net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/pci.h>
  39
  40 #include <asm/uaccess.h>
  41 #include <asm/system.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/sock.h>
  51 #include <net/pkt_sched.h>
  52 #include <net/fib_rules.h>
  53 #include <net/rtnetlink.h>
  54 #include <net/net_namespace.h>
  55
  56 struct rtnl_link {
  57         rtnl_doit_func          doit;
  58         rtnl_dumpit_func        dumpit;
  59         rtnl_calcit_func        calcit;
  60 };
  61
  62 static DEFINE_MUTEX(rtnl_mutex);
  63 static u16 min_ifinfo_dump_size;
  64
  65 void rtnl_lock(void)
  66 {
  67         mutex_lock(&rtnl_mutex);
  68 }
  69 EXPORT_SYMBOL(rtnl_lock);
  70
  71 void __rtnl_unlock(void)
  72 {
  73         mutex_unlock(&rtnl_mutex);
  74 }
  75
  76 void rtnl_unlock(void)
  77 {
  78         /* This fellow will unlock it for us. */
  79         netdev_run_todo();
  80 }
  81 EXPORT_SYMBOL(rtnl_unlock);
  82
  83 int rtnl_trylock(void)
  84 {
  85         return mutex_trylock(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_trylock);
  88
  89 int rtnl_is_locked(void)
  90 {
  91         return mutex_is_locked(&rtnl_mutex);
  92 }
  93 EXPORT_SYMBOL(rtnl_is_locked);
  94
  95 #ifdef CONFIG_PROVE_LOCKING
  96 int lockdep_rtnl_is_held(void)
  97 {
  98         return lockdep_is_held(&rtnl_mutex);
  99 }
 100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105 static inline int rtm_msgindex(int msgtype)
 106 {
 107         int msgindex = msgtype - RTM_BASE;
 108
 109         /*
 110          * msgindex < 0 implies someone tried to register a netlink
 111          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112          * the message type has not been added to linux/rtnetlink.h
 113          */
 114         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116         return msgindex;
 117 }
 118
 119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120 {
 121         struct rtnl_link *tab;
 122
 123         if (protocol <= RTNL_FAMILY_MAX)
 124                 tab = rtnl_msg_handlers[protocol];
 125         else
 126                 tab = NULL;
 127
 128         if (tab == NULL || tab[msgindex].doit == NULL)
 129                 tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131         return tab ? tab[msgindex].doit : NULL;
 132 }
 133
 134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135 {
 136         struct rtnl_link *tab;
 137
 138         if (protocol <= RTNL_FAMILY_MAX)
 139                 tab = rtnl_msg_handlers[protocol];
 140         else
 141                 tab = NULL;
 142
 143         if (tab == NULL || tab[msgindex].dumpit == NULL)
 144                 tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146         return tab ? tab[msgindex].dumpit : NULL;
 147 }
 148
 149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150 {
 151         struct rtnl_link *tab;
 152
 153         if (protocol <= RTNL_FAMILY_MAX)
 154                 tab = rtnl_msg_handlers[protocol];
 155         else
 156                 tab = NULL;
 157
 158         if (tab == NULL || tab[msgindex].calcit == NULL)
 159                 tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161         return tab ? tab[msgindex].calcit : NULL;
 162 }
 163
 164 /**
 165  * __rtnl_register - Register a rtnetlink message type
 166  * @protocol: Protocol family or PF_UNSPEC
 167  * @msgtype: rtnetlink message type
 168  * @doit: Function pointer called for each request message
 169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170  * @calcit: Function pointer to calc size of dump message
 171  *
 172  * Registers the specified function pointers (at least one of them has
 173  * to be non-NULL) to be called whenever a request message for the
 174  * specified protocol family and message type is received.
 175  *
 176  * The special protocol family PF_UNSPEC may be used to define fallback
 177  * function pointers for the case when no entry for the specific protocol
 178  * family exists.
 179  *
 180  * Returns 0 on success or a negative error code.
 181  */
 182 int __rtnl_register(int protocol, int msgtype,
 183                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184                     rtnl_calcit_func calcit)
 185 {
 186         struct rtnl_link *tab;
 187         int msgindex;
 188
 189         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190         msgindex = rtm_msgindex(msgtype);
 191
 192         tab = rtnl_msg_handlers[protocol];
 193         if (tab == NULL) {
 194                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195                 if (tab == NULL)
 196                         return -ENOBUFS;
 197
 198                 rtnl_msg_handlers[protocol] = tab;
 199         }
 200
 201         if (doit)
 202                 tab[msgindex].doit = doit;
 203
 204         if (dumpit)
 205                 tab[msgindex].dumpit = dumpit;
 206
 207         if (calcit)
 208                 tab[msgindex].calcit = calcit;
 209
 210         return 0;
 211 }
 212 EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214 /**
 215  * rtnl_register - Register a rtnetlink message type
 216  *
 217  * Identical to __rtnl_register() but panics on failure. This is useful
 218  * as failure of this function is very unlikely, it can only happen due
 219  * to lack of memory when allocating the chain to store all message
 220  * handlers for a protocol. Meant for use in init functions where lack
 221  * of memory implies no sense in continuing.
 222  */
 223 void rtnl_register(int protocol, int msgtype,
 224                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225                    rtnl_calcit_func calcit)
 226 {
 227         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228                 panic("Unable to register rtnetlink message handler, "
 229                       "protocol = %d, message type = %d\n",
 230                       protocol, msgtype);
 231 }
 232 EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234 /**
 235  * rtnl_unregister - Unregister a rtnetlink message type
 236  * @protocol: Protocol family or PF_UNSPEC
 237  * @msgtype: rtnetlink message type
 238  *
 239  * Returns 0 on success or a negative error code.
 240  */
 241 int rtnl_unregister(int protocol, int msgtype)
 242 {
 243         int msgindex;
 244
 245         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246         msgindex = rtm_msgindex(msgtype);
 247
 248         if (rtnl_msg_handlers[protocol] == NULL)
 249                 return -ENOENT;
 250
 251         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254         return 0;
 255 }
 256 EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258 /**
 259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260  * @protocol : Protocol family or PF_UNSPEC
 261  *
 262  * Identical to calling rtnl_unregster() for all registered message types
 263  * of a certain protocol family.
 264  */
 265 void rtnl_unregister_all(int protocol)
 266 {
 267         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269         kfree(rtnl_msg_handlers[protocol]);
 270         rtnl_msg_handlers[protocol] = NULL;
 271 }
 272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274 static LIST_HEAD(link_ops);
 275
 276 /**
 277  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 278  * @ops: struct rtnl_link_ops * to register
 279  *
 280  * The caller must hold the rtnl_mutex. This function should be used
 281  * by drivers that create devices during module initialization. It
 282  * must be called before registering the devices.
 283  *
 284  * Returns 0 on success or a negative error code.
 285  */
 286 int __rtnl_link_register(struct rtnl_link_ops *ops)
 287 {
 288         if (!ops->dellink)
 289                 ops->dellink = unregister_netdevice_queue;
 290
 291         list_add_tail(&ops->list, &link_ops);
 292         return 0;
 293 }
 294 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 295
 296 /**
 297  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 298  * @ops: struct rtnl_link_ops * to register
 299  *
 300  * Returns 0 on success or a negative error code.
 301  */
 302 int rtnl_link_register(struct rtnl_link_ops *ops)
 303 {
 304         int err;
 305
 306         rtnl_lock();
 307         err = __rtnl_link_register(ops);
 308         rtnl_unlock();
 309         return err;
 310 }
 311 EXPORT_SYMBOL_GPL(rtnl_link_register);
 312
 313 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 314 {
 315         struct net_device *dev;
 316         LIST_HEAD(list_kill);
 317
 318         for_each_netdev(net, dev) {
 319                 if (dev->rtnl_link_ops == ops)
 320                         ops->dellink(dev, &list_kill);
 321         }
 322         unregister_netdevice_many(&list_kill);
 323 }
 324
 325 /**
 326  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 327  * @ops: struct rtnl_link_ops * to unregister
 328  *
 329  * The caller must hold the rtnl_mutex.
 330  */
 331 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 332 {
 333         struct net *net;
 334
 335         for_each_net(net) {
 336                 __rtnl_kill_links(net, ops);
 337         }
 338         list_del(&ops->list);
 339 }
 340 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 341
 342 /**
 343  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 344  * @ops: struct rtnl_link_ops * to unregister
 345  */
 346 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 347 {
 348         rtnl_lock();
 349         __rtnl_link_unregister(ops);
 350         rtnl_unlock();
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 353
 354 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 355 {
 356         const struct rtnl_link_ops *ops;
 357
 358         list_for_each_entry(ops, &link_ops, list) {
 359                 if (!strcmp(ops->kind, kind))
 360                         return ops;
 361         }
 362         return NULL;
 363 }
 364
 365 static size_t rtnl_link_get_size(const struct net_device *dev)
 366 {
 367         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 368         size_t size;
 369
 370         if (!ops)
 371                 return 0;
 372
 373         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 374                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 375
 376         if (ops->get_size)
 377                 /* IFLA_INFO_DATA + nested data */
 378                 size += nla_total_size(sizeof(struct nlattr)) +
 379                         ops->get_size(dev);
 380
 381         if (ops->get_xstats_size)
 382                 /* IFLA_INFO_XSTATS */
 383                 size += nla_total_size(ops->get_xstats_size(dev));
 384
 385         return size;
 386 }
 387
 388 static LIST_HEAD(rtnl_af_ops);
 389
 390 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 391 {
 392         const struct rtnl_af_ops *ops;
 393
 394         list_for_each_entry(ops, &rtnl_af_ops, list) {
 395                 if (ops->family == family)
 396                         return ops;
 397         }
 398
 399         return NULL;
 400 }
 401
 402 /**
 403  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 404  * @ops: struct rtnl_af_ops * to register
 405  *
 406  * The caller must hold the rtnl_mutex.
 407  *
 408  * Returns 0 on success or a negative error code.
 409  */
 410 int __rtnl_af_register(struct rtnl_af_ops *ops)
 411 {
 412         list_add_tail(&ops->list, &rtnl_af_ops);
 413         return 0;
 414 }
 415 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 416
 417 /**
 418  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 419  * @ops: struct rtnl_af_ops * to register
 420  *
 421  * Returns 0 on success or a negative error code.
 422  */
 423 int rtnl_af_register(struct rtnl_af_ops *ops)
 424 {
 425         int err;
 426
 427         rtnl_lock();
 428         err = __rtnl_af_register(ops);
 429         rtnl_unlock();
 430         return err;
 431 }
 432 EXPORT_SYMBOL_GPL(rtnl_af_register);
 433
 434 /**
 435  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 436  * @ops: struct rtnl_af_ops * to unregister
 437  *
 438  * The caller must hold the rtnl_mutex.
 439  */
 440 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 441 {
 442         list_del(&ops->list);
 443 }
 444 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 445
 446 /**
 447  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 448  * @ops: struct rtnl_af_ops * to unregister
 449  */
 450 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 451 {
 452         rtnl_lock();
 453         __rtnl_af_unregister(ops);
 454         rtnl_unlock();
 455 }
 456 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 457
 458 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 459 {
 460         struct rtnl_af_ops *af_ops;
 461         size_t size;
 462
 463         /* IFLA_AF_SPEC */
 464         size = nla_total_size(sizeof(struct nlattr));
 465
 466         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 467                 if (af_ops->get_link_af_size) {
 468                         /* AF_* + nested data */
 469                         size += nla_total_size(sizeof(struct nlattr)) +
 470                                 af_ops->get_link_af_size(dev);
 471                 }
 472         }
 473
 474         return size;
 475 }
 476
 477 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 478 {
 479         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 480         struct nlattr *linkinfo, *data;
 481         int err = -EMSGSIZE;
 482
 483         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 484         if (linkinfo == NULL)
 485                 goto out;
 486
 487         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 488                 goto err_cancel_link;
 489         if (ops->fill_xstats) {
 490                 err = ops->fill_xstats(skb, dev);
 491                 if (err < 0)
 492                         goto err_cancel_link;
 493         }
 494         if (ops->fill_info) {
 495                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 496                 if (data == NULL)
 497                         goto err_cancel_link;
 498                 err = ops->fill_info(skb, dev);
 499                 if (err < 0)
 500                         goto err_cancel_data;
 501                 nla_nest_end(skb, data);
 502         }
 503
 504         nla_nest_end(skb, linkinfo);
 505         return 0;
 506
 507 err_cancel_data:
 508         nla_nest_cancel(skb, data);
 509 err_cancel_link:
 510         nla_nest_cancel(skb, linkinfo);
 511 out:
 512         return err;
 513 }
 514
 515 static const int rtm_min[RTM_NR_FAMILIES] =
 516 {
 517         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 518         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 519         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 520         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 521         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 522         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 523         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 524         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 525         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 526         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 527 };
 528
 529 static const int rta_max[RTM_NR_FAMILIES] =
 530 {
 531         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 532         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 533         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 534         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 535         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 536         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 537         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 538         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 539 };
 540
 541 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 542 {
 543         struct rtattr *rta;
 544         int size = RTA_LENGTH(attrlen);
 545
 546         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 547         rta->rta_type = attrtype;
 548         rta->rta_len = size;
 549         memcpy(RTA_DATA(rta), data, attrlen);
 550         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 551 }
 552 EXPORT_SYMBOL(__rta_fill);
 553
 554 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 555 {
 556         struct sock *rtnl = net->rtnl;
 557         int err = 0;
 558
 559         NETLINK_CB(skb).dst_group = group;
 560         if (echo)
 561                 atomic_inc(&skb->users);
 562         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 563         if (echo)
 564                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 565         return err;
 566 }
 567
 568 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 569 {
 570         struct sock *rtnl = net->rtnl;
 571
 572         return nlmsg_unicast(rtnl, skb, pid);
 573 }
 574 EXPORT_SYMBOL(rtnl_unicast);
 575
 576 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 577                  struct nlmsghdr *nlh, gfp_t flags)
 578 {
 579         struct sock *rtnl = net->rtnl;
 580         int report = 0;
 581
 582         if (nlh)
 583                 report = nlmsg_report(nlh);
 584
 585         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 586 }
 587 EXPORT_SYMBOL(rtnl_notify);
 588
 589 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 590 {
 591         struct sock *rtnl = net->rtnl;
 592
 593         netlink_set_err(rtnl, 0, group, error);
 594 }
 595 EXPORT_SYMBOL(rtnl_set_sk_err);
 596
 597 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 598 {
 599         struct nlattr *mx;
 600         int i, valid = 0;
 601
 602         mx = nla_nest_start(skb, RTA_METRICS);
 603         if (mx == NULL)
 604                 return -ENOBUFS;
 605
 606         for (i = 0; i < RTAX_MAX; i++) {
 607                 if (metrics[i]) {
 608                         valid++;
 609                         NLA_PUT_U32(skb, i+1, metrics[i]);
 610                 }
 611         }
 612
 613         if (!valid) {
 614                 nla_nest_cancel(skb, mx);
 615                 return 0;
 616         }
 617
 618         return nla_nest_end(skb, mx);
 619
 620 nla_put_failure:
 621         nla_nest_cancel(skb, mx);
 622         return -EMSGSIZE;
 623 }
 624 EXPORT_SYMBOL(rtnetlink_put_metrics);
 625
 626 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 627                        u32 ts, u32 tsage, long expires, u32 error)
 628 {
 629         struct rta_cacheinfo ci = {
 630                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 631                 .rta_used = dst->__use,
 632                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 633                 .rta_error = error,
 634                 .rta_id =  id,
 635                 .rta_ts = ts,
 636                 .rta_tsage = tsage,
 637         };
 638
 639         if (expires)
 640                 ci.rta_expires = jiffies_to_clock_t(expires);
 641
 642         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 643 }
 644 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 645
 646 static void set_operstate(struct net_device *dev, unsigned char transition)
 647 {
 648         unsigned char operstate = dev->operstate;
 649
 650         switch (transition) {
 651         case IF_OPER_UP:
 652                 if ((operstate == IF_OPER_DORMANT ||
 653                      operstate == IF_OPER_UNKNOWN) &&
 654                     !netif_dormant(dev))
 655                         operstate = IF_OPER_UP;
 656                 break;
 657
 658         case IF_OPER_DORMANT:
 659                 if (operstate == IF_OPER_UP ||
 660                     operstate == IF_OPER_UNKNOWN)
 661                         operstate = IF_OPER_DORMANT;
 662                 break;
 663         }
 664
 665         if (dev->operstate != operstate) {
 666                 write_lock_bh(&dev_base_lock);
 667                 dev->operstate = operstate;
 668                 write_unlock_bh(&dev_base_lock);
 669                 netdev_state_change(dev);
 670         }
 671 }
 672
 673 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 674 {
 675         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 676                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 677 }
 678
 679 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 680                                            const struct ifinfomsg *ifm)
 681 {
 682         unsigned int flags = ifm->ifi_flags;
 683
 684         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 685         if (ifm->ifi_change)
 686                 flags = (flags & ifm->ifi_change) |
 687                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 688
 689         return flags;
 690 }
 691
 692 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 693                                  const struct rtnl_link_stats64 *b)
 694 {
 695         a->rx_packets = b->rx_packets;
 696         a->tx_packets = b->tx_packets;
 697         a->rx_bytes = b->rx_bytes;
 698         a->tx_bytes = b->tx_bytes;
 699         a->rx_errors = b->rx_errors;
 700         a->tx_errors = b->tx_errors;
 701         a->rx_dropped = b->rx_dropped;
 702         a->tx_dropped = b->tx_dropped;
 703
 704         a->multicast = b->multicast;
 705         a->collisions = b->collisions;
 706
 707         a->rx_length_errors = b->rx_length_errors;
 708         a->rx_over_errors = b->rx_over_errors;
 709         a->rx_crc_errors = b->rx_crc_errors;
 710         a->rx_frame_errors = b->rx_frame_errors;
 711         a->rx_fifo_errors = b->rx_fifo_errors;
 712         a->rx_missed_errors = b->rx_missed_errors;
 713
 714         a->tx_aborted_errors = b->tx_aborted_errors;
 715         a->tx_carrier_errors = b->tx_carrier_errors;
 716         a->tx_fifo_errors = b->tx_fifo_errors;
 717         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 718         a->tx_window_errors = b->tx_window_errors;
 719
 720         a->rx_compressed = b->rx_compressed;
 721         a->tx_compressed = b->tx_compressed;
 722 }
 723
 724 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 725 {
 726         memcpy(v, b, sizeof(*b));
 727 }
 728
 729 /* All VF info */
 730 static inline int rtnl_vfinfo_size(const struct net_device *dev)
 731 {
 732         if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
 733
 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 {
 774         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 775                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 776                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 777                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 778                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 779                + nla_total_size(sizeof(struct rtnl_link_stats))
 780                + nla_total_size(sizeof(struct rtnl_link_stats64))
 781                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 782                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 783                + nla_total_size(4) /* IFLA_TXQLEN */
 784                + nla_total_size(4) /* IFLA_WEIGHT */
 785                + nla_total_size(4) /* IFLA_MTU */
 786                + nla_total_size(4) /* IFLA_LINK */
 787                + nla_total_size(4) /* IFLA_MASTER */
 788                + nla_total_size(1) /* IFLA_OPERSTATE */
 789                + nla_total_size(1) /* IFLA_LINKMODE */
 790                + nla_total_size(4) /* IFLA_NUM_VF */
 791                + rtnl_vfinfo_size(dev) /* 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                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 813                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 814                 if (err == -EMSGSIZE)
 815                         goto nla_put_failure;
 816                 if (err) {
 817                         nla_nest_cancel(skb, vf_port);
 818                         continue;
 819                 }
 820                 nla_nest_end(skb, vf_port);
 821         }
 822
 823         nla_nest_end(skb, vf_ports);
 824
 825         return 0;
 826
 827 nla_put_failure:
 828         nla_nest_cancel(skb, vf_ports);
 829         return -EMSGSIZE;
 830 }
 831
 832 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 833 {
 834         struct nlattr *port_self;
 835         int err;
 836
 837         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 838         if (!port_self)
 839                 return -EMSGSIZE;
 840
 841         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 842         if (err) {
 843                 nla_nest_cancel(skb, port_self);
 844                 return (err == -EMSGSIZE) ? err : 0;
 845         }
 846
 847         nla_nest_end(skb, port_self);
 848
 849         return 0;
 850 }
 851
 852 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 853 {
 854         int err;
 855
 856         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 857                 return 0;
 858
 859         err = rtnl_port_self_fill(skb, dev);
 860         if (err)
 861                 return err;
 862
 863         if (dev_num_vf(dev->dev.parent)) {
 864                 err = rtnl_vf_ports_fill(skb, dev);
 865                 if (err)
 866                         return err;
 867         }
 868
 869         return 0;
 870 }
 871
 872 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 873                             int type, u32 pid, u32 seq, u32 change,
 874                             unsigned int flags)
 875 {
 876         struct ifinfomsg *ifm;
 877         struct nlmsghdr *nlh;
 878         struct rtnl_link_stats64 temp;
 879         const struct rtnl_link_stats64 *stats;
 880         struct nlattr *attr, *af_spec;
 881         struct rtnl_af_ops *af_ops;
 882
 883         ASSERT_RTNL();
 884         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 885         if (nlh == NULL)
 886                 return -EMSGSIZE;
 887
 888         ifm = nlmsg_data(nlh);
 889         ifm->ifi_family = AF_UNSPEC;
 890         ifm->__ifi_pad = 0;
 891         ifm->ifi_type = dev->type;
 892         ifm->ifi_index = dev->ifindex;
 893         ifm->ifi_flags = dev_get_flags(dev);
 894         ifm->ifi_change = change;
 895
 896         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 897         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 898         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 899                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 900         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 901         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 902         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 903
 904         if (dev->ifindex != dev->iflink)
 905                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 906
 907         if (dev->master)
 908                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 909
 910         if (dev->qdisc)
 911                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 912
 913         if (dev->ifalias)
 914                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 915
 916         if (1) {
 917                 struct rtnl_link_ifmap map = {
 918                         .mem_start   = dev->mem_start,
 919                         .mem_end     = dev->mem_end,
 920                         .base_addr   = dev->base_addr,
 921                         .irq         = dev->irq,
 922                         .dma         = dev->dma,
 923                         .port        = dev->if_port,
 924                 };
 925                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 926         }
 927
 928         if (dev->addr_len) {
 929                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 930                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 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)
 948                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 949
 950         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
 951                 int i;
 952
 953                 struct nlattr *vfinfo, *vf;
 954                 int num_vfs = dev_num_vf(dev->dev.parent);
 955
 956                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 957                 if (!vfinfo)
 958                         goto nla_put_failure;
 959                 for (i = 0; i < num_vfs; i++) {
 960                         struct ifla_vf_info ivi;
 961                         struct ifla_vf_mac vf_mac;
 962                         struct ifla_vf_vlan vf_vlan;
 963                         struct ifla_vf_tx_rate vf_tx_rate;
 964                         struct ifla_vf_spoofchk vf_spoofchk;
 965
 966                         /*
 967                          * Not all SR-IOV capable drivers support the
 968                          * spoofcheck query.  Preset to -1 so the user
 969                          * space tool can detect that the driver didn't
 970                          * report anything.
 971                          */
 972                         ivi.spoofchk = -1;
 973                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 974                                 break;
 975                         vf_mac.vf =
 976                                 vf_vlan.vf =
 977                                 vf_tx_rate.vf =
 978                                 vf_spoofchk.vf = ivi.vf;
 979
 980                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 981                         vf_vlan.vlan = ivi.vlan;
 982                         vf_vlan.qos = ivi.qos;
 983                         vf_tx_rate.rate = ivi.tx_rate;
 984                         vf_spoofchk.setting = ivi.spoofchk;
 985                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 986                         if (!vf) {
 987                                 nla_nest_cancel(skb, vfinfo);
 988                                 goto nla_put_failure;
 989                         }
 990                         NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
 991                         NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
 992                         NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 993                                 &vf_tx_rate);
 994                         NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 995                                 &vf_spoofchk);
 996                         nla_nest_end(skb, vf);
 997                 }
 998                 nla_nest_end(skb, vfinfo);
 999         }
1000
1001         if (rtnl_port_fill(skb, dev))
1002                 goto nla_put_failure;
1003
1004         if (dev->rtnl_link_ops) {
1005                 if (rtnl_link_fill(skb, dev) < 0)
1006                         goto nla_put_failure;
1007         }
1008
1009         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1010                 goto nla_put_failure;
1011
1012         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1013                 if (af_ops->fill_link_af) {
1014                         struct nlattr *af;
1015                         int err;
1016
1017                         if (!(af = nla_nest_start(skb, af_ops->family)))
1018                                 goto nla_put_failure;
1019
1020                         err = af_ops->fill_link_af(skb, dev);
1021
1022                         /*
1023                          * Caller may return ENODATA to indicate that there
1024                          * was no data to be dumped. This is not an error, it
1025                          * means we should trim the attribute header and
1026                          * continue.
1027                          */
1028                         if (err == -ENODATA)
1029                                 nla_nest_cancel(skb, af);
1030                         else if (err < 0)
1031                                 goto nla_put_failure;
1032
1033                         nla_nest_end(skb, af);
1034                 }
1035         }
1036
1037         nla_nest_end(skb, af_spec);
1038
1039         return nlmsg_end(skb, nlh);
1040
1041 nla_put_failure:
1042         nlmsg_cancel(skb, nlh);
1043         return -EMSGSIZE;
1044 }
1045
1046 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1047 {
1048         struct net *net = sock_net(skb->sk);
1049         int h, s_h;
1050         int idx = 0, s_idx;
1051         struct net_device *dev;
1052         struct hlist_head *head;
1053         struct hlist_node *node;
1054
1055         s_h = cb->args[0];
1056         s_idx = cb->args[1];
1057
1058         rcu_read_lock();
1059         cb->seq = net->dev_base_seq;
1060
1061         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1062                 idx = 0;
1063                 head = &net->dev_index_head[h];
1064                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1065                         if (idx < s_idx)
1066                                 goto cont;
1067                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1068                                              NETLINK_CB(cb->skb).pid,
1069                                              cb->nlh->nlmsg_seq, 0,
1070                                              NLM_F_MULTI) <= 0)
1071                                 goto out;
1072
1073                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1074 cont:
1075                         idx++;
1076                 }
1077         }
1078 out:
1079         rcu_read_unlock();
1080         cb->args[1] = idx;
1081         cb->args[0] = h;
1082
1083         return skb->len;
1084 }
1085
1086 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1087         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1088         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1089         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1090         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1091         [IFLA_MTU]              = { .type = NLA_U32 },
1092         [IFLA_LINK]             = { .type = NLA_U32 },
1093         [IFLA_MASTER]           = { .type = NLA_U32 },
1094         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1095         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1096         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1097         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1098         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1099         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1100         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1101         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1102         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1103         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1104         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1105         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1106 };
1107 EXPORT_SYMBOL(ifla_policy);
1108
1109 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1110         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1111         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1112 };
1113
1114 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1115         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1116 };
1117
1118 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1119         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1120                                     .len = sizeof(struct ifla_vf_mac) },
1121         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1122                                     .len = sizeof(struct ifla_vf_vlan) },
1123         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1124                                     .len = sizeof(struct ifla_vf_tx_rate) },
1125         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1126                                     .len = sizeof(struct ifla_vf_spoofchk) },
1127 };
1128
1129 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1130         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1131         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1132                                     .len = PORT_PROFILE_MAX },
1133         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1134                                     .len = sizeof(struct ifla_port_vsi)},
1135         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1136                                       .len = PORT_UUID_MAX },
1137         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1138                                     .len = PORT_UUID_MAX },
1139         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1140         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1141 };
1142
1143 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1144 {
1145         struct net *net;
1146         /* Examine the link attributes and figure out which
1147          * network namespace we are talking about.
1148          */
1149         if (tb[IFLA_NET_NS_PID])
1150                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1151         else if (tb[IFLA_NET_NS_FD])
1152                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1153         else
1154                 net = get_net(src_net);
1155         return net;
1156 }
1157 EXPORT_SYMBOL(rtnl_link_get_net);
1158
1159 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1160 {
1161         if (dev) {
1162                 if (tb[IFLA_ADDRESS] &&
1163                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1164                         return -EINVAL;
1165
1166                 if (tb[IFLA_BROADCAST] &&
1167                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1168                         return -EINVAL;
1169         }
1170
1171         if (tb[IFLA_AF_SPEC]) {
1172                 struct nlattr *af;
1173                 int rem, err;
1174
1175                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1176                         const struct rtnl_af_ops *af_ops;
1177
1178                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1179                                 return -EAFNOSUPPORT;
1180
1181                         if (!af_ops->set_link_af)
1182                                 return -EOPNOTSUPP;
1183
1184                         if (af_ops->validate_link_af) {
1185                                 err = af_ops->validate_link_af(dev, af);
1186                                 if (err < 0)
1187                                         return err;
1188                         }
1189                 }
1190         }
1191
1192         return 0;
1193 }
1194
1195 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1196 {
1197         int rem, err = -EINVAL;
1198         struct nlattr *vf;
1199         const struct net_device_ops *ops = dev->netdev_ops;
1200
1201         nla_for_each_nested(vf, attr, rem) {
1202                 switch (nla_type(vf)) {
1203                 case IFLA_VF_MAC: {
1204                         struct ifla_vf_mac *ivm;
1205                         ivm = nla_data(vf);
1206                         err = -EOPNOTSUPP;
1207                         if (ops->ndo_set_vf_mac)
1208                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1209                                                           ivm->mac);
1210                         break;
1211                 }
1212                 case IFLA_VF_VLAN: {
1213                         struct ifla_vf_vlan *ivv;
1214                         ivv = nla_data(vf);
1215                         err = -EOPNOTSUPP;
1216                         if (ops->ndo_set_vf_vlan)
1217                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1218                                                            ivv->vlan,
1219                                                            ivv->qos);
1220                         break;
1221                 }
1222                 case IFLA_VF_TX_RATE: {
1223                         struct ifla_vf_tx_rate *ivt;
1224                         ivt = nla_data(vf);
1225                         err = -EOPNOTSUPP;
1226                         if (ops->ndo_set_vf_tx_rate)
1227                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1228                                                               ivt->rate);
1229                         break;
1230                 }
1231                 case IFLA_VF_SPOOFCHK: {
1232                         struct ifla_vf_spoofchk *ivs;
1233                         ivs = nla_data(vf);
1234                         err = -EOPNOTSUPP;
1235                         if (ops->ndo_set_vf_spoofchk)
1236                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1237                                                                ivs->setting);
1238                         break;
1239                 }
1240                 default:
1241                         err = -EINVAL;
1242                         break;
1243                 }
1244                 if (err)
1245                         break;
1246         }
1247         return err;
1248 }
1249
1250 static int do_set_master(struct net_device *dev, int ifindex)
1251 {
1252         struct net_device *master_dev;
1253         const struct net_device_ops *ops;
1254         int err;
1255
1256         if (dev->master) {
1257                 if (dev->master->ifindex == ifindex)
1258                         return 0;
1259                 ops = dev->master->netdev_ops;
1260                 if (ops->ndo_del_slave) {
1261                         err = ops->ndo_del_slave(dev->master, dev);
1262                         if (err)
1263                                 return err;
1264                 } else {
1265                         return -EOPNOTSUPP;
1266                 }
1267         }
1268
1269         if (ifindex) {
1270                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1271                 if (!master_dev)
1272                         return -EINVAL;
1273                 ops = master_dev->netdev_ops;
1274                 if (ops->ndo_add_slave) {
1275                         err = ops->ndo_add_slave(master_dev, dev);
1276                         if (err)
1277                                 return err;
1278                 } else {
1279                         return -EOPNOTSUPP;
1280                 }
1281         }
1282         return 0;
1283 }
1284
1285 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1286                       struct nlattr **tb, char *ifname, int modified)
1287 {
1288         const struct net_device_ops *ops = dev->netdev_ops;
1289         int send_addr_notify = 0;
1290         int err;
1291
1292         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1293                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1294                 if (IS_ERR(net)) {
1295                         err = PTR_ERR(net);
1296                         goto errout;
1297                 }
1298                 err = dev_change_net_namespace(dev, net, ifname);
1299                 put_net(net);
1300                 if (err)
1301                         goto errout;
1302                 modified = 1;
1303         }
1304
1305         if (tb[IFLA_MAP]) {
1306                 struct rtnl_link_ifmap *u_map;
1307                 struct ifmap k_map;
1308
1309                 if (!ops->ndo_set_config) {
1310                         err = -EOPNOTSUPP;
1311                         goto errout;
1312                 }
1313
1314                 if (!netif_device_present(dev)) {
1315                         err = -ENODEV;
1316                         goto errout;
1317                 }
1318
1319                 u_map = nla_data(tb[IFLA_MAP]);
1320                 k_map.mem_start = (unsigned long) u_map->mem_start;
1321                 k_map.mem_end = (unsigned long) u_map->mem_end;
1322                 k_map.base_addr = (unsigned short) u_map->base_addr;
1323                 k_map.irq = (unsigned char) u_map->irq;
1324                 k_map.dma = (unsigned char) u_map->dma;
1325                 k_map.port = (unsigned char) u_map->port;
1326
1327                 err = ops->ndo_set_config(dev, &k_map);
1328                 if (err < 0)
1329                         goto errout;
1330
1331                 modified = 1;
1332         }
1333
1334         if (tb[IFLA_ADDRESS]) {
1335                 struct sockaddr *sa;
1336                 int len;
1337
1338                 if (!ops->ndo_set_mac_address) {
1339                         err = -EOPNOTSUPP;
1340                         goto errout;
1341                 }
1342
1343                 if (!netif_device_present(dev)) {
1344                         err = -ENODEV;
1345                         goto errout;
1346                 }
1347
1348                 len = sizeof(sa_family_t) + dev->addr_len;
1349                 sa = kmalloc(len, GFP_KERNEL);
1350                 if (!sa) {
1351                         err = -ENOMEM;
1352                         goto errout;
1353                 }
1354                 sa->sa_family = dev->type;
1355                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1356                        dev->addr_len);
1357                 err = ops->ndo_set_mac_address(dev, sa);
1358                 kfree(sa);
1359                 if (err)
1360                         goto errout;
1361                 send_addr_notify = 1;
1362                 modified = 1;
1363                 add_device_randomness(dev->dev_addr, dev->addr_len);
1364         }
1365
1366         if (tb[IFLA_MTU]) {
1367                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1368                 if (err < 0)
1369                         goto errout;
1370                 modified = 1;
1371         }
1372
1373         if (tb[IFLA_GROUP]) {
1374                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1375                 modified = 1;
1376         }
1377
1378         /*
1379          * Interface selected by interface index but interface
1380          * name provided implies that a name change has been
1381          * requested.
1382          */
1383         if (ifm->ifi_index > 0 && ifname[0]) {
1384                 err = dev_change_name(dev, ifname);
1385                 if (err < 0)
1386                         goto errout;
1387                 modified = 1;
1388         }
1389
1390         if (tb[IFLA_IFALIAS]) {
1391                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1392                                     nla_len(tb[IFLA_IFALIAS]));
1393                 if (err < 0)
1394                         goto errout;
1395                 modified = 1;
1396         }
1397
1398         if (tb[IFLA_BROADCAST]) {
1399                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1400                 send_addr_notify = 1;
1401         }
1402
1403         if (ifm->ifi_flags || ifm->ifi_change) {
1404                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1405                 if (err < 0)
1406                         goto errout;
1407         }
1408
1409         if (tb[IFLA_MASTER]) {
1410                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1411                 if (err)
1412                         goto errout;
1413                 modified = 1;
1414         }
1415
1416         if (tb[IFLA_TXQLEN])
1417                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1418
1419         if (tb[IFLA_OPERSTATE])
1420                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1421
1422         if (tb[IFLA_LINKMODE]) {
1423                 write_lock_bh(&dev_base_lock);
1424                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1425                 write_unlock_bh(&dev_base_lock);
1426         }
1427
1428         if (tb[IFLA_VFINFO_LIST]) {
1429                 struct nlattr *attr;
1430                 int rem;
1431                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1432                         if (nla_type(attr) != IFLA_VF_INFO) {
1433                                 err = -EINVAL;
1434                                 goto errout;
1435                         }
1436                         err = do_setvfinfo(dev, attr);
1437                         if (err < 0)
1438                                 goto errout;
1439                         modified = 1;
1440                 }
1441         }
1442         err = 0;
1443
1444         if (tb[IFLA_VF_PORTS]) {
1445                 struct nlattr *port[IFLA_PORT_MAX+1];
1446                 struct nlattr *attr;
1447                 int vf;
1448                 int rem;
1449
1450                 err = -EOPNOTSUPP;
1451                 if (!ops->ndo_set_vf_port)
1452                         goto errout;
1453
1454                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1455                         if (nla_type(attr) != IFLA_VF_PORT)
1456                                 continue;
1457                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1458                                 attr, ifla_port_policy);
1459                         if (err < 0)
1460                                 goto errout;
1461                         if (!port[IFLA_PORT_VF]) {
1462                                 err = -EOPNOTSUPP;
1463                                 goto errout;
1464                         }
1465                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1466                         err = ops->ndo_set_vf_port(dev, vf, port);
1467                         if (err < 0)
1468                                 goto errout;
1469                         modified = 1;
1470                 }
1471         }
1472         err = 0;
1473
1474         if (tb[IFLA_PORT_SELF]) {
1475                 struct nlattr *port[IFLA_PORT_MAX+1];
1476
1477                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1478                         tb[IFLA_PORT_SELF], ifla_port_policy);
1479                 if (err < 0)
1480                         goto errout;
1481
1482                 err = -EOPNOTSUPP;
1483                 if (ops->ndo_set_vf_port)
1484                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1485                 if (err < 0)
1486                         goto errout;
1487                 modified = 1;
1488         }
1489
1490         if (tb[IFLA_AF_SPEC]) {
1491                 struct nlattr *af;
1492                 int rem;
1493
1494                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1495                         const struct rtnl_af_ops *af_ops;
1496
1497                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1498                                 BUG();
1499
1500                         err = af_ops->set_link_af(dev, af);
1501                         if (err < 0)
1502                                 goto errout;
1503
1504                         modified = 1;
1505                 }
1506         }
1507         err = 0;
1508
1509 errout:
1510         if (err < 0 && modified && net_ratelimit())
1511                 printk(KERN_WARNING "A link change request failed with "
1512                        "some changes committed already. Interface %s may "
1513                        "have been left with an inconsistent configuration, "
1514                        "please check.\n", dev->name);
1515
1516         if (send_addr_notify)
1517                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1518         return err;
1519 }
1520
1521 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1522 {
1523         struct net *net = sock_net(skb->sk);
1524         struct ifinfomsg *ifm;
1525         struct net_device *dev;
1526         int err;
1527         struct nlattr *tb[IFLA_MAX+1];
1528         char ifname[IFNAMSIZ];
1529
1530         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1531         if (err < 0)
1532                 goto errout;
1533
1534         if (tb[IFLA_IFNAME])
1535                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1536         else
1537                 ifname[0] = '\0';
1538
1539         err = -EINVAL;
1540         ifm = nlmsg_data(nlh);
1541         if (ifm->ifi_index > 0)
1542                 dev = __dev_get_by_index(net, ifm->ifi_index);
1543         else if (tb[IFLA_IFNAME])
1544                 dev = __dev_get_by_name(net, ifname);
1545         else
1546                 goto errout;
1547
1548         if (dev == NULL) {
1549                 err = -ENODEV;
1550                 goto errout;
1551         }
1552
1553         err = validate_linkmsg(dev, tb);
1554         if (err < 0)
1555                 goto errout;
1556
1557         err = do_setlink(dev, ifm, tb, ifname, 0);
1558 errout:
1559         return err;
1560 }
1561
1562 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1563 {
1564         struct net *net = sock_net(skb->sk);
1565         const struct rtnl_link_ops *ops;
1566         struct net_device *dev;
1567         struct ifinfomsg *ifm;
1568         char ifname[IFNAMSIZ];
1569         struct nlattr *tb[IFLA_MAX+1];
1570         int err;
1571         LIST_HEAD(list_kill);
1572
1573         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1574         if (err < 0)
1575                 return err;
1576
1577         if (tb[IFLA_IFNAME])
1578                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1579
1580         ifm = nlmsg_data(nlh);
1581         if (ifm->ifi_index > 0)
1582                 dev = __dev_get_by_index(net, ifm->ifi_index);
1583         else if (tb[IFLA_IFNAME])
1584                 dev = __dev_get_by_name(net, ifname);
1585         else
1586                 return -EINVAL;
1587
1588         if (!dev)
1589                 return -ENODEV;
1590
1591         ops = dev->rtnl_link_ops;
1592         if (!ops)
1593                 return -EOPNOTSUPP;
1594
1595         ops->dellink(dev, &list_kill);
1596         unregister_netdevice_many(&list_kill);
1597         list_del(&list_kill);
1598         return 0;
1599 }
1600
1601 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1602 {
1603         unsigned int old_flags;
1604         int err;
1605
1606         old_flags = dev->flags;
1607         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1608                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1609                 if (err < 0)
1610                         return err;
1611         }
1612
1613         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1614         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1615
1616         __dev_notify_flags(dev, old_flags);
1617         return 0;
1618 }
1619 EXPORT_SYMBOL(rtnl_configure_link);
1620
1621 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1622         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1623 {
1624         int err;
1625         struct net_device *dev;
1626         unsigned int num_queues = 1;
1627         unsigned int real_num_queues = 1;
1628
1629         if (ops->get_tx_queues) {
1630                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1631                                          &real_num_queues);
1632                 if (err)
1633                         goto err;
1634         }
1635         err = -ENOMEM;
1636         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1637         if (!dev)
1638                 goto err;
1639
1640         dev_net_set(dev, net);
1641         dev->rtnl_link_ops = ops;
1642         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1643
1644         if (tb[IFLA_MTU])
1645                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1646         if (tb[IFLA_ADDRESS])
1647                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1648                                 nla_len(tb[IFLA_ADDRESS]));
1649         if (tb[IFLA_BROADCAST])
1650                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1651                                 nla_len(tb[IFLA_BROADCAST]));
1652         if (tb[IFLA_TXQLEN])
1653                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1654         if (tb[IFLA_OPERSTATE])
1655                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1656         if (tb[IFLA_LINKMODE])
1657                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1658         if (tb[IFLA_GROUP])
1659                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1660
1661         return dev;
1662
1663 err:
1664         return ERR_PTR(err);
1665 }
1666 EXPORT_SYMBOL(rtnl_create_link);
1667
1668 static int rtnl_group_changelink(struct net *net, int group,
1669                 struct ifinfomsg *ifm,
1670                 struct nlattr **tb)
1671 {
1672         struct net_device *dev;
1673         int err;
1674
1675         for_each_netdev(net, dev) {
1676                 if (dev->group == group) {
1677                         err = do_setlink(dev, ifm, tb, NULL, 0);
1678                         if (err < 0)
1679                                 return err;
1680                 }
1681         }
1682
1683         return 0;
1684 }
1685
1686 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1687 {
1688         struct net *net = sock_net(skb->sk);
1689         const struct rtnl_link_ops *ops;
1690         struct net_device *dev;
1691         struct ifinfomsg *ifm;
1692         char kind[MODULE_NAME_LEN];
1693         char ifname[IFNAMSIZ];
1694         struct nlattr *tb[IFLA_MAX+1];
1695         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1696         int err;
1697
1698 #ifdef CONFIG_MODULES
1699 replay:
1700 #endif
1701         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1702         if (err < 0)
1703                 return err;
1704
1705         if (tb[IFLA_IFNAME])
1706                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1707         else
1708                 ifname[0] = '\0';
1709
1710         ifm = nlmsg_data(nlh);
1711         if (ifm->ifi_index > 0)
1712                 dev = __dev_get_by_index(net, ifm->ifi_index);
1713         else {
1714                 if (ifname[0])
1715                         dev = __dev_get_by_name(net, ifname);
1716                 else
1717                         dev = NULL;
1718         }
1719
1720         err = validate_linkmsg(dev, tb);
1721         if (err < 0)
1722                 return err;
1723
1724         if (tb[IFLA_LINKINFO]) {
1725                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1726                                        tb[IFLA_LINKINFO], ifla_info_policy);
1727                 if (err < 0)
1728                         return err;
1729         } else
1730                 memset(linkinfo, 0, sizeof(linkinfo));
1731
1732         if (linkinfo[IFLA_INFO_KIND]) {
1733                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1734                 ops = rtnl_link_ops_get(kind);
1735         } else {
1736                 kind[0] = '\0';
1737                 ops = NULL;
1738         }
1739
1740         if (1) {
1741                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1742                 struct net *dest_net;
1743
1744                 if (ops) {
1745                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1746                                 err = nla_parse_nested(attr, ops->maxtype,
1747                                                        linkinfo[IFLA_INFO_DATA],
1748                                                        ops->policy);
1749                                 if (err < 0)
1750                                         return err;
1751                                 data = attr;
1752                         }
1753                         if (ops->validate) {
1754                                 err = ops->validate(tb, data);
1755                                 if (err < 0)
1756                                         return err;
1757                         }
1758                 }
1759
1760                 if (dev) {
1761                         int modified = 0;
1762
1763                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1764                                 return -EEXIST;
1765                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1766                                 return -EOPNOTSUPP;
1767
1768                         if (linkinfo[IFLA_INFO_DATA]) {
1769                                 if (!ops || ops != dev->rtnl_link_ops ||
1770                                     !ops->changelink)
1771                                         return -EOPNOTSUPP;
1772
1773                                 err = ops->changelink(dev, tb, data);
1774                                 if (err < 0)
1775                                         return err;
1776                                 modified = 1;
1777                         }
1778
1779                         return do_setlink(dev, ifm, tb, ifname, modified);
1780                 }
1781
1782                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1783                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1784                                 return rtnl_group_changelink(net,
1785                                                 nla_get_u32(tb[IFLA_GROUP]),
1786                                                 ifm, tb);
1787                         return -ENODEV;
1788                 }
1789
1790                 if (ifm->ifi_index)
1791                         return -EOPNOTSUPP;
1792                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1793                         return -EOPNOTSUPP;
1794
1795                 if (!ops) {
1796 #ifdef CONFIG_MODULES
1797                         if (kind[0]) {
1798                                 __rtnl_unlock();
1799                                 request_module("rtnl-link-%s", kind);
1800                                 rtnl_lock();
1801                                 ops = rtnl_link_ops_get(kind);
1802                                 if (ops)
1803                                         goto replay;
1804                         }
1805 #endif
1806                         return -EOPNOTSUPP;
1807                 }
1808
1809                 if (!ifname[0])
1810                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1811
1812                 dest_net = rtnl_link_get_net(net, tb);
1813                 if (IS_ERR(dest_net))
1814                         return PTR_ERR(dest_net);
1815
1816                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1817
1818                 if (IS_ERR(dev))
1819                         err = PTR_ERR(dev);
1820                 else if (ops->newlink)
1821                         err = ops->newlink(net, dev, tb, data);
1822                 else
1823                         err = register_netdevice(dev);
1824
1825                 if (err < 0 && !IS_ERR(dev))
1826                         free_netdev(dev);
1827                 if (err < 0)
1828                         goto out;
1829
1830                 err = rtnl_configure_link(dev, ifm);
1831                 if (err < 0)
1832                         unregister_netdevice(dev);
1833 out:
1834                 put_net(dest_net);
1835                 return err;
1836         }
1837 }
1838
1839 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1840 {
1841         struct net *net = sock_net(skb->sk);
1842         struct ifinfomsg *ifm;
1843         char ifname[IFNAMSIZ];
1844         struct nlattr *tb[IFLA_MAX+1];
1845         struct net_device *dev = NULL;
1846         struct sk_buff *nskb;
1847         int err;
1848
1849         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1850         if (err < 0)
1851                 return err;
1852
1853         if (tb[IFLA_IFNAME])
1854                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1855
1856         ifm = nlmsg_data(nlh);
1857         if (ifm->ifi_index > 0)
1858                 dev = __dev_get_by_index(net, ifm->ifi_index);
1859         else if (tb[IFLA_IFNAME])
1860                 dev = __dev_get_by_name(net, ifname);
1861         else
1862                 return -EINVAL;
1863
1864         if (dev == NULL)
1865                 return -ENODEV;
1866
1867         nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1868         if (nskb == NULL)
1869                 return -ENOBUFS;
1870
1871         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1872                                nlh->nlmsg_seq, 0, 0);
1873         if (err < 0) {
1874                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1875                 WARN_ON(err == -EMSGSIZE);
1876                 kfree_skb(nskb);
1877         } else
1878                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1879
1880         return err;
1881 }
1882
1883 static u16 rtnl_calcit(struct sk_buff *skb)
1884 {
1885         return min_ifinfo_dump_size;
1886 }
1887
1888 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1889 {
1890         int idx;
1891         int s_idx = cb->family;
1892
1893         if (s_idx == 0)
1894                 s_idx = 1;
1895         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1896                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1897                 if (idx < s_idx || idx == PF_PACKET)
1898                         continue;
1899                 if (rtnl_msg_handlers[idx] == NULL ||
1900                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1901                         continue;
1902                 if (idx > s_idx)
1903                         memset(&cb->args[0], 0, sizeof(cb->args));
1904                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1905                         break;
1906         }
1907         cb->family = idx;
1908
1909         return skb->len;
1910 }
1911
1912 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1913 {
1914         struct net *net = dev_net(dev);
1915         struct sk_buff *skb;
1916         int err = -ENOBUFS;
1917         size_t if_info_size;
1918
1919         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1920         if (skb == NULL)
1921                 goto errout;
1922
1923         min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1924
1925         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1926         if (err < 0) {
1927                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1928                 WARN_ON(err == -EMSGSIZE);
1929                 kfree_skb(skb);
1930                 goto errout;
1931         }
1932         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1933         return;
1934 errout:
1935         if (err < 0)
1936                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1937 }
1938
1939 /* Protected by RTNL sempahore.  */
1940 static struct rtattr **rta_buf;
1941 static int rtattr_max;
1942
1943 /* Process one rtnetlink message. */
1944
1945 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1946 {
1947         struct net *net = sock_net(skb->sk);
1948         rtnl_doit_func doit;
1949         int sz_idx, kind;
1950         int min_len;
1951         int family;
1952         int type;
1953         int err;
1954
1955         type = nlh->nlmsg_type;
1956         if (type > RTM_MAX)
1957                 return -EOPNOTSUPP;
1958
1959         type -= RTM_BASE;
1960
1961         /* All the messages must have at least 1 byte length */
1962         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1963                 return 0;
1964
1965         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1966         sz_idx = type>>2;
1967         kind = type&3;
1968
1969         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1970                 return -EPERM;
1971
1972         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1973                 struct sock *rtnl;
1974                 rtnl_dumpit_func dumpit;
1975                 rtnl_calcit_func calcit;
1976                 u16 min_dump_alloc = 0;
1977
1978                 dumpit = rtnl_get_dumpit(family, type);
1979                 if (dumpit == NULL)
1980                         return -EOPNOTSUPP;
1981                 calcit = rtnl_get_calcit(family, type);
1982                 if (calcit)
1983                         min_dump_alloc = calcit(skb);
1984
1985                 __rtnl_unlock();
1986                 rtnl = net->rtnl;
1987                 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
1988                                          NULL, min_dump_alloc);
1989                 rtnl_lock();
1990                 return err;
1991         }
1992
1993         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1994
1995         min_len = rtm_min[sz_idx];
1996         if (nlh->nlmsg_len < min_len)
1997                 return -EINVAL;
1998
1999         if (nlh->nlmsg_len > min_len) {
2000                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2001                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2002
2003                 while (RTA_OK(attr, attrlen)) {
2004                         unsigned flavor = attr->rta_type;
2005                         if (flavor) {
2006                                 if (flavor > rta_max[sz_idx])
2007                                         return -EINVAL;
2008                                 rta_buf[flavor-1] = attr;
2009                         }
2010                         attr = RTA_NEXT(attr, attrlen);
2011                 }
2012         }
2013
2014         doit = rtnl_get_doit(family, type);
2015         if (doit == NULL)
2016                 return -EOPNOTSUPP;
2017
2018         return doit(skb, nlh, (void *)&rta_buf[0]);
2019 }
2020
2021 static void rtnetlink_rcv(struct sk_buff *skb)
2022 {
2023         rtnl_lock();
2024         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2025         rtnl_unlock();
2026 }
2027
2028 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2029 {
2030         struct net_device *dev = ptr;
2031
2032         switch (event) {
2033         case NETDEV_UP:
2034         case NETDEV_DOWN:
2035         case NETDEV_PRE_UP:
2036         case NETDEV_POST_INIT:
2037         case NETDEV_REGISTER:
2038         case NETDEV_CHANGE:
2039         case NETDEV_PRE_TYPE_CHANGE:
2040         case NETDEV_GOING_DOWN:
2041         case NETDEV_UNREGISTER:
2042         case NETDEV_UNREGISTER_BATCH:
2043         case NETDEV_RELEASE:
2044         case NETDEV_JOIN:
2045                 break;
2046         default:
2047                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2048                 break;
2049         }
2050         return NOTIFY_DONE;
2051 }
2052
2053 static struct notifier_block rtnetlink_dev_notifier = {
2054         .notifier_call  = rtnetlink_event,
2055 };
2056
2057
2058 static int __net_init rtnetlink_net_init(struct net *net)
2059 {
2060         struct sock *sk;
2061         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2062                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2063         if (!sk)
2064                 return -ENOMEM;
2065         net->rtnl = sk;
2066         return 0;
2067 }
2068
2069 static void __net_exit rtnetlink_net_exit(struct net *net)
2070 {
2071         netlink_kernel_release(net->rtnl);
2072         net->rtnl = NULL;
2073 }
2074
2075 static struct pernet_operations rtnetlink_net_ops = {
2076         .init = rtnetlink_net_init,
2077         .exit = rtnetlink_net_exit,
2078 };
2079
2080 void __init rtnetlink_init(void)
2081 {
2082         int i;
2083
2084         rtattr_max = 0;
2085         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2086                 if (rta_max[i] > rtattr_max)
2087                         rtattr_max = rta_max[i];
2088         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2089         if (!rta_buf)
2090                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2091
2092         if (register_pernet_subsys(&rtnetlink_net_ops))
2093                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2094
2095         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2096         register_netdevice_notifier(&rtnetlink_dev_notifier);
2097
2098         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2099                       rtnl_dump_ifinfo, rtnl_calcit);
2100         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2101         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2102         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2103
2104         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2105         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2106 }
2107