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