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 };
  60
  61 static DEFINE_MUTEX(rtnl_mutex);
  62
  63 void rtnl_lock(void)
  64 {
  65         mutex_lock(&rtnl_mutex);
  66 }
  67 EXPORT_SYMBOL(rtnl_lock);
  68
  69 void __rtnl_unlock(void)
  70 {
  71         mutex_unlock(&rtnl_mutex);
  72 }
  73
  74 void rtnl_unlock(void)
  75 {
  76         /* This fellow will unlock it for us. */
  77         netdev_run_todo();
  78 }
  79 EXPORT_SYMBOL(rtnl_unlock);
  80
  81 int rtnl_trylock(void)
  82 {
  83         return mutex_trylock(&rtnl_mutex);
  84 }
  85 EXPORT_SYMBOL(rtnl_trylock);
  86
  87 int rtnl_is_locked(void)
  88 {
  89         return mutex_is_locked(&rtnl_mutex);
  90 }
  91 EXPORT_SYMBOL(rtnl_is_locked);
  92
  93 #ifdef CONFIG_PROVE_LOCKING
  94 int lockdep_rtnl_is_held(void)
  95 {
  96         return lockdep_is_held(&rtnl_mutex);
  97 }
  98 EXPORT_SYMBOL(lockdep_rtnl_is_held);
  99 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 100
 101 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 102
 103 static inline int rtm_msgindex(int msgtype)
 104 {
 105         int msgindex = msgtype - RTM_BASE;
 106
 107         /*
 108          * msgindex < 0 implies someone tried to register a netlink
 109          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 110          * the message type has not been added to linux/rtnetlink.h
 111          */
 112         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 113
 114         return msgindex;
 115 }
 116
 117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 118 {
 119         struct rtnl_link *tab;
 120
 121         if (protocol <= RTNL_FAMILY_MAX)
 122                 tab = rtnl_msg_handlers[protocol];
 123         else
 124                 tab = NULL;
 125
 126         if (tab == NULL || tab[msgindex].doit == NULL)
 127                 tab = rtnl_msg_handlers[PF_UNSPEC];
 128
 129         return tab ? tab[msgindex].doit : NULL;
 130 }
 131
 132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 133 {
 134         struct rtnl_link *tab;
 135
 136         if (protocol <= RTNL_FAMILY_MAX)
 137                 tab = rtnl_msg_handlers[protocol];
 138         else
 139                 tab = NULL;
 140
 141         if (tab == NULL || tab[msgindex].dumpit == NULL)
 142                 tab = rtnl_msg_handlers[PF_UNSPEC];
 143
 144         return tab ? tab[msgindex].dumpit : NULL;
 145 }
 146
 147 /**
 148  * __rtnl_register - Register a rtnetlink message type
 149  * @protocol: Protocol family or PF_UNSPEC
 150  * @msgtype: rtnetlink message type
 151  * @doit: Function pointer called for each request message
 152  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 153  *
 154  * Registers the specified function pointers (at least one of them has
 155  * to be non-NULL) to be called whenever a request message for the
 156  * specified protocol family and message type is received.
 157  *
 158  * The special protocol family PF_UNSPEC may be used to define fallback
 159  * function pointers for the case when no entry for the specific protocol
 160  * family exists.
 161  *
 162  * Returns 0 on success or a negative error code.
 163  */
 164 int __rtnl_register(int protocol, int msgtype,
 165                     rtnl_doit_func doit, rtnl_dumpit_func dumpit)
 166 {
 167         struct rtnl_link *tab;
 168         int msgindex;
 169
 170         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 171         msgindex = rtm_msgindex(msgtype);
 172
 173         tab = rtnl_msg_handlers[protocol];
 174         if (tab == NULL) {
 175                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 176                 if (tab == NULL)
 177                         return -ENOBUFS;
 178
 179                 rtnl_msg_handlers[protocol] = tab;
 180         }
 181
 182         if (doit)
 183                 tab[msgindex].doit = doit;
 184
 185         if (dumpit)
 186                 tab[msgindex].dumpit = dumpit;
 187
 188         return 0;
 189 }
 190 EXPORT_SYMBOL_GPL(__rtnl_register);
 191
 192 /**
 193  * rtnl_register - Register a rtnetlink message type
 194  *
 195  * Identical to __rtnl_register() but panics on failure. This is useful
 196  * as failure of this function is very unlikely, it can only happen due
 197  * to lack of memory when allocating the chain to store all message
 198  * handlers for a protocol. Meant for use in init functions where lack
 199  * of memory implies no sense in continuing.
 200  */
 201 void rtnl_register(int protocol, int msgtype,
 202                    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
 203 {
 204         if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
 205                 panic("Unable to register rtnetlink message handler, "
 206                       "protocol = %d, message type = %d\n",
 207                       protocol, msgtype);
 208 }
 209 EXPORT_SYMBOL_GPL(rtnl_register);
 210
 211 /**
 212  * rtnl_unregister - Unregister a rtnetlink message type
 213  * @protocol: Protocol family or PF_UNSPEC
 214  * @msgtype: rtnetlink message type
 215  *
 216  * Returns 0 on success or a negative error code.
 217  */
 218 int rtnl_unregister(int protocol, int msgtype)
 219 {
 220         int msgindex;
 221
 222         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 223         msgindex = rtm_msgindex(msgtype);
 224
 225         if (rtnl_msg_handlers[protocol] == NULL)
 226                 return -ENOENT;
 227
 228         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 229         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 230
 231         return 0;
 232 }
 233 EXPORT_SYMBOL_GPL(rtnl_unregister);
 234
 235 /**
 236  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 237  * @protocol : Protocol family or PF_UNSPEC
 238  *
 239  * Identical to calling rtnl_unregster() for all registered message types
 240  * of a certain protocol family.
 241  */
 242 void rtnl_unregister_all(int protocol)
 243 {
 244         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 245
 246         kfree(rtnl_msg_handlers[protocol]);
 247         rtnl_msg_handlers[protocol] = NULL;
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 250
 251 static LIST_HEAD(link_ops);
 252
 253 /**
 254  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 255  * @ops: struct rtnl_link_ops * to register
 256  *
 257  * The caller must hold the rtnl_mutex. This function should be used
 258  * by drivers that create devices during module initialization. It
 259  * must be called before registering the devices.
 260  *
 261  * Returns 0 on success or a negative error code.
 262  */
 263 int __rtnl_link_register(struct rtnl_link_ops *ops)
 264 {
 265         if (!ops->dellink)
 266                 ops->dellink = unregister_netdevice_queue;
 267
 268         list_add_tail(&ops->list, &link_ops);
 269         return 0;
 270 }
 271 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 272
 273 /**
 274  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 275  * @ops: struct rtnl_link_ops * to register
 276  *
 277  * Returns 0 on success or a negative error code.
 278  */
 279 int rtnl_link_register(struct rtnl_link_ops *ops)
 280 {
 281         int err;
 282
 283         rtnl_lock();
 284         err = __rtnl_link_register(ops);
 285         rtnl_unlock();
 286         return err;
 287 }
 288 EXPORT_SYMBOL_GPL(rtnl_link_register);
 289
 290 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 291 {
 292         struct net_device *dev;
 293         LIST_HEAD(list_kill);
 294
 295         for_each_netdev(net, dev) {
 296                 if (dev->rtnl_link_ops == ops)
 297                         ops->dellink(dev, &list_kill);
 298         }
 299         unregister_netdevice_many(&list_kill);
 300 }
 301
 302 /**
 303  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 304  * @ops: struct rtnl_link_ops * to unregister
 305  *
 306  * The caller must hold the rtnl_mutex.
 307  */
 308 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 309 {
 310         struct net *net;
 311
 312         for_each_net(net) {
 313                 __rtnl_kill_links(net, ops);
 314         }
 315         list_del(&ops->list);
 316 }
 317 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 318
 319 /**
 320  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 321  * @ops: struct rtnl_link_ops * to unregister
 322  */
 323 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 324 {
 325         rtnl_lock();
 326         __rtnl_link_unregister(ops);
 327         rtnl_unlock();
 328 }
 329 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 330
 331 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 332 {
 333         const struct rtnl_link_ops *ops;
 334
 335         list_for_each_entry(ops, &link_ops, list) {
 336                 if (!strcmp(ops->kind, kind))
 337                         return ops;
 338         }
 339         return NULL;
 340 }
 341
 342 static size_t rtnl_link_get_size(const struct net_device *dev)
 343 {
 344         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 345         size_t size;
 346
 347         if (!ops)
 348                 return 0;
 349
 350         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 351                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 352
 353         if (ops->get_size)
 354                 /* IFLA_INFO_DATA + nested data */
 355                 size += nla_total_size(sizeof(struct nlattr)) +
 356                         ops->get_size(dev);
 357
 358         if (ops->get_xstats_size)
 359                 /* IFLA_INFO_XSTATS */
 360                 size += nla_total_size(ops->get_xstats_size(dev));
 361
 362         return size;
 363 }
 364
 365 static LIST_HEAD(rtnl_af_ops);
 366
 367 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 368 {
 369         const struct rtnl_af_ops *ops;
 370
 371         list_for_each_entry(ops, &rtnl_af_ops, list) {
 372                 if (ops->family == family)
 373                         return ops;
 374         }
 375
 376         return NULL;
 377 }
 378
 379 /**
 380  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 381  * @ops: struct rtnl_af_ops * to register
 382  *
 383  * The caller must hold the rtnl_mutex.
 384  *
 385  * Returns 0 on success or a negative error code.
 386  */
 387 int __rtnl_af_register(struct rtnl_af_ops *ops)
 388 {
 389         list_add_tail(&ops->list, &rtnl_af_ops);
 390         return 0;
 391 }
 392 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 393
 394 /**
 395  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 396  * @ops: struct rtnl_af_ops * to register
 397  *
 398  * Returns 0 on success or a negative error code.
 399  */
 400 int rtnl_af_register(struct rtnl_af_ops *ops)
 401 {
 402         int err;
 403
 404         rtnl_lock();
 405         err = __rtnl_af_register(ops);
 406         rtnl_unlock();
 407         return err;
 408 }
 409 EXPORT_SYMBOL_GPL(rtnl_af_register);
 410
 411 /**
 412  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 413  * @ops: struct rtnl_af_ops * to unregister
 414  *
 415  * The caller must hold the rtnl_mutex.
 416  */
 417 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 418 {
 419         list_del(&ops->list);
 420 }
 421 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 422
 423 /**
 424  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 425  * @ops: struct rtnl_af_ops * to unregister
 426  */
 427 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 428 {
 429         rtnl_lock();
 430         __rtnl_af_unregister(ops);
 431         rtnl_unlock();
 432 }
 433 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 434
 435 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 436 {
 437         struct rtnl_af_ops *af_ops;
 438         size_t size;
 439
 440         /* IFLA_AF_SPEC */
 441         size = nla_total_size(sizeof(struct nlattr));
 442
 443         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 444                 if (af_ops->get_link_af_size) {
 445                         /* AF_* + nested data */
 446                         size += nla_total_size(sizeof(struct nlattr)) +
 447                                 af_ops->get_link_af_size(dev);
 448                 }
 449         }
 450
 451         return size;
 452 }
 453
 454 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 455 {
 456         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 457         struct nlattr *linkinfo, *data;
 458         int err = -EMSGSIZE;
 459
 460         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 461         if (linkinfo == NULL)
 462                 goto out;
 463
 464         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 465                 goto err_cancel_link;
 466         if (ops->fill_xstats) {
 467                 err = ops->fill_xstats(skb, dev);
 468                 if (err < 0)
 469                         goto err_cancel_link;
 470         }
 471         if (ops->fill_info) {
 472                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 473                 if (data == NULL)
 474                         goto err_cancel_link;
 475                 err = ops->fill_info(skb, dev);
 476                 if (err < 0)
 477                         goto err_cancel_data;
 478                 nla_nest_end(skb, data);
 479         }
 480
 481         nla_nest_end(skb, linkinfo);
 482         return 0;
 483
 484 err_cancel_data:
 485         nla_nest_cancel(skb, data);
 486 err_cancel_link:
 487         nla_nest_cancel(skb, linkinfo);
 488 out:
 489         return err;
 490 }
 491
 492 static const int rtm_min[RTM_NR_FAMILIES] =
 493 {
 494         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 495         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 496         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 497         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 498         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 499         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 500         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 501         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 502         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 503         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 504 };
 505
 506 static const int rta_max[RTM_NR_FAMILIES] =
 507 {
 508         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 509         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 510         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 511         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 512         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 513         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 514         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 515         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 516 };
 517
 518 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 519 {
 520         struct rtattr *rta;
 521         int size = RTA_LENGTH(attrlen);
 522
 523         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 524         rta->rta_type = attrtype;
 525         rta->rta_len = size;
 526         memcpy(RTA_DATA(rta), data, attrlen);
 527         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 528 }
 529 EXPORT_SYMBOL(__rta_fill);
 530
 531 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 532 {
 533         struct sock *rtnl = net->rtnl;
 534         int err = 0;
 535
 536         NETLINK_CB(skb).dst_group = group;
 537         if (echo)
 538                 atomic_inc(&skb->users);
 539         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 540         if (echo)
 541                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 542         return err;
 543 }
 544
 545 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 546 {
 547         struct sock *rtnl = net->rtnl;
 548
 549         return nlmsg_unicast(rtnl, skb, pid);
 550 }
 551 EXPORT_SYMBOL(rtnl_unicast);
 552
 553 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 554                  struct nlmsghdr *nlh, gfp_t flags)
 555 {
 556         struct sock *rtnl = net->rtnl;
 557         int report = 0;
 558
 559         if (nlh)
 560                 report = nlmsg_report(nlh);
 561
 562         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 563 }
 564 EXPORT_SYMBOL(rtnl_notify);
 565
 566 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 567 {
 568         struct sock *rtnl = net->rtnl;
 569
 570         netlink_set_err(rtnl, 0, group, error);
 571 }
 572 EXPORT_SYMBOL(rtnl_set_sk_err);
 573
 574 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 575 {
 576         struct nlattr *mx;
 577         int i, valid = 0;
 578
 579         mx = nla_nest_start(skb, RTA_METRICS);
 580         if (mx == NULL)
 581                 return -ENOBUFS;
 582
 583         for (i = 0; i < RTAX_MAX; i++) {
 584                 if (metrics[i]) {
 585                         valid++;
 586                         NLA_PUT_U32(skb, i+1, metrics[i]);
 587                 }
 588         }
 589
 590         if (!valid) {
 591                 nla_nest_cancel(skb, mx);
 592                 return 0;
 593         }
 594
 595         return nla_nest_end(skb, mx);
 596
 597 nla_put_failure:
 598         nla_nest_cancel(skb, mx);
 599         return -EMSGSIZE;
 600 }
 601 EXPORT_SYMBOL(rtnetlink_put_metrics);
 602
 603 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 604                        u32 ts, u32 tsage, long expires, u32 error)
 605 {
 606         struct rta_cacheinfo ci = {
 607                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 608                 .rta_used = dst->__use,
 609                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 610                 .rta_error = error,
 611                 .rta_id =  id,
 612                 .rta_ts = ts,
 613                 .rta_tsage = tsage,
 614         };
 615
 616         if (expires)
 617                 ci.rta_expires = jiffies_to_clock_t(expires);
 618
 619         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 620 }
 621 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 622
 623 static void set_operstate(struct net_device *dev, unsigned char transition)
 624 {
 625         unsigned char operstate = dev->operstate;
 626
 627         switch (transition) {
 628         case IF_OPER_UP:
 629                 if ((operstate == IF_OPER_DORMANT ||
 630                      operstate == IF_OPER_UNKNOWN) &&
 631                     !netif_dormant(dev))
 632                         operstate = IF_OPER_UP;
 633                 break;
 634
 635         case IF_OPER_DORMANT:
 636                 if (operstate == IF_OPER_UP ||
 637                     operstate == IF_OPER_UNKNOWN)
 638                         operstate = IF_OPER_DORMANT;
 639                 break;
 640         }
 641
 642         if (dev->operstate != operstate) {
 643                 write_lock_bh(&dev_base_lock);
 644                 dev->operstate = operstate;
 645                 write_unlock_bh(&dev_base_lock);
 646                 netdev_state_change(dev);
 647         }
 648 }
 649
 650 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 651                                            const struct ifinfomsg *ifm)
 652 {
 653         unsigned int flags = ifm->ifi_flags;
 654
 655         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 656         if (ifm->ifi_change)
 657                 flags = (flags & ifm->ifi_change) |
 658                         (dev->flags & ~ifm->ifi_change);
 659
 660         return flags;
 661 }
 662
 663 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 664                                  const struct rtnl_link_stats64 *b)
 665 {
 666         a->rx_packets = b->rx_packets;
 667         a->tx_packets = b->tx_packets;
 668         a->rx_bytes = b->rx_bytes;
 669         a->tx_bytes = b->tx_bytes;
 670         a->rx_errors = b->rx_errors;
 671         a->tx_errors = b->tx_errors;
 672         a->rx_dropped = b->rx_dropped;
 673         a->tx_dropped = b->tx_dropped;
 674
 675         a->multicast = b->multicast;
 676         a->collisions = b->collisions;
 677
 678         a->rx_length_errors = b->rx_length_errors;
 679         a->rx_over_errors = b->rx_over_errors;
 680         a->rx_crc_errors = b->rx_crc_errors;
 681         a->rx_frame_errors = b->rx_frame_errors;
 682         a->rx_fifo_errors = b->rx_fifo_errors;
 683         a->rx_missed_errors = b->rx_missed_errors;
 684
 685         a->tx_aborted_errors = b->tx_aborted_errors;
 686         a->tx_carrier_errors = b->tx_carrier_errors;
 687         a->tx_fifo_errors = b->tx_fifo_errors;
 688         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 689         a->tx_window_errors = b->tx_window_errors;
 690
 691         a->rx_compressed = b->rx_compressed;
 692         a->tx_compressed = b->tx_compressed;
 693 }
 694
 695 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 696 {
 697         memcpy(v, b, sizeof(*b));
 698 }
 699
 700 /* All VF info */
 701 static inline int rtnl_vfinfo_size(const struct net_device *dev)
 702 {
 703         if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
 704
 705                 int num_vfs = dev_num_vf(dev->dev.parent);
 706                 size_t size = nla_total_size(sizeof(struct nlattr));
 707                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 708                 size += num_vfs *
 709                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 710                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 711                          nla_total_size(sizeof(struct ifla_vf_tx_rate)));
 712                 return size;
 713         } else
 714                 return 0;
 715 }
 716
 717 static size_t rtnl_port_size(const struct net_device *dev)
 718 {
 719         size_t port_size = nla_total_size(4)            /* PORT_VF */
 720                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 721                 + nla_total_size(sizeof(struct ifla_port_vsi))
 722                                                         /* PORT_VSI_TYPE */
 723                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 724                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 725                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 726                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 727         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 728         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 729                 + port_size;
 730         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 731                 + port_size;
 732
 733         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 734                 return 0;
 735         if (dev_num_vf(dev->dev.parent))
 736                 return port_self_size + vf_ports_size +
 737                         vf_port_size * dev_num_vf(dev->dev.parent);
 738         else
 739                 return port_self_size;
 740 }
 741
 742 static noinline size_t if_nlmsg_size(const struct net_device *dev)
 743 {
 744         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 745                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 746                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 747                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 748                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 749                + nla_total_size(sizeof(struct rtnl_link_stats))
 750                + nla_total_size(sizeof(struct rtnl_link_stats64))
 751                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 752                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 753                + nla_total_size(4) /* IFLA_TXQLEN */
 754                + nla_total_size(4) /* IFLA_WEIGHT */
 755                + nla_total_size(4) /* IFLA_MTU */
 756                + nla_total_size(4) /* IFLA_LINK */
 757                + nla_total_size(4) /* IFLA_MASTER */
 758                + nla_total_size(1) /* IFLA_OPERSTATE */
 759                + nla_total_size(1) /* IFLA_LINKMODE */
 760                + nla_total_size(4) /* IFLA_NUM_VF */
 761                + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
 762                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 763                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 764                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 765 }
 766
 767 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 768 {
 769         struct nlattr *vf_ports;
 770         struct nlattr *vf_port;
 771         int vf;
 772         int err;
 773
 774         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 775         if (!vf_ports)
 776                 return -EMSGSIZE;
 777
 778         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 779                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 780                 if (!vf_port)
 781                         goto nla_put_failure;
 782                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 783                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 784                 if (err == -EMSGSIZE)
 785                         goto nla_put_failure;
 786                 if (err) {
 787                         nla_nest_cancel(skb, vf_port);
 788                         continue;
 789                 }
 790                 nla_nest_end(skb, vf_port);
 791         }
 792
 793         nla_nest_end(skb, vf_ports);
 794
 795         return 0;
 796
 797 nla_put_failure:
 798         nla_nest_cancel(skb, vf_ports);
 799         return -EMSGSIZE;
 800 }
 801
 802 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 803 {
 804         struct nlattr *port_self;
 805         int err;
 806
 807         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 808         if (!port_self)
 809                 return -EMSGSIZE;
 810
 811         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 812         if (err) {
 813                 nla_nest_cancel(skb, port_self);
 814                 return (err == -EMSGSIZE) ? err : 0;
 815         }
 816
 817         nla_nest_end(skb, port_self);
 818
 819         return 0;
 820 }
 821
 822 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 823 {
 824         int err;
 825
 826         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 827                 return 0;
 828
 829         err = rtnl_port_self_fill(skb, dev);
 830         if (err)
 831                 return err;
 832
 833         if (dev_num_vf(dev->dev.parent)) {
 834                 err = rtnl_vf_ports_fill(skb, dev);
 835                 if (err)
 836                         return err;
 837         }
 838
 839         return 0;
 840 }
 841
 842 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 843                             int type, u32 pid, u32 seq, u32 change,
 844                             unsigned int flags)
 845 {
 846         struct ifinfomsg *ifm;
 847         struct nlmsghdr *nlh;
 848         struct rtnl_link_stats64 temp;
 849         const struct rtnl_link_stats64 *stats;
 850         struct nlattr *attr, *af_spec;
 851         struct rtnl_af_ops *af_ops;
 852
 853         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 854         if (nlh == NULL)
 855                 return -EMSGSIZE;
 856
 857         ifm = nlmsg_data(nlh);
 858         ifm->ifi_family = AF_UNSPEC;
 859         ifm->__ifi_pad = 0;
 860         ifm->ifi_type = dev->type;
 861         ifm->ifi_index = dev->ifindex;
 862         ifm->ifi_flags = dev_get_flags(dev);
 863         ifm->ifi_change = change;
 864
 865         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 866         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 867         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 868                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 869         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 870         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 871         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 872
 873         if (dev->ifindex != dev->iflink)
 874                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 875
 876         if (dev->master)
 877                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 878
 879         if (dev->qdisc)
 880                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 881
 882         if (dev->ifalias)
 883                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 884
 885         if (1) {
 886                 struct rtnl_link_ifmap map = {
 887                         .mem_start   = dev->mem_start,
 888                         .mem_end     = dev->mem_end,
 889                         .base_addr   = dev->base_addr,
 890                         .irq         = dev->irq,
 891                         .dma         = dev->dma,
 892                         .port        = dev->if_port,
 893                 };
 894                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 895         }
 896
 897         if (dev->addr_len) {
 898                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 899                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 900         }
 901
 902         attr = nla_reserve(skb, IFLA_STATS,
 903                         sizeof(struct rtnl_link_stats));
 904         if (attr == NULL)
 905                 goto nla_put_failure;
 906
 907         stats = dev_get_stats(dev, &temp);
 908         copy_rtnl_link_stats(nla_data(attr), stats);
 909
 910         attr = nla_reserve(skb, IFLA_STATS64,
 911                         sizeof(struct rtnl_link_stats64));
 912         if (attr == NULL)
 913                 goto nla_put_failure;
 914         copy_rtnl_link_stats64(nla_data(attr), stats);
 915
 916         if (dev->dev.parent)
 917                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 918
 919         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
 920                 int i;
 921
 922                 struct nlattr *vfinfo, *vf;
 923                 int num_vfs = dev_num_vf(dev->dev.parent);
 924
 925                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 926                 if (!vfinfo)
 927                         goto nla_put_failure;
 928                 for (i = 0; i < num_vfs; i++) {
 929                         struct ifla_vf_info ivi;
 930                         struct ifla_vf_mac vf_mac;
 931                         struct ifla_vf_vlan vf_vlan;
 932                         struct ifla_vf_tx_rate vf_tx_rate;
 933                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 934                                 break;
 935                         vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf;
 936                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 937                         vf_vlan.vlan = ivi.vlan;
 938                         vf_vlan.qos = ivi.qos;
 939                         vf_tx_rate.rate = ivi.tx_rate;
 940                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 941                         if (!vf) {
 942                                 nla_nest_cancel(skb, vfinfo);
 943                                 goto nla_put_failure;
 944                         }
 945                         NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
 946                         NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
 947                         NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate);
 948                         nla_nest_end(skb, vf);
 949                 }
 950                 nla_nest_end(skb, vfinfo);
 951         }
 952
 953         if (rtnl_port_fill(skb, dev))
 954                 goto nla_put_failure;
 955
 956         if (dev->rtnl_link_ops) {
 957                 if (rtnl_link_fill(skb, dev) < 0)
 958                         goto nla_put_failure;
 959         }
 960
 961         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
 962                 goto nla_put_failure;
 963
 964         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 965                 if (af_ops->fill_link_af) {
 966                         struct nlattr *af;
 967                         int err;
 968
 969                         if (!(af = nla_nest_start(skb, af_ops->family)))
 970                                 goto nla_put_failure;
 971
 972                         err = af_ops->fill_link_af(skb, dev);
 973
 974                         /*
 975                          * Caller may return ENODATA to indicate that there
 976                          * was no data to be dumped. This is not an error, it
 977                          * means we should trim the attribute header and
 978                          * continue.
 979                          */
 980                         if (err == -ENODATA)
 981                                 nla_nest_cancel(skb, af);
 982                         else if (err < 0)
 983                                 goto nla_put_failure;
 984
 985                         nla_nest_end(skb, af);
 986                 }
 987         }
 988
 989         nla_nest_end(skb, af_spec);
 990
 991         return nlmsg_end(skb, nlh);
 992
 993 nla_put_failure:
 994         nlmsg_cancel(skb, nlh);
 995         return -EMSGSIZE;
 996 }
 997
 998 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 999 {
1000         struct net *net = sock_net(skb->sk);
1001         int h, s_h;
1002         int idx = 0, s_idx;
1003         struct net_device *dev;
1004         struct hlist_head *head;
1005         struct hlist_node *node;
1006
1007         s_h = cb->args[0];
1008         s_idx = cb->args[1];
1009
1010         rcu_read_lock();
1011         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1012                 idx = 0;
1013                 head = &net->dev_index_head[h];
1014                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1015                         if (idx < s_idx)
1016                                 goto cont;
1017                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1018                                              NETLINK_CB(cb->skb).pid,
1019                                              cb->nlh->nlmsg_seq, 0,
1020                                              NLM_F_MULTI) <= 0)
1021                                 goto out;
1022 cont:
1023                         idx++;
1024                 }
1025         }
1026 out:
1027         rcu_read_unlock();
1028         cb->args[1] = idx;
1029         cb->args[0] = h;
1030
1031         return skb->len;
1032 }
1033
1034 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1035         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1036         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1037         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1038         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1039         [IFLA_MTU]              = { .type = NLA_U32 },
1040         [IFLA_LINK]             = { .type = NLA_U32 },
1041         [IFLA_MASTER]           = { .type = NLA_U32 },
1042         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1043         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1044         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1045         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1046         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1047         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1048         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1049         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1050         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1051         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1052         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1053 };
1054 EXPORT_SYMBOL(ifla_policy);
1055
1056 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1057         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1058         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1059 };
1060
1061 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1062         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1063 };
1064
1065 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1066         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1067                                     .len = sizeof(struct ifla_vf_mac) },
1068         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1069                                     .len = sizeof(struct ifla_vf_vlan) },
1070         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1071                                     .len = sizeof(struct ifla_vf_tx_rate) },
1072 };
1073
1074 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1075         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1076         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1077                                     .len = PORT_PROFILE_MAX },
1078         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1079                                     .len = sizeof(struct ifla_port_vsi)},
1080         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1081                                       .len = PORT_UUID_MAX },
1082         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1083                                     .len = PORT_UUID_MAX },
1084         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1085         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1086 };
1087
1088 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1089 {
1090         struct net *net;
1091         /* Examine the link attributes and figure out which
1092          * network namespace we are talking about.
1093          */
1094         if (tb[IFLA_NET_NS_PID])
1095                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1096         else
1097                 net = get_net(src_net);
1098         return net;
1099 }
1100 EXPORT_SYMBOL(rtnl_link_get_net);
1101
1102 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1103 {
1104         if (dev) {
1105                 if (tb[IFLA_ADDRESS] &&
1106                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1107                         return -EINVAL;
1108
1109                 if (tb[IFLA_BROADCAST] &&
1110                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1111                         return -EINVAL;
1112         }
1113
1114         if (tb[IFLA_AF_SPEC]) {
1115                 struct nlattr *af;
1116                 int rem, err;
1117
1118                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1119                         const struct rtnl_af_ops *af_ops;
1120
1121                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1122                                 return -EAFNOSUPPORT;
1123
1124                         if (!af_ops->set_link_af)
1125                                 return -EOPNOTSUPP;
1126
1127                         if (af_ops->validate_link_af) {
1128                                 err = af_ops->validate_link_af(dev, af);
1129                                 if (err < 0)
1130                                         return err;
1131                         }
1132                 }
1133         }
1134
1135         return 0;
1136 }
1137
1138 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1139 {
1140         int rem, err = -EINVAL;
1141         struct nlattr *vf;
1142         const struct net_device_ops *ops = dev->netdev_ops;
1143
1144         nla_for_each_nested(vf, attr, rem) {
1145                 switch (nla_type(vf)) {
1146                 case IFLA_VF_MAC: {
1147                         struct ifla_vf_mac *ivm;
1148                         ivm = nla_data(vf);
1149                         err = -EOPNOTSUPP;
1150                         if (ops->ndo_set_vf_mac)
1151                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1152                                                           ivm->mac);
1153                         break;
1154                 }
1155                 case IFLA_VF_VLAN: {
1156                         struct ifla_vf_vlan *ivv;
1157                         ivv = nla_data(vf);
1158                         err = -EOPNOTSUPP;
1159                         if (ops->ndo_set_vf_vlan)
1160                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1161                                                            ivv->vlan,
1162                                                            ivv->qos);
1163                         break;
1164                 }
1165                 case IFLA_VF_TX_RATE: {
1166                         struct ifla_vf_tx_rate *ivt;
1167                         ivt = nla_data(vf);
1168                         err = -EOPNOTSUPP;
1169                         if (ops->ndo_set_vf_tx_rate)
1170                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1171                                                               ivt->rate);
1172                         break;
1173                 }
1174                 default:
1175                         err = -EINVAL;
1176                         break;
1177                 }
1178                 if (err)
1179                         break;
1180         }
1181         return err;
1182 }
1183
1184 static int do_set_master(struct net_device *dev, int ifindex)
1185 {
1186         struct net_device *master_dev;
1187         const struct net_device_ops *ops;
1188         int err;
1189
1190         if (dev->master) {
1191                 if (dev->master->ifindex == ifindex)
1192                         return 0;
1193                 ops = dev->master->netdev_ops;
1194                 if (ops->ndo_del_slave) {
1195                         err = ops->ndo_del_slave(dev->master, dev);
1196                         if (err)
1197                                 return err;
1198                 } else {
1199                         return -EOPNOTSUPP;
1200                 }
1201         }
1202
1203         if (ifindex) {
1204                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1205                 if (!master_dev)
1206                         return -EINVAL;
1207                 ops = master_dev->netdev_ops;
1208                 if (ops->ndo_add_slave) {
1209                         err = ops->ndo_add_slave(master_dev, dev);
1210                         if (err)
1211                                 return err;
1212                 } else {
1213                         return -EOPNOTSUPP;
1214                 }
1215         }
1216         return 0;
1217 }
1218
1219 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1220                       struct nlattr **tb, char *ifname, int modified)
1221 {
1222         const struct net_device_ops *ops = dev->netdev_ops;
1223         int send_addr_notify = 0;
1224         int err;
1225
1226         if (tb[IFLA_NET_NS_PID]) {
1227                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1228                 if (IS_ERR(net)) {
1229                         err = PTR_ERR(net);
1230                         goto errout;
1231                 }
1232                 err = dev_change_net_namespace(dev, net, ifname);
1233                 put_net(net);
1234                 if (err)
1235                         goto errout;
1236                 modified = 1;
1237         }
1238
1239         if (tb[IFLA_MAP]) {
1240                 struct rtnl_link_ifmap *u_map;
1241                 struct ifmap k_map;
1242
1243                 if (!ops->ndo_set_config) {
1244                         err = -EOPNOTSUPP;
1245                         goto errout;
1246                 }
1247
1248                 if (!netif_device_present(dev)) {
1249                         err = -ENODEV;
1250                         goto errout;
1251                 }
1252
1253                 u_map = nla_data(tb[IFLA_MAP]);
1254                 k_map.mem_start = (unsigned long) u_map->mem_start;
1255                 k_map.mem_end = (unsigned long) u_map->mem_end;
1256                 k_map.base_addr = (unsigned short) u_map->base_addr;
1257                 k_map.irq = (unsigned char) u_map->irq;
1258                 k_map.dma = (unsigned char) u_map->dma;
1259                 k_map.port = (unsigned char) u_map->port;
1260
1261                 err = ops->ndo_set_config(dev, &k_map);
1262                 if (err < 0)
1263                         goto errout;
1264
1265                 modified = 1;
1266         }
1267
1268         if (tb[IFLA_ADDRESS]) {
1269                 struct sockaddr *sa;
1270                 int len;
1271
1272                 if (!ops->ndo_set_mac_address) {
1273                         err = -EOPNOTSUPP;
1274                         goto errout;
1275                 }
1276
1277                 if (!netif_device_present(dev)) {
1278                         err = -ENODEV;
1279                         goto errout;
1280                 }
1281
1282                 len = sizeof(sa_family_t) + dev->addr_len;
1283                 sa = kmalloc(len, GFP_KERNEL);
1284                 if (!sa) {
1285                         err = -ENOMEM;
1286                         goto errout;
1287                 }
1288                 sa->sa_family = dev->type;
1289                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1290                        dev->addr_len);
1291                 err = ops->ndo_set_mac_address(dev, sa);
1292                 kfree(sa);
1293                 if (err)
1294                         goto errout;
1295                 send_addr_notify = 1;
1296                 modified = 1;
1297         }
1298
1299         if (tb[IFLA_MTU]) {
1300                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1301                 if (err < 0)
1302                         goto errout;
1303                 modified = 1;
1304         }
1305
1306         if (tb[IFLA_GROUP]) {
1307                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1308                 modified = 1;
1309         }
1310
1311         /*
1312          * Interface selected by interface index but interface
1313          * name provided implies that a name change has been
1314          * requested.
1315          */
1316         if (ifm->ifi_index > 0 && ifname[0]) {
1317                 err = dev_change_name(dev, ifname);
1318                 if (err < 0)
1319                         goto errout;
1320                 modified = 1;
1321         }
1322
1323         if (tb[IFLA_IFALIAS]) {
1324                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1325                                     nla_len(tb[IFLA_IFALIAS]));
1326                 if (err < 0)
1327                         goto errout;
1328                 modified = 1;
1329         }
1330
1331         if (tb[IFLA_BROADCAST]) {
1332                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1333                 send_addr_notify = 1;
1334         }
1335
1336         if (ifm->ifi_flags || ifm->ifi_change) {
1337                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1338                 if (err < 0)
1339                         goto errout;
1340         }
1341
1342         if (tb[IFLA_MASTER]) {
1343                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1344                 if (err)
1345                         goto errout;
1346                 modified = 1;
1347         }
1348
1349         if (tb[IFLA_TXQLEN])
1350                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1351
1352         if (tb[IFLA_OPERSTATE])
1353                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1354
1355         if (tb[IFLA_LINKMODE]) {
1356                 write_lock_bh(&dev_base_lock);
1357                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1358                 write_unlock_bh(&dev_base_lock);
1359         }
1360
1361         if (tb[IFLA_VFINFO_LIST]) {
1362                 struct nlattr *attr;
1363                 int rem;
1364                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1365                         if (nla_type(attr) != IFLA_VF_INFO) {
1366                                 err = -EINVAL;
1367                                 goto errout;
1368                         }
1369                         err = do_setvfinfo(dev, attr);
1370                         if (err < 0)
1371                                 goto errout;
1372                         modified = 1;
1373                 }
1374         }
1375         err = 0;
1376
1377         if (tb[IFLA_VF_PORTS]) {
1378                 struct nlattr *port[IFLA_PORT_MAX+1];
1379                 struct nlattr *attr;
1380                 int vf;
1381                 int rem;
1382
1383                 err = -EOPNOTSUPP;
1384                 if (!ops->ndo_set_vf_port)
1385                         goto errout;
1386
1387                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1388                         if (nla_type(attr) != IFLA_VF_PORT)
1389                                 continue;
1390                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1391                                 attr, ifla_port_policy);
1392                         if (err < 0)
1393                                 goto errout;
1394                         if (!port[IFLA_PORT_VF]) {
1395                                 err = -EOPNOTSUPP;
1396                                 goto errout;
1397                         }
1398                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1399                         err = ops->ndo_set_vf_port(dev, vf, port);
1400                         if (err < 0)
1401                                 goto errout;
1402                         modified = 1;
1403                 }
1404         }
1405         err = 0;
1406
1407         if (tb[IFLA_PORT_SELF]) {
1408                 struct nlattr *port[IFLA_PORT_MAX+1];
1409
1410                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1411                         tb[IFLA_PORT_SELF], ifla_port_policy);
1412                 if (err < 0)
1413                         goto errout;
1414
1415                 err = -EOPNOTSUPP;
1416                 if (ops->ndo_set_vf_port)
1417                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1418                 if (err < 0)
1419                         goto errout;
1420                 modified = 1;
1421         }
1422
1423         if (tb[IFLA_AF_SPEC]) {
1424                 struct nlattr *af;
1425                 int rem;
1426
1427                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1428                         const struct rtnl_af_ops *af_ops;
1429
1430                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1431                                 BUG();
1432
1433                         err = af_ops->set_link_af(dev, af);
1434                         if (err < 0)
1435                                 goto errout;
1436
1437                         modified = 1;
1438                 }
1439         }
1440         err = 0;
1441
1442 errout:
1443         if (err < 0 && modified && net_ratelimit())
1444                 printk(KERN_WARNING "A link change request failed with "
1445                        "some changes committed already. Interface %s may "
1446                        "have been left with an inconsistent configuration, "
1447                        "please check.\n", dev->name);
1448
1449         if (send_addr_notify)
1450                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1451         return err;
1452 }
1453
1454 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1455 {
1456         struct net *net = sock_net(skb->sk);
1457         struct ifinfomsg *ifm;
1458         struct net_device *dev;
1459         int err;
1460         struct nlattr *tb[IFLA_MAX+1];
1461         char ifname[IFNAMSIZ];
1462
1463         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1464         if (err < 0)
1465                 goto errout;
1466
1467         if (tb[IFLA_IFNAME])
1468                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1469         else
1470                 ifname[0] = '\0';
1471
1472         err = -EINVAL;
1473         ifm = nlmsg_data(nlh);
1474         if (ifm->ifi_index > 0)
1475                 dev = __dev_get_by_index(net, ifm->ifi_index);
1476         else if (tb[IFLA_IFNAME])
1477                 dev = __dev_get_by_name(net, ifname);
1478         else
1479                 goto errout;
1480
1481         if (dev == NULL) {
1482                 err = -ENODEV;
1483                 goto errout;
1484         }
1485
1486         err = validate_linkmsg(dev, tb);
1487         if (err < 0)
1488                 goto errout;
1489
1490         err = do_setlink(dev, ifm, tb, ifname, 0);
1491 errout:
1492         return err;
1493 }
1494
1495 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1496 {
1497         struct net *net = sock_net(skb->sk);
1498         const struct rtnl_link_ops *ops;
1499         struct net_device *dev;
1500         struct ifinfomsg *ifm;
1501         char ifname[IFNAMSIZ];
1502         struct nlattr *tb[IFLA_MAX+1];
1503         int err;
1504         LIST_HEAD(list_kill);
1505
1506         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1507         if (err < 0)
1508                 return err;
1509
1510         if (tb[IFLA_IFNAME])
1511                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1512
1513         ifm = nlmsg_data(nlh);
1514         if (ifm->ifi_index > 0)
1515                 dev = __dev_get_by_index(net, ifm->ifi_index);
1516         else if (tb[IFLA_IFNAME])
1517                 dev = __dev_get_by_name(net, ifname);
1518         else
1519                 return -EINVAL;
1520
1521         if (!dev)
1522                 return -ENODEV;
1523
1524         ops = dev->rtnl_link_ops;
1525         if (!ops)
1526                 return -EOPNOTSUPP;
1527
1528         ops->dellink(dev, &list_kill);
1529         unregister_netdevice_many(&list_kill);
1530         list_del(&list_kill);
1531         return 0;
1532 }
1533
1534 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1535 {
1536         unsigned int old_flags;
1537         int err;
1538
1539         old_flags = dev->flags;
1540         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1541                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1542                 if (err < 0)
1543                         return err;
1544         }
1545
1546         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1547         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1548
1549         __dev_notify_flags(dev, old_flags);
1550         return 0;
1551 }
1552 EXPORT_SYMBOL(rtnl_configure_link);
1553
1554 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1555         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1556 {
1557         int err;
1558         struct net_device *dev;
1559         unsigned int num_queues = 1;
1560         unsigned int real_num_queues = 1;
1561
1562         if (ops->get_tx_queues) {
1563                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1564                                          &real_num_queues);
1565                 if (err)
1566                         goto err;
1567         }
1568         err = -ENOMEM;
1569         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1570         if (!dev)
1571                 goto err;
1572
1573         dev_net_set(dev, net);
1574         dev->rtnl_link_ops = ops;
1575         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1576         dev->real_num_tx_queues = real_num_queues;
1577
1578         if (tb[IFLA_MTU])
1579                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1580         if (tb[IFLA_ADDRESS])
1581                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1582                                 nla_len(tb[IFLA_ADDRESS]));
1583         if (tb[IFLA_BROADCAST])
1584                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1585                                 nla_len(tb[IFLA_BROADCAST]));
1586         if (tb[IFLA_TXQLEN])
1587                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1588         if (tb[IFLA_OPERSTATE])
1589                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1590         if (tb[IFLA_LINKMODE])
1591                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1592         if (tb[IFLA_GROUP])
1593                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1594
1595         return dev;
1596
1597 err:
1598         return ERR_PTR(err);
1599 }
1600 EXPORT_SYMBOL(rtnl_create_link);
1601
1602 static int rtnl_group_changelink(struct net *net, int group,
1603                 struct ifinfomsg *ifm,
1604                 struct nlattr **tb)
1605 {
1606         struct net_device *dev;
1607         int err;
1608
1609         for_each_netdev(net, dev) {
1610                 if (dev->group == group) {
1611                         err = do_setlink(dev, ifm, tb, NULL, 0);
1612                         if (err < 0)
1613                                 return err;
1614                 }
1615         }
1616
1617         return 0;
1618 }
1619
1620 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1621 {
1622         struct net *net = sock_net(skb->sk);
1623         const struct rtnl_link_ops *ops;
1624         struct net_device *dev;
1625         struct ifinfomsg *ifm;
1626         char kind[MODULE_NAME_LEN];
1627         char ifname[IFNAMSIZ];
1628         struct nlattr *tb[IFLA_MAX+1];
1629         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1630         int err;
1631
1632 #ifdef CONFIG_MODULES
1633 replay:
1634 #endif
1635         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1636         if (err < 0)
1637                 return err;
1638
1639         if (tb[IFLA_IFNAME])
1640                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1641         else
1642                 ifname[0] = '\0';
1643
1644         ifm = nlmsg_data(nlh);
1645         if (ifm->ifi_index > 0)
1646                 dev = __dev_get_by_index(net, ifm->ifi_index);
1647         else {
1648                 if (ifname[0])
1649                         dev = __dev_get_by_name(net, ifname);
1650                 else
1651                         dev = NULL;
1652         }
1653
1654         err = validate_linkmsg(dev, tb);
1655         if (err < 0)
1656                 return err;
1657
1658         if (tb[IFLA_LINKINFO]) {
1659                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1660                                        tb[IFLA_LINKINFO], ifla_info_policy);
1661                 if (err < 0)
1662                         return err;
1663         } else
1664                 memset(linkinfo, 0, sizeof(linkinfo));
1665
1666         if (linkinfo[IFLA_INFO_KIND]) {
1667                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1668                 ops = rtnl_link_ops_get(kind);
1669         } else {
1670                 kind[0] = '\0';
1671                 ops = NULL;
1672         }
1673
1674         if (1) {
1675                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1676                 struct net *dest_net;
1677
1678                 if (ops) {
1679                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1680                                 err = nla_parse_nested(attr, ops->maxtype,
1681                                                        linkinfo[IFLA_INFO_DATA],
1682                                                        ops->policy);
1683                                 if (err < 0)
1684                                         return err;
1685                                 data = attr;
1686                         }
1687                         if (ops->validate) {
1688                                 err = ops->validate(tb, data);
1689                                 if (err < 0)
1690                                         return err;
1691                         }
1692                 }
1693
1694                 if (dev) {
1695                         int modified = 0;
1696
1697                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1698                                 return -EEXIST;
1699                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1700                                 return -EOPNOTSUPP;
1701
1702                         if (linkinfo[IFLA_INFO_DATA]) {
1703                                 if (!ops || ops != dev->rtnl_link_ops ||
1704                                     !ops->changelink)
1705                                         return -EOPNOTSUPP;
1706
1707                                 err = ops->changelink(dev, tb, data);
1708                                 if (err < 0)
1709                                         return err;
1710                                 modified = 1;
1711                         }
1712
1713                         return do_setlink(dev, ifm, tb, ifname, modified);
1714                 }
1715
1716                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1717                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1718                                 return rtnl_group_changelink(net,
1719                                                 nla_get_u32(tb[IFLA_GROUP]),
1720                                                 ifm, tb);
1721                         return -ENODEV;
1722                 }
1723
1724                 if (ifm->ifi_index)
1725                         return -EOPNOTSUPP;
1726                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1727                         return -EOPNOTSUPP;
1728
1729                 if (!ops) {
1730 #ifdef CONFIG_MODULES
1731                         if (kind[0]) {
1732                                 __rtnl_unlock();
1733                                 request_module("rtnl-link-%s", kind);
1734                                 rtnl_lock();
1735                                 ops = rtnl_link_ops_get(kind);
1736                                 if (ops)
1737                                         goto replay;
1738                         }
1739 #endif
1740                         return -EOPNOTSUPP;
1741                 }
1742
1743                 if (!ifname[0])
1744                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1745
1746                 dest_net = rtnl_link_get_net(net, tb);
1747                 if (IS_ERR(dest_net))
1748                         return PTR_ERR(dest_net);
1749
1750                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1751
1752                 if (IS_ERR(dev))
1753                         err = PTR_ERR(dev);
1754                 else if (ops->newlink)
1755                         err = ops->newlink(net, dev, tb, data);
1756                 else
1757                         err = register_netdevice(dev);
1758
1759                 if (err < 0 && !IS_ERR(dev))
1760                         free_netdev(dev);
1761                 if (err < 0)
1762                         goto out;
1763
1764                 err = rtnl_configure_link(dev, ifm);
1765                 if (err < 0)
1766                         unregister_netdevice(dev);
1767 out:
1768                 put_net(dest_net);
1769                 return err;
1770         }
1771 }
1772
1773 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1774 {
1775         struct net *net = sock_net(skb->sk);
1776         struct ifinfomsg *ifm;
1777         char ifname[IFNAMSIZ];
1778         struct nlattr *tb[IFLA_MAX+1];
1779         struct net_device *dev = NULL;
1780         struct sk_buff *nskb;
1781         int err;
1782
1783         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1784         if (err < 0)
1785                 return err;
1786
1787         if (tb[IFLA_IFNAME])
1788                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1789
1790         ifm = nlmsg_data(nlh);
1791         if (ifm->ifi_index > 0)
1792                 dev = __dev_get_by_index(net, ifm->ifi_index);
1793         else if (tb[IFLA_IFNAME])
1794                 dev = __dev_get_by_name(net, ifname);
1795         else
1796                 return -EINVAL;
1797
1798         if (dev == NULL)
1799                 return -ENODEV;
1800
1801         nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1802         if (nskb == NULL)
1803                 return -ENOBUFS;
1804
1805         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1806                                nlh->nlmsg_seq, 0, 0);
1807         if (err < 0) {
1808                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1809                 WARN_ON(err == -EMSGSIZE);
1810                 kfree_skb(nskb);
1811         } else
1812                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1813
1814         return err;
1815 }
1816
1817 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1818 {
1819         int idx;
1820         int s_idx = cb->family;
1821
1822         if (s_idx == 0)
1823                 s_idx = 1;
1824         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1825                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1826                 if (idx < s_idx || idx == PF_PACKET)
1827                         continue;
1828                 if (rtnl_msg_handlers[idx] == NULL ||
1829                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1830                         continue;
1831                 if (idx > s_idx)
1832                         memset(&cb->args[0], 0, sizeof(cb->args));
1833                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1834                         break;
1835         }
1836         cb->family = idx;
1837
1838         return skb->len;
1839 }
1840
1841 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1842 {
1843         struct net *net = dev_net(dev);
1844         struct sk_buff *skb;
1845         int err = -ENOBUFS;
1846
1847         skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1848         if (skb == NULL)
1849                 goto errout;
1850
1851         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1852         if (err < 0) {
1853                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1854                 WARN_ON(err == -EMSGSIZE);
1855                 kfree_skb(skb);
1856                 goto errout;
1857         }
1858         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1859         return;
1860 errout:
1861         if (err < 0)
1862                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1863 }
1864
1865 /* Protected by RTNL sempahore.  */
1866 static struct rtattr **rta_buf;
1867 static int rtattr_max;
1868
1869 /* Process one rtnetlink message. */
1870
1871 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1872 {
1873         struct net *net = sock_net(skb->sk);
1874         rtnl_doit_func doit;
1875         int sz_idx, kind;
1876         int min_len;
1877         int family;
1878         int type;
1879
1880         type = nlh->nlmsg_type;
1881         if (type > RTM_MAX)
1882                 return -EOPNOTSUPP;
1883
1884         type -= RTM_BASE;
1885
1886         /* All the messages must have at least 1 byte length */
1887         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1888                 return 0;
1889
1890         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1891         sz_idx = type>>2;
1892         kind = type&3;
1893
1894         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1895                 return -EPERM;
1896
1897         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1898                 struct sock *rtnl;
1899                 rtnl_dumpit_func dumpit;
1900
1901                 dumpit = rtnl_get_dumpit(family, type);
1902                 if (dumpit == NULL)
1903                         return -EOPNOTSUPP;
1904
1905                 rtnl = net->rtnl;
1906                 return netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1907         }
1908
1909         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1910
1911         min_len = rtm_min[sz_idx];
1912         if (nlh->nlmsg_len < min_len)
1913                 return -EINVAL;
1914
1915         if (nlh->nlmsg_len > min_len) {
1916                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1917                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1918
1919                 while (RTA_OK(attr, attrlen)) {
1920                         unsigned flavor = attr->rta_type;
1921                         if (flavor) {
1922                                 if (flavor > rta_max[sz_idx])
1923                                         return -EINVAL;
1924                                 rta_buf[flavor-1] = attr;
1925                         }
1926                         attr = RTA_NEXT(attr, attrlen);
1927                 }
1928         }
1929
1930         doit = rtnl_get_doit(family, type);
1931         if (doit == NULL)
1932                 return -EOPNOTSUPP;
1933
1934         return doit(skb, nlh, (void *)&rta_buf[0]);
1935 }
1936
1937 static void rtnetlink_rcv(struct sk_buff *skb)
1938 {
1939         rtnl_lock();
1940         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1941         rtnl_unlock();
1942 }
1943
1944 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1945 {
1946         struct net_device *dev = ptr;
1947
1948         switch (event) {
1949         case NETDEV_UP:
1950         case NETDEV_DOWN:
1951         case NETDEV_PRE_UP:
1952         case NETDEV_POST_INIT:
1953         case NETDEV_REGISTER:
1954         case NETDEV_CHANGE:
1955         case NETDEV_PRE_TYPE_CHANGE:
1956         case NETDEV_GOING_DOWN:
1957         case NETDEV_UNREGISTER:
1958         case NETDEV_UNREGISTER_BATCH:
1959         case NETDEV_RELEASE:
1960         case NETDEV_JOIN:
1961                 break;
1962         default:
1963                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1964                 break;
1965         }
1966         return NOTIFY_DONE;
1967 }
1968
1969 static struct notifier_block rtnetlink_dev_notifier = {
1970         .notifier_call  = rtnetlink_event,
1971 };
1972
1973
1974 static int __net_init rtnetlink_net_init(struct net *net)
1975 {
1976         struct sock *sk;
1977         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1978                                    rtnetlink_rcv, NULL, THIS_MODULE);
1979         if (!sk)
1980                 return -ENOMEM;
1981         net->rtnl = sk;
1982         return 0;
1983 }
1984
1985 static void __net_exit rtnetlink_net_exit(struct net *net)
1986 {
1987         netlink_kernel_release(net->rtnl);
1988         net->rtnl = NULL;
1989 }
1990
1991 static struct pernet_operations rtnetlink_net_ops = {
1992         .init = rtnetlink_net_init,
1993         .exit = rtnetlink_net_exit,
1994 };
1995
1996 void __init rtnetlink_init(void)
1997 {
1998         int i;
1999
2000         rtattr_max = 0;
2001         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2002                 if (rta_max[i] > rtattr_max)
2003                         rtattr_max = rta_max[i];
2004         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2005         if (!rta_buf)
2006                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2007
2008         if (register_pernet_subsys(&rtnetlink_net_ops))
2009                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2010
2011         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2012         register_netdevice_notifier(&rtnetlink_dev_notifier);
2013
2014         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
2015         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
2016         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
2017         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
2018
2019         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
2020         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
2021 }
2022