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/pci.h>
  39
  40 #include <asm/uaccess.h>
  41 #include <asm/system.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/sock.h>
  51 #include <net/pkt_sched.h>
  52 #include <net/fib_rules.h>
  53 #include <net/rtnetlink.h>
  54 #include <net/net_namespace.h>
  55
  56 struct rtnl_link {
  57         rtnl_doit_func          doit;
  58         rtnl_dumpit_func        dumpit;
  59         rtnl_calcit_func        calcit;
  60 };
  61
  62 static DEFINE_MUTEX(rtnl_mutex);
  63 static u16 min_ifinfo_dump_size;
  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 /**
 277  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 278  * @ops: struct rtnl_link_ops * to register
 279  *
 280  * The caller must hold the rtnl_mutex. This function should be used
 281  * by drivers that create devices during module initialization. It
 282  * must be called before registering the devices.
 283  *
 284  * Returns 0 on success or a negative error code.
 285  */
 286 int __rtnl_link_register(struct rtnl_link_ops *ops)
 287 {
 288         if (!ops->dellink)
 289                 ops->dellink = unregister_netdevice_queue;
 290
 291         list_add_tail(&ops->list, &link_ops);
 292         return 0;
 293 }
 294 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 295
 296 /**
 297  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 298  * @ops: struct rtnl_link_ops * to register
 299  *
 300  * Returns 0 on success or a negative error code.
 301  */
 302 int rtnl_link_register(struct rtnl_link_ops *ops)
 303 {
 304         int err;
 305
 306         rtnl_lock();
 307         err = __rtnl_link_register(ops);
 308         rtnl_unlock();
 309         return err;
 310 }
 311 EXPORT_SYMBOL_GPL(rtnl_link_register);
 312
 313 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 314 {
 315         struct net_device *dev;
 316         LIST_HEAD(list_kill);
 317
 318         for_each_netdev(net, dev) {
 319                 if (dev->rtnl_link_ops == ops)
 320                         ops->dellink(dev, &list_kill);
 321         }
 322         unregister_netdevice_many(&list_kill);
 323 }
 324
 325 /**
 326  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 327  * @ops: struct rtnl_link_ops * to unregister
 328  *
 329  * The caller must hold the rtnl_mutex.
 330  */
 331 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 332 {
 333         struct net *net;
 334
 335         for_each_net(net) {
 336                 __rtnl_kill_links(net, ops);
 337         }
 338         list_del(&ops->list);
 339 }
 340 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 341
 342 /**
 343  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 344  * @ops: struct rtnl_link_ops * to unregister
 345  */
 346 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 347 {
 348         rtnl_lock();
 349         __rtnl_link_unregister(ops);
 350         rtnl_unlock();
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 353
 354 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 355 {
 356         const struct rtnl_link_ops *ops;
 357
 358         list_for_each_entry(ops, &link_ops, list) {
 359                 if (!strcmp(ops->kind, kind))
 360                         return ops;
 361         }
 362         return NULL;
 363 }
 364
 365 static size_t rtnl_link_get_size(const struct net_device *dev)
 366 {
 367         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 368         size_t size;
 369
 370         if (!ops)
 371                 return 0;
 372
 373         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 374                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 375
 376         if (ops->get_size)
 377                 /* IFLA_INFO_DATA + nested data */
 378                 size += nla_total_size(sizeof(struct nlattr)) +
 379                         ops->get_size(dev);
 380
 381         if (ops->get_xstats_size)
 382                 /* IFLA_INFO_XSTATS */
 383                 size += nla_total_size(ops->get_xstats_size(dev));
 384
 385         return size;
 386 }
 387
 388 static LIST_HEAD(rtnl_af_ops);
 389
 390 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 391 {
 392         const struct rtnl_af_ops *ops;
 393
 394         list_for_each_entry(ops, &rtnl_af_ops, list) {
 395                 if (ops->family == family)
 396                         return ops;
 397         }
 398
 399         return NULL;
 400 }
 401
 402 /**
 403  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 404  * @ops: struct rtnl_af_ops * to register
 405  *
 406  * The caller must hold the rtnl_mutex.
 407  *
 408  * Returns 0 on success or a negative error code.
 409  */
 410 int __rtnl_af_register(struct rtnl_af_ops *ops)
 411 {
 412         list_add_tail(&ops->list, &rtnl_af_ops);
 413         return 0;
 414 }
 415 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 416
 417 /**
 418  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 419  * @ops: struct rtnl_af_ops * to register
 420  *
 421  * Returns 0 on success or a negative error code.
 422  */
 423 int rtnl_af_register(struct rtnl_af_ops *ops)
 424 {
 425         int err;
 426
 427         rtnl_lock();
 428         err = __rtnl_af_register(ops);
 429         rtnl_unlock();
 430         return err;
 431 }
 432 EXPORT_SYMBOL_GPL(rtnl_af_register);
 433
 434 /**
 435  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 436  * @ops: struct rtnl_af_ops * to unregister
 437  *
 438  * The caller must hold the rtnl_mutex.
 439  */
 440 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 441 {
 442         list_del(&ops->list);
 443 }
 444 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 445
 446 /**
 447  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 448  * @ops: struct rtnl_af_ops * to unregister
 449  */
 450 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 451 {
 452         rtnl_lock();
 453         __rtnl_af_unregister(ops);
 454         rtnl_unlock();
 455 }
 456 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 457
 458 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 459 {
 460         struct rtnl_af_ops *af_ops;
 461         size_t size;
 462
 463         /* IFLA_AF_SPEC */
 464         size = nla_total_size(sizeof(struct nlattr));
 465
 466         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 467                 if (af_ops->get_link_af_size) {
 468                         /* AF_* + nested data */
 469                         size += nla_total_size(sizeof(struct nlattr)) +
 470                                 af_ops->get_link_af_size(dev);
 471                 }
 472         }
 473
 474         return size;
 475 }
 476
 477 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 478 {
 479         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 480         struct nlattr *linkinfo, *data;
 481         int err = -EMSGSIZE;
 482
 483         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 484         if (linkinfo == NULL)
 485                 goto out;
 486
 487         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 488                 goto err_cancel_link;
 489         if (ops->fill_xstats) {
 490                 err = ops->fill_xstats(skb, dev);
 491                 if (err < 0)
 492                         goto err_cancel_link;
 493         }
 494         if (ops->fill_info) {
 495                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 496                 if (data == NULL)
 497                         goto err_cancel_link;
 498                 err = ops->fill_info(skb, dev);
 499                 if (err < 0)
 500                         goto err_cancel_data;
 501                 nla_nest_end(skb, data);
 502         }
 503
 504         nla_nest_end(skb, linkinfo);
 505         return 0;
 506
 507 err_cancel_data:
 508         nla_nest_cancel(skb, data);
 509 err_cancel_link:
 510         nla_nest_cancel(skb, linkinfo);
 511 out:
 512         return err;
 513 }
 514
 515 static const int rtm_min[RTM_NR_FAMILIES] =
 516 {
 517         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 518         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 519         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 520         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 521         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 522         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 523         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 524         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 525         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 526         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 527 };
 528
 529 static const int rta_max[RTM_NR_FAMILIES] =
 530 {
 531         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 532         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 533         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 534         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 535         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 536         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 537         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 538         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 539 };
 540
 541 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 542 {
 543         struct rtattr *rta;
 544         int size = RTA_LENGTH(attrlen);
 545
 546         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 547         rta->rta_type = attrtype;
 548         rta->rta_len = size;
 549         memcpy(RTA_DATA(rta), data, attrlen);
 550         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 551 }
 552 EXPORT_SYMBOL(__rta_fill);
 553
 554 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 555 {
 556         struct sock *rtnl = net->rtnl;
 557         int err = 0;
 558
 559         NETLINK_CB(skb).dst_group = group;
 560         if (echo)
 561                 atomic_inc(&skb->users);
 562         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 563         if (echo)
 564                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 565         return err;
 566 }
 567
 568 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 569 {
 570         struct sock *rtnl = net->rtnl;
 571
 572         return nlmsg_unicast(rtnl, skb, pid);
 573 }
 574 EXPORT_SYMBOL(rtnl_unicast);
 575
 576 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 577                  struct nlmsghdr *nlh, gfp_t flags)
 578 {
 579         struct sock *rtnl = net->rtnl;
 580         int report = 0;
 581
 582         if (nlh)
 583                 report = nlmsg_report(nlh);
 584
 585         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 586 }
 587 EXPORT_SYMBOL(rtnl_notify);
 588
 589 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 590 {
 591         struct sock *rtnl = net->rtnl;
 592
 593         netlink_set_err(rtnl, 0, group, error);
 594 }
 595 EXPORT_SYMBOL(rtnl_set_sk_err);
 596
 597 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 598 {
 599         struct nlattr *mx;
 600         int i, valid = 0;
 601
 602         mx = nla_nest_start(skb, RTA_METRICS);
 603         if (mx == NULL)
 604                 return -ENOBUFS;
 605
 606         for (i = 0; i < RTAX_MAX; i++) {
 607                 if (metrics[i]) {
 608                         valid++;
 609                         NLA_PUT_U32(skb, i+1, metrics[i]);
 610                 }
 611         }
 612
 613         if (!valid) {
 614                 nla_nest_cancel(skb, mx);
 615                 return 0;
 616         }
 617
 618         return nla_nest_end(skb, mx);
 619
 620 nla_put_failure:
 621         nla_nest_cancel(skb, mx);
 622         return -EMSGSIZE;
 623 }
 624 EXPORT_SYMBOL(rtnetlink_put_metrics);
 625
 626 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 627                        u32 ts, u32 tsage, long expires, u32 error)
 628 {
 629         struct rta_cacheinfo ci = {
 630                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 631                 .rta_used = dst->__use,
 632                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 633                 .rta_error = error,
 634                 .rta_id =  id,
 635                 .rta_ts = ts,
 636                 .rta_tsage = tsage,
 637         };
 638
 639         if (expires)
 640                 ci.rta_expires = jiffies_to_clock_t(expires);
 641
 642         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 643 }
 644 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 645
 646 static void set_operstate(struct net_device *dev, unsigned char transition)
 647 {
 648         unsigned char operstate = dev->operstate;
 649
 650         switch (transition) {
 651         case IF_OPER_UP:
 652                 if ((operstate == IF_OPER_DORMANT ||
 653                      operstate == IF_OPER_UNKNOWN) &&
 654                     !netif_dormant(dev))
 655                         operstate = IF_OPER_UP;
 656                 break;
 657
 658         case IF_OPER_DORMANT:
 659                 if (operstate == IF_OPER_UP ||
 660                     operstate == IF_OPER_UNKNOWN)
 661                         operstate = IF_OPER_DORMANT;
 662                 break;
 663         }
 664
 665         if (dev->operstate != operstate) {
 666                 write_lock_bh(&dev_base_lock);
 667                 dev->operstate = operstate;
 668                 write_unlock_bh(&dev_base_lock);
 669                 netdev_state_change(dev);
 670         }
 671 }
 672
 673 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 674                                            const struct ifinfomsg *ifm)
 675 {
 676         unsigned int flags = ifm->ifi_flags;
 677
 678         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 679         if (ifm->ifi_change)
 680                 flags = (flags & ifm->ifi_change) |
 681                         (dev->flags & ~ifm->ifi_change);
 682
 683         return flags;
 684 }
 685
 686 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 687                                  const struct rtnl_link_stats64 *b)
 688 {
 689         a->rx_packets = b->rx_packets;
 690         a->tx_packets = b->tx_packets;
 691         a->rx_bytes = b->rx_bytes;
 692         a->tx_bytes = b->tx_bytes;
 693         a->rx_errors = b->rx_errors;
 694         a->tx_errors = b->tx_errors;
 695         a->rx_dropped = b->rx_dropped;
 696         a->tx_dropped = b->tx_dropped;
 697
 698         a->multicast = b->multicast;
 699         a->collisions = b->collisions;
 700
 701         a->rx_length_errors = b->rx_length_errors;
 702         a->rx_over_errors = b->rx_over_errors;
 703         a->rx_crc_errors = b->rx_crc_errors;
 704         a->rx_frame_errors = b->rx_frame_errors;
 705         a->rx_fifo_errors = b->rx_fifo_errors;
 706         a->rx_missed_errors = b->rx_missed_errors;
 707
 708         a->tx_aborted_errors = b->tx_aborted_errors;
 709         a->tx_carrier_errors = b->tx_carrier_errors;
 710         a->tx_fifo_errors = b->tx_fifo_errors;
 711         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 712         a->tx_window_errors = b->tx_window_errors;
 713
 714         a->rx_compressed = b->rx_compressed;
 715         a->tx_compressed = b->tx_compressed;
 716 }
 717
 718 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 719 {
 720         memcpy(v, b, sizeof(*b));
 721 }
 722
 723 /* All VF info */
 724 static inline int rtnl_vfinfo_size(const struct net_device *dev)
 725 {
 726         if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
 727
 728                 int num_vfs = dev_num_vf(dev->dev.parent);
 729                 size_t size = nla_total_size(sizeof(struct nlattr));
 730                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 731                 size += num_vfs *
 732                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 733                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 734                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 735                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 736                 return size;
 737         } else
 738                 return 0;
 739 }
 740
 741 static size_t rtnl_port_size(const struct net_device *dev)
 742 {
 743         size_t port_size = nla_total_size(4)            /* PORT_VF */
 744                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 745                 + nla_total_size(sizeof(struct ifla_port_vsi))
 746                                                         /* PORT_VSI_TYPE */
 747                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 748                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 749                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 750                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 751         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 752         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 753                 + port_size;
 754         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 755                 + port_size;
 756
 757         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 758                 return 0;
 759         if (dev_num_vf(dev->dev.parent))
 760                 return port_self_size + vf_ports_size +
 761                         vf_port_size * dev_num_vf(dev->dev.parent);
 762         else
 763                 return port_self_size;
 764 }
 765
 766 static noinline size_t if_nlmsg_size(const struct net_device *dev)
 767 {
 768         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 769                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 770                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 771                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 772                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 773                + nla_total_size(sizeof(struct rtnl_link_stats))
 774                + nla_total_size(sizeof(struct rtnl_link_stats64))
 775                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 776                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 777                + nla_total_size(4) /* IFLA_TXQLEN */
 778                + nla_total_size(4) /* IFLA_WEIGHT */
 779                + nla_total_size(4) /* IFLA_MTU */
 780                + nla_total_size(4) /* IFLA_LINK */
 781                + nla_total_size(4) /* IFLA_MASTER */
 782                + nla_total_size(1) /* IFLA_OPERSTATE */
 783                + nla_total_size(1) /* IFLA_LINKMODE */
 784                + nla_total_size(4) /* IFLA_NUM_VF */
 785                + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
 786                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 787                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 788                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 789 }
 790
 791 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 792 {
 793         struct nlattr *vf_ports;
 794         struct nlattr *vf_port;
 795         int vf;
 796         int err;
 797
 798         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 799         if (!vf_ports)
 800                 return -EMSGSIZE;
 801
 802         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 803                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 804                 if (!vf_port)
 805                         goto nla_put_failure;
 806                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 807                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 808                 if (err == -EMSGSIZE)
 809                         goto nla_put_failure;
 810                 if (err) {
 811                         nla_nest_cancel(skb, vf_port);
 812                         continue;
 813                 }
 814                 nla_nest_end(skb, vf_port);
 815         }
 816
 817         nla_nest_end(skb, vf_ports);
 818
 819         return 0;
 820
 821 nla_put_failure:
 822         nla_nest_cancel(skb, vf_ports);
 823         return -EMSGSIZE;
 824 }
 825
 826 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 827 {
 828         struct nlattr *port_self;
 829         int err;
 830
 831         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 832         if (!port_self)
 833                 return -EMSGSIZE;
 834
 835         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 836         if (err) {
 837                 nla_nest_cancel(skb, port_self);
 838                 return (err == -EMSGSIZE) ? err : 0;
 839         }
 840
 841         nla_nest_end(skb, port_self);
 842
 843         return 0;
 844 }
 845
 846 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 847 {
 848         int err;
 849
 850         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 851                 return 0;
 852
 853         err = rtnl_port_self_fill(skb, dev);
 854         if (err)
 855                 return err;
 856
 857         if (dev_num_vf(dev->dev.parent)) {
 858                 err = rtnl_vf_ports_fill(skb, dev);
 859                 if (err)
 860                         return err;
 861         }
 862
 863         return 0;
 864 }
 865
 866 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 867                             int type, u32 pid, u32 seq, u32 change,
 868                             unsigned int flags)
 869 {
 870         struct ifinfomsg *ifm;
 871         struct nlmsghdr *nlh;
 872         struct rtnl_link_stats64 temp;
 873         const struct rtnl_link_stats64 *stats;
 874         struct nlattr *attr, *af_spec;
 875         struct rtnl_af_ops *af_ops;
 876
 877         ASSERT_RTNL();
 878         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 879         if (nlh == NULL)
 880                 return -EMSGSIZE;
 881
 882         ifm = nlmsg_data(nlh);
 883         ifm->ifi_family = AF_UNSPEC;
 884         ifm->__ifi_pad = 0;
 885         ifm->ifi_type = dev->type;
 886         ifm->ifi_index = dev->ifindex;
 887         ifm->ifi_flags = dev_get_flags(dev);
 888         ifm->ifi_change = change;
 889
 890         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 891         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 892         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 893                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 894         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 895         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 896         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 897
 898         if (dev->ifindex != dev->iflink)
 899                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 900
 901         if (dev->master)
 902                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 903
 904         if (dev->qdisc)
 905                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 906
 907         if (dev->ifalias)
 908                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 909
 910         if (1) {
 911                 struct rtnl_link_ifmap map = {
 912                         .mem_start   = dev->mem_start,
 913                         .mem_end     = dev->mem_end,
 914                         .base_addr   = dev->base_addr,
 915                         .irq         = dev->irq,
 916                         .dma         = dev->dma,
 917                         .port        = dev->if_port,
 918                 };
 919                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 920         }
 921
 922         if (dev->addr_len) {
 923                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 924                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 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)
 942                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 943
 944         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
 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                         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                         nla_nest_end(skb, vf);
 991                 }
 992                 nla_nest_end(skb, vfinfo);
 993         }
 994
 995         if (rtnl_port_fill(skb, dev))
 996                 goto nla_put_failure;
 997
 998         if (dev->rtnl_link_ops) {
 999                 if (rtnl_link_fill(skb, dev) < 0)
1000                         goto nla_put_failure;
1001         }
1002
1003         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1004                 goto nla_put_failure;
1005
1006         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1007                 if (af_ops->fill_link_af) {
1008                         struct nlattr *af;
1009                         int err;
1010
1011                         if (!(af = nla_nest_start(skb, af_ops->family)))
1012                                 goto nla_put_failure;
1013
1014                         err = af_ops->fill_link_af(skb, dev);
1015
1016                         /*
1017                          * Caller may return ENODATA to indicate that there
1018                          * was no data to be dumped. This is not an error, it
1019                          * means we should trim the attribute header and
1020                          * continue.
1021                          */
1022                         if (err == -ENODATA)
1023                                 nla_nest_cancel(skb, af);
1024                         else if (err < 0)
1025                                 goto nla_put_failure;
1026
1027                         nla_nest_end(skb, af);
1028                 }
1029         }
1030
1031         nla_nest_end(skb, af_spec);
1032
1033         return nlmsg_end(skb, nlh);
1034
1035 nla_put_failure:
1036         nlmsg_cancel(skb, nlh);
1037         return -EMSGSIZE;
1038 }
1039
1040 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1041 {
1042         struct net *net = sock_net(skb->sk);
1043         int h, s_h;
1044         int idx = 0, s_idx;
1045         struct net_device *dev;
1046         struct hlist_head *head;
1047         struct hlist_node *node;
1048
1049         s_h = cb->args[0];
1050         s_idx = cb->args[1];
1051
1052         rcu_read_lock();
1053         cb->seq = net->dev_base_seq;
1054
1055         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1056                 idx = 0;
1057                 head = &net->dev_index_head[h];
1058                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1059                         if (idx < s_idx)
1060                                 goto cont;
1061                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1062                                              NETLINK_CB(cb->skb).pid,
1063                                              cb->nlh->nlmsg_seq, 0,
1064                                              NLM_F_MULTI) <= 0)
1065                                 goto out;
1066
1067                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1068 cont:
1069                         idx++;
1070                 }
1071         }
1072 out:
1073         rcu_read_unlock();
1074         cb->args[1] = idx;
1075         cb->args[0] = h;
1076
1077         return skb->len;
1078 }
1079
1080 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1081         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1082         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1083         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1084         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1085         [IFLA_MTU]              = { .type = NLA_U32 },
1086         [IFLA_LINK]             = { .type = NLA_U32 },
1087         [IFLA_MASTER]           = { .type = NLA_U32 },
1088         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1089         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1090         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1091         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1092         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1093         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1094         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1095         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1096         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1097         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1098         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1099         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1100 };
1101 EXPORT_SYMBOL(ifla_policy);
1102
1103 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1104         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1105         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1106 };
1107
1108 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1109         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1110 };
1111
1112 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1113         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1114                                     .len = sizeof(struct ifla_vf_mac) },
1115         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1116                                     .len = sizeof(struct ifla_vf_vlan) },
1117         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1118                                     .len = sizeof(struct ifla_vf_tx_rate) },
1119 };
1120
1121 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1122         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1123         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1124                                     .len = PORT_PROFILE_MAX },
1125         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1126                                     .len = sizeof(struct ifla_port_vsi)},
1127         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1128                                       .len = PORT_UUID_MAX },
1129         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1130                                     .len = PORT_UUID_MAX },
1131         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1132         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1133 };
1134
1135 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1136 {
1137         struct net *net;
1138         /* Examine the link attributes and figure out which
1139          * network namespace we are talking about.
1140          */
1141         if (tb[IFLA_NET_NS_PID])
1142                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1143         else if (tb[IFLA_NET_NS_FD])
1144                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1145         else
1146                 net = get_net(src_net);
1147         return net;
1148 }
1149 EXPORT_SYMBOL(rtnl_link_get_net);
1150
1151 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1152 {
1153         if (dev) {
1154                 if (tb[IFLA_ADDRESS] &&
1155                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1156                         return -EINVAL;
1157
1158                 if (tb[IFLA_BROADCAST] &&
1159                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1160                         return -EINVAL;
1161         }
1162
1163         if (tb[IFLA_AF_SPEC]) {
1164                 struct nlattr *af;
1165                 int rem, err;
1166
1167                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1168                         const struct rtnl_af_ops *af_ops;
1169
1170                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1171                                 return -EAFNOSUPPORT;
1172
1173                         if (!af_ops->set_link_af)
1174                                 return -EOPNOTSUPP;
1175
1176                         if (af_ops->validate_link_af) {
1177                                 err = af_ops->validate_link_af(dev, af);
1178                                 if (err < 0)
1179                                         return err;
1180                         }
1181                 }
1182         }
1183
1184         return 0;
1185 }
1186
1187 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1188 {
1189         int rem, err = -EINVAL;
1190         struct nlattr *vf;
1191         const struct net_device_ops *ops = dev->netdev_ops;
1192
1193         nla_for_each_nested(vf, attr, rem) {
1194                 switch (nla_type(vf)) {
1195                 case IFLA_VF_MAC: {
1196                         struct ifla_vf_mac *ivm;
1197                         ivm = nla_data(vf);
1198                         err = -EOPNOTSUPP;
1199                         if (ops->ndo_set_vf_mac)
1200                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1201                                                           ivm->mac);
1202                         break;
1203                 }
1204                 case IFLA_VF_VLAN: {
1205                         struct ifla_vf_vlan *ivv;
1206                         ivv = nla_data(vf);
1207                         err = -EOPNOTSUPP;
1208                         if (ops->ndo_set_vf_vlan)
1209                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1210                                                            ivv->vlan,
1211                                                            ivv->qos);
1212                         break;
1213                 }
1214                 case IFLA_VF_TX_RATE: {
1215                         struct ifla_vf_tx_rate *ivt;
1216                         ivt = nla_data(vf);
1217                         err = -EOPNOTSUPP;
1218                         if (ops->ndo_set_vf_tx_rate)
1219                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1220                                                               ivt->rate);
1221                         break;
1222                 }
1223                 case IFLA_VF_SPOOFCHK: {
1224                         struct ifla_vf_spoofchk *ivs;
1225                         ivs = nla_data(vf);
1226                         err = -EOPNOTSUPP;
1227                         if (ops->ndo_set_vf_spoofchk)
1228                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1229                                                                ivs->setting);
1230                         break;
1231                 }
1232                 default:
1233                         err = -EINVAL;
1234                         break;
1235                 }
1236                 if (err)
1237                         break;
1238         }
1239         return err;
1240 }
1241
1242 static int do_set_master(struct net_device *dev, int ifindex)
1243 {
1244         struct net_device *master_dev;
1245         const struct net_device_ops *ops;
1246         int err;
1247
1248         if (dev->master) {
1249                 if (dev->master->ifindex == ifindex)
1250                         return 0;
1251                 ops = dev->master->netdev_ops;
1252                 if (ops->ndo_del_slave) {
1253                         err = ops->ndo_del_slave(dev->master, dev);
1254                         if (err)
1255                                 return err;
1256                 } else {
1257                         return -EOPNOTSUPP;
1258                 }
1259         }
1260
1261         if (ifindex) {
1262                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1263                 if (!master_dev)
1264                         return -EINVAL;
1265                 ops = master_dev->netdev_ops;
1266                 if (ops->ndo_add_slave) {
1267                         err = ops->ndo_add_slave(master_dev, dev);
1268                         if (err)
1269                                 return err;
1270                 } else {
1271                         return -EOPNOTSUPP;
1272                 }
1273         }
1274         return 0;
1275 }
1276
1277 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1278                       struct nlattr **tb, char *ifname, int modified)
1279 {
1280         const struct net_device_ops *ops = dev->netdev_ops;
1281         int send_addr_notify = 0;
1282         int err;
1283
1284         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1285                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1286                 if (IS_ERR(net)) {
1287                         err = PTR_ERR(net);
1288                         goto errout;
1289                 }
1290                 err = dev_change_net_namespace(dev, net, ifname);
1291                 put_net(net);
1292                 if (err)
1293                         goto errout;
1294                 modified = 1;
1295         }
1296
1297         if (tb[IFLA_MAP]) {
1298                 struct rtnl_link_ifmap *u_map;
1299                 struct ifmap k_map;
1300
1301                 if (!ops->ndo_set_config) {
1302                         err = -EOPNOTSUPP;
1303                         goto errout;
1304                 }
1305
1306                 if (!netif_device_present(dev)) {
1307                         err = -ENODEV;
1308                         goto errout;
1309                 }
1310
1311                 u_map = nla_data(tb[IFLA_MAP]);
1312                 k_map.mem_start = (unsigned long) u_map->mem_start;
1313                 k_map.mem_end = (unsigned long) u_map->mem_end;
1314                 k_map.base_addr = (unsigned short) u_map->base_addr;
1315                 k_map.irq = (unsigned char) u_map->irq;
1316                 k_map.dma = (unsigned char) u_map->dma;
1317                 k_map.port = (unsigned char) u_map->port;
1318
1319                 err = ops->ndo_set_config(dev, &k_map);
1320                 if (err < 0)
1321                         goto errout;
1322
1323                 modified = 1;
1324         }
1325
1326         if (tb[IFLA_ADDRESS]) {
1327                 struct sockaddr *sa;
1328                 int len;
1329
1330                 if (!ops->ndo_set_mac_address) {
1331                         err = -EOPNOTSUPP;
1332                         goto errout;
1333                 }
1334
1335                 if (!netif_device_present(dev)) {
1336                         err = -ENODEV;
1337                         goto errout;
1338                 }
1339
1340                 len = sizeof(sa_family_t) + dev->addr_len;
1341                 sa = kmalloc(len, GFP_KERNEL);
1342                 if (!sa) {
1343                         err = -ENOMEM;
1344                         goto errout;
1345                 }
1346                 sa->sa_family = dev->type;
1347                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1348                        dev->addr_len);
1349                 err = ops->ndo_set_mac_address(dev, sa);
1350                 kfree(sa);
1351                 if (err)
1352                         goto errout;
1353                 send_addr_notify = 1;
1354                 modified = 1;
1355         }
1356
1357         if (tb[IFLA_MTU]) {
1358                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1359                 if (err < 0)
1360                         goto errout;
1361                 modified = 1;
1362         }
1363
1364         if (tb[IFLA_GROUP]) {
1365                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1366                 modified = 1;
1367         }
1368
1369         /*
1370          * Interface selected by interface index but interface
1371          * name provided implies that a name change has been
1372          * requested.
1373          */
1374         if (ifm->ifi_index > 0 && ifname[0]) {
1375                 err = dev_change_name(dev, ifname);
1376                 if (err < 0)
1377                         goto errout;
1378                 modified = 1;
1379         }
1380
1381         if (tb[IFLA_IFALIAS]) {
1382                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1383                                     nla_len(tb[IFLA_IFALIAS]));
1384                 if (err < 0)
1385                         goto errout;
1386                 modified = 1;
1387         }
1388
1389         if (tb[IFLA_BROADCAST]) {
1390                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1391                 send_addr_notify = 1;
1392         }
1393
1394         if (ifm->ifi_flags || ifm->ifi_change) {
1395                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1396                 if (err < 0)
1397                         goto errout;
1398         }
1399
1400         if (tb[IFLA_MASTER]) {
1401                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1402                 if (err)
1403                         goto errout;
1404                 modified = 1;
1405         }
1406
1407         if (tb[IFLA_TXQLEN])
1408                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1409
1410         if (tb[IFLA_OPERSTATE])
1411                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1412
1413         if (tb[IFLA_LINKMODE]) {
1414                 write_lock_bh(&dev_base_lock);
1415                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1416                 write_unlock_bh(&dev_base_lock);
1417         }
1418
1419         if (tb[IFLA_VFINFO_LIST]) {
1420                 struct nlattr *attr;
1421                 int rem;
1422                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1423                         if (nla_type(attr) != IFLA_VF_INFO) {
1424                                 err = -EINVAL;
1425                                 goto errout;
1426                         }
1427                         err = do_setvfinfo(dev, attr);
1428                         if (err < 0)
1429                                 goto errout;
1430                         modified = 1;
1431                 }
1432         }
1433         err = 0;
1434
1435         if (tb[IFLA_VF_PORTS]) {
1436                 struct nlattr *port[IFLA_PORT_MAX+1];
1437                 struct nlattr *attr;
1438                 int vf;
1439                 int rem;
1440
1441                 err = -EOPNOTSUPP;
1442                 if (!ops->ndo_set_vf_port)
1443                         goto errout;
1444
1445                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1446                         if (nla_type(attr) != IFLA_VF_PORT)
1447                                 continue;
1448                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1449                                 attr, ifla_port_policy);
1450                         if (err < 0)
1451                                 goto errout;
1452                         if (!port[IFLA_PORT_VF]) {
1453                                 err = -EOPNOTSUPP;
1454                                 goto errout;
1455                         }
1456                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1457                         err = ops->ndo_set_vf_port(dev, vf, port);
1458                         if (err < 0)
1459                                 goto errout;
1460                         modified = 1;
1461                 }
1462         }
1463         err = 0;
1464
1465         if (tb[IFLA_PORT_SELF]) {
1466                 struct nlattr *port[IFLA_PORT_MAX+1];
1467
1468                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1469                         tb[IFLA_PORT_SELF], ifla_port_policy);
1470                 if (err < 0)
1471                         goto errout;
1472
1473                 err = -EOPNOTSUPP;
1474                 if (ops->ndo_set_vf_port)
1475                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1476                 if (err < 0)
1477                         goto errout;
1478                 modified = 1;
1479         }
1480
1481         if (tb[IFLA_AF_SPEC]) {
1482                 struct nlattr *af;
1483                 int rem;
1484
1485                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1486                         const struct rtnl_af_ops *af_ops;
1487
1488                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1489                                 BUG();
1490
1491                         err = af_ops->set_link_af(dev, af);
1492                         if (err < 0)
1493                                 goto errout;
1494
1495                         modified = 1;
1496                 }
1497         }
1498         err = 0;
1499
1500 errout:
1501         if (err < 0 && modified && net_ratelimit())
1502                 printk(KERN_WARNING "A link change request failed with "
1503                        "some changes committed already. Interface %s may "
1504                        "have been left with an inconsistent configuration, "
1505                        "please check.\n", dev->name);
1506
1507         if (send_addr_notify)
1508                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1509         return err;
1510 }
1511
1512 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1513 {
1514         struct net *net = sock_net(skb->sk);
1515         struct ifinfomsg *ifm;
1516         struct net_device *dev;
1517         int err;
1518         struct nlattr *tb[IFLA_MAX+1];
1519         char ifname[IFNAMSIZ];
1520
1521         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1522         if (err < 0)
1523                 goto errout;
1524
1525         if (tb[IFLA_IFNAME])
1526                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1527         else
1528                 ifname[0] = '\0';
1529
1530         err = -EINVAL;
1531         ifm = nlmsg_data(nlh);
1532         if (ifm->ifi_index > 0)
1533                 dev = __dev_get_by_index(net, ifm->ifi_index);
1534         else if (tb[IFLA_IFNAME])
1535                 dev = __dev_get_by_name(net, ifname);
1536         else
1537                 goto errout;
1538
1539         if (dev == NULL) {
1540                 err = -ENODEV;
1541                 goto errout;
1542         }
1543
1544         err = validate_linkmsg(dev, tb);
1545         if (err < 0)
1546                 goto errout;
1547
1548         err = do_setlink(dev, ifm, tb, ifname, 0);
1549 errout:
1550         return err;
1551 }
1552
1553 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1554 {
1555         struct net *net = sock_net(skb->sk);
1556         const struct rtnl_link_ops *ops;
1557         struct net_device *dev;
1558         struct ifinfomsg *ifm;
1559         char ifname[IFNAMSIZ];
1560         struct nlattr *tb[IFLA_MAX+1];
1561         int err;
1562         LIST_HEAD(list_kill);
1563
1564         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1565         if (err < 0)
1566                 return err;
1567
1568         if (tb[IFLA_IFNAME])
1569                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1570
1571         ifm = nlmsg_data(nlh);
1572         if (ifm->ifi_index > 0)
1573                 dev = __dev_get_by_index(net, ifm->ifi_index);
1574         else if (tb[IFLA_IFNAME])
1575                 dev = __dev_get_by_name(net, ifname);
1576         else
1577                 return -EINVAL;
1578
1579         if (!dev)
1580                 return -ENODEV;
1581
1582         ops = dev->rtnl_link_ops;
1583         if (!ops)
1584                 return -EOPNOTSUPP;
1585
1586         ops->dellink(dev, &list_kill);
1587         unregister_netdevice_many(&list_kill);
1588         list_del(&list_kill);
1589         return 0;
1590 }
1591
1592 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1593 {
1594         unsigned int old_flags;
1595         int err;
1596
1597         old_flags = dev->flags;
1598         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1599                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1600                 if (err < 0)
1601                         return err;
1602         }
1603
1604         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1605         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1606
1607         __dev_notify_flags(dev, old_flags);
1608         return 0;
1609 }
1610 EXPORT_SYMBOL(rtnl_configure_link);
1611
1612 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1613         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1614 {
1615         int err;
1616         struct net_device *dev;
1617         unsigned int num_queues = 1;
1618         unsigned int real_num_queues = 1;
1619
1620         if (ops->get_tx_queues) {
1621                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1622                                          &real_num_queues);
1623                 if (err)
1624                         goto err;
1625         }
1626         err = -ENOMEM;
1627         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1628         if (!dev)
1629                 goto err;
1630
1631         dev_net_set(dev, net);
1632         dev->rtnl_link_ops = ops;
1633         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1634
1635         if (tb[IFLA_MTU])
1636                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1637         if (tb[IFLA_ADDRESS])
1638                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1639                                 nla_len(tb[IFLA_ADDRESS]));
1640         if (tb[IFLA_BROADCAST])
1641                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1642                                 nla_len(tb[IFLA_BROADCAST]));
1643         if (tb[IFLA_TXQLEN])
1644                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1645         if (tb[IFLA_OPERSTATE])
1646                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1647         if (tb[IFLA_LINKMODE])
1648                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1649         if (tb[IFLA_GROUP])
1650                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1651
1652         return dev;
1653
1654 err:
1655         return ERR_PTR(err);
1656 }
1657 EXPORT_SYMBOL(rtnl_create_link);
1658
1659 static int rtnl_group_changelink(struct net *net, int group,
1660                 struct ifinfomsg *ifm,
1661                 struct nlattr **tb)
1662 {
1663         struct net_device *dev;
1664         int err;
1665
1666         for_each_netdev(net, dev) {
1667                 if (dev->group == group) {
1668                         err = do_setlink(dev, ifm, tb, NULL, 0);
1669                         if (err < 0)
1670                                 return err;
1671                 }
1672         }
1673
1674         return 0;
1675 }
1676
1677 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1678 {
1679         struct net *net = sock_net(skb->sk);
1680         const struct rtnl_link_ops *ops;
1681         struct net_device *dev;
1682         struct ifinfomsg *ifm;
1683         char kind[MODULE_NAME_LEN];
1684         char ifname[IFNAMSIZ];
1685         struct nlattr *tb[IFLA_MAX+1];
1686         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1687         int err;
1688
1689 #ifdef CONFIG_MODULES
1690 replay:
1691 #endif
1692         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1693         if (err < 0)
1694                 return err;
1695
1696         if (tb[IFLA_IFNAME])
1697                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1698         else
1699                 ifname[0] = '\0';
1700
1701         ifm = nlmsg_data(nlh);
1702         if (ifm->ifi_index > 0)
1703                 dev = __dev_get_by_index(net, ifm->ifi_index);
1704         else {
1705                 if (ifname[0])
1706                         dev = __dev_get_by_name(net, ifname);
1707                 else
1708                         dev = NULL;
1709         }
1710
1711         err = validate_linkmsg(dev, tb);
1712         if (err < 0)
1713                 return err;
1714
1715         if (tb[IFLA_LINKINFO]) {
1716                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1717                                        tb[IFLA_LINKINFO], ifla_info_policy);
1718                 if (err < 0)
1719                         return err;
1720         } else
1721                 memset(linkinfo, 0, sizeof(linkinfo));
1722
1723         if (linkinfo[IFLA_INFO_KIND]) {
1724                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1725                 ops = rtnl_link_ops_get(kind);
1726         } else {
1727                 kind[0] = '\0';
1728                 ops = NULL;
1729         }
1730
1731         if (1) {
1732                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1733                 struct net *dest_net;
1734
1735                 if (ops) {
1736                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1737                                 err = nla_parse_nested(attr, ops->maxtype,
1738                                                        linkinfo[IFLA_INFO_DATA],
1739                                                        ops->policy);
1740                                 if (err < 0)
1741                                         return err;
1742                                 data = attr;
1743                         }
1744                         if (ops->validate) {
1745                                 err = ops->validate(tb, data);
1746                                 if (err < 0)
1747                                         return err;
1748                         }
1749                 }
1750
1751                 if (dev) {
1752                         int modified = 0;
1753
1754                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1755                                 return -EEXIST;
1756                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1757                                 return -EOPNOTSUPP;
1758
1759                         if (linkinfo[IFLA_INFO_DATA]) {
1760                                 if (!ops || ops != dev->rtnl_link_ops ||
1761                                     !ops->changelink)
1762                                         return -EOPNOTSUPP;
1763
1764                                 err = ops->changelink(dev, tb, data);
1765                                 if (err < 0)
1766                                         return err;
1767                                 modified = 1;
1768                         }
1769
1770                         return do_setlink(dev, ifm, tb, ifname, modified);
1771                 }
1772
1773                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1774                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1775                                 return rtnl_group_changelink(net,
1776                                                 nla_get_u32(tb[IFLA_GROUP]),
1777                                                 ifm, tb);
1778                         return -ENODEV;
1779                 }
1780
1781                 if (ifm->ifi_index)
1782                         return -EOPNOTSUPP;
1783                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1784                         return -EOPNOTSUPP;
1785
1786                 if (!ops) {
1787 #ifdef CONFIG_MODULES
1788                         if (kind[0]) {
1789                                 __rtnl_unlock();
1790                                 request_module("rtnl-link-%s", kind);
1791                                 rtnl_lock();
1792                                 ops = rtnl_link_ops_get(kind);
1793                                 if (ops)
1794                                         goto replay;
1795                         }
1796 #endif
1797                         return -EOPNOTSUPP;
1798                 }
1799
1800                 if (!ifname[0])
1801                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1802
1803                 dest_net = rtnl_link_get_net(net, tb);
1804                 if (IS_ERR(dest_net))
1805                         return PTR_ERR(dest_net);
1806
1807                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1808
1809                 if (IS_ERR(dev))
1810                         err = PTR_ERR(dev);
1811                 else if (ops->newlink)
1812                         err = ops->newlink(net, dev, tb, data);
1813                 else
1814                         err = register_netdevice(dev);
1815
1816                 if (err < 0 && !IS_ERR(dev))
1817                         free_netdev(dev);
1818                 if (err < 0)
1819                         goto out;
1820
1821                 err = rtnl_configure_link(dev, ifm);
1822                 if (err < 0)
1823                         unregister_netdevice(dev);
1824 out:
1825                 put_net(dest_net);
1826                 return err;
1827         }
1828 }
1829
1830 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1831 {
1832         struct net *net = sock_net(skb->sk);
1833         struct ifinfomsg *ifm;
1834         char ifname[IFNAMSIZ];
1835         struct nlattr *tb[IFLA_MAX+1];
1836         struct net_device *dev = NULL;
1837         struct sk_buff *nskb;
1838         int err;
1839
1840         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1841         if (err < 0)
1842                 return err;
1843
1844         if (tb[IFLA_IFNAME])
1845                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1846
1847         ifm = nlmsg_data(nlh);
1848         if (ifm->ifi_index > 0)
1849                 dev = __dev_get_by_index(net, ifm->ifi_index);
1850         else if (tb[IFLA_IFNAME])
1851                 dev = __dev_get_by_name(net, ifname);
1852         else
1853                 return -EINVAL;
1854
1855         if (dev == NULL)
1856                 return -ENODEV;
1857
1858         nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1859         if (nskb == NULL)
1860                 return -ENOBUFS;
1861
1862         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1863                                nlh->nlmsg_seq, 0, 0);
1864         if (err < 0) {
1865                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1866                 WARN_ON(err == -EMSGSIZE);
1867                 kfree_skb(nskb);
1868         } else
1869                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1870
1871         return err;
1872 }
1873
1874 static u16 rtnl_calcit(struct sk_buff *skb)
1875 {
1876         return min_ifinfo_dump_size;
1877 }
1878
1879 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1880 {
1881         int idx;
1882         int s_idx = cb->family;
1883
1884         if (s_idx == 0)
1885                 s_idx = 1;
1886         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1887                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1888                 if (idx < s_idx || idx == PF_PACKET)
1889                         continue;
1890                 if (rtnl_msg_handlers[idx] == NULL ||
1891                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1892                         continue;
1893                 if (idx > s_idx)
1894                         memset(&cb->args[0], 0, sizeof(cb->args));
1895                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1896                         break;
1897         }
1898         cb->family = idx;
1899
1900         return skb->len;
1901 }
1902
1903 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1904 {
1905         struct net *net = dev_net(dev);
1906         struct sk_buff *skb;
1907         int err = -ENOBUFS;
1908         size_t if_info_size;
1909
1910         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1911         if (skb == NULL)
1912                 goto errout;
1913
1914         min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1915
1916         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1917         if (err < 0) {
1918                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1919                 WARN_ON(err == -EMSGSIZE);
1920                 kfree_skb(skb);
1921                 goto errout;
1922         }
1923         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1924         return;
1925 errout:
1926         if (err < 0)
1927                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1928 }
1929
1930 /* Protected by RTNL sempahore.  */
1931 static struct rtattr **rta_buf;
1932 static int rtattr_max;
1933
1934 /* Process one rtnetlink message. */
1935
1936 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1937 {
1938         struct net *net = sock_net(skb->sk);
1939         rtnl_doit_func doit;
1940         int sz_idx, kind;
1941         int min_len;
1942         int family;
1943         int type;
1944         int err;
1945
1946         type = nlh->nlmsg_type;
1947         if (type > RTM_MAX)
1948                 return -EOPNOTSUPP;
1949
1950         type -= RTM_BASE;
1951
1952         /* All the messages must have at least 1 byte length */
1953         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1954                 return 0;
1955
1956         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1957         sz_idx = type>>2;
1958         kind = type&3;
1959
1960         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1961                 return -EPERM;
1962
1963         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1964                 struct sock *rtnl;
1965                 rtnl_dumpit_func dumpit;
1966                 rtnl_calcit_func calcit;
1967                 u16 min_dump_alloc = 0;
1968
1969                 dumpit = rtnl_get_dumpit(family, type);
1970                 if (dumpit == NULL)
1971                         return -EOPNOTSUPP;
1972                 calcit = rtnl_get_calcit(family, type);
1973                 if (calcit)
1974                         min_dump_alloc = calcit(skb);
1975
1976                 __rtnl_unlock();
1977                 rtnl = net->rtnl;
1978                 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
1979                                          NULL, min_dump_alloc);
1980                 rtnl_lock();
1981                 return err;
1982         }
1983
1984         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1985
1986         min_len = rtm_min[sz_idx];
1987         if (nlh->nlmsg_len < min_len)
1988                 return -EINVAL;
1989
1990         if (nlh->nlmsg_len > min_len) {
1991                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1992                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1993
1994                 while (RTA_OK(attr, attrlen)) {
1995                         unsigned flavor = attr->rta_type;
1996                         if (flavor) {
1997                                 if (flavor > rta_max[sz_idx])
1998                                         return -EINVAL;
1999                                 rta_buf[flavor-1] = attr;
2000                         }
2001                         attr = RTA_NEXT(attr, attrlen);
2002                 }
2003         }
2004
2005         doit = rtnl_get_doit(family, type);
2006         if (doit == NULL)
2007                 return -EOPNOTSUPP;
2008
2009         return doit(skb, nlh, (void *)&rta_buf[0]);
2010 }
2011
2012 static void rtnetlink_rcv(struct sk_buff *skb)
2013 {
2014         rtnl_lock();
2015         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2016         rtnl_unlock();
2017 }
2018
2019 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2020 {
2021         struct net_device *dev = ptr;
2022
2023         switch (event) {
2024         case NETDEV_UP:
2025         case NETDEV_DOWN:
2026         case NETDEV_PRE_UP:
2027         case NETDEV_POST_INIT:
2028         case NETDEV_REGISTER:
2029         case NETDEV_CHANGE:
2030         case NETDEV_PRE_TYPE_CHANGE:
2031         case NETDEV_GOING_DOWN:
2032         case NETDEV_UNREGISTER:
2033         case NETDEV_UNREGISTER_BATCH:
2034         case NETDEV_RELEASE:
2035         case NETDEV_JOIN:
2036                 break;
2037         default:
2038                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2039                 break;
2040         }
2041         return NOTIFY_DONE;
2042 }
2043
2044 static struct notifier_block rtnetlink_dev_notifier = {
2045         .notifier_call  = rtnetlink_event,
2046 };
2047
2048
2049 static int __net_init rtnetlink_net_init(struct net *net)
2050 {
2051         struct sock *sk;
2052         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2053                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2054         if (!sk)
2055                 return -ENOMEM;
2056         net->rtnl = sk;
2057         return 0;
2058 }
2059
2060 static void __net_exit rtnetlink_net_exit(struct net *net)
2061 {
2062         netlink_kernel_release(net->rtnl);
2063         net->rtnl = NULL;
2064 }
2065
2066 static struct pernet_operations rtnetlink_net_ops = {
2067         .init = rtnetlink_net_init,
2068         .exit = rtnetlink_net_exit,
2069 };
2070
2071 void __init rtnetlink_init(void)
2072 {
2073         int i;
2074
2075         rtattr_max = 0;
2076         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2077                 if (rta_max[i] > rtattr_max)
2078                         rtattr_max = rta_max[i];
2079         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2080         if (!rta_buf)
2081                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2082
2083         if (register_pernet_subsys(&rtnetlink_net_ops))
2084                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2085
2086         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2087         register_netdevice_notifier(&rtnetlink_dev_notifier);
2088
2089         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2090                       rtnl_dump_ifinfo, rtnl_calcit);
2091         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2092         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2093         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2094
2095         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2096         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2097 }
2098