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 /**
 276  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 277  * @ops: struct rtnl_link_ops * to register
 278  *
 279  * The caller must hold the rtnl_mutex. This function should be used
 280  * by drivers that create devices during module initialization. It
 281  * must be called before registering the devices.
 282  *
 283  * Returns 0 on success or a negative error code.
 284  */
 285 int __rtnl_link_register(struct rtnl_link_ops *ops)
 286 {
 287         if (!ops->dellink)
 288                 ops->dellink = unregister_netdevice_queue;
 289
 290         list_add_tail(&ops->list, &link_ops);
 291         return 0;
 292 }
 293 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 294
 295 /**
 296  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 297  * @ops: struct rtnl_link_ops * to register
 298  *
 299  * Returns 0 on success or a negative error code.
 300  */
 301 int rtnl_link_register(struct rtnl_link_ops *ops)
 302 {
 303         int err;
 304
 305         rtnl_lock();
 306         err = __rtnl_link_register(ops);
 307         rtnl_unlock();
 308         return err;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_link_register);
 311
 312 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 313 {
 314         struct net_device *dev;
 315         LIST_HEAD(list_kill);
 316
 317         for_each_netdev(net, dev) {
 318                 if (dev->rtnl_link_ops == ops)
 319                         ops->dellink(dev, &list_kill);
 320         }
 321         unregister_netdevice_many(&list_kill);
 322 }
 323
 324 /**
 325  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 326  * @ops: struct rtnl_link_ops * to unregister
 327  *
 328  * The caller must hold the rtnl_mutex.
 329  */
 330 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 331 {
 332         struct net *net;
 333
 334         for_each_net(net) {
 335                 __rtnl_kill_links(net, ops);
 336         }
 337         list_del(&ops->list);
 338 }
 339 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 340
 341 /**
 342  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 343  * @ops: struct rtnl_link_ops * to unregister
 344  */
 345 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 346 {
 347         rtnl_lock();
 348         __rtnl_link_unregister(ops);
 349         rtnl_unlock();
 350 }
 351 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 352
 353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 354 {
 355         const struct rtnl_link_ops *ops;
 356
 357         list_for_each_entry(ops, &link_ops, list) {
 358                 if (!strcmp(ops->kind, kind))
 359                         return ops;
 360         }
 361         return NULL;
 362 }
 363
 364 static size_t rtnl_link_get_size(const struct net_device *dev)
 365 {
 366         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 367         size_t size;
 368
 369         if (!ops)
 370                 return 0;
 371
 372         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 373                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 374
 375         if (ops->get_size)
 376                 /* IFLA_INFO_DATA + nested data */
 377                 size += nla_total_size(sizeof(struct nlattr)) +
 378                         ops->get_size(dev);
 379
 380         if (ops->get_xstats_size)
 381                 /* IFLA_INFO_XSTATS */
 382                 size += nla_total_size(ops->get_xstats_size(dev));
 383
 384         return size;
 385 }
 386
 387 static LIST_HEAD(rtnl_af_ops);
 388
 389 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 390 {
 391         const struct rtnl_af_ops *ops;
 392
 393         list_for_each_entry(ops, &rtnl_af_ops, list) {
 394                 if (ops->family == family)
 395                         return ops;
 396         }
 397
 398         return NULL;
 399 }
 400
 401 /**
 402  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 403  * @ops: struct rtnl_af_ops * to register
 404  *
 405  * The caller must hold the rtnl_mutex.
 406  *
 407  * Returns 0 on success or a negative error code.
 408  */
 409 int __rtnl_af_register(struct rtnl_af_ops *ops)
 410 {
 411         list_add_tail(&ops->list, &rtnl_af_ops);
 412         return 0;
 413 }
 414 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 415
 416 /**
 417  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 418  * @ops: struct rtnl_af_ops * to register
 419  *
 420  * Returns 0 on success or a negative error code.
 421  */
 422 int rtnl_af_register(struct rtnl_af_ops *ops)
 423 {
 424         int err;
 425
 426         rtnl_lock();
 427         err = __rtnl_af_register(ops);
 428         rtnl_unlock();
 429         return err;
 430 }
 431 EXPORT_SYMBOL_GPL(rtnl_af_register);
 432
 433 /**
 434  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 435  * @ops: struct rtnl_af_ops * to unregister
 436  *
 437  * The caller must hold the rtnl_mutex.
 438  */
 439 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 440 {
 441         list_del(&ops->list);
 442 }
 443 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 444
 445 /**
 446  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 447  * @ops: struct rtnl_af_ops * to unregister
 448  */
 449 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 450 {
 451         rtnl_lock();
 452         __rtnl_af_unregister(ops);
 453         rtnl_unlock();
 454 }
 455 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 456
 457 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 458 {
 459         struct rtnl_af_ops *af_ops;
 460         size_t size;
 461
 462         /* IFLA_AF_SPEC */
 463         size = nla_total_size(sizeof(struct nlattr));
 464
 465         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 466                 if (af_ops->get_link_af_size) {
 467                         /* AF_* + nested data */
 468                         size += nla_total_size(sizeof(struct nlattr)) +
 469                                 af_ops->get_link_af_size(dev);
 470                 }
 471         }
 472
 473         return size;
 474 }
 475
 476 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 477 {
 478         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 479         struct nlattr *linkinfo, *data;
 480         int err = -EMSGSIZE;
 481
 482         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 483         if (linkinfo == NULL)
 484                 goto out;
 485
 486         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 487                 goto err_cancel_link;
 488         if (ops->fill_xstats) {
 489                 err = ops->fill_xstats(skb, dev);
 490                 if (err < 0)
 491                         goto err_cancel_link;
 492         }
 493         if (ops->fill_info) {
 494                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 495                 if (data == NULL)
 496                         goto err_cancel_link;
 497                 err = ops->fill_info(skb, dev);
 498                 if (err < 0)
 499                         goto err_cancel_data;
 500                 nla_nest_end(skb, data);
 501         }
 502
 503         nla_nest_end(skb, linkinfo);
 504         return 0;
 505
 506 err_cancel_data:
 507         nla_nest_cancel(skb, data);
 508 err_cancel_link:
 509         nla_nest_cancel(skb, linkinfo);
 510 out:
 511         return err;
 512 }
 513
 514 static const int rtm_min[RTM_NR_FAMILIES] =
 515 {
 516         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 517         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 518         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 519         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 520         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 521         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 522         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 523         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 524         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 525         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 526 };
 527
 528 static const int rta_max[RTM_NR_FAMILIES] =
 529 {
 530         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 531         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 532         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 533         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 534         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 535         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 536         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 537         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 538 };
 539
 540 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 541 {
 542         struct rtattr *rta;
 543         int size = RTA_LENGTH(attrlen);
 544
 545         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 546         rta->rta_type = attrtype;
 547         rta->rta_len = size;
 548         memcpy(RTA_DATA(rta), data, attrlen);
 549         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 550 }
 551 EXPORT_SYMBOL(__rta_fill);
 552
 553 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 554 {
 555         struct sock *rtnl = net->rtnl;
 556         int err = 0;
 557
 558         NETLINK_CB(skb).dst_group = group;
 559         if (echo)
 560                 atomic_inc(&skb->users);
 561         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 562         if (echo)
 563                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 564         return err;
 565 }
 566
 567 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 568 {
 569         struct sock *rtnl = net->rtnl;
 570
 571         return nlmsg_unicast(rtnl, skb, pid);
 572 }
 573 EXPORT_SYMBOL(rtnl_unicast);
 574
 575 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 576                  struct nlmsghdr *nlh, gfp_t flags)
 577 {
 578         struct sock *rtnl = net->rtnl;
 579         int report = 0;
 580
 581         if (nlh)
 582                 report = nlmsg_report(nlh);
 583
 584         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 585 }
 586 EXPORT_SYMBOL(rtnl_notify);
 587
 588 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 589 {
 590         struct sock *rtnl = net->rtnl;
 591
 592         netlink_set_err(rtnl, 0, group, error);
 593 }
 594 EXPORT_SYMBOL(rtnl_set_sk_err);
 595
 596 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 597 {
 598         struct nlattr *mx;
 599         int i, valid = 0;
 600
 601         mx = nla_nest_start(skb, RTA_METRICS);
 602         if (mx == NULL)
 603                 return -ENOBUFS;
 604
 605         for (i = 0; i < RTAX_MAX; i++) {
 606                 if (metrics[i]) {
 607                         valid++;
 608                         NLA_PUT_U32(skb, i+1, metrics[i]);
 609                 }
 610         }
 611
 612         if (!valid) {
 613                 nla_nest_cancel(skb, mx);
 614                 return 0;
 615         }
 616
 617         return nla_nest_end(skb, mx);
 618
 619 nla_put_failure:
 620         nla_nest_cancel(skb, mx);
 621         return -EMSGSIZE;
 622 }
 623 EXPORT_SYMBOL(rtnetlink_put_metrics);
 624
 625 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 626                        u32 ts, u32 tsage, long expires, u32 error)
 627 {
 628         struct rta_cacheinfo ci = {
 629                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 630                 .rta_used = dst->__use,
 631                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 632                 .rta_error = error,
 633                 .rta_id =  id,
 634                 .rta_ts = ts,
 635                 .rta_tsage = tsage,
 636         };
 637
 638         if (expires)
 639                 ci.rta_expires = jiffies_to_clock_t(expires);
 640
 641         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 642 }
 643 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 644
 645 static void set_operstate(struct net_device *dev, unsigned char transition)
 646 {
 647         unsigned char operstate = dev->operstate;
 648
 649         switch (transition) {
 650         case IF_OPER_UP:
 651                 if ((operstate == IF_OPER_DORMANT ||
 652                      operstate == IF_OPER_UNKNOWN) &&
 653                     !netif_dormant(dev))
 654                         operstate = IF_OPER_UP;
 655                 break;
 656
 657         case IF_OPER_DORMANT:
 658                 if (operstate == IF_OPER_UP ||
 659                     operstate == IF_OPER_UNKNOWN)
 660                         operstate = IF_OPER_DORMANT;
 661                 break;
 662         }
 663
 664         if (dev->operstate != operstate) {
 665                 write_lock_bh(&dev_base_lock);
 666                 dev->operstate = operstate;
 667                 write_unlock_bh(&dev_base_lock);
 668                 netdev_state_change(dev);
 669         }
 670 }
 671
 672 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 673 {
 674         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 675                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 676 }
 677
 678 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 679                                            const struct ifinfomsg *ifm)
 680 {
 681         unsigned int flags = ifm->ifi_flags;
 682
 683         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 684         if (ifm->ifi_change)
 685                 flags = (flags & ifm->ifi_change) |
 686                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 687
 688         return flags;
 689 }
 690
 691 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 692                                  const struct rtnl_link_stats64 *b)
 693 {
 694         a->rx_packets = b->rx_packets;
 695         a->tx_packets = b->tx_packets;
 696         a->rx_bytes = b->rx_bytes;
 697         a->tx_bytes = b->tx_bytes;
 698         a->rx_errors = b->rx_errors;
 699         a->tx_errors = b->tx_errors;
 700         a->rx_dropped = b->rx_dropped;
 701         a->tx_dropped = b->tx_dropped;
 702
 703         a->multicast = b->multicast;
 704         a->collisions = b->collisions;
 705
 706         a->rx_length_errors = b->rx_length_errors;
 707         a->rx_over_errors = b->rx_over_errors;
 708         a->rx_crc_errors = b->rx_crc_errors;
 709         a->rx_frame_errors = b->rx_frame_errors;
 710         a->rx_fifo_errors = b->rx_fifo_errors;
 711         a->rx_missed_errors = b->rx_missed_errors;
 712
 713         a->tx_aborted_errors = b->tx_aborted_errors;
 714         a->tx_carrier_errors = b->tx_carrier_errors;
 715         a->tx_fifo_errors = b->tx_fifo_errors;
 716         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 717         a->tx_window_errors = b->tx_window_errors;
 718
 719         a->rx_compressed = b->rx_compressed;
 720         a->tx_compressed = b->tx_compressed;
 721 }
 722
 723 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 724 {
 725         memcpy(v, b, sizeof(*b));
 726 }
 727
 728 /* All VF info */
 729 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 730                                    u32 ext_filter_mask)
 731 {
 732         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 733             (ext_filter_mask & RTEXT_FILTER_VF)) {
 734                 int num_vfs = dev_num_vf(dev->dev.parent);
 735                 size_t size = nla_total_size(sizeof(struct nlattr));
 736                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 737                 size += num_vfs *
 738                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 739                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 740                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 741                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 742                 return size;
 743         } else
 744                 return 0;
 745 }
 746
 747 static size_t rtnl_port_size(const struct net_device *dev)
 748 {
 749         size_t port_size = nla_total_size(4)            /* PORT_VF */
 750                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 751                 + nla_total_size(sizeof(struct ifla_port_vsi))
 752                                                         /* PORT_VSI_TYPE */
 753                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 754                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 755                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 756                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 757         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 758         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 759                 + port_size;
 760         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 761                 + port_size;
 762
 763         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 764                 return 0;
 765         if (dev_num_vf(dev->dev.parent))
 766                 return port_self_size + vf_ports_size +
 767                         vf_port_size * dev_num_vf(dev->dev.parent);
 768         else
 769                 return port_self_size;
 770 }
 771
 772 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 773                                      u32 ext_filter_mask)
 774 {
 775         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 776                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 777                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 778                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 779                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 780                + nla_total_size(sizeof(struct rtnl_link_stats))
 781                + nla_total_size(sizeof(struct rtnl_link_stats64))
 782                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 783                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 784                + nla_total_size(4) /* IFLA_TXQLEN */
 785                + nla_total_size(4) /* IFLA_WEIGHT */
 786                + nla_total_size(4) /* IFLA_MTU */
 787                + nla_total_size(4) /* IFLA_LINK */
 788                + nla_total_size(4) /* IFLA_MASTER */
 789                + nla_total_size(1) /* IFLA_OPERSTATE */
 790                + nla_total_size(1) /* IFLA_LINKMODE */
 791                + nla_total_size(ext_filter_mask
 792                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 793                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 794                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 795                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 796                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 797 }
 798
 799 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 800 {
 801         struct nlattr *vf_ports;
 802         struct nlattr *vf_port;
 803         int vf;
 804         int err;
 805
 806         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 807         if (!vf_ports)
 808                 return -EMSGSIZE;
 809
 810         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 811                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 812                 if (!vf_port)
 813                         goto nla_put_failure;
 814                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 815                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 816                 if (err == -EMSGSIZE)
 817                         goto nla_put_failure;
 818                 if (err) {
 819                         nla_nest_cancel(skb, vf_port);
 820                         continue;
 821                 }
 822                 nla_nest_end(skb, vf_port);
 823         }
 824
 825         nla_nest_end(skb, vf_ports);
 826
 827         return 0;
 828
 829 nla_put_failure:
 830         nla_nest_cancel(skb, vf_ports);
 831         return -EMSGSIZE;
 832 }
 833
 834 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 835 {
 836         struct nlattr *port_self;
 837         int err;
 838
 839         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 840         if (!port_self)
 841                 return -EMSGSIZE;
 842
 843         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 844         if (err) {
 845                 nla_nest_cancel(skb, port_self);
 846                 return (err == -EMSGSIZE) ? err : 0;
 847         }
 848
 849         nla_nest_end(skb, port_self);
 850
 851         return 0;
 852 }
 853
 854 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 855 {
 856         int err;
 857
 858         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 859                 return 0;
 860
 861         err = rtnl_port_self_fill(skb, dev);
 862         if (err)
 863                 return err;
 864
 865         if (dev_num_vf(dev->dev.parent)) {
 866                 err = rtnl_vf_ports_fill(skb, dev);
 867                 if (err)
 868                         return err;
 869         }
 870
 871         return 0;
 872 }
 873
 874 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 875                             int type, u32 pid, u32 seq, u32 change,
 876                             unsigned int flags, u32 ext_filter_mask)
 877 {
 878         struct ifinfomsg *ifm;
 879         struct nlmsghdr *nlh;
 880         struct rtnl_link_stats64 temp;
 881         const struct rtnl_link_stats64 *stats;
 882         struct nlattr *attr, *af_spec;
 883         struct rtnl_af_ops *af_ops;
 884
 885         ASSERT_RTNL();
 886         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 887         if (nlh == NULL)
 888                 return -EMSGSIZE;
 889
 890         ifm = nlmsg_data(nlh);
 891         ifm->ifi_family = AF_UNSPEC;
 892         ifm->__ifi_pad = 0;
 893         ifm->ifi_type = dev->type;
 894         ifm->ifi_index = dev->ifindex;
 895         ifm->ifi_flags = dev_get_flags(dev);
 896         ifm->ifi_change = change;
 897
 898         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 899         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 900         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 901                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 902         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 903         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 904         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 905
 906         if (dev->ifindex != dev->iflink)
 907                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 908
 909         if (dev->master)
 910                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 911
 912         if (dev->qdisc)
 913                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 914
 915         if (dev->ifalias)
 916                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 917
 918         if (1) {
 919                 struct rtnl_link_ifmap map = {
 920                         .mem_start   = dev->mem_start,
 921                         .mem_end     = dev->mem_end,
 922                         .base_addr   = dev->base_addr,
 923                         .irq         = dev->irq,
 924                         .dma         = dev->dma,
 925                         .port        = dev->if_port,
 926                 };
 927                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 928         }
 929
 930         if (dev->addr_len) {
 931                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 932                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 933         }
 934
 935         attr = nla_reserve(skb, IFLA_STATS,
 936                         sizeof(struct rtnl_link_stats));
 937         if (attr == NULL)
 938                 goto nla_put_failure;
 939
 940         stats = dev_get_stats(dev, &temp);
 941         copy_rtnl_link_stats(nla_data(attr), stats);
 942
 943         attr = nla_reserve(skb, IFLA_STATS64,
 944                         sizeof(struct rtnl_link_stats64));
 945         if (attr == NULL)
 946                 goto nla_put_failure;
 947         copy_rtnl_link_stats64(nla_data(attr), stats);
 948
 949         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF))
 950                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 951
 952         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 953             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 954                 int i;
 955
 956                 struct nlattr *vfinfo, *vf;
 957                 int num_vfs = dev_num_vf(dev->dev.parent);
 958
 959                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 960                 if (!vfinfo)
 961                         goto nla_put_failure;
 962                 for (i = 0; i < num_vfs; i++) {
 963                         struct ifla_vf_info ivi;
 964                         struct ifla_vf_mac vf_mac;
 965                         struct ifla_vf_vlan vf_vlan;
 966                         struct ifla_vf_tx_rate vf_tx_rate;
 967                         struct ifla_vf_spoofchk vf_spoofchk;
 968
 969                         /*
 970                          * Not all SR-IOV capable drivers support the
 971                          * spoofcheck query.  Preset to -1 so the user
 972                          * space tool can detect that the driver didn't
 973                          * report anything.
 974                          */
 975                         ivi.spoofchk = -1;
 976                         memset(ivi.mac, 0, sizeof(ivi.mac));
 977                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 978                                 break;
 979                         vf_mac.vf =
 980                                 vf_vlan.vf =
 981                                 vf_tx_rate.vf =
 982                                 vf_spoofchk.vf = ivi.vf;
 983
 984                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 985                         vf_vlan.vlan = ivi.vlan;
 986                         vf_vlan.qos = ivi.qos;
 987                         vf_tx_rate.rate = ivi.tx_rate;
 988                         vf_spoofchk.setting = ivi.spoofchk;
 989                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 990                         if (!vf) {
 991                                 nla_nest_cancel(skb, vfinfo);
 992                                 goto nla_put_failure;
 993                         }
 994                         NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
 995                         NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
 996                         NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 997                                 &vf_tx_rate);
 998                         NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 999                                 &vf_spoofchk);
1000                         nla_nest_end(skb, vf);
1001                 }
1002                 nla_nest_end(skb, vfinfo);
1003         }
1004
1005         if (rtnl_port_fill(skb, dev))
1006                 goto nla_put_failure;
1007
1008         if (dev->rtnl_link_ops) {
1009                 if (rtnl_link_fill(skb, dev) < 0)
1010                         goto nla_put_failure;
1011         }
1012
1013         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1014                 goto nla_put_failure;
1015
1016         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1017                 if (af_ops->fill_link_af) {
1018                         struct nlattr *af;
1019                         int err;
1020
1021                         if (!(af = nla_nest_start(skb, af_ops->family)))
1022                                 goto nla_put_failure;
1023
1024                         err = af_ops->fill_link_af(skb, dev);
1025
1026                         /*
1027                          * Caller may return ENODATA to indicate that there
1028                          * was no data to be dumped. This is not an error, it
1029                          * means we should trim the attribute header and
1030                          * continue.
1031                          */
1032                         if (err == -ENODATA)
1033                                 nla_nest_cancel(skb, af);
1034                         else if (err < 0)
1035                                 goto nla_put_failure;
1036
1037                         nla_nest_end(skb, af);
1038                 }
1039         }
1040
1041         nla_nest_end(skb, af_spec);
1042
1043         return nlmsg_end(skb, nlh);
1044
1045 nla_put_failure:
1046         nlmsg_cancel(skb, nlh);
1047         return -EMSGSIZE;
1048 }
1049
1050 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1051 {
1052         struct net *net = sock_net(skb->sk);
1053         int h, s_h;
1054         int idx = 0, s_idx;
1055         struct net_device *dev;
1056         struct hlist_head *head;
1057         struct hlist_node *node;
1058         struct nlattr *tb[IFLA_MAX+1];
1059         u32 ext_filter_mask = 0;
1060         int err;
1061
1062         s_h = cb->args[0];
1063         s_idx = cb->args[1];
1064
1065         rcu_read_lock();
1066         cb->seq = net->dev_base_seq;
1067
1068         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1069                         ifla_policy) >= 0) {
1070
1071                 if (tb[IFLA_EXT_MASK])
1072                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1073         }
1074
1075         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1076                 idx = 0;
1077                 head = &net->dev_index_head[h];
1078                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1079                         if (idx < s_idx)
1080                                 goto cont;
1081                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1082                                                NETLINK_CB(cb->skb).pid,
1083                                                cb->nlh->nlmsg_seq, 0,
1084                                                NLM_F_MULTI,
1085                                                ext_filter_mask);
1086                         /* If we ran out of room on the first message,
1087                          * we're in trouble
1088                          */
1089                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1090
1091                         if (err <= 0)
1092                                 goto out;
1093
1094                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1095 cont:
1096                         idx++;
1097                 }
1098         }
1099 out:
1100         rcu_read_unlock();
1101         cb->args[1] = idx;
1102         cb->args[0] = h;
1103
1104         return skb->len;
1105 }
1106
1107 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1108         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1109         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1110         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1111         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1112         [IFLA_MTU]              = { .type = NLA_U32 },
1113         [IFLA_LINK]             = { .type = NLA_U32 },
1114         [IFLA_MASTER]           = { .type = NLA_U32 },
1115         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1116         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1117         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1118         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1119         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1120         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1121         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1122         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1123         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1124         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1125         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1126         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1127         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1128 };
1129 EXPORT_SYMBOL(ifla_policy);
1130
1131 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1132         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1133         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1134 };
1135
1136 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1137         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1138 };
1139
1140 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1141         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1142                                     .len = sizeof(struct ifla_vf_mac) },
1143         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1144                                     .len = sizeof(struct ifla_vf_vlan) },
1145         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1146                                     .len = sizeof(struct ifla_vf_tx_rate) },
1147         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1148                                     .len = sizeof(struct ifla_vf_spoofchk) },
1149 };
1150
1151 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1152         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1153         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1154                                     .len = PORT_PROFILE_MAX },
1155         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1156                                     .len = sizeof(struct ifla_port_vsi)},
1157         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1158                                       .len = PORT_UUID_MAX },
1159         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1160                                     .len = PORT_UUID_MAX },
1161         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1162         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1163 };
1164
1165 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1166 {
1167         struct net *net;
1168         /* Examine the link attributes and figure out which
1169          * network namespace we are talking about.
1170          */
1171         if (tb[IFLA_NET_NS_PID])
1172                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1173         else if (tb[IFLA_NET_NS_FD])
1174                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1175         else
1176                 net = get_net(src_net);
1177         return net;
1178 }
1179 EXPORT_SYMBOL(rtnl_link_get_net);
1180
1181 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1182 {
1183         if (dev) {
1184                 if (tb[IFLA_ADDRESS] &&
1185                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1186                         return -EINVAL;
1187
1188                 if (tb[IFLA_BROADCAST] &&
1189                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1190                         return -EINVAL;
1191         }
1192
1193         if (tb[IFLA_AF_SPEC]) {
1194                 struct nlattr *af;
1195                 int rem, err;
1196
1197                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1198                         const struct rtnl_af_ops *af_ops;
1199
1200                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1201                                 return -EAFNOSUPPORT;
1202
1203                         if (!af_ops->set_link_af)
1204                                 return -EOPNOTSUPP;
1205
1206                         if (af_ops->validate_link_af) {
1207                                 err = af_ops->validate_link_af(dev, af);
1208                                 if (err < 0)
1209                                         return err;
1210                         }
1211                 }
1212         }
1213
1214         return 0;
1215 }
1216
1217 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1218 {
1219         int rem, err = -EINVAL;
1220         struct nlattr *vf;
1221         const struct net_device_ops *ops = dev->netdev_ops;
1222
1223         nla_for_each_nested(vf, attr, rem) {
1224                 switch (nla_type(vf)) {
1225                 case IFLA_VF_MAC: {
1226                         struct ifla_vf_mac *ivm;
1227                         ivm = nla_data(vf);
1228                         err = -EOPNOTSUPP;
1229                         if (ops->ndo_set_vf_mac)
1230                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1231                                                           ivm->mac);
1232                         break;
1233                 }
1234                 case IFLA_VF_VLAN: {
1235                         struct ifla_vf_vlan *ivv;
1236                         ivv = nla_data(vf);
1237                         err = -EOPNOTSUPP;
1238                         if (ops->ndo_set_vf_vlan)
1239                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1240                                                            ivv->vlan,
1241                                                            ivv->qos);
1242                         break;
1243                 }
1244                 case IFLA_VF_TX_RATE: {
1245                         struct ifla_vf_tx_rate *ivt;
1246                         ivt = nla_data(vf);
1247                         err = -EOPNOTSUPP;
1248                         if (ops->ndo_set_vf_tx_rate)
1249                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1250                                                               ivt->rate);
1251                         break;
1252                 }
1253                 case IFLA_VF_SPOOFCHK: {
1254                         struct ifla_vf_spoofchk *ivs;
1255                         ivs = nla_data(vf);
1256                         err = -EOPNOTSUPP;
1257                         if (ops->ndo_set_vf_spoofchk)
1258                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1259                                                                ivs->setting);
1260                         break;
1261                 }
1262                 default:
1263                         err = -EINVAL;
1264                         break;
1265                 }
1266                 if (err)
1267                         break;
1268         }
1269         return err;
1270 }
1271
1272 static int do_set_master(struct net_device *dev, int ifindex)
1273 {
1274         struct net_device *master_dev;
1275         const struct net_device_ops *ops;
1276         int err;
1277
1278         if (dev->master) {
1279                 if (dev->master->ifindex == ifindex)
1280                         return 0;
1281                 ops = dev->master->netdev_ops;
1282                 if (ops->ndo_del_slave) {
1283                         err = ops->ndo_del_slave(dev->master, dev);
1284                         if (err)
1285                                 return err;
1286                 } else {
1287                         return -EOPNOTSUPP;
1288                 }
1289         }
1290
1291         if (ifindex) {
1292                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1293                 if (!master_dev)
1294                         return -EINVAL;
1295                 ops = master_dev->netdev_ops;
1296                 if (ops->ndo_add_slave) {
1297                         err = ops->ndo_add_slave(master_dev, dev);
1298                         if (err)
1299                                 return err;
1300                 } else {
1301                         return -EOPNOTSUPP;
1302                 }
1303         }
1304         return 0;
1305 }
1306
1307 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1308                       struct nlattr **tb, char *ifname, int modified)
1309 {
1310         const struct net_device_ops *ops = dev->netdev_ops;
1311         int send_addr_notify = 0;
1312         int err;
1313
1314         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1315                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1316                 if (IS_ERR(net)) {
1317                         err = PTR_ERR(net);
1318                         goto errout;
1319                 }
1320                 err = dev_change_net_namespace(dev, net, ifname);
1321                 put_net(net);
1322                 if (err)
1323                         goto errout;
1324                 modified = 1;
1325         }
1326
1327         if (tb[IFLA_MAP]) {
1328                 struct rtnl_link_ifmap *u_map;
1329                 struct ifmap k_map;
1330
1331                 if (!ops->ndo_set_config) {
1332                         err = -EOPNOTSUPP;
1333                         goto errout;
1334                 }
1335
1336                 if (!netif_device_present(dev)) {
1337                         err = -ENODEV;
1338                         goto errout;
1339                 }
1340
1341                 u_map = nla_data(tb[IFLA_MAP]);
1342                 k_map.mem_start = (unsigned long) u_map->mem_start;
1343                 k_map.mem_end = (unsigned long) u_map->mem_end;
1344                 k_map.base_addr = (unsigned short) u_map->base_addr;
1345                 k_map.irq = (unsigned char) u_map->irq;
1346                 k_map.dma = (unsigned char) u_map->dma;
1347                 k_map.port = (unsigned char) u_map->port;
1348
1349                 err = ops->ndo_set_config(dev, &k_map);
1350                 if (err < 0)
1351                         goto errout;
1352
1353                 modified = 1;
1354         }
1355
1356         if (tb[IFLA_ADDRESS]) {
1357                 struct sockaddr *sa;
1358                 int len;
1359
1360                 if (!ops->ndo_set_mac_address) {
1361                         err = -EOPNOTSUPP;
1362                         goto errout;
1363                 }
1364
1365                 if (!netif_device_present(dev)) {
1366                         err = -ENODEV;
1367                         goto errout;
1368                 }
1369
1370                 len = sizeof(sa_family_t) + dev->addr_len;
1371                 sa = kmalloc(len, GFP_KERNEL);
1372                 if (!sa) {
1373                         err = -ENOMEM;
1374                         goto errout;
1375                 }
1376                 sa->sa_family = dev->type;
1377                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1378                        dev->addr_len);
1379                 err = ops->ndo_set_mac_address(dev, sa);
1380                 kfree(sa);
1381                 if (err)
1382                         goto errout;
1383                 send_addr_notify = 1;
1384                 modified = 1;
1385                 add_device_randomness(dev->dev_addr, dev->addr_len);
1386         }
1387
1388         if (tb[IFLA_MTU]) {
1389                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1390                 if (err < 0)
1391                         goto errout;
1392                 modified = 1;
1393         }
1394
1395         if (tb[IFLA_GROUP]) {
1396                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1397                 modified = 1;
1398         }
1399
1400         /*
1401          * Interface selected by interface index but interface
1402          * name provided implies that a name change has been
1403          * requested.
1404          */
1405         if (ifm->ifi_index > 0 && ifname[0]) {
1406                 err = dev_change_name(dev, ifname);
1407                 if (err < 0)
1408                         goto errout;
1409                 modified = 1;
1410         }
1411
1412         if (tb[IFLA_IFALIAS]) {
1413                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1414                                     nla_len(tb[IFLA_IFALIAS]));
1415                 if (err < 0)
1416                         goto errout;
1417                 modified = 1;
1418         }
1419
1420         if (tb[IFLA_BROADCAST]) {
1421                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1422                 send_addr_notify = 1;
1423         }
1424
1425         if (ifm->ifi_flags || ifm->ifi_change) {
1426                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1427                 if (err < 0)
1428                         goto errout;
1429         }
1430
1431         if (tb[IFLA_MASTER]) {
1432                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1433                 if (err)
1434                         goto errout;
1435                 modified = 1;
1436         }
1437
1438         if (tb[IFLA_TXQLEN])
1439                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1440
1441         if (tb[IFLA_OPERSTATE])
1442                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1443
1444         if (tb[IFLA_LINKMODE]) {
1445                 write_lock_bh(&dev_base_lock);
1446                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1447                 write_unlock_bh(&dev_base_lock);
1448         }
1449
1450         if (tb[IFLA_VFINFO_LIST]) {
1451                 struct nlattr *attr;
1452                 int rem;
1453                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1454                         if (nla_type(attr) != IFLA_VF_INFO) {
1455                                 err = -EINVAL;
1456                                 goto errout;
1457                         }
1458                         err = do_setvfinfo(dev, attr);
1459                         if (err < 0)
1460                                 goto errout;
1461                         modified = 1;
1462                 }
1463         }
1464         err = 0;
1465
1466         if (tb[IFLA_VF_PORTS]) {
1467                 struct nlattr *port[IFLA_PORT_MAX+1];
1468                 struct nlattr *attr;
1469                 int vf;
1470                 int rem;
1471
1472                 err = -EOPNOTSUPP;
1473                 if (!ops->ndo_set_vf_port)
1474                         goto errout;
1475
1476                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1477                         if (nla_type(attr) != IFLA_VF_PORT)
1478                                 continue;
1479                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1480                                 attr, ifla_port_policy);
1481                         if (err < 0)
1482                                 goto errout;
1483                         if (!port[IFLA_PORT_VF]) {
1484                                 err = -EOPNOTSUPP;
1485                                 goto errout;
1486                         }
1487                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1488                         err = ops->ndo_set_vf_port(dev, vf, port);
1489                         if (err < 0)
1490                                 goto errout;
1491                         modified = 1;
1492                 }
1493         }
1494         err = 0;
1495
1496         if (tb[IFLA_PORT_SELF]) {
1497                 struct nlattr *port[IFLA_PORT_MAX+1];
1498
1499                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1500                         tb[IFLA_PORT_SELF], ifla_port_policy);
1501                 if (err < 0)
1502                         goto errout;
1503
1504                 err = -EOPNOTSUPP;
1505                 if (ops->ndo_set_vf_port)
1506                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1507                 if (err < 0)
1508                         goto errout;
1509                 modified = 1;
1510         }
1511
1512         if (tb[IFLA_AF_SPEC]) {
1513                 struct nlattr *af;
1514                 int rem;
1515
1516                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1517                         const struct rtnl_af_ops *af_ops;
1518
1519                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1520                                 BUG();
1521
1522                         err = af_ops->set_link_af(dev, af);
1523                         if (err < 0)
1524                                 goto errout;
1525
1526                         modified = 1;
1527                 }
1528         }
1529         err = 0;
1530
1531 errout:
1532         if (err < 0 && modified && net_ratelimit())
1533                 printk(KERN_WARNING "A link change request failed with "
1534                        "some changes committed already. Interface %s may "
1535                        "have been left with an inconsistent configuration, "
1536                        "please check.\n", dev->name);
1537
1538         if (send_addr_notify)
1539                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1540         return err;
1541 }
1542
1543 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1544 {
1545         struct net *net = sock_net(skb->sk);
1546         struct ifinfomsg *ifm;
1547         struct net_device *dev;
1548         int err;
1549         struct nlattr *tb[IFLA_MAX+1];
1550         char ifname[IFNAMSIZ];
1551
1552         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1553         if (err < 0)
1554                 goto errout;
1555
1556         if (tb[IFLA_IFNAME])
1557                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1558         else
1559                 ifname[0] = '\0';
1560
1561         err = -EINVAL;
1562         ifm = nlmsg_data(nlh);
1563         if (ifm->ifi_index > 0)
1564                 dev = __dev_get_by_index(net, ifm->ifi_index);
1565         else if (tb[IFLA_IFNAME])
1566                 dev = __dev_get_by_name(net, ifname);
1567         else
1568                 goto errout;
1569
1570         if (dev == NULL) {
1571                 err = -ENODEV;
1572                 goto errout;
1573         }
1574
1575         err = validate_linkmsg(dev, tb);
1576         if (err < 0)
1577                 goto errout;
1578
1579         err = do_setlink(dev, ifm, tb, ifname, 0);
1580 errout:
1581         return err;
1582 }
1583
1584 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1585 {
1586         struct net *net = sock_net(skb->sk);
1587         const struct rtnl_link_ops *ops;
1588         struct net_device *dev;
1589         struct ifinfomsg *ifm;
1590         char ifname[IFNAMSIZ];
1591         struct nlattr *tb[IFLA_MAX+1];
1592         int err;
1593         LIST_HEAD(list_kill);
1594
1595         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1596         if (err < 0)
1597                 return err;
1598
1599         if (tb[IFLA_IFNAME])
1600                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1601
1602         ifm = nlmsg_data(nlh);
1603         if (ifm->ifi_index > 0)
1604                 dev = __dev_get_by_index(net, ifm->ifi_index);
1605         else if (tb[IFLA_IFNAME])
1606                 dev = __dev_get_by_name(net, ifname);
1607         else
1608                 return -EINVAL;
1609
1610         if (!dev)
1611                 return -ENODEV;
1612
1613         ops = dev->rtnl_link_ops;
1614         if (!ops)
1615                 return -EOPNOTSUPP;
1616
1617         ops->dellink(dev, &list_kill);
1618         unregister_netdevice_many(&list_kill);
1619         list_del(&list_kill);
1620         return 0;
1621 }
1622
1623 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1624 {
1625         unsigned int old_flags;
1626         int err;
1627
1628         old_flags = dev->flags;
1629         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1630                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1631                 if (err < 0)
1632                         return err;
1633         }
1634
1635         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1636         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1637
1638         __dev_notify_flags(dev, old_flags);
1639         return 0;
1640 }
1641 EXPORT_SYMBOL(rtnl_configure_link);
1642
1643 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1644         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1645 {
1646         int err;
1647         struct net_device *dev;
1648         unsigned int num_queues = 1;
1649         unsigned int real_num_queues = 1;
1650
1651         if (ops->get_tx_queues) {
1652                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1653                                          &real_num_queues);
1654                 if (err)
1655                         goto err;
1656         }
1657         err = -ENOMEM;
1658         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1659         if (!dev)
1660                 goto err;
1661
1662         dev_net_set(dev, net);
1663         dev->rtnl_link_ops = ops;
1664         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1665
1666         if (tb[IFLA_MTU])
1667                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1668         if (tb[IFLA_ADDRESS])
1669                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1670                                 nla_len(tb[IFLA_ADDRESS]));
1671         if (tb[IFLA_BROADCAST])
1672                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1673                                 nla_len(tb[IFLA_BROADCAST]));
1674         if (tb[IFLA_TXQLEN])
1675                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1676         if (tb[IFLA_OPERSTATE])
1677                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1678         if (tb[IFLA_LINKMODE])
1679                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1680         if (tb[IFLA_GROUP])
1681                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1682
1683         return dev;
1684
1685 err:
1686         return ERR_PTR(err);
1687 }
1688 EXPORT_SYMBOL(rtnl_create_link);
1689
1690 static int rtnl_group_changelink(struct net *net, int group,
1691                 struct ifinfomsg *ifm,
1692                 struct nlattr **tb)
1693 {
1694         struct net_device *dev;
1695         int err;
1696
1697         for_each_netdev(net, dev) {
1698                 if (dev->group == group) {
1699                         err = do_setlink(dev, ifm, tb, NULL, 0);
1700                         if (err < 0)
1701                                 return err;
1702                 }
1703         }
1704
1705         return 0;
1706 }
1707
1708 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1709 {
1710         struct net *net = sock_net(skb->sk);
1711         const struct rtnl_link_ops *ops;
1712         struct net_device *dev;
1713         struct ifinfomsg *ifm;
1714         char kind[MODULE_NAME_LEN];
1715         char ifname[IFNAMSIZ];
1716         struct nlattr *tb[IFLA_MAX+1];
1717         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1718         int err;
1719
1720 #ifdef CONFIG_MODULES
1721 replay:
1722 #endif
1723         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1724         if (err < 0)
1725                 return err;
1726
1727         if (tb[IFLA_IFNAME])
1728                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1729         else
1730                 ifname[0] = '\0';
1731
1732         ifm = nlmsg_data(nlh);
1733         if (ifm->ifi_index > 0)
1734                 dev = __dev_get_by_index(net, ifm->ifi_index);
1735         else {
1736                 if (ifname[0])
1737                         dev = __dev_get_by_name(net, ifname);
1738                 else
1739                         dev = NULL;
1740         }
1741
1742         err = validate_linkmsg(dev, tb);
1743         if (err < 0)
1744                 return err;
1745
1746         if (tb[IFLA_LINKINFO]) {
1747                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1748                                        tb[IFLA_LINKINFO], ifla_info_policy);
1749                 if (err < 0)
1750                         return err;
1751         } else
1752                 memset(linkinfo, 0, sizeof(linkinfo));
1753
1754         if (linkinfo[IFLA_INFO_KIND]) {
1755                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1756                 ops = rtnl_link_ops_get(kind);
1757         } else {
1758                 kind[0] = '\0';
1759                 ops = NULL;
1760         }
1761
1762         if (1) {
1763                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1764                 struct net *dest_net;
1765
1766                 if (ops) {
1767                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1768                                 err = nla_parse_nested(attr, ops->maxtype,
1769                                                        linkinfo[IFLA_INFO_DATA],
1770                                                        ops->policy);
1771                                 if (err < 0)
1772                                         return err;
1773                                 data = attr;
1774                         }
1775                         if (ops->validate) {
1776                                 err = ops->validate(tb, data);
1777                                 if (err < 0)
1778                                         return err;
1779                         }
1780                 }
1781
1782                 if (dev) {
1783                         int modified = 0;
1784
1785                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1786                                 return -EEXIST;
1787                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1788                                 return -EOPNOTSUPP;
1789
1790                         if (linkinfo[IFLA_INFO_DATA]) {
1791                                 if (!ops || ops != dev->rtnl_link_ops ||
1792                                     !ops->changelink)
1793                                         return -EOPNOTSUPP;
1794
1795                                 err = ops->changelink(dev, tb, data);
1796                                 if (err < 0)
1797                                         return err;
1798                                 modified = 1;
1799                         }
1800
1801                         return do_setlink(dev, ifm, tb, ifname, modified);
1802                 }
1803
1804                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1805                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1806                                 return rtnl_group_changelink(net,
1807                                                 nla_get_u32(tb[IFLA_GROUP]),
1808                                                 ifm, tb);
1809                         return -ENODEV;
1810                 }
1811
1812                 if (ifm->ifi_index)
1813                         return -EOPNOTSUPP;
1814                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1815                         return -EOPNOTSUPP;
1816
1817                 if (!ops) {
1818 #ifdef CONFIG_MODULES
1819                         if (kind[0]) {
1820                                 __rtnl_unlock();
1821                                 request_module("rtnl-link-%s", kind);
1822                                 rtnl_lock();
1823                                 ops = rtnl_link_ops_get(kind);
1824                                 if (ops)
1825                                         goto replay;
1826                         }
1827 #endif
1828                         return -EOPNOTSUPP;
1829                 }
1830
1831                 if (!ifname[0])
1832                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1833
1834                 dest_net = rtnl_link_get_net(net, tb);
1835                 if (IS_ERR(dest_net))
1836                         return PTR_ERR(dest_net);
1837
1838                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1839
1840                 if (IS_ERR(dev))
1841                         err = PTR_ERR(dev);
1842                 else if (ops->newlink)
1843                         err = ops->newlink(net, dev, tb, data);
1844                 else
1845                         err = register_netdevice(dev);
1846
1847                 if (err < 0 && !IS_ERR(dev))
1848                         free_netdev(dev);
1849                 if (err < 0)
1850                         goto out;
1851
1852                 err = rtnl_configure_link(dev, ifm);
1853                 if (err < 0)
1854                         unregister_netdevice(dev);
1855 out:
1856                 put_net(dest_net);
1857                 return err;
1858         }
1859 }
1860
1861 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1862 {
1863         struct net *net = sock_net(skb->sk);
1864         struct ifinfomsg *ifm;
1865         char ifname[IFNAMSIZ];
1866         struct nlattr *tb[IFLA_MAX+1];
1867         struct net_device *dev = NULL;
1868         struct sk_buff *nskb;
1869         int err;
1870         u32 ext_filter_mask = 0;
1871
1872         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1873         if (err < 0)
1874                 return err;
1875
1876         if (tb[IFLA_IFNAME])
1877                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1878
1879         if (tb[IFLA_EXT_MASK])
1880                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1881
1882         ifm = nlmsg_data(nlh);
1883         if (ifm->ifi_index > 0)
1884                 dev = __dev_get_by_index(net, ifm->ifi_index);
1885         else if (tb[IFLA_IFNAME])
1886                 dev = __dev_get_by_name(net, ifname);
1887         else
1888                 return -EINVAL;
1889
1890         if (dev == NULL)
1891                 return -ENODEV;
1892
1893         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1894         if (nskb == NULL)
1895                 return -ENOBUFS;
1896
1897         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1898                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1899         if (err < 0) {
1900                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1901                 WARN_ON(err == -EMSGSIZE);
1902                 kfree_skb(nskb);
1903         } else
1904                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1905
1906         return err;
1907 }
1908
1909 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1910 {
1911         struct net *net = sock_net(skb->sk);
1912         struct net_device *dev;
1913         struct nlattr *tb[IFLA_MAX+1];
1914         u32 ext_filter_mask = 0;
1915         u16 min_ifinfo_dump_size = 0;
1916
1917         if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1918                         ifla_policy) >= 0) {
1919                 if (tb[IFLA_EXT_MASK])
1920                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1921         }
1922
1923         if (!ext_filter_mask)
1924                 return NLMSG_GOODSIZE;
1925         /*
1926          * traverse the list of net devices and compute the minimum
1927          * buffer size based upon the filter mask.
1928          */
1929         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1930                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1931                                              if_nlmsg_size(dev,
1932                                                            ext_filter_mask));
1933         }
1934
1935         return min_ifinfo_dump_size;
1936 }
1937
1938 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1939 {
1940         int idx;
1941         int s_idx = cb->family;
1942
1943         if (s_idx == 0)
1944                 s_idx = 1;
1945         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1946                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1947                 if (idx < s_idx || idx == PF_PACKET)
1948                         continue;
1949                 if (rtnl_msg_handlers[idx] == NULL ||
1950                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1951                         continue;
1952                 if (idx > s_idx)
1953                         memset(&cb->args[0], 0, sizeof(cb->args));
1954                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1955                         break;
1956         }
1957         cb->family = idx;
1958
1959         return skb->len;
1960 }
1961
1962 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1963 {
1964         struct net *net = dev_net(dev);
1965         struct sk_buff *skb;
1966         int err = -ENOBUFS;
1967         size_t if_info_size;
1968
1969         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1970         if (skb == NULL)
1971                 goto errout;
1972
1973         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1974         if (err < 0) {
1975                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1976                 WARN_ON(err == -EMSGSIZE);
1977                 kfree_skb(skb);
1978                 goto errout;
1979         }
1980         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1981         return;
1982 errout:
1983         if (err < 0)
1984                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1985 }
1986
1987 /* Protected by RTNL sempahore.  */
1988 static struct rtattr **rta_buf;
1989 static int rtattr_max;
1990
1991 /* Process one rtnetlink message. */
1992
1993 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1994 {
1995         struct net *net = sock_net(skb->sk);
1996         rtnl_doit_func doit;
1997         int sz_idx, kind;
1998         int min_len;
1999         int family;
2000         int type;
2001         int err;
2002
2003         type = nlh->nlmsg_type;
2004         if (type > RTM_MAX)
2005                 return -EOPNOTSUPP;
2006
2007         type -= RTM_BASE;
2008
2009         /* All the messages must have at least 1 byte length */
2010         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2011                 return 0;
2012
2013         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2014         sz_idx = type>>2;
2015         kind = type&3;
2016
2017         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
2018                 return -EPERM;
2019
2020         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2021                 struct sock *rtnl;
2022                 rtnl_dumpit_func dumpit;
2023                 rtnl_calcit_func calcit;
2024                 u16 min_dump_alloc = 0;
2025
2026                 dumpit = rtnl_get_dumpit(family, type);
2027                 if (dumpit == NULL)
2028                         return -EOPNOTSUPP;
2029                 calcit = rtnl_get_calcit(family, type);
2030                 if (calcit)
2031                         min_dump_alloc = calcit(skb, nlh);
2032
2033                 __rtnl_unlock();
2034                 rtnl = net->rtnl;
2035                 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
2036                                          NULL, min_dump_alloc);
2037                 rtnl_lock();
2038                 return err;
2039         }
2040
2041         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2042
2043         min_len = rtm_min[sz_idx];
2044         if (nlh->nlmsg_len < min_len)
2045                 return -EINVAL;
2046
2047         if (nlh->nlmsg_len > min_len) {
2048                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2049                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2050
2051                 while (RTA_OK(attr, attrlen)) {
2052                         unsigned int flavor = attr->rta_type & NLA_TYPE_MASK;
2053                         if (flavor) {
2054                                 if (flavor > rta_max[sz_idx])
2055                                         return -EINVAL;
2056                                 rta_buf[flavor-1] = attr;
2057                         }
2058                         attr = RTA_NEXT(attr, attrlen);
2059                 }
2060         }
2061
2062         doit = rtnl_get_doit(family, type);
2063         if (doit == NULL)
2064                 return -EOPNOTSUPP;
2065
2066         return doit(skb, nlh, (void *)&rta_buf[0]);
2067 }
2068
2069 static void rtnetlink_rcv(struct sk_buff *skb)
2070 {
2071         rtnl_lock();
2072         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2073         rtnl_unlock();
2074 }
2075
2076 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2077 {
2078         struct net_device *dev = ptr;
2079
2080         switch (event) {
2081         case NETDEV_UP:
2082         case NETDEV_DOWN:
2083         case NETDEV_PRE_UP:
2084         case NETDEV_POST_INIT:
2085         case NETDEV_REGISTER:
2086         case NETDEV_CHANGE:
2087         case NETDEV_PRE_TYPE_CHANGE:
2088         case NETDEV_GOING_DOWN:
2089         case NETDEV_UNREGISTER:
2090         case NETDEV_UNREGISTER_BATCH:
2091         case NETDEV_RELEASE:
2092         case NETDEV_JOIN:
2093                 break;
2094         default:
2095                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2096                 break;
2097         }
2098         return NOTIFY_DONE;
2099 }
2100
2101 static struct notifier_block rtnetlink_dev_notifier = {
2102         .notifier_call  = rtnetlink_event,
2103 };
2104
2105
2106 static int __net_init rtnetlink_net_init(struct net *net)
2107 {
2108         struct sock *sk;
2109         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2110                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2111         if (!sk)
2112                 return -ENOMEM;
2113         net->rtnl = sk;
2114         return 0;
2115 }
2116
2117 static void __net_exit rtnetlink_net_exit(struct net *net)
2118 {
2119         netlink_kernel_release(net->rtnl);
2120         net->rtnl = NULL;
2121 }
2122
2123 static struct pernet_operations rtnetlink_net_ops = {
2124         .init = rtnetlink_net_init,
2125         .exit = rtnetlink_net_exit,
2126 };
2127
2128 void __init rtnetlink_init(void)
2129 {
2130         int i;
2131
2132         rtattr_max = 0;
2133         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2134                 if (rta_max[i] > rtattr_max)
2135                         rtattr_max = rta_max[i];
2136         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2137         if (!rta_buf)
2138                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2139
2140         if (register_pernet_subsys(&rtnetlink_net_ops))
2141                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2142
2143         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2144         register_netdevice_notifier(&rtnetlink_dev_notifier);
2145
2146         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2147                       rtnl_dump_ifinfo, rtnl_calcit);
2148         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2149         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2150         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2151
2152         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2153         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2154 }
2155