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                              u32 ext_filter_mask)
 749 {
 750         size_t port_size = nla_total_size(4)            /* PORT_VF */
 751                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 752                 + nla_total_size(sizeof(struct ifla_port_vsi))
 753                                                         /* PORT_VSI_TYPE */
 754                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 755                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 756                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 757                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 758         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 759         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 760                 + port_size;
 761         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 762                 + port_size;
 763
 764         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 765             !(ext_filter_mask & RTEXT_FILTER_VF))
 766                 return 0;
 767         if (dev_num_vf(dev->dev.parent))
 768                 return port_self_size + vf_ports_size +
 769                         vf_port_size * dev_num_vf(dev->dev.parent);
 770         else
 771                 return port_self_size;
 772 }
 773
 774 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 775                                      u32 ext_filter_mask)
 776 {
 777         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 778                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 779                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 780                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 781                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 782                + nla_total_size(sizeof(struct rtnl_link_stats))
 783                + nla_total_size(sizeof(struct rtnl_link_stats64))
 784                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 785                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 786                + nla_total_size(4) /* IFLA_TXQLEN */
 787                + nla_total_size(4) /* IFLA_WEIGHT */
 788                + nla_total_size(4) /* IFLA_MTU */
 789                + nla_total_size(4) /* IFLA_LINK */
 790                + nla_total_size(4) /* IFLA_MASTER */
 791                + nla_total_size(1) /* IFLA_OPERSTATE */
 792                + nla_total_size(1) /* IFLA_LINKMODE */
 793                + nla_total_size(ext_filter_mask
 794                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 795                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 796                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 797                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 798                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 799 }
 800
 801 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 802 {
 803         struct nlattr *vf_ports;
 804         struct nlattr *vf_port;
 805         int vf;
 806         int err;
 807
 808         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 809         if (!vf_ports)
 810                 return -EMSGSIZE;
 811
 812         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 813                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 814                 if (!vf_port)
 815                         goto nla_put_failure;
 816                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 817                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 818                 if (err == -EMSGSIZE)
 819                         goto nla_put_failure;
 820                 if (err) {
 821                         nla_nest_cancel(skb, vf_port);
 822                         continue;
 823                 }
 824                 nla_nest_end(skb, vf_port);
 825         }
 826
 827         nla_nest_end(skb, vf_ports);
 828
 829         return 0;
 830
 831 nla_put_failure:
 832         nla_nest_cancel(skb, vf_ports);
 833         return -EMSGSIZE;
 834 }
 835
 836 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 837 {
 838         struct nlattr *port_self;
 839         int err;
 840
 841         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 842         if (!port_self)
 843                 return -EMSGSIZE;
 844
 845         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 846         if (err) {
 847                 nla_nest_cancel(skb, port_self);
 848                 return (err == -EMSGSIZE) ? err : 0;
 849         }
 850
 851         nla_nest_end(skb, port_self);
 852
 853         return 0;
 854 }
 855
 856 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 857                           u32 ext_filter_mask)
 858 {
 859         int err;
 860
 861         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 862             !(ext_filter_mask & RTEXT_FILTER_VF))
 863                 return 0;
 864
 865         err = rtnl_port_self_fill(skb, dev);
 866         if (err)
 867                 return err;
 868
 869         if (dev_num_vf(dev->dev.parent)) {
 870                 err = rtnl_vf_ports_fill(skb, dev);
 871                 if (err)
 872                         return err;
 873         }
 874
 875         return 0;
 876 }
 877
 878 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 879                             int type, u32 pid, u32 seq, u32 change,
 880                             unsigned int flags, u32 ext_filter_mask)
 881 {
 882         struct ifinfomsg *ifm;
 883         struct nlmsghdr *nlh;
 884         struct rtnl_link_stats64 temp;
 885         const struct rtnl_link_stats64 *stats;
 886         struct nlattr *attr, *af_spec;
 887         struct rtnl_af_ops *af_ops;
 888
 889         ASSERT_RTNL();
 890         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 891         if (nlh == NULL)
 892                 return -EMSGSIZE;
 893
 894         ifm = nlmsg_data(nlh);
 895         ifm->ifi_family = AF_UNSPEC;
 896         ifm->__ifi_pad = 0;
 897         ifm->ifi_type = dev->type;
 898         ifm->ifi_index = dev->ifindex;
 899         ifm->ifi_flags = dev_get_flags(dev);
 900         ifm->ifi_change = change;
 901
 902         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 903         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 904         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 905                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 906         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 907         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 908         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 909
 910         if (dev->ifindex != dev->iflink)
 911                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 912
 913         if (dev->master)
 914                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 915
 916         if (dev->qdisc)
 917                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 918
 919         if (dev->ifalias)
 920                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 921
 922         if (1) {
 923                 struct rtnl_link_ifmap map = {
 924                         .mem_start   = dev->mem_start,
 925                         .mem_end     = dev->mem_end,
 926                         .base_addr   = dev->base_addr,
 927                         .irq         = dev->irq,
 928                         .dma         = dev->dma,
 929                         .port        = dev->if_port,
 930                 };
 931                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 932         }
 933
 934         if (dev->addr_len) {
 935                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 936                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 937         }
 938
 939         attr = nla_reserve(skb, IFLA_STATS,
 940                         sizeof(struct rtnl_link_stats));
 941         if (attr == NULL)
 942                 goto nla_put_failure;
 943
 944         stats = dev_get_stats(dev, &temp);
 945         copy_rtnl_link_stats(nla_data(attr), stats);
 946
 947         attr = nla_reserve(skb, IFLA_STATS64,
 948                         sizeof(struct rtnl_link_stats64));
 949         if (attr == NULL)
 950                 goto nla_put_failure;
 951         copy_rtnl_link_stats64(nla_data(attr), stats);
 952
 953         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF))
 954                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 955
 956         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 957             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 958                 int i;
 959
 960                 struct nlattr *vfinfo, *vf;
 961                 int num_vfs = dev_num_vf(dev->dev.parent);
 962
 963                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 964                 if (!vfinfo)
 965                         goto nla_put_failure;
 966                 for (i = 0; i < num_vfs; i++) {
 967                         struct ifla_vf_info ivi;
 968                         struct ifla_vf_mac vf_mac;
 969                         struct ifla_vf_vlan vf_vlan;
 970                         struct ifla_vf_tx_rate vf_tx_rate;
 971                         struct ifla_vf_spoofchk vf_spoofchk;
 972
 973                         /*
 974                          * Not all SR-IOV capable drivers support the
 975                          * spoofcheck query.  Preset to -1 so the user
 976                          * space tool can detect that the driver didn't
 977                          * report anything.
 978                          */
 979                         ivi.spoofchk = -1;
 980                         memset(ivi.mac, 0, sizeof(ivi.mac));
 981                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 982                                 break;
 983                         vf_mac.vf =
 984                                 vf_vlan.vf =
 985                                 vf_tx_rate.vf =
 986                                 vf_spoofchk.vf = ivi.vf;
 987
 988                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 989                         vf_vlan.vlan = ivi.vlan;
 990                         vf_vlan.qos = ivi.qos;
 991                         vf_tx_rate.rate = ivi.tx_rate;
 992                         vf_spoofchk.setting = ivi.spoofchk;
 993                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 994                         if (!vf) {
 995                                 nla_nest_cancel(skb, vfinfo);
 996                                 goto nla_put_failure;
 997                         }
 998                         NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
 999                         NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
1000                         NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1001                                 &vf_tx_rate);
1002                         NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1003                                 &vf_spoofchk);
1004                         nla_nest_end(skb, vf);
1005                 }
1006                 nla_nest_end(skb, vfinfo);
1007         }
1008
1009         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1010                 goto nla_put_failure;
1011
1012         if (dev->rtnl_link_ops) {
1013                 if (rtnl_link_fill(skb, dev) < 0)
1014                         goto nla_put_failure;
1015         }
1016
1017         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1018                 goto nla_put_failure;
1019
1020         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1021                 if (af_ops->fill_link_af) {
1022                         struct nlattr *af;
1023                         int err;
1024
1025                         if (!(af = nla_nest_start(skb, af_ops->family)))
1026                                 goto nla_put_failure;
1027
1028                         err = af_ops->fill_link_af(skb, dev);
1029
1030                         /*
1031                          * Caller may return ENODATA to indicate that there
1032                          * was no data to be dumped. This is not an error, it
1033                          * means we should trim the attribute header and
1034                          * continue.
1035                          */
1036                         if (err == -ENODATA)
1037                                 nla_nest_cancel(skb, af);
1038                         else if (err < 0)
1039                                 goto nla_put_failure;
1040
1041                         nla_nest_end(skb, af);
1042                 }
1043         }
1044
1045         nla_nest_end(skb, af_spec);
1046
1047         return nlmsg_end(skb, nlh);
1048
1049 nla_put_failure:
1050         nlmsg_cancel(skb, nlh);
1051         return -EMSGSIZE;
1052 }
1053
1054 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1055 {
1056         struct net *net = sock_net(skb->sk);
1057         int h, s_h;
1058         int idx = 0, s_idx;
1059         struct net_device *dev;
1060         struct hlist_head *head;
1061         struct hlist_node *node;
1062         struct nlattr *tb[IFLA_MAX+1];
1063         u32 ext_filter_mask = 0;
1064         int err;
1065         int hdrlen;
1066
1067         s_h = cb->args[0];
1068         s_idx = cb->args[1];
1069
1070         rcu_read_lock();
1071         cb->seq = net->dev_base_seq;
1072
1073         /* A hack to preserve kernel<->userspace interface.
1074          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1075          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1076          * what iproute2 < v3.9.0 used.
1077          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1078          * attribute, its netlink message is shorter than struct ifinfomsg.
1079          */
1080         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1081                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1082
1083         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1084
1085                 if (tb[IFLA_EXT_MASK])
1086                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1087         }
1088
1089         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1090                 idx = 0;
1091                 head = &net->dev_index_head[h];
1092                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1093                         if (idx < s_idx)
1094                                 goto cont;
1095                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1096                                                NETLINK_CB(cb->skb).pid,
1097                                                cb->nlh->nlmsg_seq, 0,
1098                                                NLM_F_MULTI,
1099                                                ext_filter_mask);
1100                         /* If we ran out of room on the first message,
1101                          * we're in trouble
1102                          */
1103                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1104
1105                         if (err <= 0)
1106                                 goto out;
1107
1108                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1109 cont:
1110                         idx++;
1111                 }
1112         }
1113 out:
1114         rcu_read_unlock();
1115         cb->args[1] = idx;
1116         cb->args[0] = h;
1117
1118         return skb->len;
1119 }
1120
1121 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1122         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1123         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1124         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1125         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1126         [IFLA_MTU]              = { .type = NLA_U32 },
1127         [IFLA_LINK]             = { .type = NLA_U32 },
1128         [IFLA_MASTER]           = { .type = NLA_U32 },
1129         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1130         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1131         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1132         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1133         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1134         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1135         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1136         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1137         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1138         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1139         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1140         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1141         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1142 };
1143 EXPORT_SYMBOL(ifla_policy);
1144
1145 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1146         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1147         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1148 };
1149
1150 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1151         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1152 };
1153
1154 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1155         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1156                                     .len = sizeof(struct ifla_vf_mac) },
1157         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1158                                     .len = sizeof(struct ifla_vf_vlan) },
1159         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1160                                     .len = sizeof(struct ifla_vf_tx_rate) },
1161         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1162                                     .len = sizeof(struct ifla_vf_spoofchk) },
1163 };
1164
1165 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1166         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1167         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1168                                     .len = PORT_PROFILE_MAX },
1169         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1170                                     .len = sizeof(struct ifla_port_vsi)},
1171         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1172                                       .len = PORT_UUID_MAX },
1173         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1174                                     .len = PORT_UUID_MAX },
1175         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1176         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1177 };
1178
1179 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1180 {
1181         struct net *net;
1182         /* Examine the link attributes and figure out which
1183          * network namespace we are talking about.
1184          */
1185         if (tb[IFLA_NET_NS_PID])
1186                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1187         else if (tb[IFLA_NET_NS_FD])
1188                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1189         else
1190                 net = get_net(src_net);
1191         return net;
1192 }
1193 EXPORT_SYMBOL(rtnl_link_get_net);
1194
1195 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1196 {
1197         if (dev) {
1198                 if (tb[IFLA_ADDRESS] &&
1199                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1200                         return -EINVAL;
1201
1202                 if (tb[IFLA_BROADCAST] &&
1203                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1204                         return -EINVAL;
1205         }
1206
1207         if (tb[IFLA_AF_SPEC]) {
1208                 struct nlattr *af;
1209                 int rem, err;
1210
1211                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1212                         const struct rtnl_af_ops *af_ops;
1213
1214                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1215                                 return -EAFNOSUPPORT;
1216
1217                         if (!af_ops->set_link_af)
1218                                 return -EOPNOTSUPP;
1219
1220                         if (af_ops->validate_link_af) {
1221                                 err = af_ops->validate_link_af(dev, af);
1222                                 if (err < 0)
1223                                         return err;
1224                         }
1225                 }
1226         }
1227
1228         return 0;
1229 }
1230
1231 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1232 {
1233         int rem, err = -EINVAL;
1234         struct nlattr *vf;
1235         const struct net_device_ops *ops = dev->netdev_ops;
1236
1237         nla_for_each_nested(vf, attr, rem) {
1238                 switch (nla_type(vf)) {
1239                 case IFLA_VF_MAC: {
1240                         struct ifla_vf_mac *ivm;
1241                         ivm = nla_data(vf);
1242                         err = -EOPNOTSUPP;
1243                         if (ops->ndo_set_vf_mac)
1244                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1245                                                           ivm->mac);
1246                         break;
1247                 }
1248                 case IFLA_VF_VLAN: {
1249                         struct ifla_vf_vlan *ivv;
1250                         ivv = nla_data(vf);
1251                         err = -EOPNOTSUPP;
1252                         if (ops->ndo_set_vf_vlan)
1253                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1254                                                            ivv->vlan,
1255                                                            ivv->qos);
1256                         break;
1257                 }
1258                 case IFLA_VF_TX_RATE: {
1259                         struct ifla_vf_tx_rate *ivt;
1260                         ivt = nla_data(vf);
1261                         err = -EOPNOTSUPP;
1262                         if (ops->ndo_set_vf_tx_rate)
1263                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1264                                                               ivt->rate);
1265                         break;
1266                 }
1267                 case IFLA_VF_SPOOFCHK: {
1268                         struct ifla_vf_spoofchk *ivs;
1269                         ivs = nla_data(vf);
1270                         err = -EOPNOTSUPP;
1271                         if (ops->ndo_set_vf_spoofchk)
1272                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1273                                                                ivs->setting);
1274                         break;
1275                 }
1276                 default:
1277                         err = -EINVAL;
1278                         break;
1279                 }
1280                 if (err)
1281                         break;
1282         }
1283         return err;
1284 }
1285
1286 static int do_set_master(struct net_device *dev, int ifindex)
1287 {
1288         struct net_device *master_dev;
1289         const struct net_device_ops *ops;
1290         int err;
1291
1292         if (dev->master) {
1293                 if (dev->master->ifindex == ifindex)
1294                         return 0;
1295                 ops = dev->master->netdev_ops;
1296                 if (ops->ndo_del_slave) {
1297                         err = ops->ndo_del_slave(dev->master, dev);
1298                         if (err)
1299                                 return err;
1300                 } else {
1301                         return -EOPNOTSUPP;
1302                 }
1303         }
1304
1305         if (ifindex) {
1306                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1307                 if (!master_dev)
1308                         return -EINVAL;
1309                 ops = master_dev->netdev_ops;
1310                 if (ops->ndo_add_slave) {
1311                         err = ops->ndo_add_slave(master_dev, dev);
1312                         if (err)
1313                                 return err;
1314                 } else {
1315                         return -EOPNOTSUPP;
1316                 }
1317         }
1318         return 0;
1319 }
1320
1321 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1322                       struct nlattr **tb, char *ifname, int modified)
1323 {
1324         const struct net_device_ops *ops = dev->netdev_ops;
1325         int send_addr_notify = 0;
1326         int err;
1327
1328         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1329                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1330                 if (IS_ERR(net)) {
1331                         err = PTR_ERR(net);
1332                         goto errout;
1333                 }
1334                 err = dev_change_net_namespace(dev, net, ifname);
1335                 put_net(net);
1336                 if (err)
1337                         goto errout;
1338                 modified = 1;
1339         }
1340
1341         if (tb[IFLA_MAP]) {
1342                 struct rtnl_link_ifmap *u_map;
1343                 struct ifmap k_map;
1344
1345                 if (!ops->ndo_set_config) {
1346                         err = -EOPNOTSUPP;
1347                         goto errout;
1348                 }
1349
1350                 if (!netif_device_present(dev)) {
1351                         err = -ENODEV;
1352                         goto errout;
1353                 }
1354
1355                 u_map = nla_data(tb[IFLA_MAP]);
1356                 k_map.mem_start = (unsigned long) u_map->mem_start;
1357                 k_map.mem_end = (unsigned long) u_map->mem_end;
1358                 k_map.base_addr = (unsigned short) u_map->base_addr;
1359                 k_map.irq = (unsigned char) u_map->irq;
1360                 k_map.dma = (unsigned char) u_map->dma;
1361                 k_map.port = (unsigned char) u_map->port;
1362
1363                 err = ops->ndo_set_config(dev, &k_map);
1364                 if (err < 0)
1365                         goto errout;
1366
1367                 modified = 1;
1368         }
1369
1370         if (tb[IFLA_ADDRESS]) {
1371                 struct sockaddr *sa;
1372                 int len;
1373
1374                 if (!ops->ndo_set_mac_address) {
1375                         err = -EOPNOTSUPP;
1376                         goto errout;
1377                 }
1378
1379                 if (!netif_device_present(dev)) {
1380                         err = -ENODEV;
1381                         goto errout;
1382                 }
1383
1384                 len = sizeof(sa_family_t) + dev->addr_len;
1385                 sa = kmalloc(len, GFP_KERNEL);
1386                 if (!sa) {
1387                         err = -ENOMEM;
1388                         goto errout;
1389                 }
1390                 sa->sa_family = dev->type;
1391                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1392                        dev->addr_len);
1393                 err = ops->ndo_set_mac_address(dev, sa);
1394                 kfree(sa);
1395                 if (err)
1396                         goto errout;
1397                 send_addr_notify = 1;
1398                 modified = 1;
1399                 add_device_randomness(dev->dev_addr, dev->addr_len);
1400         }
1401
1402         if (tb[IFLA_MTU]) {
1403                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1404                 if (err < 0)
1405                         goto errout;
1406                 modified = 1;
1407         }
1408
1409         if (tb[IFLA_GROUP]) {
1410                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1411                 modified = 1;
1412         }
1413
1414         /*
1415          * Interface selected by interface index but interface
1416          * name provided implies that a name change has been
1417          * requested.
1418          */
1419         if (ifm->ifi_index > 0 && ifname[0]) {
1420                 err = dev_change_name(dev, ifname);
1421                 if (err < 0)
1422                         goto errout;
1423                 modified = 1;
1424         }
1425
1426         if (tb[IFLA_IFALIAS]) {
1427                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1428                                     nla_len(tb[IFLA_IFALIAS]));
1429                 if (err < 0)
1430                         goto errout;
1431                 modified = 1;
1432         }
1433
1434         if (tb[IFLA_BROADCAST]) {
1435                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1436                 send_addr_notify = 1;
1437         }
1438
1439         if (ifm->ifi_flags || ifm->ifi_change) {
1440                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1441                 if (err < 0)
1442                         goto errout;
1443         }
1444
1445         if (tb[IFLA_MASTER]) {
1446                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1447                 if (err)
1448                         goto errout;
1449                 modified = 1;
1450         }
1451
1452         if (tb[IFLA_TXQLEN])
1453                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1454
1455         if (tb[IFLA_OPERSTATE])
1456                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1457
1458         if (tb[IFLA_LINKMODE]) {
1459                 write_lock_bh(&dev_base_lock);
1460                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1461                 write_unlock_bh(&dev_base_lock);
1462         }
1463
1464         if (tb[IFLA_VFINFO_LIST]) {
1465                 struct nlattr *attr;
1466                 int rem;
1467                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1468                         if (nla_type(attr) != IFLA_VF_INFO) {
1469                                 err = -EINVAL;
1470                                 goto errout;
1471                         }
1472                         err = do_setvfinfo(dev, attr);
1473                         if (err < 0)
1474                                 goto errout;
1475                         modified = 1;
1476                 }
1477         }
1478         err = 0;
1479
1480         if (tb[IFLA_VF_PORTS]) {
1481                 struct nlattr *port[IFLA_PORT_MAX+1];
1482                 struct nlattr *attr;
1483                 int vf;
1484                 int rem;
1485
1486                 err = -EOPNOTSUPP;
1487                 if (!ops->ndo_set_vf_port)
1488                         goto errout;
1489
1490                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1491                         if (nla_type(attr) != IFLA_VF_PORT)
1492                                 continue;
1493                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1494                                 attr, ifla_port_policy);
1495                         if (err < 0)
1496                                 goto errout;
1497                         if (!port[IFLA_PORT_VF]) {
1498                                 err = -EOPNOTSUPP;
1499                                 goto errout;
1500                         }
1501                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1502                         err = ops->ndo_set_vf_port(dev, vf, port);
1503                         if (err < 0)
1504                                 goto errout;
1505                         modified = 1;
1506                 }
1507         }
1508         err = 0;
1509
1510         if (tb[IFLA_PORT_SELF]) {
1511                 struct nlattr *port[IFLA_PORT_MAX+1];
1512
1513                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1514                         tb[IFLA_PORT_SELF], ifla_port_policy);
1515                 if (err < 0)
1516                         goto errout;
1517
1518                 err = -EOPNOTSUPP;
1519                 if (ops->ndo_set_vf_port)
1520                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1521                 if (err < 0)
1522                         goto errout;
1523                 modified = 1;
1524         }
1525
1526         if (tb[IFLA_AF_SPEC]) {
1527                 struct nlattr *af;
1528                 int rem;
1529
1530                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1531                         const struct rtnl_af_ops *af_ops;
1532
1533                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1534                                 BUG();
1535
1536                         err = af_ops->set_link_af(dev, af);
1537                         if (err < 0)
1538                                 goto errout;
1539
1540                         modified = 1;
1541                 }
1542         }
1543         err = 0;
1544
1545 errout:
1546         if (err < 0 && modified && net_ratelimit())
1547                 printk(KERN_WARNING "A link change request failed with "
1548                        "some changes committed already. Interface %s may "
1549                        "have been left with an inconsistent configuration, "
1550                        "please check.\n", dev->name);
1551
1552         if (send_addr_notify)
1553                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1554         return err;
1555 }
1556
1557 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1558 {
1559         struct net *net = sock_net(skb->sk);
1560         struct ifinfomsg *ifm;
1561         struct net_device *dev;
1562         int err;
1563         struct nlattr *tb[IFLA_MAX+1];
1564         char ifname[IFNAMSIZ];
1565
1566         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1567         if (err < 0)
1568                 goto errout;
1569
1570         if (tb[IFLA_IFNAME])
1571                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1572         else
1573                 ifname[0] = '\0';
1574
1575         err = -EINVAL;
1576         ifm = nlmsg_data(nlh);
1577         if (ifm->ifi_index > 0)
1578                 dev = __dev_get_by_index(net, ifm->ifi_index);
1579         else if (tb[IFLA_IFNAME])
1580                 dev = __dev_get_by_name(net, ifname);
1581         else
1582                 goto errout;
1583
1584         if (dev == NULL) {
1585                 err = -ENODEV;
1586                 goto errout;
1587         }
1588
1589         err = validate_linkmsg(dev, tb);
1590         if (err < 0)
1591                 goto errout;
1592
1593         err = do_setlink(dev, ifm, tb, ifname, 0);
1594 errout:
1595         return err;
1596 }
1597
1598 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1599 {
1600         struct net *net = sock_net(skb->sk);
1601         const struct rtnl_link_ops *ops;
1602         struct net_device *dev;
1603         struct ifinfomsg *ifm;
1604         char ifname[IFNAMSIZ];
1605         struct nlattr *tb[IFLA_MAX+1];
1606         int err;
1607         LIST_HEAD(list_kill);
1608
1609         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1610         if (err < 0)
1611                 return err;
1612
1613         if (tb[IFLA_IFNAME])
1614                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1615
1616         ifm = nlmsg_data(nlh);
1617         if (ifm->ifi_index > 0)
1618                 dev = __dev_get_by_index(net, ifm->ifi_index);
1619         else if (tb[IFLA_IFNAME])
1620                 dev = __dev_get_by_name(net, ifname);
1621         else
1622                 return -EINVAL;
1623
1624         if (!dev)
1625                 return -ENODEV;
1626
1627         ops = dev->rtnl_link_ops;
1628         if (!ops)
1629                 return -EOPNOTSUPP;
1630
1631         ops->dellink(dev, &list_kill);
1632         unregister_netdevice_many(&list_kill);
1633         list_del(&list_kill);
1634         return 0;
1635 }
1636
1637 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1638 {
1639         unsigned int old_flags;
1640         int err;
1641
1642         old_flags = dev->flags;
1643         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1644                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1645                 if (err < 0)
1646                         return err;
1647         }
1648
1649         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1650         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1651
1652         __dev_notify_flags(dev, old_flags);
1653         return 0;
1654 }
1655 EXPORT_SYMBOL(rtnl_configure_link);
1656
1657 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1658         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1659 {
1660         int err;
1661         struct net_device *dev;
1662         unsigned int num_queues = 1;
1663         unsigned int real_num_queues = 1;
1664
1665         if (ops->get_tx_queues) {
1666                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1667                                          &real_num_queues);
1668                 if (err)
1669                         goto err;
1670         }
1671         err = -ENOMEM;
1672         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1673         if (!dev)
1674                 goto err;
1675
1676         dev_net_set(dev, net);
1677         dev->rtnl_link_ops = ops;
1678         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1679
1680         if (tb[IFLA_MTU])
1681                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1682         if (tb[IFLA_ADDRESS])
1683                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1684                                 nla_len(tb[IFLA_ADDRESS]));
1685         if (tb[IFLA_BROADCAST])
1686                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1687                                 nla_len(tb[IFLA_BROADCAST]));
1688         if (tb[IFLA_TXQLEN])
1689                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1690         if (tb[IFLA_OPERSTATE])
1691                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1692         if (tb[IFLA_LINKMODE])
1693                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1694         if (tb[IFLA_GROUP])
1695                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1696
1697         return dev;
1698
1699 err:
1700         return ERR_PTR(err);
1701 }
1702 EXPORT_SYMBOL(rtnl_create_link);
1703
1704 static int rtnl_group_changelink(struct net *net, int group,
1705                 struct ifinfomsg *ifm,
1706                 struct nlattr **tb)
1707 {
1708         struct net_device *dev;
1709         int err;
1710
1711         for_each_netdev(net, dev) {
1712                 if (dev->group == group) {
1713                         err = do_setlink(dev, ifm, tb, NULL, 0);
1714                         if (err < 0)
1715                                 return err;
1716                 }
1717         }
1718
1719         return 0;
1720 }
1721
1722 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1723 {
1724         struct net *net = sock_net(skb->sk);
1725         const struct rtnl_link_ops *ops;
1726         struct net_device *dev;
1727         struct ifinfomsg *ifm;
1728         char kind[MODULE_NAME_LEN];
1729         char ifname[IFNAMSIZ];
1730         struct nlattr *tb[IFLA_MAX+1];
1731         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1732         int err;
1733
1734 #ifdef CONFIG_MODULES
1735 replay:
1736 #endif
1737         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1738         if (err < 0)
1739                 return err;
1740
1741         if (tb[IFLA_IFNAME])
1742                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1743         else
1744                 ifname[0] = '\0';
1745
1746         ifm = nlmsg_data(nlh);
1747         if (ifm->ifi_index > 0)
1748                 dev = __dev_get_by_index(net, ifm->ifi_index);
1749         else {
1750                 if (ifname[0])
1751                         dev = __dev_get_by_name(net, ifname);
1752                 else
1753                         dev = NULL;
1754         }
1755
1756         err = validate_linkmsg(dev, tb);
1757         if (err < 0)
1758                 return err;
1759
1760         if (tb[IFLA_LINKINFO]) {
1761                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1762                                        tb[IFLA_LINKINFO], ifla_info_policy);
1763                 if (err < 0)
1764                         return err;
1765         } else
1766                 memset(linkinfo, 0, sizeof(linkinfo));
1767
1768         if (linkinfo[IFLA_INFO_KIND]) {
1769                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1770                 ops = rtnl_link_ops_get(kind);
1771         } else {
1772                 kind[0] = '\0';
1773                 ops = NULL;
1774         }
1775
1776         if (1) {
1777                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1778                 struct net *dest_net;
1779
1780                 if (ops) {
1781                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1782                                 err = nla_parse_nested(attr, ops->maxtype,
1783                                                        linkinfo[IFLA_INFO_DATA],
1784                                                        ops->policy);
1785                                 if (err < 0)
1786                                         return err;
1787                                 data = attr;
1788                         }
1789                         if (ops->validate) {
1790                                 err = ops->validate(tb, data);
1791                                 if (err < 0)
1792                                         return err;
1793                         }
1794                 }
1795
1796                 if (dev) {
1797                         int modified = 0;
1798
1799                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1800                                 return -EEXIST;
1801                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1802                                 return -EOPNOTSUPP;
1803
1804                         if (linkinfo[IFLA_INFO_DATA]) {
1805                                 if (!ops || ops != dev->rtnl_link_ops ||
1806                                     !ops->changelink)
1807                                         return -EOPNOTSUPP;
1808
1809                                 err = ops->changelink(dev, tb, data);
1810                                 if (err < 0)
1811                                         return err;
1812                                 modified = 1;
1813                         }
1814
1815                         return do_setlink(dev, ifm, tb, ifname, modified);
1816                 }
1817
1818                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1819                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1820                                 return rtnl_group_changelink(net,
1821                                                 nla_get_u32(tb[IFLA_GROUP]),
1822                                                 ifm, tb);
1823                         return -ENODEV;
1824                 }
1825
1826                 if (ifm->ifi_index)
1827                         return -EOPNOTSUPP;
1828                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1829                         return -EOPNOTSUPP;
1830
1831                 if (!ops) {
1832 #ifdef CONFIG_MODULES
1833                         if (kind[0]) {
1834                                 __rtnl_unlock();
1835                                 request_module("rtnl-link-%s", kind);
1836                                 rtnl_lock();
1837                                 ops = rtnl_link_ops_get(kind);
1838                                 if (ops)
1839                                         goto replay;
1840                         }
1841 #endif
1842                         return -EOPNOTSUPP;
1843                 }
1844
1845                 if (!ifname[0])
1846                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1847
1848                 dest_net = rtnl_link_get_net(net, tb);
1849                 if (IS_ERR(dest_net))
1850                         return PTR_ERR(dest_net);
1851
1852                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1853
1854                 if (IS_ERR(dev))
1855                         err = PTR_ERR(dev);
1856                 else if (ops->newlink)
1857                         err = ops->newlink(net, dev, tb, data);
1858                 else
1859                         err = register_netdevice(dev);
1860
1861                 if (err < 0 && !IS_ERR(dev))
1862                         free_netdev(dev);
1863                 if (err < 0)
1864                         goto out;
1865
1866                 err = rtnl_configure_link(dev, ifm);
1867                 if (err < 0)
1868                         unregister_netdevice(dev);
1869 out:
1870                 put_net(dest_net);
1871                 return err;
1872         }
1873 }
1874
1875 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1876 {
1877         struct net *net = sock_net(skb->sk);
1878         struct ifinfomsg *ifm;
1879         char ifname[IFNAMSIZ];
1880         struct nlattr *tb[IFLA_MAX+1];
1881         struct net_device *dev = NULL;
1882         struct sk_buff *nskb;
1883         int err;
1884         u32 ext_filter_mask = 0;
1885
1886         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1887         if (err < 0)
1888                 return err;
1889
1890         if (tb[IFLA_IFNAME])
1891                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1892
1893         if (tb[IFLA_EXT_MASK])
1894                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1895
1896         ifm = nlmsg_data(nlh);
1897         if (ifm->ifi_index > 0)
1898                 dev = __dev_get_by_index(net, ifm->ifi_index);
1899         else if (tb[IFLA_IFNAME])
1900                 dev = __dev_get_by_name(net, ifname);
1901         else
1902                 return -EINVAL;
1903
1904         if (dev == NULL)
1905                 return -ENODEV;
1906
1907         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1908         if (nskb == NULL)
1909                 return -ENOBUFS;
1910
1911         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1912                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1913         if (err < 0) {
1914                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1915                 WARN_ON(err == -EMSGSIZE);
1916                 kfree_skb(nskb);
1917         } else
1918                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1919
1920         return err;
1921 }
1922
1923 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1924 {
1925         struct net *net = sock_net(skb->sk);
1926         struct net_device *dev;
1927         struct nlattr *tb[IFLA_MAX+1];
1928         u32 ext_filter_mask = 0;
1929         u16 min_ifinfo_dump_size = 0;
1930         int hdrlen;
1931
1932         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
1933         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1934                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1935
1936         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1937                 if (tb[IFLA_EXT_MASK])
1938                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1939         }
1940
1941         if (!ext_filter_mask)
1942                 return NLMSG_GOODSIZE;
1943         /*
1944          * traverse the list of net devices and compute the minimum
1945          * buffer size based upon the filter mask.
1946          */
1947         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1948                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1949                                              if_nlmsg_size(dev,
1950                                                            ext_filter_mask));
1951         }
1952
1953         return min_ifinfo_dump_size;
1954 }
1955
1956 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1957 {
1958         int idx;
1959         int s_idx = cb->family;
1960
1961         if (s_idx == 0)
1962                 s_idx = 1;
1963         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1964                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1965                 if (idx < s_idx || idx == PF_PACKET)
1966                         continue;
1967                 if (rtnl_msg_handlers[idx] == NULL ||
1968                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1969                         continue;
1970                 if (idx > s_idx)
1971                         memset(&cb->args[0], 0, sizeof(cb->args));
1972                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1973                         break;
1974         }
1975         cb->family = idx;
1976
1977         return skb->len;
1978 }
1979
1980 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1981 {
1982         struct net *net = dev_net(dev);
1983         struct sk_buff *skb;
1984         int err = -ENOBUFS;
1985         size_t if_info_size;
1986
1987         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1988         if (skb == NULL)
1989                 goto errout;
1990
1991         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1992         if (err < 0) {
1993                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1994                 WARN_ON(err == -EMSGSIZE);
1995                 kfree_skb(skb);
1996                 goto errout;
1997         }
1998         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1999         return;
2000 errout:
2001         if (err < 0)
2002                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2003 }
2004
2005 /* Protected by RTNL sempahore.  */
2006 static struct rtattr **rta_buf;
2007 static int rtattr_max;
2008
2009 /* Process one rtnetlink message. */
2010
2011 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2012 {
2013         struct net *net = sock_net(skb->sk);
2014         rtnl_doit_func doit;
2015         int sz_idx, kind;
2016         int min_len;
2017         int family;
2018         int type;
2019         int err;
2020
2021         type = nlh->nlmsg_type;
2022         if (type > RTM_MAX)
2023                 return -EOPNOTSUPP;
2024
2025         type -= RTM_BASE;
2026
2027         /* All the messages must have at least 1 byte length */
2028         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2029                 return 0;
2030
2031         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2032         sz_idx = type>>2;
2033         kind = type&3;
2034
2035         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
2036                 return -EPERM;
2037
2038         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2039                 struct sock *rtnl;
2040                 rtnl_dumpit_func dumpit;
2041                 rtnl_calcit_func calcit;
2042                 u16 min_dump_alloc = 0;
2043
2044                 dumpit = rtnl_get_dumpit(family, type);
2045                 if (dumpit == NULL)
2046                         return -EOPNOTSUPP;
2047                 calcit = rtnl_get_calcit(family, type);
2048                 if (calcit)
2049                         min_dump_alloc = calcit(skb, nlh);
2050
2051                 __rtnl_unlock();
2052                 rtnl = net->rtnl;
2053                 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
2054                                          NULL, min_dump_alloc);
2055                 rtnl_lock();
2056                 return err;
2057         }
2058
2059         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2060
2061         min_len = rtm_min[sz_idx];
2062         if (nlh->nlmsg_len < min_len)
2063                 return -EINVAL;
2064
2065         if (nlh->nlmsg_len > min_len) {
2066                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2067                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2068
2069                 while (RTA_OK(attr, attrlen)) {
2070                         unsigned int flavor = attr->rta_type & NLA_TYPE_MASK;
2071                         if (flavor) {
2072                                 if (flavor > rta_max[sz_idx])
2073                                         return -EINVAL;
2074                                 rta_buf[flavor-1] = attr;
2075                         }
2076                         attr = RTA_NEXT(attr, attrlen);
2077                 }
2078         }
2079
2080         doit = rtnl_get_doit(family, type);
2081         if (doit == NULL)
2082                 return -EOPNOTSUPP;
2083
2084         return doit(skb, nlh, (void *)&rta_buf[0]);
2085 }
2086
2087 static void rtnetlink_rcv(struct sk_buff *skb)
2088 {
2089         rtnl_lock();
2090         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2091         rtnl_unlock();
2092 }
2093
2094 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2095 {
2096         struct net_device *dev = ptr;
2097
2098         switch (event) {
2099         case NETDEV_UP:
2100         case NETDEV_DOWN:
2101         case NETDEV_PRE_UP:
2102         case NETDEV_POST_INIT:
2103         case NETDEV_REGISTER:
2104         case NETDEV_CHANGE:
2105         case NETDEV_PRE_TYPE_CHANGE:
2106         case NETDEV_GOING_DOWN:
2107         case NETDEV_UNREGISTER:
2108         case NETDEV_UNREGISTER_BATCH:
2109         case NETDEV_RELEASE:
2110         case NETDEV_JOIN:
2111                 break;
2112         default:
2113                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2114                 break;
2115         }
2116         return NOTIFY_DONE;
2117 }
2118
2119 static struct notifier_block rtnetlink_dev_notifier = {
2120         .notifier_call  = rtnetlink_event,
2121 };
2122
2123
2124 static int __net_init rtnetlink_net_init(struct net *net)
2125 {
2126         struct sock *sk;
2127         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2128                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2129         if (!sk)
2130                 return -ENOMEM;
2131         net->rtnl = sk;
2132         return 0;
2133 }
2134
2135 static void __net_exit rtnetlink_net_exit(struct net *net)
2136 {
2137         netlink_kernel_release(net->rtnl);
2138         net->rtnl = NULL;
2139 }
2140
2141 static struct pernet_operations rtnetlink_net_ops = {
2142         .init = rtnetlink_net_init,
2143         .exit = rtnetlink_net_exit,
2144 };
2145
2146 void __init rtnetlink_init(void)
2147 {
2148         int i;
2149
2150         rtattr_max = 0;
2151         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2152                 if (rta_max[i] > rtattr_max)
2153                         rtattr_max = rta_max[i];
2154         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2155         if (!rta_buf)
2156                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2157
2158         if (register_pernet_subsys(&rtnetlink_net_ops))
2159                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2160
2161         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2162         register_netdevice_notifier(&rtnetlink_dev_notifier);
2163
2164         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2165                       rtnl_dump_ifinfo, rtnl_calcit);
2166         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2167         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2168         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2169
2170         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2171         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2172 }
2173