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
  64 void rtnl_lock(void)
  65 {
  66         mutex_lock(&rtnl_mutex);
  67 }
  68 EXPORT_SYMBOL(rtnl_lock);
  69
  70 void __rtnl_unlock(void)
  71 {
  72         mutex_unlock(&rtnl_mutex);
  73 }
  74
  75 void rtnl_unlock(void)
  76 {
  77         /* This fellow will unlock it for us. */
  78         netdev_run_todo();
  79 }
  80 EXPORT_SYMBOL(rtnl_unlock);
  81
  82 int rtnl_trylock(void)
  83 {
  84         return mutex_trylock(&rtnl_mutex);
  85 }
  86 EXPORT_SYMBOL(rtnl_trylock);
  87
  88 int rtnl_is_locked(void)
  89 {
  90         return mutex_is_locked(&rtnl_mutex);
  91 }
  92 EXPORT_SYMBOL(rtnl_is_locked);
  93
  94 #ifdef CONFIG_PROVE_LOCKING
  95 int lockdep_rtnl_is_held(void)
  96 {
  97         return lockdep_is_held(&rtnl_mutex);
  98 }
  99 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 100 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 101
 102 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 103
 104 static inline int rtm_msgindex(int msgtype)
 105 {
 106         int msgindex = msgtype - RTM_BASE;
 107
 108         /*
 109          * msgindex < 0 implies someone tried to register a netlink
 110          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 111          * the message type has not been added to linux/rtnetlink.h
 112          */
 113         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 114
 115         return msgindex;
 116 }
 117
 118 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 119 {
 120         struct rtnl_link *tab;
 121
 122         if (protocol <= RTNL_FAMILY_MAX)
 123                 tab = rtnl_msg_handlers[protocol];
 124         else
 125                 tab = NULL;
 126
 127         if (tab == NULL || tab[msgindex].doit == NULL)
 128                 tab = rtnl_msg_handlers[PF_UNSPEC];
 129
 130         return tab ? tab[msgindex].doit : NULL;
 131 }
 132
 133 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 134 {
 135         struct rtnl_link *tab;
 136
 137         if (protocol <= RTNL_FAMILY_MAX)
 138                 tab = rtnl_msg_handlers[protocol];
 139         else
 140                 tab = NULL;
 141
 142         if (tab == NULL || tab[msgindex].dumpit == NULL)
 143                 tab = rtnl_msg_handlers[PF_UNSPEC];
 144
 145         return tab ? tab[msgindex].dumpit : NULL;
 146 }
 147
 148 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 149 {
 150         struct rtnl_link *tab;
 151
 152         if (protocol <= RTNL_FAMILY_MAX)
 153                 tab = rtnl_msg_handlers[protocol];
 154         else
 155                 tab = NULL;
 156
 157         if (tab == NULL || tab[msgindex].calcit == NULL)
 158                 tab = rtnl_msg_handlers[PF_UNSPEC];
 159
 160         return tab ? tab[msgindex].calcit : NULL;
 161 }
 162
 163 /**
 164  * __rtnl_register - Register a rtnetlink message type
 165  * @protocol: Protocol family or PF_UNSPEC
 166  * @msgtype: rtnetlink message type
 167  * @doit: Function pointer called for each request message
 168  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 169  * @calcit: Function pointer to calc size of dump message
 170  *
 171  * Registers the specified function pointers (at least one of them has
 172  * to be non-NULL) to be called whenever a request message for the
 173  * specified protocol family and message type is received.
 174  *
 175  * The special protocol family PF_UNSPEC may be used to define fallback
 176  * function pointers for the case when no entry for the specific protocol
 177  * family exists.
 178  *
 179  * Returns 0 on success or a negative error code.
 180  */
 181 int __rtnl_register(int protocol, int msgtype,
 182                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 183                     rtnl_calcit_func calcit)
 184 {
 185         struct rtnl_link *tab;
 186         int msgindex;
 187
 188         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 189         msgindex = rtm_msgindex(msgtype);
 190
 191         tab = rtnl_msg_handlers[protocol];
 192         if (tab == NULL) {
 193                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 194                 if (tab == NULL)
 195                         return -ENOBUFS;
 196
 197                 rtnl_msg_handlers[protocol] = tab;
 198         }
 199
 200         if (doit)
 201                 tab[msgindex].doit = doit;
 202
 203         if (dumpit)
 204                 tab[msgindex].dumpit = dumpit;
 205
 206         if (calcit)
 207                 tab[msgindex].calcit = calcit;
 208
 209         return 0;
 210 }
 211 EXPORT_SYMBOL_GPL(__rtnl_register);
 212
 213 /**
 214  * rtnl_register - Register a rtnetlink message type
 215  *
 216  * Identical to __rtnl_register() but panics on failure. This is useful
 217  * as failure of this function is very unlikely, it can only happen due
 218  * to lack of memory when allocating the chain to store all message
 219  * handlers for a protocol. Meant for use in init functions where lack
 220  * of memory implies no sense in continuing.
 221  */
 222 void rtnl_register(int protocol, int msgtype,
 223                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 224                    rtnl_calcit_func calcit)
 225 {
 226         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 227                 panic("Unable to register rtnetlink message handler, "
 228                       "protocol = %d, message type = %d\n",
 229                       protocol, msgtype);
 230 }
 231 EXPORT_SYMBOL_GPL(rtnl_register);
 232
 233 /**
 234  * rtnl_unregister - Unregister a rtnetlink message type
 235  * @protocol: Protocol family or PF_UNSPEC
 236  * @msgtype: rtnetlink message type
 237  *
 238  * Returns 0 on success or a negative error code.
 239  */
 240 int rtnl_unregister(int protocol, int msgtype)
 241 {
 242         int msgindex;
 243
 244         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 245         msgindex = rtm_msgindex(msgtype);
 246
 247         if (rtnl_msg_handlers[protocol] == NULL)
 248                 return -ENOENT;
 249
 250         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 251         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 252
 253         return 0;
 254 }
 255 EXPORT_SYMBOL_GPL(rtnl_unregister);
 256
 257 /**
 258  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 259  * @protocol : Protocol family or PF_UNSPEC
 260  *
 261  * Identical to calling rtnl_unregster() for all registered message types
 262  * of a certain protocol family.
 263  */
 264 void rtnl_unregister_all(int protocol)
 265 {
 266         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 267
 268         kfree(rtnl_msg_handlers[protocol]);
 269         rtnl_msg_handlers[protocol] = NULL;
 270 }
 271 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 272
 273 static LIST_HEAD(link_ops);
 274
 275 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 276 {
 277         const struct rtnl_link_ops *ops;
 278
 279         list_for_each_entry(ops, &link_ops, list) {
 280                 if (!strcmp(ops->kind, kind))
 281                         return ops;
 282         }
 283         return NULL;
 284 }
 285
 286 /**
 287  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 288  * @ops: struct rtnl_link_ops * to register
 289  *
 290  * The caller must hold the rtnl_mutex. This function should be used
 291  * by drivers that create devices during module initialization. It
 292  * must be called before registering the devices.
 293  *
 294  * Returns 0 on success or a negative error code.
 295  */
 296 int __rtnl_link_register(struct rtnl_link_ops *ops)
 297 {
 298         if (rtnl_link_ops_get(ops->kind))
 299                 return -EEXIST;
 300
 301         if (!ops->dellink)
 302                 ops->dellink = unregister_netdevice_queue;
 303
 304         list_add_tail(&ops->list, &link_ops);
 305         return 0;
 306 }
 307 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 308
 309 /**
 310  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 311  * @ops: struct rtnl_link_ops * to register
 312  *
 313  * Returns 0 on success or a negative error code.
 314  */
 315 int rtnl_link_register(struct rtnl_link_ops *ops)
 316 {
 317         int err;
 318
 319         rtnl_lock();
 320         err = __rtnl_link_register(ops);
 321         rtnl_unlock();
 322         return err;
 323 }
 324 EXPORT_SYMBOL_GPL(rtnl_link_register);
 325
 326 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 327 {
 328         struct net_device *dev;
 329         LIST_HEAD(list_kill);
 330
 331         for_each_netdev(net, dev) {
 332                 if (dev->rtnl_link_ops == ops)
 333                         ops->dellink(dev, &list_kill);
 334         }
 335         unregister_netdevice_many(&list_kill);
 336 }
 337
 338 /**
 339  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 340  * @ops: struct rtnl_link_ops * to unregister
 341  *
 342  * The caller must hold the rtnl_mutex.
 343  */
 344 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 345 {
 346         struct net *net;
 347
 348         for_each_net(net) {
 349                 __rtnl_kill_links(net, ops);
 350         }
 351         list_del(&ops->list);
 352 }
 353 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 354
 355 /**
 356  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 357  * @ops: struct rtnl_link_ops * to unregister
 358  */
 359 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 360 {
 361         rtnl_lock();
 362         __rtnl_link_unregister(ops);
 363         rtnl_unlock();
 364 }
 365 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 366
 367 static size_t rtnl_link_get_size(const struct net_device *dev)
 368 {
 369         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 370         size_t size;
 371
 372         if (!ops)
 373                 return 0;
 374
 375         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 376                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 377
 378         if (ops->get_size)
 379                 /* IFLA_INFO_DATA + nested data */
 380                 size += nla_total_size(sizeof(struct nlattr)) +
 381                         ops->get_size(dev);
 382
 383         if (ops->get_xstats_size)
 384                 /* IFLA_INFO_XSTATS */
 385                 size += nla_total_size(ops->get_xstats_size(dev));
 386
 387         return size;
 388 }
 389
 390 static LIST_HEAD(rtnl_af_ops);
 391
 392 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 393 {
 394         const struct rtnl_af_ops *ops;
 395
 396         list_for_each_entry(ops, &rtnl_af_ops, list) {
 397                 if (ops->family == family)
 398                         return ops;
 399         }
 400
 401         return NULL;
 402 }
 403
 404 /**
 405  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 406  * @ops: struct rtnl_af_ops * to register
 407  *
 408  * The caller must hold the rtnl_mutex.
 409  *
 410  * Returns 0 on success or a negative error code.
 411  */
 412 int __rtnl_af_register(struct rtnl_af_ops *ops)
 413 {
 414         list_add_tail(&ops->list, &rtnl_af_ops);
 415         return 0;
 416 }
 417 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 418
 419 /**
 420  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 421  * @ops: struct rtnl_af_ops * to register
 422  *
 423  * Returns 0 on success or a negative error code.
 424  */
 425 int rtnl_af_register(struct rtnl_af_ops *ops)
 426 {
 427         int err;
 428
 429         rtnl_lock();
 430         err = __rtnl_af_register(ops);
 431         rtnl_unlock();
 432         return err;
 433 }
 434 EXPORT_SYMBOL_GPL(rtnl_af_register);
 435
 436 /**
 437  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 438  * @ops: struct rtnl_af_ops * to unregister
 439  *
 440  * The caller must hold the rtnl_mutex.
 441  */
 442 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 443 {
 444         list_del(&ops->list);
 445 }
 446 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 447
 448 /**
 449  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 450  * @ops: struct rtnl_af_ops * to unregister
 451  */
 452 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 453 {
 454         rtnl_lock();
 455         __rtnl_af_unregister(ops);
 456         rtnl_unlock();
 457 }
 458 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 459
 460 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 461 {
 462         struct rtnl_af_ops *af_ops;
 463         size_t size;
 464
 465         /* IFLA_AF_SPEC */
 466         size = nla_total_size(sizeof(struct nlattr));
 467
 468         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 469                 if (af_ops->get_link_af_size) {
 470                         /* AF_* + nested data */
 471                         size += nla_total_size(sizeof(struct nlattr)) +
 472                                 af_ops->get_link_af_size(dev);
 473                 }
 474         }
 475
 476         return size;
 477 }
 478
 479 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 480 {
 481         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 482         struct nlattr *linkinfo, *data;
 483         int err = -EMSGSIZE;
 484
 485         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 486         if (linkinfo == NULL)
 487                 goto out;
 488
 489         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 490                 goto err_cancel_link;
 491         if (ops->fill_xstats) {
 492                 err = ops->fill_xstats(skb, dev);
 493                 if (err < 0)
 494                         goto err_cancel_link;
 495         }
 496         if (ops->fill_info) {
 497                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 498                 if (data == NULL)
 499                         goto err_cancel_link;
 500                 err = ops->fill_info(skb, dev);
 501                 if (err < 0)
 502                         goto err_cancel_data;
 503                 nla_nest_end(skb, data);
 504         }
 505
 506         nla_nest_end(skb, linkinfo);
 507         return 0;
 508
 509 err_cancel_data:
 510         nla_nest_cancel(skb, data);
 511 err_cancel_link:
 512         nla_nest_cancel(skb, linkinfo);
 513 out:
 514         return err;
 515 }
 516
 517 static const int rtm_min[RTM_NR_FAMILIES] =
 518 {
 519         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 520         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 521         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 522         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 523         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 524         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 527         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 528         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529 };
 530
 531 static const int rta_max[RTM_NR_FAMILIES] =
 532 {
 533         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 534         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 535         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 536         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 537         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 538         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 539         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 540         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 541 };
 542
 543 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 544 {
 545         struct rtattr *rta;
 546         int size = RTA_LENGTH(attrlen);
 547
 548         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 549         rta->rta_type = attrtype;
 550         rta->rta_len = size;
 551         memcpy(RTA_DATA(rta), data, attrlen);
 552         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 553 }
 554 EXPORT_SYMBOL(__rta_fill);
 555
 556 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 557 {
 558         struct sock *rtnl = net->rtnl;
 559         int err = 0;
 560
 561         NETLINK_CB(skb).dst_group = group;
 562         if (echo)
 563                 atomic_inc(&skb->users);
 564         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 565         if (echo)
 566                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 567         return err;
 568 }
 569
 570 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 571 {
 572         struct sock *rtnl = net->rtnl;
 573
 574         return nlmsg_unicast(rtnl, skb, pid);
 575 }
 576 EXPORT_SYMBOL(rtnl_unicast);
 577
 578 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 579                  struct nlmsghdr *nlh, gfp_t flags)
 580 {
 581         struct sock *rtnl = net->rtnl;
 582         int report = 0;
 583
 584         if (nlh)
 585                 report = nlmsg_report(nlh);
 586
 587         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 588 }
 589 EXPORT_SYMBOL(rtnl_notify);
 590
 591 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 592 {
 593         struct sock *rtnl = net->rtnl;
 594
 595         netlink_set_err(rtnl, 0, group, error);
 596 }
 597 EXPORT_SYMBOL(rtnl_set_sk_err);
 598
 599 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 600 {
 601         struct nlattr *mx;
 602         int i, valid = 0;
 603
 604         mx = nla_nest_start(skb, RTA_METRICS);
 605         if (mx == NULL)
 606                 return -ENOBUFS;
 607
 608         for (i = 0; i < RTAX_MAX; i++) {
 609                 if (metrics[i]) {
 610                         valid++;
 611                         if (nla_put_u32(skb, i+1, metrics[i]))
 612                                 goto nla_put_failure;
 613                 }
 614         }
 615
 616         if (!valid) {
 617                 nla_nest_cancel(skb, mx);
 618                 return 0;
 619         }
 620
 621         return nla_nest_end(skb, mx);
 622
 623 nla_put_failure:
 624         nla_nest_cancel(skb, mx);
 625         return -EMSGSIZE;
 626 }
 627 EXPORT_SYMBOL(rtnetlink_put_metrics);
 628
 629 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 630                        u32 ts, u32 tsage, long expires, u32 error)
 631 {
 632         struct rta_cacheinfo ci = {
 633                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 634                 .rta_used = dst->__use,
 635                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 636                 .rta_error = error,
 637                 .rta_id =  id,
 638                 .rta_ts = ts,
 639                 .rta_tsage = tsage,
 640         };
 641
 642         if (expires)
 643                 ci.rta_expires = jiffies_to_clock_t(expires);
 644
 645         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 646 }
 647 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 648
 649 static void set_operstate(struct net_device *dev, unsigned char transition)
 650 {
 651         unsigned char operstate = dev->operstate;
 652
 653         switch (transition) {
 654         case IF_OPER_UP:
 655                 if ((operstate == IF_OPER_DORMANT ||
 656                      operstate == IF_OPER_UNKNOWN) &&
 657                     !netif_dormant(dev))
 658                         operstate = IF_OPER_UP;
 659                 break;
 660
 661         case IF_OPER_DORMANT:
 662                 if (operstate == IF_OPER_UP ||
 663                     operstate == IF_OPER_UNKNOWN)
 664                         operstate = IF_OPER_DORMANT;
 665                 break;
 666         }
 667
 668         if (dev->operstate != operstate) {
 669                 write_lock_bh(&dev_base_lock);
 670                 dev->operstate = operstate;
 671                 write_unlock_bh(&dev_base_lock);
 672                 netdev_state_change(dev);
 673         }
 674 }
 675
 676 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 677                                            const struct ifinfomsg *ifm)
 678 {
 679         unsigned int flags = ifm->ifi_flags;
 680
 681         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 682         if (ifm->ifi_change)
 683                 flags = (flags & ifm->ifi_change) |
 684                         (dev->flags & ~ifm->ifi_change);
 685
 686         return flags;
 687 }
 688
 689 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 690                                  const struct rtnl_link_stats64 *b)
 691 {
 692         a->rx_packets = b->rx_packets;
 693         a->tx_packets = b->tx_packets;
 694         a->rx_bytes = b->rx_bytes;
 695         a->tx_bytes = b->tx_bytes;
 696         a->rx_errors = b->rx_errors;
 697         a->tx_errors = b->tx_errors;
 698         a->rx_dropped = b->rx_dropped;
 699         a->tx_dropped = b->tx_dropped;
 700
 701         a->multicast = b->multicast;
 702         a->collisions = b->collisions;
 703
 704         a->rx_length_errors = b->rx_length_errors;
 705         a->rx_over_errors = b->rx_over_errors;
 706         a->rx_crc_errors = b->rx_crc_errors;
 707         a->rx_frame_errors = b->rx_frame_errors;
 708         a->rx_fifo_errors = b->rx_fifo_errors;
 709         a->rx_missed_errors = b->rx_missed_errors;
 710
 711         a->tx_aborted_errors = b->tx_aborted_errors;
 712         a->tx_carrier_errors = b->tx_carrier_errors;
 713         a->tx_fifo_errors = b->tx_fifo_errors;
 714         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 715         a->tx_window_errors = b->tx_window_errors;
 716
 717         a->rx_compressed = b->rx_compressed;
 718         a->tx_compressed = b->tx_compressed;
 719 }
 720
 721 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 722 {
 723         memcpy(v, b, sizeof(*b));
 724 }
 725
 726 /* All VF info */
 727 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 728                                    u32 ext_filter_mask)
 729 {
 730         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 731             (ext_filter_mask & RTEXT_FILTER_VF)) {
 732                 int num_vfs = dev_num_vf(dev->dev.parent);
 733                 size_t size = nla_total_size(sizeof(struct nlattr));
 734                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 735                 size += num_vfs *
 736                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 737                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 738                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 739                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 740                 return size;
 741         } else
 742                 return 0;
 743 }
 744
 745 static size_t rtnl_port_size(const struct net_device *dev)
 746 {
 747         size_t port_size = nla_total_size(4)            /* PORT_VF */
 748                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 749                 + nla_total_size(sizeof(struct ifla_port_vsi))
 750                                                         /* PORT_VSI_TYPE */
 751                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 752                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 753                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 754                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 755         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 756         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 757                 + port_size;
 758         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 759                 + port_size;
 760
 761         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 762                 return 0;
 763         if (dev_num_vf(dev->dev.parent))
 764                 return port_self_size + vf_ports_size +
 765                         vf_port_size * dev_num_vf(dev->dev.parent);
 766         else
 767                 return port_self_size;
 768 }
 769
 770 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 771                                      u32 ext_filter_mask)
 772 {
 773         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 774                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 775                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 776                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 777                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 778                + nla_total_size(sizeof(struct rtnl_link_stats))
 779                + nla_total_size(sizeof(struct rtnl_link_stats64))
 780                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 781                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 782                + nla_total_size(4) /* IFLA_TXQLEN */
 783                + nla_total_size(4) /* IFLA_WEIGHT */
 784                + nla_total_size(4) /* IFLA_MTU */
 785                + nla_total_size(4) /* IFLA_LINK */
 786                + nla_total_size(4) /* IFLA_MASTER */
 787                + nla_total_size(1) /* IFLA_OPERSTATE */
 788                + nla_total_size(1) /* IFLA_LINKMODE */
 789                + nla_total_size(ext_filter_mask
 790                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 791                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 792                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 793                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 794                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 795 }
 796
 797 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 798 {
 799         struct nlattr *vf_ports;
 800         struct nlattr *vf_port;
 801         int vf;
 802         int err;
 803
 804         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 805         if (!vf_ports)
 806                 return -EMSGSIZE;
 807
 808         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 809                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 810                 if (!vf_port)
 811                         goto nla_put_failure;
 812                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 813                         goto nla_put_failure;
 814                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 815                 if (err == -EMSGSIZE)
 816                         goto nla_put_failure;
 817                 if (err) {
 818                         nla_nest_cancel(skb, vf_port);
 819                         continue;
 820                 }
 821                 nla_nest_end(skb, vf_port);
 822         }
 823
 824         nla_nest_end(skb, vf_ports);
 825
 826         return 0;
 827
 828 nla_put_failure:
 829         nla_nest_cancel(skb, vf_ports);
 830         return -EMSGSIZE;
 831 }
 832
 833 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 834 {
 835         struct nlattr *port_self;
 836         int err;
 837
 838         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 839         if (!port_self)
 840                 return -EMSGSIZE;
 841
 842         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 843         if (err) {
 844                 nla_nest_cancel(skb, port_self);
 845                 return (err == -EMSGSIZE) ? err : 0;
 846         }
 847
 848         nla_nest_end(skb, port_self);
 849
 850         return 0;
 851 }
 852
 853 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 854 {
 855         int err;
 856
 857         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 858                 return 0;
 859
 860         err = rtnl_port_self_fill(skb, dev);
 861         if (err)
 862                 return err;
 863
 864         if (dev_num_vf(dev->dev.parent)) {
 865                 err = rtnl_vf_ports_fill(skb, dev);
 866                 if (err)
 867                         return err;
 868         }
 869
 870         return 0;
 871 }
 872
 873 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 874                             int type, u32 pid, u32 seq, u32 change,
 875                             unsigned int flags, u32 ext_filter_mask)
 876 {
 877         struct ifinfomsg *ifm;
 878         struct nlmsghdr *nlh;
 879         struct rtnl_link_stats64 temp;
 880         const struct rtnl_link_stats64 *stats;
 881         struct nlattr *attr, *af_spec;
 882         struct rtnl_af_ops *af_ops;
 883
 884         ASSERT_RTNL();
 885         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 886         if (nlh == NULL)
 887                 return -EMSGSIZE;
 888
 889         ifm = nlmsg_data(nlh);
 890         ifm->ifi_family = AF_UNSPEC;
 891         ifm->__ifi_pad = 0;
 892         ifm->ifi_type = dev->type;
 893         ifm->ifi_index = dev->ifindex;
 894         ifm->ifi_flags = dev_get_flags(dev);
 895         ifm->ifi_change = change;
 896
 897         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 898             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 899             nla_put_u8(skb, IFLA_OPERSTATE,
 900                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 901             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 902             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 903             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 904             (dev->ifindex != dev->iflink &&
 905              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 906             (dev->master &&
 907              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 908             (dev->qdisc &&
 909              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 910             (dev->ifalias &&
 911              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 912                 goto nla_put_failure;
 913
 914         if (1) {
 915                 struct rtnl_link_ifmap map = {
 916                         .mem_start   = dev->mem_start,
 917                         .mem_end     = dev->mem_end,
 918                         .base_addr   = dev->base_addr,
 919                         .irq         = dev->irq,
 920                         .dma         = dev->dma,
 921                         .port        = dev->if_port,
 922                 };
 923                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 924                         goto nla_put_failure;
 925         }
 926
 927         if (dev->addr_len) {
 928                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 929                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 930                         goto nla_put_failure;
 931         }
 932
 933         attr = nla_reserve(skb, IFLA_STATS,
 934                         sizeof(struct rtnl_link_stats));
 935         if (attr == NULL)
 936                 goto nla_put_failure;
 937
 938         stats = dev_get_stats(dev, &temp);
 939         copy_rtnl_link_stats(nla_data(attr), stats);
 940
 941         attr = nla_reserve(skb, IFLA_STATS64,
 942                         sizeof(struct rtnl_link_stats64));
 943         if (attr == NULL)
 944                 goto nla_put_failure;
 945         copy_rtnl_link_stats64(nla_data(attr), stats);
 946
 947         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 948             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 949                 goto nla_put_failure;
 950
 951         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 952             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 953                 int i;
 954
 955                 struct nlattr *vfinfo, *vf;
 956                 int num_vfs = dev_num_vf(dev->dev.parent);
 957
 958                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 959                 if (!vfinfo)
 960                         goto nla_put_failure;
 961                 for (i = 0; i < num_vfs; i++) {
 962                         struct ifla_vf_info ivi;
 963                         struct ifla_vf_mac vf_mac;
 964                         struct ifla_vf_vlan vf_vlan;
 965                         struct ifla_vf_tx_rate vf_tx_rate;
 966                         struct ifla_vf_spoofchk vf_spoofchk;
 967
 968                         /*
 969                          * Not all SR-IOV capable drivers support the
 970                          * spoofcheck query.  Preset to -1 so the user
 971                          * space tool can detect that the driver didn't
 972                          * report anything.
 973                          */
 974                         ivi.spoofchk = -1;
 975                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 976                                 break;
 977                         vf_mac.vf =
 978                                 vf_vlan.vf =
 979                                 vf_tx_rate.vf =
 980                                 vf_spoofchk.vf = ivi.vf;
 981
 982                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 983                         vf_vlan.vlan = ivi.vlan;
 984                         vf_vlan.qos = ivi.qos;
 985                         vf_tx_rate.rate = ivi.tx_rate;
 986                         vf_spoofchk.setting = ivi.spoofchk;
 987                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 988                         if (!vf) {
 989                                 nla_nest_cancel(skb, vfinfo);
 990                                 goto nla_put_failure;
 991                         }
 992                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 993                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 994                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 995                                     &vf_tx_rate) ||
 996                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 997                                     &vf_spoofchk))
 998                                 goto nla_put_failure;
 999                         nla_nest_end(skb, vf);
1000                 }
1001                 nla_nest_end(skb, vfinfo);
1002         }
1003
1004         if (rtnl_port_fill(skb, dev))
1005                 goto nla_put_failure;
1006
1007         if (dev->rtnl_link_ops) {
1008                 if (rtnl_link_fill(skb, dev) < 0)
1009                         goto nla_put_failure;
1010         }
1011
1012         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1013                 goto nla_put_failure;
1014
1015         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1016                 if (af_ops->fill_link_af) {
1017                         struct nlattr *af;
1018                         int err;
1019
1020                         if (!(af = nla_nest_start(skb, af_ops->family)))
1021                                 goto nla_put_failure;
1022
1023                         err = af_ops->fill_link_af(skb, dev);
1024
1025                         /*
1026                          * Caller may return ENODATA to indicate that there
1027                          * was no data to be dumped. This is not an error, it
1028                          * means we should trim the attribute header and
1029                          * continue.
1030                          */
1031                         if (err == -ENODATA)
1032                                 nla_nest_cancel(skb, af);
1033                         else if (err < 0)
1034                                 goto nla_put_failure;
1035
1036                         nla_nest_end(skb, af);
1037                 }
1038         }
1039
1040         nla_nest_end(skb, af_spec);
1041
1042         return nlmsg_end(skb, nlh);
1043
1044 nla_put_failure:
1045         nlmsg_cancel(skb, nlh);
1046         return -EMSGSIZE;
1047 }
1048
1049 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1050 {
1051         struct net *net = sock_net(skb->sk);
1052         int h, s_h;
1053         int idx = 0, s_idx;
1054         struct net_device *dev;
1055         struct hlist_head *head;
1056         struct hlist_node *node;
1057         struct nlattr *tb[IFLA_MAX+1];
1058         u32 ext_filter_mask = 0;
1059
1060         s_h = cb->args[0];
1061         s_idx = cb->args[1];
1062
1063         rcu_read_lock();
1064         cb->seq = net->dev_base_seq;
1065
1066         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1067                         ifla_policy) >= 0) {
1068
1069                 if (tb[IFLA_EXT_MASK])
1070                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1071         }
1072
1073         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1074                 idx = 0;
1075                 head = &net->dev_index_head[h];
1076                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1077                         if (idx < s_idx)
1078                                 goto cont;
1079                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1080                                              NETLINK_CB(cb->skb).pid,
1081                                              cb->nlh->nlmsg_seq, 0,
1082                                              NLM_F_MULTI,
1083                                              ext_filter_mask) <= 0)
1084                                 goto out;
1085
1086                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1087 cont:
1088                         idx++;
1089                 }
1090         }
1091 out:
1092         rcu_read_unlock();
1093         cb->args[1] = idx;
1094         cb->args[0] = h;
1095
1096         return skb->len;
1097 }
1098
1099 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1100         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1101         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1102         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1103         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1104         [IFLA_MTU]              = { .type = NLA_U32 },
1105         [IFLA_LINK]             = { .type = NLA_U32 },
1106         [IFLA_MASTER]           = { .type = NLA_U32 },
1107         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1108         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1109         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1110         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1111         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1112         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1113         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1114         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1115         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1116         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1117         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1118         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1119         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1120 };
1121 EXPORT_SYMBOL(ifla_policy);
1122
1123 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1124         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1125         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1126 };
1127
1128 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1129         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1130 };
1131
1132 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1133         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1134                                     .len = sizeof(struct ifla_vf_mac) },
1135         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1136                                     .len = sizeof(struct ifla_vf_vlan) },
1137         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1138                                     .len = sizeof(struct ifla_vf_tx_rate) },
1139         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1140                                     .len = sizeof(struct ifla_vf_spoofchk) },
1141 };
1142
1143 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1144         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1145         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1146                                     .len = PORT_PROFILE_MAX },
1147         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1148                                     .len = sizeof(struct ifla_port_vsi)},
1149         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1150                                       .len = PORT_UUID_MAX },
1151         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1152                                     .len = PORT_UUID_MAX },
1153         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1154         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1155 };
1156
1157 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1158 {
1159         struct net *net;
1160         /* Examine the link attributes and figure out which
1161          * network namespace we are talking about.
1162          */
1163         if (tb[IFLA_NET_NS_PID])
1164                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1165         else if (tb[IFLA_NET_NS_FD])
1166                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1167         else
1168                 net = get_net(src_net);
1169         return net;
1170 }
1171 EXPORT_SYMBOL(rtnl_link_get_net);
1172
1173 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1174 {
1175         if (dev) {
1176                 if (tb[IFLA_ADDRESS] &&
1177                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1178                         return -EINVAL;
1179
1180                 if (tb[IFLA_BROADCAST] &&
1181                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1182                         return -EINVAL;
1183         }
1184
1185         if (tb[IFLA_AF_SPEC]) {
1186                 struct nlattr *af;
1187                 int rem, err;
1188
1189                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1190                         const struct rtnl_af_ops *af_ops;
1191
1192                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1193                                 return -EAFNOSUPPORT;
1194
1195                         if (!af_ops->set_link_af)
1196                                 return -EOPNOTSUPP;
1197
1198                         if (af_ops->validate_link_af) {
1199                                 err = af_ops->validate_link_af(dev, af);
1200                                 if (err < 0)
1201                                         return err;
1202                         }
1203                 }
1204         }
1205
1206         return 0;
1207 }
1208
1209 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1210 {
1211         int rem, err = -EINVAL;
1212         struct nlattr *vf;
1213         const struct net_device_ops *ops = dev->netdev_ops;
1214
1215         nla_for_each_nested(vf, attr, rem) {
1216                 switch (nla_type(vf)) {
1217                 case IFLA_VF_MAC: {
1218                         struct ifla_vf_mac *ivm;
1219                         ivm = nla_data(vf);
1220                         err = -EOPNOTSUPP;
1221                         if (ops->ndo_set_vf_mac)
1222                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1223                                                           ivm->mac);
1224                         break;
1225                 }
1226                 case IFLA_VF_VLAN: {
1227                         struct ifla_vf_vlan *ivv;
1228                         ivv = nla_data(vf);
1229                         err = -EOPNOTSUPP;
1230                         if (ops->ndo_set_vf_vlan)
1231                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1232                                                            ivv->vlan,
1233                                                            ivv->qos);
1234                         break;
1235                 }
1236                 case IFLA_VF_TX_RATE: {
1237                         struct ifla_vf_tx_rate *ivt;
1238                         ivt = nla_data(vf);
1239                         err = -EOPNOTSUPP;
1240                         if (ops->ndo_set_vf_tx_rate)
1241                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1242                                                               ivt->rate);
1243                         break;
1244                 }
1245                 case IFLA_VF_SPOOFCHK: {
1246                         struct ifla_vf_spoofchk *ivs;
1247                         ivs = nla_data(vf);
1248                         err = -EOPNOTSUPP;
1249                         if (ops->ndo_set_vf_spoofchk)
1250                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1251                                                                ivs->setting);
1252                         break;
1253                 }
1254                 default:
1255                         err = -EINVAL;
1256                         break;
1257                 }
1258                 if (err)
1259                         break;
1260         }
1261         return err;
1262 }
1263
1264 static int do_set_master(struct net_device *dev, int ifindex)
1265 {
1266         struct net_device *master_dev;
1267         const struct net_device_ops *ops;
1268         int err;
1269
1270         if (dev->master) {
1271                 if (dev->master->ifindex == ifindex)
1272                         return 0;
1273                 ops = dev->master->netdev_ops;
1274                 if (ops->ndo_del_slave) {
1275                         err = ops->ndo_del_slave(dev->master, dev);
1276                         if (err)
1277                                 return err;
1278                 } else {
1279                         return -EOPNOTSUPP;
1280                 }
1281         }
1282
1283         if (ifindex) {
1284                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1285                 if (!master_dev)
1286                         return -EINVAL;
1287                 ops = master_dev->netdev_ops;
1288                 if (ops->ndo_add_slave) {
1289                         err = ops->ndo_add_slave(master_dev, dev);
1290                         if (err)
1291                                 return err;
1292                 } else {
1293                         return -EOPNOTSUPP;
1294                 }
1295         }
1296         return 0;
1297 }
1298
1299 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1300                       struct nlattr **tb, char *ifname, int modified)
1301 {
1302         const struct net_device_ops *ops = dev->netdev_ops;
1303         int send_addr_notify = 0;
1304         int err;
1305
1306         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1307                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1308                 if (IS_ERR(net)) {
1309                         err = PTR_ERR(net);
1310                         goto errout;
1311                 }
1312                 err = dev_change_net_namespace(dev, net, ifname);
1313                 put_net(net);
1314                 if (err)
1315                         goto errout;
1316                 modified = 1;
1317         }
1318
1319         if (tb[IFLA_MAP]) {
1320                 struct rtnl_link_ifmap *u_map;
1321                 struct ifmap k_map;
1322
1323                 if (!ops->ndo_set_config) {
1324                         err = -EOPNOTSUPP;
1325                         goto errout;
1326                 }
1327
1328                 if (!netif_device_present(dev)) {
1329                         err = -ENODEV;
1330                         goto errout;
1331                 }
1332
1333                 u_map = nla_data(tb[IFLA_MAP]);
1334                 k_map.mem_start = (unsigned long) u_map->mem_start;
1335                 k_map.mem_end = (unsigned long) u_map->mem_end;
1336                 k_map.base_addr = (unsigned short) u_map->base_addr;
1337                 k_map.irq = (unsigned char) u_map->irq;
1338                 k_map.dma = (unsigned char) u_map->dma;
1339                 k_map.port = (unsigned char) u_map->port;
1340
1341                 err = ops->ndo_set_config(dev, &k_map);
1342                 if (err < 0)
1343                         goto errout;
1344
1345                 modified = 1;
1346         }
1347
1348         if (tb[IFLA_ADDRESS]) {
1349                 struct sockaddr *sa;
1350                 int len;
1351
1352                 if (!ops->ndo_set_mac_address) {
1353                         err = -EOPNOTSUPP;
1354                         goto errout;
1355                 }
1356
1357                 if (!netif_device_present(dev)) {
1358                         err = -ENODEV;
1359                         goto errout;
1360                 }
1361
1362                 len = sizeof(sa_family_t) + dev->addr_len;
1363                 sa = kmalloc(len, GFP_KERNEL);
1364                 if (!sa) {
1365                         err = -ENOMEM;
1366                         goto errout;
1367                 }
1368                 sa->sa_family = dev->type;
1369                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1370                        dev->addr_len);
1371                 err = ops->ndo_set_mac_address(dev, sa);
1372                 kfree(sa);
1373                 if (err)
1374                         goto errout;
1375                 send_addr_notify = 1;
1376                 modified = 1;
1377         }
1378
1379         if (tb[IFLA_MTU]) {
1380                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1381                 if (err < 0)
1382                         goto errout;
1383                 modified = 1;
1384         }
1385
1386         if (tb[IFLA_GROUP]) {
1387                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1388                 modified = 1;
1389         }
1390
1391         /*
1392          * Interface selected by interface index but interface
1393          * name provided implies that a name change has been
1394          * requested.
1395          */
1396         if (ifm->ifi_index > 0 && ifname[0]) {
1397                 err = dev_change_name(dev, ifname);
1398                 if (err < 0)
1399                         goto errout;
1400                 modified = 1;
1401         }
1402
1403         if (tb[IFLA_IFALIAS]) {
1404                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1405                                     nla_len(tb[IFLA_IFALIAS]));
1406                 if (err < 0)
1407                         goto errout;
1408                 modified = 1;
1409         }
1410
1411         if (tb[IFLA_BROADCAST]) {
1412                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1413                 send_addr_notify = 1;
1414         }
1415
1416         if (ifm->ifi_flags || ifm->ifi_change) {
1417                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1418                 if (err < 0)
1419                         goto errout;
1420         }
1421
1422         if (tb[IFLA_MASTER]) {
1423                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1424                 if (err)
1425                         goto errout;
1426                 modified = 1;
1427         }
1428
1429         if (tb[IFLA_TXQLEN])
1430                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1431
1432         if (tb[IFLA_OPERSTATE])
1433                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1434
1435         if (tb[IFLA_LINKMODE]) {
1436                 write_lock_bh(&dev_base_lock);
1437                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1438                 write_unlock_bh(&dev_base_lock);
1439         }
1440
1441         if (tb[IFLA_VFINFO_LIST]) {
1442                 struct nlattr *attr;
1443                 int rem;
1444                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1445                         if (nla_type(attr) != IFLA_VF_INFO) {
1446                                 err = -EINVAL;
1447                                 goto errout;
1448                         }
1449                         err = do_setvfinfo(dev, attr);
1450                         if (err < 0)
1451                                 goto errout;
1452                         modified = 1;
1453                 }
1454         }
1455         err = 0;
1456
1457         if (tb[IFLA_VF_PORTS]) {
1458                 struct nlattr *port[IFLA_PORT_MAX+1];
1459                 struct nlattr *attr;
1460                 int vf;
1461                 int rem;
1462
1463                 err = -EOPNOTSUPP;
1464                 if (!ops->ndo_set_vf_port)
1465                         goto errout;
1466
1467                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1468                         if (nla_type(attr) != IFLA_VF_PORT)
1469                                 continue;
1470                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1471                                 attr, ifla_port_policy);
1472                         if (err < 0)
1473                                 goto errout;
1474                         if (!port[IFLA_PORT_VF]) {
1475                                 err = -EOPNOTSUPP;
1476                                 goto errout;
1477                         }
1478                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1479                         err = ops->ndo_set_vf_port(dev, vf, port);
1480                         if (err < 0)
1481                                 goto errout;
1482                         modified = 1;
1483                 }
1484         }
1485         err = 0;
1486
1487         if (tb[IFLA_PORT_SELF]) {
1488                 struct nlattr *port[IFLA_PORT_MAX+1];
1489
1490                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1491                         tb[IFLA_PORT_SELF], ifla_port_policy);
1492                 if (err < 0)
1493                         goto errout;
1494
1495                 err = -EOPNOTSUPP;
1496                 if (ops->ndo_set_vf_port)
1497                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1498                 if (err < 0)
1499                         goto errout;
1500                 modified = 1;
1501         }
1502
1503         if (tb[IFLA_AF_SPEC]) {
1504                 struct nlattr *af;
1505                 int rem;
1506
1507                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1508                         const struct rtnl_af_ops *af_ops;
1509
1510                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1511                                 BUG();
1512
1513                         err = af_ops->set_link_af(dev, af);
1514                         if (err < 0)
1515                                 goto errout;
1516
1517                         modified = 1;
1518                 }
1519         }
1520         err = 0;
1521
1522 errout:
1523         if (err < 0 && modified && net_ratelimit())
1524                 printk(KERN_WARNING "A link change request failed with "
1525                        "some changes committed already. Interface %s may "
1526                        "have been left with an inconsistent configuration, "
1527                        "please check.\n", dev->name);
1528
1529         if (send_addr_notify)
1530                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1531
1532         return err;
1533 }
1534
1535 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1536 {
1537         struct net *net = sock_net(skb->sk);
1538         struct ifinfomsg *ifm;
1539         struct net_device *dev;
1540         int err;
1541         struct nlattr *tb[IFLA_MAX+1];
1542         char ifname[IFNAMSIZ];
1543
1544         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1545         if (err < 0)
1546                 goto errout;
1547
1548         if (tb[IFLA_IFNAME])
1549                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1550         else
1551                 ifname[0] = '\0';
1552
1553         err = -EINVAL;
1554         ifm = nlmsg_data(nlh);
1555         if (ifm->ifi_index > 0)
1556                 dev = __dev_get_by_index(net, ifm->ifi_index);
1557         else if (tb[IFLA_IFNAME])
1558                 dev = __dev_get_by_name(net, ifname);
1559         else
1560                 goto errout;
1561
1562         if (dev == NULL) {
1563                 err = -ENODEV;
1564                 goto errout;
1565         }
1566
1567         err = validate_linkmsg(dev, tb);
1568         if (err < 0)
1569                 goto errout;
1570
1571         err = do_setlink(dev, ifm, tb, ifname, 0);
1572 errout:
1573         return err;
1574 }
1575
1576 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1577 {
1578         struct net *net = sock_net(skb->sk);
1579         const struct rtnl_link_ops *ops;
1580         struct net_device *dev;
1581         struct ifinfomsg *ifm;
1582         char ifname[IFNAMSIZ];
1583         struct nlattr *tb[IFLA_MAX+1];
1584         int err;
1585         LIST_HEAD(list_kill);
1586
1587         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1588         if (err < 0)
1589                 return err;
1590
1591         if (tb[IFLA_IFNAME])
1592                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1593
1594         ifm = nlmsg_data(nlh);
1595         if (ifm->ifi_index > 0)
1596                 dev = __dev_get_by_index(net, ifm->ifi_index);
1597         else if (tb[IFLA_IFNAME])
1598                 dev = __dev_get_by_name(net, ifname);
1599         else
1600                 return -EINVAL;
1601
1602         if (!dev)
1603                 return -ENODEV;
1604
1605         ops = dev->rtnl_link_ops;
1606         if (!ops)
1607                 return -EOPNOTSUPP;
1608
1609         ops->dellink(dev, &list_kill);
1610         unregister_netdevice_many(&list_kill);
1611         list_del(&list_kill);
1612         return 0;
1613 }
1614
1615 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1616 {
1617         unsigned int old_flags;
1618         int err;
1619
1620         old_flags = dev->flags;
1621         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1622                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1623                 if (err < 0)
1624                         return err;
1625         }
1626
1627         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1628         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1629
1630         __dev_notify_flags(dev, old_flags);
1631         return 0;
1632 }
1633 EXPORT_SYMBOL(rtnl_configure_link);
1634
1635 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1636         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1637 {
1638         int err;
1639         struct net_device *dev;
1640         unsigned int num_queues = 1;
1641         unsigned int real_num_queues = 1;
1642
1643         if (ops->get_tx_queues) {
1644                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1645                                          &real_num_queues);
1646                 if (err)
1647                         goto err;
1648         }
1649         err = -ENOMEM;
1650         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1651         if (!dev)
1652                 goto err;
1653
1654         dev_net_set(dev, net);
1655         dev->rtnl_link_ops = ops;
1656         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1657
1658         if (tb[IFLA_MTU])
1659                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1660         if (tb[IFLA_ADDRESS])
1661                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1662                                 nla_len(tb[IFLA_ADDRESS]));
1663         if (tb[IFLA_BROADCAST])
1664                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1665                                 nla_len(tb[IFLA_BROADCAST]));
1666         if (tb[IFLA_TXQLEN])
1667                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1668         if (tb[IFLA_OPERSTATE])
1669                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1670         if (tb[IFLA_LINKMODE])
1671                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1672         if (tb[IFLA_GROUP])
1673                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1674
1675         return dev;
1676
1677 err:
1678         return ERR_PTR(err);
1679 }
1680 EXPORT_SYMBOL(rtnl_create_link);
1681
1682 static int rtnl_group_changelink(struct net *net, int group,
1683                 struct ifinfomsg *ifm,
1684                 struct nlattr **tb)
1685 {
1686         struct net_device *dev;
1687         int err;
1688
1689         for_each_netdev(net, dev) {
1690                 if (dev->group == group) {
1691                         err = do_setlink(dev, ifm, tb, NULL, 0);
1692                         if (err < 0)
1693                                 return err;
1694                 }
1695         }
1696
1697         return 0;
1698 }
1699
1700 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1701 {
1702         struct net *net = sock_net(skb->sk);
1703         const struct rtnl_link_ops *ops;
1704         struct net_device *dev;
1705         struct ifinfomsg *ifm;
1706         char kind[MODULE_NAME_LEN];
1707         char ifname[IFNAMSIZ];
1708         struct nlattr *tb[IFLA_MAX+1];
1709         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1710         int err;
1711
1712 #ifdef CONFIG_MODULES
1713 replay:
1714 #endif
1715         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1716         if (err < 0)
1717                 return err;
1718
1719         if (tb[IFLA_IFNAME])
1720                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1721         else
1722                 ifname[0] = '\0';
1723
1724         ifm = nlmsg_data(nlh);
1725         if (ifm->ifi_index > 0)
1726                 dev = __dev_get_by_index(net, ifm->ifi_index);
1727         else {
1728                 if (ifname[0])
1729                         dev = __dev_get_by_name(net, ifname);
1730                 else
1731                         dev = NULL;
1732         }
1733
1734         err = validate_linkmsg(dev, tb);
1735         if (err < 0)
1736                 return err;
1737
1738         if (tb[IFLA_LINKINFO]) {
1739                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1740                                        tb[IFLA_LINKINFO], ifla_info_policy);
1741                 if (err < 0)
1742                         return err;
1743         } else
1744                 memset(linkinfo, 0, sizeof(linkinfo));
1745
1746         if (linkinfo[IFLA_INFO_KIND]) {
1747                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1748                 ops = rtnl_link_ops_get(kind);
1749         } else {
1750                 kind[0] = '\0';
1751                 ops = NULL;
1752         }
1753
1754         if (1) {
1755                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1756                 struct net *dest_net;
1757
1758                 if (ops) {
1759                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1760                                 err = nla_parse_nested(attr, ops->maxtype,
1761                                                        linkinfo[IFLA_INFO_DATA],
1762                                                        ops->policy);
1763                                 if (err < 0)
1764                                         return err;
1765                                 data = attr;
1766                         }
1767                         if (ops->validate) {
1768                                 err = ops->validate(tb, data);
1769                                 if (err < 0)
1770                                         return err;
1771                         }
1772                 }
1773
1774                 if (dev) {
1775                         int modified = 0;
1776
1777                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1778                                 return -EEXIST;
1779                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1780                                 return -EOPNOTSUPP;
1781
1782                         if (linkinfo[IFLA_INFO_DATA]) {
1783                                 if (!ops || ops != dev->rtnl_link_ops ||
1784                                     !ops->changelink)
1785                                         return -EOPNOTSUPP;
1786
1787                                 err = ops->changelink(dev, tb, data);
1788                                 if (err < 0)
1789                                         return err;
1790                                 modified = 1;
1791                         }
1792
1793                         return do_setlink(dev, ifm, tb, ifname, modified);
1794                 }
1795
1796                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1797                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1798                                 return rtnl_group_changelink(net,
1799                                                 nla_get_u32(tb[IFLA_GROUP]),
1800                                                 ifm, tb);
1801                         return -ENODEV;
1802                 }
1803
1804                 if (ifm->ifi_index)
1805                         return -EOPNOTSUPP;
1806                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1807                         return -EOPNOTSUPP;
1808
1809                 if (!ops) {
1810 #ifdef CONFIG_MODULES
1811                         if (kind[0]) {
1812                                 __rtnl_unlock();
1813                                 request_module("rtnl-link-%s", kind);
1814                                 rtnl_lock();
1815                                 ops = rtnl_link_ops_get(kind);
1816                                 if (ops)
1817                                         goto replay;
1818                         }
1819 #endif
1820                         return -EOPNOTSUPP;
1821                 }
1822
1823                 if (!ifname[0])
1824                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1825
1826                 dest_net = rtnl_link_get_net(net, tb);
1827                 if (IS_ERR(dest_net))
1828                         return PTR_ERR(dest_net);
1829
1830                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1831
1832                 if (IS_ERR(dev))
1833                         err = PTR_ERR(dev);
1834                 else if (ops->newlink)
1835                         err = ops->newlink(net, dev, tb, data);
1836                 else
1837                         err = register_netdevice(dev);
1838
1839                 if (err < 0 && !IS_ERR(dev))
1840                         free_netdev(dev);
1841                 if (err < 0)
1842                         goto out;
1843
1844                 err = rtnl_configure_link(dev, ifm);
1845                 if (err < 0)
1846                         unregister_netdevice(dev);
1847 out:
1848                 put_net(dest_net);
1849                 return err;
1850         }
1851 }
1852
1853 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1854 {
1855         struct net *net = sock_net(skb->sk);
1856         struct ifinfomsg *ifm;
1857         char ifname[IFNAMSIZ];
1858         struct nlattr *tb[IFLA_MAX+1];
1859         struct net_device *dev = NULL;
1860         struct sk_buff *nskb;
1861         int err;
1862         u32 ext_filter_mask = 0;
1863
1864         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1865         if (err < 0)
1866                 return err;
1867
1868         if (tb[IFLA_IFNAME])
1869                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1870
1871         if (tb[IFLA_EXT_MASK])
1872                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1873
1874         ifm = nlmsg_data(nlh);
1875         if (ifm->ifi_index > 0)
1876                 dev = __dev_get_by_index(net, ifm->ifi_index);
1877         else if (tb[IFLA_IFNAME])
1878                 dev = __dev_get_by_name(net, ifname);
1879         else
1880                 return -EINVAL;
1881
1882         if (dev == NULL)
1883                 return -ENODEV;
1884
1885         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1886         if (nskb == NULL)
1887                 return -ENOBUFS;
1888
1889         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1890                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1891         if (err < 0) {
1892                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1893                 WARN_ON(err == -EMSGSIZE);
1894                 kfree_skb(nskb);
1895         } else
1896                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1897
1898         return err;
1899 }
1900
1901 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1902 {
1903         struct net *net = sock_net(skb->sk);
1904         struct net_device *dev;
1905         struct nlattr *tb[IFLA_MAX+1];
1906         u32 ext_filter_mask = 0;
1907         u16 min_ifinfo_dump_size = 0;
1908
1909         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1910                         ifla_policy) >= 0) {
1911                 if (tb[IFLA_EXT_MASK])
1912                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1913         }
1914
1915         if (!ext_filter_mask)
1916                 return NLMSG_GOODSIZE;
1917         /*
1918          * traverse the list of net devices and compute the minimum
1919          * buffer size based upon the filter mask.
1920          */
1921         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1922                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1923                                              if_nlmsg_size(dev,
1924                                                            ext_filter_mask));
1925         }
1926
1927         return min_ifinfo_dump_size;
1928 }
1929
1930 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1931 {
1932         int idx;
1933         int s_idx = cb->family;
1934
1935         if (s_idx == 0)
1936                 s_idx = 1;
1937         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1938                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1939                 if (idx < s_idx || idx == PF_PACKET)
1940                         continue;
1941                 if (rtnl_msg_handlers[idx] == NULL ||
1942                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1943                         continue;
1944                 if (idx > s_idx)
1945                         memset(&cb->args[0], 0, sizeof(cb->args));
1946                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1947                         break;
1948         }
1949         cb->family = idx;
1950
1951         return skb->len;
1952 }
1953
1954 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1955 {
1956         struct net *net = dev_net(dev);
1957         struct sk_buff *skb;
1958         int err = -ENOBUFS;
1959         size_t if_info_size;
1960
1961         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1962         if (skb == NULL)
1963                 goto errout;
1964
1965         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1966         if (err < 0) {
1967                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1968                 WARN_ON(err == -EMSGSIZE);
1969                 kfree_skb(skb);
1970                 goto errout;
1971         }
1972         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1973         return;
1974 errout:
1975         if (err < 0)
1976                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1977 }
1978
1979 /* Protected by RTNL sempahore.  */
1980 static struct rtattr **rta_buf;
1981 static int rtattr_max;
1982
1983 /* Process one rtnetlink message. */
1984
1985 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1986 {
1987         struct net *net = sock_net(skb->sk);
1988         rtnl_doit_func doit;
1989         int sz_idx, kind;
1990         int min_len;
1991         int family;
1992         int type;
1993         int err;
1994
1995         type = nlh->nlmsg_type;
1996         if (type > RTM_MAX)
1997                 return -EOPNOTSUPP;
1998
1999         type -= RTM_BASE;
2000
2001         /* All the messages must have at least 1 byte length */
2002         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2003                 return 0;
2004
2005         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2006         sz_idx = type>>2;
2007         kind = type&3;
2008
2009         if (kind != 2 && !capable(CAP_NET_ADMIN))
2010                 return -EPERM;
2011
2012         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2013                 struct sock *rtnl;
2014                 rtnl_dumpit_func dumpit;
2015                 rtnl_calcit_func calcit;
2016                 u16 min_dump_alloc = 0;
2017
2018                 dumpit = rtnl_get_dumpit(family, type);
2019                 if (dumpit == NULL)
2020                         return -EOPNOTSUPP;
2021                 calcit = rtnl_get_calcit(family, type);
2022                 if (calcit)
2023                         min_dump_alloc = calcit(skb, nlh);
2024
2025                 __rtnl_unlock();
2026                 rtnl = net->rtnl;
2027                 {
2028                         struct netlink_dump_control c = {
2029                                 .dump           = dumpit,
2030                                 .min_dump_alloc = min_dump_alloc,
2031                         };
2032                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2033                 }
2034                 rtnl_lock();
2035                 return err;
2036         }
2037
2038         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2039
2040         min_len = rtm_min[sz_idx];
2041         if (nlh->nlmsg_len < min_len)
2042                 return -EINVAL;
2043
2044         if (nlh->nlmsg_len > min_len) {
2045                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2046                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2047
2048                 while (RTA_OK(attr, attrlen)) {
2049                         unsigned flavor = attr->rta_type;
2050                         if (flavor) {
2051                                 if (flavor > rta_max[sz_idx])
2052                                         return -EINVAL;
2053                                 rta_buf[flavor-1] = attr;
2054                         }
2055                         attr = RTA_NEXT(attr, attrlen);
2056                 }
2057         }
2058
2059         doit = rtnl_get_doit(family, type);
2060         if (doit == NULL)
2061                 return -EOPNOTSUPP;
2062
2063         return doit(skb, nlh, (void *)&rta_buf[0]);
2064 }
2065
2066 static void rtnetlink_rcv(struct sk_buff *skb)
2067 {
2068         rtnl_lock();
2069         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2070         rtnl_unlock();
2071 }
2072
2073 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2074 {
2075         struct net_device *dev = ptr;
2076
2077         switch (event) {
2078         case NETDEV_UP:
2079         case NETDEV_DOWN:
2080         case NETDEV_PRE_UP:
2081         case NETDEV_POST_INIT:
2082         case NETDEV_REGISTER:
2083         case NETDEV_CHANGE:
2084         case NETDEV_PRE_TYPE_CHANGE:
2085         case NETDEV_GOING_DOWN:
2086         case NETDEV_UNREGISTER:
2087         case NETDEV_UNREGISTER_BATCH:
2088         case NETDEV_RELEASE:
2089         case NETDEV_JOIN:
2090                 break;
2091         default:
2092                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2093                 break;
2094         }
2095         return NOTIFY_DONE;
2096 }
2097
2098 static struct notifier_block rtnetlink_dev_notifier = {
2099         .notifier_call  = rtnetlink_event,
2100 };
2101
2102
2103 static int __net_init rtnetlink_net_init(struct net *net)
2104 {
2105         struct sock *sk;
2106         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2107                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2108         if (!sk)
2109                 return -ENOMEM;
2110         net->rtnl = sk;
2111         return 0;
2112 }
2113
2114 static void __net_exit rtnetlink_net_exit(struct net *net)
2115 {
2116         netlink_kernel_release(net->rtnl);
2117         net->rtnl = NULL;
2118 }
2119
2120 static struct pernet_operations rtnetlink_net_ops = {
2121         .init = rtnetlink_net_init,
2122         .exit = rtnetlink_net_exit,
2123 };
2124
2125 void __init rtnetlink_init(void)
2126 {
2127         int i;
2128
2129         rtattr_max = 0;
2130         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2131                 if (rta_max[i] > rtattr_max)
2132                         rtattr_max = rta_max[i];
2133         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2134         if (!rta_buf)
2135                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2136
2137         if (register_pernet_subsys(&rtnetlink_net_ops))
2138                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2139
2140         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2141         register_netdevice_notifier(&rtnetlink_dev_notifier);
2142
2143         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2144                       rtnl_dump_ifinfo, rtnl_calcit);
2145         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2146         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2147         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2148
2149         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2150         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2151 }
2152