto errout;
2029         }
2030
2031         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2032         return;
2033 errout:
2034         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2035 }
2036
2037 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2038 {
2039         struct net *net = sock_net(skb->sk);
2040         struct net_device *master = NULL;
2041         struct ndmsg *ndm;
2042         struct nlattr *tb[NDA_MAX+1];
2043         struct net_device *dev;
2044         u8 *addr;
2045         int err;
2046
2047         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2048         if (err < 0)
2049                 return err;
2050
2051         ndm = nlmsg_data(nlh);
2052         if (ndm->ndm_ifindex == 0) {
2053                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2054                 return -EINVAL;
2055         }
2056
2057         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2058         if (dev == NULL) {
2059                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2060                 return -ENODEV;
2061         }
2062
2063         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2064                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2065                 return -EINVAL;
2066         }
2067
2068         addr = nla_data(tb[NDA_LLADDR]);
2069         if (!is_valid_ether_addr(addr)) {
2070                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2071                 return -EINVAL;
2072         }
2073
2074         err = -EOPNOTSUPP;
2075
2076         /* Support fdb on master device the net/bridge default case */
2077         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2078             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2079                 master = dev->master;
2080                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2081                                                       nlh->nlmsg_flags);
2082                 if (err)
2083                         goto out;
2084                 else
2085                         ndm->ndm_flags &= ~NTF_MASTER;
2086         }
2087
2088         /* Embedded bridge, macvlan, and any other device support */
2089         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2090                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2091                                                    nlh->nlmsg_flags);
2092
2093                 if (!err) {
2094                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2095                         ndm->ndm_flags &= ~NTF_SELF;
2096                 }
2097         }
2098 out:
2099         return err;
2100 }
2101
2102 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2103 {
2104         struct net *net = sock_net(skb->sk);
2105         struct ndmsg *ndm;
2106         struct nlattr *llattr;
2107         struct net_device *dev;
2108         int err = -EINVAL;
2109         __u8 *addr;
2110
2111         if (nlmsg_len(nlh) < sizeof(*ndm))
2112                 return -EINVAL;
2113
2114         ndm = nlmsg_data(nlh);
2115         if (ndm->ndm_ifindex == 0) {
2116                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2117                 return -EINVAL;
2118         }
2119
2120         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2121         if (dev == NULL) {
2122                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2123                 return -ENODEV;
2124         }
2125
2126         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2127         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2128                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2129                 return -EINVAL;
2130         }
2131
2132         addr = nla_data(llattr);
2133         err = -EOPNOTSUPP;
2134
2135         /* Support fdb on master device the net/bridge default case */
2136         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2137             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2138                 struct net_device *master = dev->master;
2139
2140                 if (master->netdev_ops->ndo_fdb_del)
2141                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2142
2143                 if (err)
2144                         goto out;
2145                 else
2146                         ndm->ndm_flags &= ~NTF_MASTER;
2147         }
2148
2149         /* Embedded bridge, macvlan, and any other device support */
2150         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2151                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2152
2153                 if (!err) {
2154                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2155                         ndm->ndm_flags &= ~NTF_SELF;
2156                 }
2157         }
2158 out:
2159         return err;
2160 }
2161
2162 static int nlmsg_populate_fdb(struct sk_buff *skb,
2163                               struct netlink_callback *cb,
2164                               struct net_device *dev,
2165                               int *idx,
2166                               struct netdev_hw_addr_list *list)
2167 {
2168         struct netdev_hw_addr *ha;
2169         int err;
2170         u32 pid, seq;
2171
2172         pid = NETLINK_CB(cb->skb).pid;
2173         seq = cb->nlh->nlmsg_seq;
2174
2175         list_for_each_entry(ha, &list->list, list) {
2176                 if (*idx < cb->args[0])
2177                         goto skip;
2178
2179                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2180                                               pid, seq, 0, NTF_SELF);
2181                 if (err < 0)
2182                         return err;
2183 skip:
2184                 *idx += 1;
2185         }
2186         return 0;
2187 }
2188
2189 /**
2190  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2191  * @nlh: netlink message header
2192  * @dev: netdevice
2193  *
2194  * Default netdevice operation to dump the existing unicast address list.
2195  * Returns zero on success.
2196  */
2197 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2198                       struct netlink_callback *cb,
2199                       struct net_device *dev,
2200                       int idx)
2201 {
2202         int err;
2203
2204         netif_addr_lock_bh(dev);
2205         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2206         if (err)
2207                 goto out;
2208         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2209 out:
2210         netif_addr_unlock_bh(dev);
2211         return idx;
2212 }
2213 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2214
2215 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2216 {
2217         int idx = 0;
2218         struct net *net = sock_net(skb->sk);
2219         struct net_device *dev;
2220
2221         rcu_read_lock();
2222         for_each_netdev_rcu(net, dev) {
2223                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2224                         struct net_device *master = dev->master;
2225                         const struct net_device_ops *ops = master->netdev_ops;
2226
2227                         if (ops->ndo_fdb_dump)
2228                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2229                 }
2230
2231                 if (dev->netdev_ops->ndo_fdb_dump)
2232                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2233         }
2234         rcu_read_unlock();
2235
2236         cb->args[0] = idx;
2237         return skb->len;
2238 }
2239
2240 /* Protected by RTNL sempahore.  */
2241 static struct rtattr **rta_buf;
2242 static int rtattr_max;
2243
2244 /* Process one rtnetlink message. */
2245
2246 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2247 {
2248         struct net *net = sock_net(skb->sk);
2249         rtnl_doit_func doit;
2250         int sz_idx, kind;
2251         int min_len;
2252         int family;
2253         int type;
2254         int err;
2255
2256         type = nlh->nlmsg_type;
2257         if (type > RTM_MAX)
2258                 return -EOPNOTSUPP;
2259
2260         type -= RTM_BASE;
2261
2262         /* All the messages must have at least 1 byte length */
2263         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2264                 return 0;
2265
2266         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2267         sz_idx = type>>2;
2268         kind = type&3;
2269
2270         if (kind != 2 && !capable(CAP_NET_ADMIN))
2271                 return -EPERM;
2272
2273         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2274                 struct sock *rtnl;
2275                 rtnl_dumpit_func dumpit;
2276                 rtnl_calcit_func calcit;
2277                 u16 min_dump_alloc = 0;
2278
2279                 dumpit = rtnl_get_dumpit(family, type);
2280                 if (dumpit == NULL)
2281                         return -EOPNOTSUPP;
2282                 calcit = rtnl_get_calcit(family, type);
2283                 if (calcit)
2284                         min_dump_alloc = calcit(skb, nlh);
2285
2286                 __rtnl_unlock();
2287                 rtnl = net->rtnl;
2288                 {
2289                         struct netlink_dump_control c = {
2290                                 .dump           = dumpit,
2291                                 .min_dump_alloc = min_dump_alloc,
2292                         };
2293                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2294                 }
2295                 rtnl_lock();
2296                 return err;
2297         }
2298
2299         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2300
2301         min_len = rtm_min[sz_idx];
2302         if (nlh->nlmsg_len < min_len)
2303                 return -EINVAL;
2304
2305         if (nlh->nlmsg_len > min_len) {
2306                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2307                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2308
2309                 while (RTA_OK(attr, attrlen)) {
2310                         unsigned int flavor = attr->rta_type;
2311                         if (flavor) {
2312                                 if (flavor > rta_max[sz_idx])
2313                                         return -EINVAL;
2314                                 rta_buf[flavor-1] = attr;
2315                         }
2316                         attr = RTA_NEXT(attr, attrlen);
2317                 }
2318         }
2319
2320         doit = rtnl_get_doit(family, type);
2321         if (doit == NULL)
2322                 return -EOPNOTSUPP;
2323
2324         return doit(skb, nlh, (void *)&rta_buf[0]);
2325 }
2326
2327 static void rtnetlink_rcv(struct sk_buff *skb)
2328 {
2329         rtnl_lock();
2330         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2331         rtnl_unlock();
2332 }
2333
2334 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2335 {
2336         struct net_device *dev = ptr;
2337
2338         switch (event) {
2339         case NETDEV_UP:
2340         case NETDEV_DOWN:
2341         case NETDEV_PRE_UP:
2342         case NETDEV_POST_INIT:
2343         case NETDEV_REGISTER:
2344         case NETDEV_CHANGE:
2345         case NETDEV_PRE_TYPE_CHANGE:
2346         case NETDEV_GOING_DOWN:
2347         case NETDEV_UNREGISTER:
2348         case NETDEV_UNREGISTER_BATCH:
2349         case NETDEV_RELEASE:
2350         case NETDEV_JOIN:
2351                 break;
2352         default:
2353                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2354                 break;
2355         }
2356         return NOTIFY_DONE;
2357 }
2358
2359 static struct notifier_block rtnetlink_dev_notifier = {
2360         .notifier_call  = rtnetlink_event,
2361 };
2362
2363
2364 static int __net_init rtnetlink_net_init(struct net *net)
2365 {
2366         struct sock *sk;
2367         struct netlink_kernel_cfg cfg = {
2368                 .groups         = RTNLGRP_MAX,
2369                 .input          = rtnetlink_rcv,
2370                 .cb_mutex       = &rtnl_mutex,
2371         };
2372
2373         sk = netlink_kernel_create(net, NETLINK_ROUTE, THIS_MODULE, &cfg);
2374         if (!sk)
2375                 return -ENOMEM;
2376         net->rtnl = sk;
2377         return 0;
2378 }
2379
2380 static void __net_exit rtnetlink_net_exit(struct net *net)
2381 {
2382         netlink_kernel_release(net->rtnl);
2383         net->rtnl = NULL;
2384 }
2385
2386 static struct pernet_operations rtnetlink_net_ops = {
2387         .init = rtnetlink_net_init,
2388         .exit = rtnetlink_net_exit,
2389 };
2390
2391 void __init rtnetlink_init(void)
2392 {
2393         int i;
2394
2395         rtattr_max = 0;
2396         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2397                 if (rta_max[i] > rtattr_max)
2398                         rtattr_max = rta_max[i];
2399         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2400         if (!rta_buf)
2401                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2402
2403         if (register_pernet_subsys(&rtnetlink_net_ops))
2404                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2405
2406         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2407         register_netdevice_notifier(&rtnetlink_dev_notifier);
2408
2409         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2410                       rtnl_dump_ifinfo, rtnl_calcit);
2411         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2412         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2413         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2414
2415         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2416         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2417
2418         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2419         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2420         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2421 }
2422