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/if_bridge.h>
  39 #include <linux/pci.h>
  40 #include <linux/etherdevice.h>
  41
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 struct rtnl_link {
  58         rtnl_doit_func          doit;
  59         rtnl_dumpit_func        dumpit;
  60         rtnl_calcit_func        calcit;
  61 };
  62
  63 static DEFINE_MUTEX(rtnl_mutex);
  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 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277 {
 278         const struct rtnl_link_ops *ops;
 279
 280         list_for_each_entry(ops, &link_ops, list) {
 281                 if (!strcmp(ops->kind, kind))
 282                         return ops;
 283         }
 284         return NULL;
 285 }
 286
 287 /**
 288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289  * @ops: struct rtnl_link_ops * to register
 290  *
 291  * The caller must hold the rtnl_mutex. This function should be used
 292  * by drivers that create devices during module initialization. It
 293  * must be called before registering the devices.
 294  *
 295  * Returns 0 on success or a negative error code.
 296  */
 297 int __rtnl_link_register(struct rtnl_link_ops *ops)
 298 {
 299         if (rtnl_link_ops_get(ops->kind))
 300                 return -EEXIST;
 301
 302         if (!ops->dellink)
 303                 ops->dellink = unregister_netdevice_queue;
 304
 305         list_add_tail(&ops->list, &link_ops);
 306         return 0;
 307 }
 308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 309
 310 /**
 311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 312  * @ops: struct rtnl_link_ops * to register
 313  *
 314  * Returns 0 on success or a negative error code.
 315  */
 316 int rtnl_link_register(struct rtnl_link_ops *ops)
 317 {
 318         int err;
 319
 320         rtnl_lock();
 321         err = __rtnl_link_register(ops);
 322         rtnl_unlock();
 323         return err;
 324 }
 325 EXPORT_SYMBOL_GPL(rtnl_link_register);
 326
 327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 328 {
 329         struct net_device *dev;
 330         LIST_HEAD(list_kill);
 331
 332         for_each_netdev(net, dev) {
 333                 if (dev->rtnl_link_ops == ops)
 334                         ops->dellink(dev, &list_kill);
 335         }
 336         unregister_netdevice_many(&list_kill);
 337 }
 338
 339 /**
 340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 341  * @ops: struct rtnl_link_ops * to unregister
 342  *
 343  * The caller must hold the rtnl_mutex.
 344  */
 345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 346 {
 347         struct net *net;
 348
 349         for_each_net(net) {
 350                 __rtnl_kill_links(net, ops);
 351         }
 352         list_del(&ops->list);
 353 }
 354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 355
 356 /**
 357  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 358  * @ops: struct rtnl_link_ops * to unregister
 359  */
 360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 361 {
 362         rtnl_lock();
 363         __rtnl_link_unregister(ops);
 364         rtnl_unlock();
 365 }
 366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 367
 368 static size_t rtnl_link_get_size(const struct net_device *dev)
 369 {
 370         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 371         size_t size;
 372
 373         if (!ops)
 374                 return 0;
 375
 376         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 377                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 378
 379         if (ops->get_size)
 380                 /* IFLA_INFO_DATA + nested data */
 381                 size += nla_total_size(sizeof(struct nlattr)) +
 382                         ops->get_size(dev);
 383
 384         if (ops->get_xstats_size)
 385                 /* IFLA_INFO_XSTATS */
 386                 size += nla_total_size(ops->get_xstats_size(dev));
 387
 388         return size;
 389 }
 390
 391 static LIST_HEAD(rtnl_af_ops);
 392
 393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 394 {
 395         const struct rtnl_af_ops *ops;
 396
 397         list_for_each_entry(ops, &rtnl_af_ops, list) {
 398                 if (ops->family == family)
 399                         return ops;
 400         }
 401
 402         return NULL;
 403 }
 404
 405 /**
 406  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 407  * @ops: struct rtnl_af_ops * to register
 408  *
 409  * The caller must hold the rtnl_mutex.
 410  *
 411  * Returns 0 on success or a negative error code.
 412  */
 413 int __rtnl_af_register(struct rtnl_af_ops *ops)
 414 {
 415         list_add_tail(&ops->list, &rtnl_af_ops);
 416         return 0;
 417 }
 418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 419
 420 /**
 421  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 422  * @ops: struct rtnl_af_ops * to register
 423  *
 424  * Returns 0 on success or a negative error code.
 425  */
 426 int rtnl_af_register(struct rtnl_af_ops *ops)
 427 {
 428         int err;
 429
 430         rtnl_lock();
 431         err = __rtnl_af_register(ops);
 432         rtnl_unlock();
 433         return err;
 434 }
 435 EXPORT_SYMBOL_GPL(rtnl_af_register);
 436
 437 /**
 438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 439  * @ops: struct rtnl_af_ops * to unregister
 440  *
 441  * The caller must hold the rtnl_mutex.
 442  */
 443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 444 {
 445         list_del(&ops->list);
 446 }
 447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 448
 449 /**
 450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 451  * @ops: struct rtnl_af_ops * to unregister
 452  */
 453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 454 {
 455         rtnl_lock();
 456         __rtnl_af_unregister(ops);
 457         rtnl_unlock();
 458 }
 459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 460
 461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 462 {
 463         struct rtnl_af_ops *af_ops;
 464         size_t size;
 465
 466         /* IFLA_AF_SPEC */
 467         size = nla_total_size(sizeof(struct nlattr));
 468
 469         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 470                 if (af_ops->get_link_af_size) {
 471                         /* AF_* + nested data */
 472                         size += nla_total_size(sizeof(struct nlattr)) +
 473                                 af_ops->get_link_af_size(dev);
 474                 }
 475         }
 476
 477         return size;
 478 }
 479
 480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 481 {
 482         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 483         struct nlattr *linkinfo, *data;
 484         int err = -EMSGSIZE;
 485
 486         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 487         if (linkinfo == NULL)
 488                 goto out;
 489
 490         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 491                 goto err_cancel_link;
 492         if (ops->fill_xstats) {
 493                 err = ops->fill_xstats(skb, dev);
 494                 if (err < 0)
 495                         goto err_cancel_link;
 496         }
 497         if (ops->fill_info) {
 498                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 499                 if (data == NULL)
 500                         goto err_cancel_link;
 501                 err = ops->fill_info(skb, dev);
 502                 if (err < 0)
 503                         goto err_cancel_data;
 504                 nla_nest_end(skb, data);
 505         }
 506
 507         nla_nest_end(skb, linkinfo);
 508         return 0;
 509
 510 err_cancel_data:
 511         nla_nest_cancel(skb, data);
 512 err_cancel_link:
 513         nla_nest_cancel(skb, linkinfo);
 514 out:
 515         return err;
 516 }
 517
 518 static const int rtm_min[RTM_NR_FAMILIES] =
 519 {
 520         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 521         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 522         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 523         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 524         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 527         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 528         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 530 };
 531
 532 static const int rta_max[RTM_NR_FAMILIES] =
 533 {
 534         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 535         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 536         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 537         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 538         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 539         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 540         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 541         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 542 };
 543
 544 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 545 {
 546         struct sock *rtnl = net->rtnl;
 547         int err = 0;
 548
 549         NETLINK_CB(skb).dst_group = group;
 550         if (echo)
 551                 atomic_inc(&skb->users);
 552         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 553         if (echo)
 554                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 555         return err;
 556 }
 557
 558 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 559 {
 560         struct sock *rtnl = net->rtnl;
 561
 562         return nlmsg_unicast(rtnl, skb, pid);
 563 }
 564 EXPORT_SYMBOL(rtnl_unicast);
 565
 566 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 567                  struct nlmsghdr *nlh, gfp_t flags)
 568 {
 569         struct sock *rtnl = net->rtnl;
 570         int report = 0;
 571
 572         if (nlh)
 573                 report = nlmsg_report(nlh);
 574
 575         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 576 }
 577 EXPORT_SYMBOL(rtnl_notify);
 578
 579 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 580 {
 581         struct sock *rtnl = net->rtnl;
 582
 583         netlink_set_err(rtnl, 0, group, error);
 584 }
 585 EXPORT_SYMBOL(rtnl_set_sk_err);
 586
 587 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 588 {
 589         struct nlattr *mx;
 590         int i, valid = 0;
 591
 592         mx = nla_nest_start(skb, RTA_METRICS);
 593         if (mx == NULL)
 594                 return -ENOBUFS;
 595
 596         for (i = 0; i < RTAX_MAX; i++) {
 597                 if (metrics[i]) {
 598                         valid++;
 599                         if (nla_put_u32(skb, i+1, metrics[i]))
 600                                 goto nla_put_failure;
 601                 }
 602         }
 603
 604         if (!valid) {
 605                 nla_nest_cancel(skb, mx);
 606                 return 0;
 607         }
 608
 609         return nla_nest_end(skb, mx);
 610
 611 nla_put_failure:
 612         nla_nest_cancel(skb, mx);
 613         return -EMSGSIZE;
 614 }
 615 EXPORT_SYMBOL(rtnetlink_put_metrics);
 616
 617 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 618                        u32 ts, u32 tsage, long expires, u32 error)
 619 {
 620         struct rta_cacheinfo ci = {
 621                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 622                 .rta_used = dst->__use,
 623                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 624                 .rta_error = error,
 625                 .rta_id =  id,
 626                 .rta_ts = ts,
 627                 .rta_tsage = tsage,
 628         };
 629
 630         if (expires)
 631                 ci.rta_expires = jiffies_to_clock_t(expires);
 632
 633         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 634 }
 635 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 636
 637 static void set_operstate(struct net_device *dev, unsigned char transition)
 638 {
 639         unsigned char operstate = dev->operstate;
 640
 641         switch (transition) {
 642         case IF_OPER_UP:
 643                 if ((operstate == IF_OPER_DORMANT ||
 644                      operstate == IF_OPER_UNKNOWN) &&
 645                     !netif_dormant(dev))
 646                         operstate = IF_OPER_UP;
 647                 break;
 648
 649         case IF_OPER_DORMANT:
 650                 if (operstate == IF_OPER_UP ||
 651                     operstate == IF_OPER_UNKNOWN)
 652                         operstate = IF_OPER_DORMANT;
 653                 break;
 654         }
 655
 656         if (dev->operstate != operstate) {
 657                 write_lock_bh(&dev_base_lock);
 658                 dev->operstate = operstate;
 659                 write_unlock_bh(&dev_base_lock);
 660                 netdev_state_change(dev);
 661         }
 662 }
 663
 664 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 665                                            const struct ifinfomsg *ifm)
 666 {
 667         unsigned int flags = ifm->ifi_flags;
 668
 669         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 670         if (ifm->ifi_change)
 671                 flags = (flags & ifm->ifi_change) |
 672                         (dev->flags & ~ifm->ifi_change);
 673
 674         return flags;
 675 }
 676
 677 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 678                                  const struct rtnl_link_stats64 *b)
 679 {
 680         a->rx_packets = b->rx_packets;
 681         a->tx_packets = b->tx_packets;
 682         a->rx_bytes = b->rx_bytes;
 683         a->tx_bytes = b->tx_bytes;
 684         a->rx_errors = b->rx_errors;
 685         a->tx_errors = b->tx_errors;
 686         a->rx_dropped = b->rx_dropped;
 687         a->tx_dropped = b->tx_dropped;
 688
 689         a->multicast = b->multicast;
 690         a->collisions = b->collisions;
 691
 692         a->rx_length_errors = b->rx_length_errors;
 693         a->rx_over_errors = b->rx_over_errors;
 694         a->rx_crc_errors = b->rx_crc_errors;
 695         a->rx_frame_errors = b->rx_frame_errors;
 696         a->rx_fifo_errors = b->rx_fifo_errors;
 697         a->rx_missed_errors = b->rx_missed_errors;
 698
 699         a->tx_aborted_errors = b->tx_aborted_errors;
 700         a->tx_carrier_errors = b->tx_carrier_errors;
 701         a->tx_fifo_errors = b->tx_fifo_errors;
 702         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 703         a->tx_window_errors = b->tx_window_errors;
 704
 705         a->rx_compressed = b->rx_compressed;
 706         a->tx_compressed = b->tx_compressed;
 707 }
 708
 709 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 710 {
 711         memcpy(v, b, sizeof(*b));
 712 }
 713
 714 /* All VF info */
 715 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 716                                    u32 ext_filter_mask)
 717 {
 718         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 719             (ext_filter_mask & RTEXT_FILTER_VF)) {
 720                 int num_vfs = dev_num_vf(dev->dev.parent);
 721                 size_t size = nla_total_size(sizeof(struct nlattr));
 722                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 723                 size += num_vfs *
 724                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 725                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 726                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 727                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 728                 return size;
 729         } else
 730                 return 0;
 731 }
 732
 733 static size_t rtnl_port_size(const struct net_device *dev)
 734 {
 735         size_t port_size = nla_total_size(4)            /* PORT_VF */
 736                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 737                 + nla_total_size(sizeof(struct ifla_port_vsi))
 738                                                         /* PORT_VSI_TYPE */
 739                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 740                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 741                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 742                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 743         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 744         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 745                 + port_size;
 746         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 747                 + port_size;
 748
 749         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 750                 return 0;
 751         if (dev_num_vf(dev->dev.parent))
 752                 return port_self_size + vf_ports_size +
 753                         vf_port_size * dev_num_vf(dev->dev.parent);
 754         else
 755                 return port_self_size;
 756 }
 757
 758 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 759                                      u32 ext_filter_mask)
 760 {
 761         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 762                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 763                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 764                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 765                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 766                + nla_total_size(sizeof(struct rtnl_link_stats))
 767                + nla_total_size(sizeof(struct rtnl_link_stats64))
 768                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 769                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 770                + nla_total_size(4) /* IFLA_TXQLEN */
 771                + nla_total_size(4) /* IFLA_WEIGHT */
 772                + nla_total_size(4) /* IFLA_MTU */
 773                + nla_total_size(4) /* IFLA_LINK */
 774                + nla_total_size(4) /* IFLA_MASTER */
 775                + nla_total_size(4) /* IFLA_PROMISCUITY */
 776                + nla_total_size(1) /* IFLA_OPERSTATE */
 777                + nla_total_size(1) /* IFLA_LINKMODE */
 778                + nla_total_size(ext_filter_mask
 779                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 780                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 781                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 782                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 783                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 784 }
 785
 786 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 787 {
 788         struct nlattr *vf_ports;
 789         struct nlattr *vf_port;
 790         int vf;
 791         int err;
 792
 793         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 794         if (!vf_ports)
 795                 return -EMSGSIZE;
 796
 797         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 798                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 799                 if (!vf_port)
 800                         goto nla_put_failure;
 801                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 802                         goto nla_put_failure;
 803                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 804                 if (err == -EMSGSIZE)
 805                         goto nla_put_failure;
 806                 if (err) {
 807                         nla_nest_cancel(skb, vf_port);
 808                         continue;
 809                 }
 810                 nla_nest_end(skb, vf_port);
 811         }
 812
 813         nla_nest_end(skb, vf_ports);
 814
 815         return 0;
 816
 817 nla_put_failure:
 818         nla_nest_cancel(skb, vf_ports);
 819         return -EMSGSIZE;
 820 }
 821
 822 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 823 {
 824         struct nlattr *port_self;
 825         int err;
 826
 827         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 828         if (!port_self)
 829                 return -EMSGSIZE;
 830
 831         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 832         if (err) {
 833                 nla_nest_cancel(skb, port_self);
 834                 return (err == -EMSGSIZE) ? err : 0;
 835         }
 836
 837         nla_nest_end(skb, port_self);
 838
 839         return 0;
 840 }
 841
 842 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 843 {
 844         int err;
 845
 846         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 847                 return 0;
 848
 849         err = rtnl_port_self_fill(skb, dev);
 850         if (err)
 851                 return err;
 852
 853         if (dev_num_vf(dev->dev.parent)) {
 854                 err = rtnl_vf_ports_fill(skb, dev);
 855                 if (err)
 856                         return err;
 857         }
 858
 859         return 0;
 860 }
 861
 862 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 863                             int type, u32 pid, u32 seq, u32 change,
 864                             unsigned int flags, u32 ext_filter_mask)
 865 {
 866         struct ifinfomsg *ifm;
 867         struct nlmsghdr *nlh;
 868         struct rtnl_link_stats64 temp;
 869         const struct rtnl_link_stats64 *stats;
 870         struct nlattr *attr, *af_spec;
 871         struct rtnl_af_ops *af_ops;
 872
 873         ASSERT_RTNL();
 874         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 875         if (nlh == NULL)
 876                 return -EMSGSIZE;
 877
 878         ifm = nlmsg_data(nlh);
 879         ifm->ifi_family = AF_UNSPEC;
 880         ifm->__ifi_pad = 0;
 881         ifm->ifi_type = dev->type;
 882         ifm->ifi_index = dev->ifindex;
 883         ifm->ifi_flags = dev_get_flags(dev);
 884         ifm->ifi_change = change;
 885
 886         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 887             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 888             nla_put_u8(skb, IFLA_OPERSTATE,
 889                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 890             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 891             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 892             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 893             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 894             (dev->ifindex != dev->iflink &&
 895              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 896             (dev->master &&
 897              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 898             (dev->qdisc &&
 899              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 900             (dev->ifalias &&
 901              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 902                 goto nla_put_failure;
 903
 904         if (1) {
 905                 struct rtnl_link_ifmap map = {
 906                         .mem_start   = dev->mem_start,
 907                         .mem_end     = dev->mem_end,
 908                         .base_addr   = dev->base_addr,
 909                         .irq         = dev->irq,
 910                         .dma         = dev->dma,
 911                         .port        = dev->if_port,
 912                 };
 913                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 914                         goto nla_put_failure;
 915         }
 916
 917         if (dev->addr_len) {
 918                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 919                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 920                         goto nla_put_failure;
 921         }
 922
 923         attr = nla_reserve(skb, IFLA_STATS,
 924                         sizeof(struct rtnl_link_stats));
 925         if (attr == NULL)
 926                 goto nla_put_failure;
 927
 928         stats = dev_get_stats(dev, &temp);
 929         copy_rtnl_link_stats(nla_data(attr), stats);
 930
 931         attr = nla_reserve(skb, IFLA_STATS64,
 932                         sizeof(struct rtnl_link_stats64));
 933         if (attr == NULL)
 934                 goto nla_put_failure;
 935         copy_rtnl_link_stats64(nla_data(attr), stats);
 936
 937         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 938             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 939                 goto nla_put_failure;
 940
 941         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 942             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 943                 int i;
 944
 945                 struct nlattr *vfinfo, *vf;
 946                 int num_vfs = dev_num_vf(dev->dev.parent);
 947
 948                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 949                 if (!vfinfo)
 950                         goto nla_put_failure;
 951                 for (i = 0; i < num_vfs; i++) {
 952                         struct ifla_vf_info ivi;
 953                         struct ifla_vf_mac vf_mac;
 954                         struct ifla_vf_vlan vf_vlan;
 955                         struct ifla_vf_tx_rate vf_tx_rate;
 956                         struct ifla_vf_spoofchk vf_spoofchk;
 957
 958                         /*
 959                          * Not all SR-IOV capable drivers support the
 960                          * spoofcheck query.  Preset to -1 so the user
 961                          * space tool can detect that the driver didn't
 962                          * report anything.
 963                          */
 964                         ivi.spoofchk = -1;
 965                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 966                                 break;
 967                         vf_mac.vf =
 968                                 vf_vlan.vf =
 969                                 vf_tx_rate.vf =
 970                                 vf_spoofchk.vf = ivi.vf;
 971
 972                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 973                         vf_vlan.vlan = ivi.vlan;
 974                         vf_vlan.qos = ivi.qos;
 975                         vf_tx_rate.rate = ivi.tx_rate;
 976                         vf_spoofchk.setting = ivi.spoofchk;
 977                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 978                         if (!vf) {
 979                                 nla_nest_cancel(skb, vfinfo);
 980                                 goto nla_put_failure;
 981                         }
 982                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 983                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 984                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 985                                     &vf_tx_rate) ||
 986                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 987                                     &vf_spoofchk))
 988                                 goto nla_put_failure;
 989                         nla_nest_end(skb, vf);
 990                 }
 991                 nla_nest_end(skb, vfinfo);
 992         }
 993
 994         if (rtnl_port_fill(skb, dev))
 995                 goto nla_put_failure;
 996
 997         if (dev->rtnl_link_ops) {
 998                 if (rtnl_link_fill(skb, dev) < 0)
 999                         goto nla_put_failure;
1000         }
1001
1002         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1003                 goto nla_put_failure;
1004
1005         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1006                 if (af_ops->fill_link_af) {
1007                         struct nlattr *af;
1008                         int err;
1009
1010                         if (!(af = nla_nest_start(skb, af_ops->family)))
1011                                 goto nla_put_failure;
1012
1013                         err = af_ops->fill_link_af(skb, dev);
1014
1015                         /*
1016                          * Caller may return ENODATA to indicate that there
1017                          * was no data to be dumped. This is not an error, it
1018                          * means we should trim the attribute header and
1019                          * continue.
1020                          */
1021                         if (err == -ENODATA)
1022                                 nla_nest_cancel(skb, af);
1023                         else if (err < 0)
1024                                 goto nla_put_failure;
1025
1026                         nla_nest_end(skb, af);
1027                 }
1028         }
1029
1030         nla_nest_end(skb, af_spec);
1031
1032         return nlmsg_end(skb, nlh);
1033
1034 nla_put_failure:
1035         nlmsg_cancel(skb, nlh);
1036         return -EMSGSIZE;
1037 }
1038
1039 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1040 {
1041         struct net *net = sock_net(skb->sk);
1042         int h, s_h;
1043         int idx = 0, s_idx;
1044         struct net_device *dev;
1045         struct hlist_head *head;
1046         struct hlist_node *node;
1047         struct nlattr *tb[IFLA_MAX+1];
1048         u32 ext_filter_mask = 0;
1049
1050         s_h = cb->args[0];
1051         s_idx = cb->args[1];
1052
1053         rcu_read_lock();
1054         cb->seq = net->dev_base_seq;
1055
1056         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1057                         ifla_policy) >= 0) {
1058
1059                 if (tb[IFLA_EXT_MASK])
1060                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1061         }
1062
1063         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1064                 idx = 0;
1065                 head = &net->dev_index_head[h];
1066                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1067                         if (idx < s_idx)
1068                                 goto cont;
1069                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1070                                              NETLINK_CB(cb->skb).pid,
1071                                              cb->nlh->nlmsg_seq, 0,
1072                                              NLM_F_MULTI,
1073                                              ext_filter_mask) <= 0)
1074                                 goto out;
1075
1076                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1077 cont:
1078                         idx++;
1079                 }
1080         }
1081 out:
1082         rcu_read_unlock();
1083         cb->args[1] = idx;
1084         cb->args[0] = h;
1085
1086         return skb->len;
1087 }
1088
1089 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1090         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1091         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1092         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1093         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1094         [IFLA_MTU]              = { .type = NLA_U32 },
1095         [IFLA_LINK]             = { .type = NLA_U32 },
1096         [IFLA_MASTER]           = { .type = NLA_U32 },
1097         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1098         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1099         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1100         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1101         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1102         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1103         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1104         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1105         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1106         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1107         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1108         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1109         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1110         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1111 };
1112 EXPORT_SYMBOL(ifla_policy);
1113
1114 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1115         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1116         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1117 };
1118
1119 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1120         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1121 };
1122
1123 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1124         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1125                                     .len = sizeof(struct ifla_vf_mac) },
1126         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1127                                     .len = sizeof(struct ifla_vf_vlan) },
1128         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1129                                     .len = sizeof(struct ifla_vf_tx_rate) },
1130         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1131                                     .len = sizeof(struct ifla_vf_spoofchk) },
1132 };
1133
1134 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1135         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1136         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1137                                     .len = PORT_PROFILE_MAX },
1138         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1139                                     .len = sizeof(struct ifla_port_vsi)},
1140         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1141                                       .len = PORT_UUID_MAX },
1142         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1143                                     .len = PORT_UUID_MAX },
1144         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1145         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1146 };
1147
1148 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1149 {
1150         struct net *net;
1151         /* Examine the link attributes and figure out which
1152          * network namespace we are talking about.
1153          */
1154         if (tb[IFLA_NET_NS_PID])
1155                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1156         else if (tb[IFLA_NET_NS_FD])
1157                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1158         else
1159                 net = get_net(src_net);
1160         return net;
1161 }
1162 EXPORT_SYMBOL(rtnl_link_get_net);
1163
1164 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1165 {
1166         if (dev) {
1167                 if (tb[IFLA_ADDRESS] &&
1168                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1169                         return -EINVAL;
1170
1171                 if (tb[IFLA_BROADCAST] &&
1172                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1173                         return -EINVAL;
1174         }
1175
1176         if (tb[IFLA_AF_SPEC]) {
1177                 struct nlattr *af;
1178                 int rem, err;
1179
1180                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1181                         const struct rtnl_af_ops *af_ops;
1182
1183                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1184                                 return -EAFNOSUPPORT;
1185
1186                         if (!af_ops->set_link_af)
1187                                 return -EOPNOTSUPP;
1188
1189                         if (af_ops->validate_link_af) {
1190                                 err = af_ops->validate_link_af(dev, af);
1191                                 if (err < 0)
1192                                         return err;
1193                         }
1194                 }
1195         }
1196
1197         return 0;
1198 }
1199
1200 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1201 {
1202         int rem, err = -EINVAL;
1203         struct nlattr *vf;
1204         const struct net_device_ops *ops = dev->netdev_ops;
1205
1206         nla_for_each_nested(vf, attr, rem) {
1207                 switch (nla_type(vf)) {
1208                 case IFLA_VF_MAC: {
1209                         struct ifla_vf_mac *ivm;
1210                         ivm = nla_data(vf);
1211                         err = -EOPNOTSUPP;
1212                         if (ops->ndo_set_vf_mac)
1213                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1214                                                           ivm->mac);
1215                         break;
1216                 }
1217                 case IFLA_VF_VLAN: {
1218                         struct ifla_vf_vlan *ivv;
1219                         ivv = nla_data(vf);
1220                         err = -EOPNOTSUPP;
1221                         if (ops->ndo_set_vf_vlan)
1222                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1223                                                            ivv->vlan,
1224                                                            ivv->qos);
1225                         break;
1226                 }
1227                 case IFLA_VF_TX_RATE: {
1228                         struct ifla_vf_tx_rate *ivt;
1229                         ivt = nla_data(vf);
1230                         err = -EOPNOTSUPP;
1231                         if (ops->ndo_set_vf_tx_rate)
1232                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1233                                                               ivt->rate);
1234                         break;
1235                 }
1236                 case IFLA_VF_SPOOFCHK: {
1237                         struct ifla_vf_spoofchk *ivs;
1238                         ivs = nla_data(vf);
1239                         err = -EOPNOTSUPP;
1240                         if (ops->ndo_set_vf_spoofchk)
1241                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1242                                                                ivs->setting);
1243                         break;
1244                 }
1245                 default:
1246                         err = -EINVAL;
1247                         break;
1248                 }
1249                 if (err)
1250                         break;
1251         }
1252         return err;
1253 }
1254
1255 static int do_set_master(struct net_device *dev, int ifindex)
1256 {
1257         struct net_device *master_dev;
1258         const struct net_device_ops *ops;
1259         int err;
1260
1261         if (dev->master) {
1262                 if (dev->master->ifindex == ifindex)
1263                         return 0;
1264                 ops = dev->master->netdev_ops;
1265                 if (ops->ndo_del_slave) {
1266                         err = ops->ndo_del_slave(dev->master, dev);
1267                         if (err)
1268                                 return err;
1269                 } else {
1270                         return -EOPNOTSUPP;
1271                 }
1272         }
1273
1274         if (ifindex) {
1275                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1276                 if (!master_dev)
1277                         return -EINVAL;
1278                 ops = master_dev->netdev_ops;
1279                 if (ops->ndo_add_slave) {
1280                         err = ops->ndo_add_slave(master_dev, dev);
1281                         if (err)
1282                                 return err;
1283                 } else {
1284                         return -EOPNOTSUPP;
1285                 }
1286         }
1287         return 0;
1288 }
1289
1290 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1291                       struct nlattr **tb, char *ifname, int modified)
1292 {
1293         const struct net_device_ops *ops = dev->netdev_ops;
1294         int send_addr_notify = 0;
1295         int err;
1296
1297         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1298                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1299                 if (IS_ERR(net)) {
1300                         err = PTR_ERR(net);
1301                         goto errout;
1302                 }
1303                 err = dev_change_net_namespace(dev, net, ifname);
1304                 put_net(net);
1305                 if (err)
1306                         goto errout;
1307                 modified = 1;
1308         }
1309
1310         if (tb[IFLA_MAP]) {
1311                 struct rtnl_link_ifmap *u_map;
1312                 struct ifmap k_map;
1313
1314                 if (!ops->ndo_set_config) {
1315                         err = -EOPNOTSUPP;
1316                         goto errout;
1317                 }
1318
1319                 if (!netif_device_present(dev)) {
1320                         err = -ENODEV;
1321                         goto errout;
1322                 }
1323
1324                 u_map = nla_data(tb[IFLA_MAP]);
1325                 k_map.mem_start = (unsigned long) u_map->mem_start;
1326                 k_map.mem_end = (unsigned long) u_map->mem_end;
1327                 k_map.base_addr = (unsigned short) u_map->base_addr;
1328                 k_map.irq = (unsigned char) u_map->irq;
1329                 k_map.dma = (unsigned char) u_map->dma;
1330                 k_map.port = (unsigned char) u_map->port;
1331
1332                 err = ops->ndo_set_config(dev, &k_map);
1333                 if (err < 0)
1334                         goto errout;
1335
1336                 modified = 1;
1337         }
1338
1339         if (tb[IFLA_ADDRESS]) {
1340                 struct sockaddr *sa;
1341                 int len;
1342
1343                 if (!ops->ndo_set_mac_address) {
1344                         err = -EOPNOTSUPP;
1345                         goto errout;
1346                 }
1347
1348                 if (!netif_device_present(dev)) {
1349                         err = -ENODEV;
1350                         goto errout;
1351                 }
1352
1353                 len = sizeof(sa_family_t) + dev->addr_len;
1354                 sa = kmalloc(len, GFP_KERNEL);
1355                 if (!sa) {
1356                         err = -ENOMEM;
1357                         goto errout;
1358                 }
1359                 sa->sa_family = dev->type;
1360                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1361                        dev->addr_len);
1362                 err = ops->ndo_set_mac_address(dev, sa);
1363                 kfree(sa);
1364                 if (err)
1365                         goto errout;
1366                 send_addr_notify = 1;
1367                 modified = 1;
1368         }
1369
1370         if (tb[IFLA_MTU]) {
1371                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1372                 if (err < 0)
1373                         goto errout;
1374                 modified = 1;
1375         }
1376
1377         if (tb[IFLA_GROUP]) {
1378                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1379                 modified = 1;
1380         }
1381
1382         /*
1383          * Interface selected by interface index but interface
1384          * name provided implies that a name change has been
1385          * requested.
1386          */
1387         if (ifm->ifi_index > 0 && ifname[0]) {
1388                 err = dev_change_name(dev, ifname);
1389                 if (err < 0)
1390                         goto errout;
1391                 modified = 1;
1392         }
1393
1394         if (tb[IFLA_IFALIAS]) {
1395                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1396                                     nla_len(tb[IFLA_IFALIAS]));
1397                 if (err < 0)
1398                         goto errout;
1399                 modified = 1;
1400         }
1401
1402         if (tb[IFLA_BROADCAST]) {
1403                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1404                 send_addr_notify = 1;
1405         }
1406
1407         if (ifm->ifi_flags || ifm->ifi_change) {
1408                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1409                 if (err < 0)
1410                         goto errout;
1411         }
1412
1413         if (tb[IFLA_MASTER]) {
1414                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1415                 if (err)
1416                         goto errout;
1417                 modified = 1;
1418         }
1419
1420         if (tb[IFLA_TXQLEN])
1421                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1422
1423         if (tb[IFLA_OPERSTATE])
1424                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1425
1426         if (tb[IFLA_LINKMODE]) {
1427                 write_lock_bh(&dev_base_lock);
1428                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1429                 write_unlock_bh(&dev_base_lock);
1430         }
1431
1432         if (tb[IFLA_VFINFO_LIST]) {
1433                 struct nlattr *attr;
1434                 int rem;
1435                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1436                         if (nla_type(attr) != IFLA_VF_INFO) {
1437                                 err = -EINVAL;
1438                                 goto errout;
1439                         }
1440                         err = do_setvfinfo(dev, attr);
1441                         if (err < 0)
1442                                 goto errout;
1443                         modified = 1;
1444                 }
1445         }
1446         err = 0;
1447
1448         if (tb[IFLA_VF_PORTS]) {
1449                 struct nlattr *port[IFLA_PORT_MAX+1];
1450                 struct nlattr *attr;
1451                 int vf;
1452                 int rem;
1453
1454                 err = -EOPNOTSUPP;
1455                 if (!ops->ndo_set_vf_port)
1456                         goto errout;
1457
1458                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1459                         if (nla_type(attr) != IFLA_VF_PORT)
1460                                 continue;
1461                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1462                                 attr, ifla_port_policy);
1463                         if (err < 0)
1464                                 goto errout;
1465                         if (!port[IFLA_PORT_VF]) {
1466                                 err = -EOPNOTSUPP;
1467                                 goto errout;
1468                         }
1469                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1470                         err = ops->ndo_set_vf_port(dev, vf, port);
1471                         if (err < 0)
1472                                 goto errout;
1473                         modified = 1;
1474                 }
1475         }
1476         err = 0;
1477
1478         if (tb[IFLA_PORT_SELF]) {
1479                 struct nlattr *port[IFLA_PORT_MAX+1];
1480
1481                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1482                         tb[IFLA_PORT_SELF], ifla_port_policy);
1483                 if (err < 0)
1484                         goto errout;
1485
1486                 err = -EOPNOTSUPP;
1487                 if (ops->ndo_set_vf_port)
1488                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1489                 if (err < 0)
1490                         goto errout;
1491                 modified = 1;
1492         }
1493
1494         if (tb[IFLA_AF_SPEC]) {
1495                 struct nlattr *af;
1496                 int rem;
1497
1498                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1499                         const struct rtnl_af_ops *af_ops;
1500
1501                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1502                                 BUG();
1503
1504                         err = af_ops->set_link_af(dev, af);
1505                         if (err < 0)
1506                                 goto errout;
1507
1508                         modified = 1;
1509                 }
1510         }
1511         err = 0;
1512
1513 errout:
1514         if (err < 0 && modified)
1515                 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1516                                      dev->name);
1517
1518         if (send_addr_notify)
1519                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1520
1521         return err;
1522 }
1523
1524 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1525 {
1526         struct net *net = sock_net(skb->sk);
1527         struct ifinfomsg *ifm;
1528         struct net_device *dev;
1529         int err;
1530         struct nlattr *tb[IFLA_MAX+1];
1531         char ifname[IFNAMSIZ];
1532
1533         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1534         if (err < 0)
1535                 goto errout;
1536
1537         if (tb[IFLA_IFNAME])
1538                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1539         else
1540                 ifname[0] = '\0';
1541
1542         err = -EINVAL;
1543         ifm = nlmsg_data(nlh);
1544         if (ifm->ifi_index > 0)
1545                 dev = __dev_get_by_index(net, ifm->ifi_index);
1546         else if (tb[IFLA_IFNAME])
1547                 dev = __dev_get_by_name(net, ifname);
1548         else
1549                 goto errout;
1550
1551         if (dev == NULL) {
1552                 err = -ENODEV;
1553                 goto errout;
1554         }
1555
1556         err = validate_linkmsg(dev, tb);
1557         if (err < 0)
1558                 goto errout;
1559
1560         err = do_setlink(dev, ifm, tb, ifname, 0);
1561 errout:
1562         return err;
1563 }
1564
1565 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1566 {
1567         struct net *net = sock_net(skb->sk);
1568         const struct rtnl_link_ops *ops;
1569         struct net_device *dev;
1570         struct ifinfomsg *ifm;
1571         char ifname[IFNAMSIZ];
1572         struct nlattr *tb[IFLA_MAX+1];
1573         int err;
1574         LIST_HEAD(list_kill);
1575
1576         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1577         if (err < 0)
1578                 return err;
1579
1580         if (tb[IFLA_IFNAME])
1581                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1582
1583         ifm = nlmsg_data(nlh);
1584         if (ifm->ifi_index > 0)
1585                 dev = __dev_get_by_index(net, ifm->ifi_index);
1586         else if (tb[IFLA_IFNAME])
1587                 dev = __dev_get_by_name(net, ifname);
1588         else
1589                 return -EINVAL;
1590
1591         if (!dev)
1592                 return -ENODEV;
1593
1594         ops = dev->rtnl_link_ops;
1595         if (!ops)
1596                 return -EOPNOTSUPP;
1597
1598         ops->dellink(dev, &list_kill);
1599         unregister_netdevice_many(&list_kill);
1600         list_del(&list_kill);
1601         return 0;
1602 }
1603
1604 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1605 {
1606         unsigned int old_flags;
1607         int err;
1608
1609         old_flags = dev->flags;
1610         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1611                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1612                 if (err < 0)
1613                         return err;
1614         }
1615
1616         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1617         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1618
1619         __dev_notify_flags(dev, old_flags);
1620         return 0;
1621 }
1622 EXPORT_SYMBOL(rtnl_configure_link);
1623
1624 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1625         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1626 {
1627         int err;
1628         struct net_device *dev;
1629         unsigned int num_queues = 1;
1630
1631         if (ops->get_tx_queues) {
1632                 err = ops->get_tx_queues(src_net, tb);
1633                 if (err < 0)
1634                         goto err;
1635                 num_queues = err;
1636         }
1637
1638         err = -ENOMEM;
1639         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1640         if (!dev)
1641                 goto err;
1642
1643         dev_net_set(dev, net);
1644         dev->rtnl_link_ops = ops;
1645         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1646
1647         if (tb[IFLA_MTU])
1648                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1649         if (tb[IFLA_ADDRESS])
1650                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1651                                 nla_len(tb[IFLA_ADDRESS]));
1652         if (tb[IFLA_BROADCAST])
1653                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1654                                 nla_len(tb[IFLA_BROADCAST]));
1655         if (tb[IFLA_TXQLEN])
1656                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1657         if (tb[IFLA_OPERSTATE])
1658                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1659         if (tb[IFLA_LINKMODE])
1660                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1661         if (tb[IFLA_GROUP])
1662                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1663
1664         return dev;
1665
1666 err:
1667         return ERR_PTR(err);
1668 }
1669 EXPORT_SYMBOL(rtnl_create_link);
1670
1671 static int rtnl_group_changelink(struct net *net, int group,
1672                 struct ifinfomsg *ifm,
1673                 struct nlattr **tb)
1674 {
1675         struct net_device *dev;
1676         int err;
1677
1678         for_each_netdev(net, dev) {
1679                 if (dev->group == group) {
1680                         err = do_setlink(dev, ifm, tb, NULL, 0);
1681                         if (err < 0)
1682                                 return err;
1683                 }
1684         }
1685
1686         return 0;
1687 }
1688
1689 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1690 {
1691         struct net *net = sock_net(skb->sk);
1692         const struct rtnl_link_ops *ops;
1693         struct net_device *dev;
1694         struct ifinfomsg *ifm;
1695         char kind[MODULE_NAME_LEN];
1696         char ifname[IFNAMSIZ];
1697         struct nlattr *tb[IFLA_MAX+1];
1698         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1699         int err;
1700
1701 #ifdef CONFIG_MODULES
1702 replay:
1703 #endif
1704         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1705         if (err < 0)
1706                 return err;
1707
1708         if (tb[IFLA_IFNAME])
1709                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1710         else
1711                 ifname[0] = '\0';
1712
1713         ifm = nlmsg_data(nlh);
1714         if (ifm->ifi_index > 0)
1715                 dev = __dev_get_by_index(net, ifm->ifi_index);
1716         else {
1717                 if (ifname[0])
1718                         dev = __dev_get_by_name(net, ifname);
1719                 else
1720                         dev = NULL;
1721         }
1722
1723         err = validate_linkmsg(dev, tb);
1724         if (err < 0)
1725                 return err;
1726
1727         if (tb[IFLA_LINKINFO]) {
1728                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1729                                        tb[IFLA_LINKINFO], ifla_info_policy);
1730                 if (err < 0)
1731                         return err;
1732         } else
1733                 memset(linkinfo, 0, sizeof(linkinfo));
1734
1735         if (linkinfo[IFLA_INFO_KIND]) {
1736                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1737                 ops = rtnl_link_ops_get(kind);
1738         } else {
1739                 kind[0] = '\0';
1740                 ops = NULL;
1741         }
1742
1743         if (1) {
1744                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1745                 struct net *dest_net;
1746
1747                 if (ops) {
1748                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1749                                 err = nla_parse_nested(attr, ops->maxtype,
1750                                                        linkinfo[IFLA_INFO_DATA],
1751                                                        ops->policy);
1752                                 if (err < 0)
1753                                         return err;
1754                                 data = attr;
1755                         }
1756                         if (ops->validate) {
1757                                 err = ops->validate(tb, data);
1758                                 if (err < 0)
1759                                         return err;
1760                         }
1761                 }
1762
1763                 if (dev) {
1764                         int modified = 0;
1765
1766                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1767                                 return -EEXIST;
1768                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1769                                 return -EOPNOTSUPP;
1770
1771                         if (linkinfo[IFLA_INFO_DATA]) {
1772                                 if (!ops || ops != dev->rtnl_link_ops ||
1773                                     !ops->changelink)
1774                                         return -EOPNOTSUPP;
1775
1776                                 err = ops->changelink(dev, tb, data);
1777                                 if (err < 0)
1778                                         return err;
1779                                 modified = 1;
1780                         }
1781
1782                         return do_setlink(dev, ifm, tb, ifname, modified);
1783                 }
1784
1785                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1786                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1787                                 return rtnl_group_changelink(net,
1788                                                 nla_get_u32(tb[IFLA_GROUP]),
1789                                                 ifm, tb);
1790                         return -ENODEV;
1791                 }
1792
1793                 if (ifm->ifi_index)
1794                         return -EOPNOTSUPP;
1795                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1796                         return -EOPNOTSUPP;
1797
1798                 if (!ops) {
1799 #ifdef CONFIG_MODULES
1800                         if (kind[0]) {
1801                                 __rtnl_unlock();
1802                                 request_module("rtnl-link-%s", kind);
1803                                 rtnl_lock();
1804                                 ops = rtnl_link_ops_get(kind);
1805                                 if (ops)
1806                                         goto replay;
1807                         }
1808 #endif
1809                         return -EOPNOTSUPP;
1810                 }
1811
1812                 if (!ifname[0])
1813                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1814
1815                 dest_net = rtnl_link_get_net(net, tb);
1816                 if (IS_ERR(dest_net))
1817                         return PTR_ERR(dest_net);
1818
1819                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1820
1821                 if (IS_ERR(dev))
1822                         err = PTR_ERR(dev);
1823                 else if (ops->newlink)
1824                         err = ops->newlink(net, dev, tb, data);
1825                 else
1826                         err = register_netdevice(dev);
1827
1828                 if (err < 0 && !IS_ERR(dev))
1829                         free_netdev(dev);
1830                 if (err < 0)
1831                         goto out;
1832
1833                 err = rtnl_configure_link(dev, ifm);
1834                 if (err < 0)
1835                         unregister_netdevice(dev);
1836 out:
1837                 put_net(dest_net);
1838                 return err;
1839         }
1840 }
1841
1842 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1843 {
1844         struct net *net = sock_net(skb->sk);
1845         struct ifinfomsg *ifm;
1846         char ifname[IFNAMSIZ];
1847         struct nlattr *tb[IFLA_MAX+1];
1848         struct net_device *dev = NULL;
1849         struct sk_buff *nskb;
1850         int err;
1851         u32 ext_filter_mask = 0;
1852
1853         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1854         if (err < 0)
1855                 return err;
1856
1857         if (tb[IFLA_IFNAME])
1858                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1859
1860         if (tb[IFLA_EXT_MASK])
1861                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1862
1863         ifm = nlmsg_data(nlh);
1864         if (ifm->ifi_index > 0)
1865                 dev = __dev_get_by_index(net, ifm->ifi_index);
1866         else if (tb[IFLA_IFNAME])
1867                 dev = __dev_get_by_name(net, ifname);
1868         else
1869                 return -EINVAL;
1870
1871         if (dev == NULL)
1872                 return -ENODEV;
1873
1874         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1875         if (nskb == NULL)
1876                 return -ENOBUFS;
1877
1878         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1879                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1880         if (err < 0) {
1881                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1882                 WARN_ON(err == -EMSGSIZE);
1883                 kfree_skb(nskb);
1884         } else
1885                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1886
1887         return err;
1888 }
1889
1890 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1891 {
1892         struct net *net = sock_net(skb->sk);
1893         struct net_device *dev;
1894         struct nlattr *tb[IFLA_MAX+1];
1895         u32 ext_filter_mask = 0;
1896         u16 min_ifinfo_dump_size = 0;
1897
1898         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1899                         ifla_policy) >= 0) {
1900                 if (tb[IFLA_EXT_MASK])
1901                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1902         }
1903
1904         if (!ext_filter_mask)
1905                 return NLMSG_GOODSIZE;
1906         /*
1907          * traverse the list of net devices and compute the minimum
1908          * buffer size based upon the filter mask.
1909          */
1910         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1911                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1912                                              if_nlmsg_size(dev,
1913                                                            ext_filter_mask));
1914         }
1915
1916         return min_ifinfo_dump_size;
1917 }
1918
1919 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1920 {
1921         int idx;
1922         int s_idx = cb->family;
1923
1924         if (s_idx == 0)
1925                 s_idx = 1;
1926         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1927                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1928                 if (idx < s_idx || idx == PF_PACKET)
1929                         continue;
1930                 if (rtnl_msg_handlers[idx] == NULL ||
1931                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1932                         continue;
1933                 if (idx > s_idx)
1934                         memset(&cb->args[0], 0, sizeof(cb->args));
1935                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1936                         break;
1937         }
1938         cb->family = idx;
1939
1940         return skb->len;
1941 }
1942
1943 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1944 {
1945         struct net *net = dev_net(dev);
1946         struct sk_buff *skb;
1947         int err = -ENOBUFS;
1948         size_t if_info_size;
1949
1950         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1951         if (skb == NULL)
1952                 goto errout;
1953
1954         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1955         if (err < 0) {
1956                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1957                 WARN_ON(err == -EMSGSIZE);
1958                 kfree_skb(skb);
1959                 goto errout;
1960         }
1961         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1962         return;
1963 errout:
1964         if (err < 0)
1965                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1966 }
1967
1968 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1969                                    struct net_device *dev,
1970                                    u8 *addr, u32 pid, u32 seq,
1971                                    int type, unsigned int flags)
1972 {
1973         struct nlmsghdr *nlh;
1974         struct ndmsg *ndm;
1975
1976         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1977         if (!nlh)
1978                 return -EMSGSIZE;
1979
1980         ndm = nlmsg_data(nlh);
1981         ndm->ndm_family  = AF_BRIDGE;
1982         ndm->ndm_pad1    = 0;
1983         ndm->ndm_pad2    = 0;
1984         ndm->ndm_flags   = flags;
1985         ndm->ndm_type    = 0;
1986         ndm->ndm_ifindex = dev->ifindex;
1987         ndm->ndm_state   = NUD_PERMANENT;
1988
1989         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
1990                 goto nla_put_failure;
1991
1992         return nlmsg_end(skb, nlh);
1993
1994 nla_put_failure:
1995         nlmsg_cancel(skb, nlh);
1996         return -EMSGSIZE;
1997 }
1998
1999 static inline size_t rtnl_fdb_nlmsg_size(void)
2000 {
2001         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2002 }
2003
2004 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2005 {
2006         struct net *net = dev_net(dev);
2007         struct sk_buff *skb;
2008         int err = -ENOBUFS;
2009
2010         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2011         if (!skb)
2012                 goto errout;
2013
2014         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2015         if (err < 0) {
2016                 kfree_skb(skb);
2017                 goto errout;
2018         }
2019
2020         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2021         return;
2022 errout:
2023         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2024 }
2025
2026 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2027 {
2028         struct net *net = sock_net(skb->sk);
2029         struct net_device *master = NULL;
2030         struct ndmsg *ndm;
2031         struct nlattr *tb[NDA_MAX+1];
2032         struct net_device *dev;
2033         u8 *addr;
2034         int err;
2035
2036         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2037         if (err < 0)
2038                 return err;
2039
2040         ndm = nlmsg_data(nlh);
2041         if (ndm->ndm_ifindex == 0) {
2042                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2043                 return -EINVAL;
2044         }
2045
2046         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2047         if (dev == NULL) {
2048                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2049                 return -ENODEV;
2050         }
2051
2052         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2053                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2054                 return -EINVAL;
2055         }
2056
2057         addr = nla_data(tb[NDA_LLADDR]);
2058         if (!is_valid_ether_addr(addr)) {
2059                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2060                 return -EINVAL;
2061         }
2062
2063         err = -EOPNOTSUPP;
2064
2065         /* Support fdb on master device the net/bridge default case */
2066         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2067             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2068                 master = dev->master;
2069                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2070                                                       nlh->nlmsg_flags);
2071                 if (err)
2072                         goto out;
2073                 else
2074                         ndm->ndm_flags &= ~NTF_MASTER;
2075         }
2076
2077         /* Embedded bridge, macvlan, and any other device support */
2078         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2079                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2080                                                    nlh->nlmsg_flags);
2081
2082                 if (!err) {
2083                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2084                         ndm->ndm_flags &= ~NTF_SELF;
2085                 }
2086         }
2087 out:
2088         return err;
2089 }
2090
2091 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2092 {
2093         struct net *net = sock_net(skb->sk);
2094         struct ndmsg *ndm;
2095         struct nlattr *llattr;
2096         struct net_device *dev;
2097         int err = -EINVAL;
2098         __u8 *addr;
2099
2100         if (nlmsg_len(nlh) < sizeof(*ndm))
2101                 return -EINVAL;
2102
2103         ndm = nlmsg_data(nlh);
2104         if (ndm->ndm_ifindex == 0) {
2105                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2106                 return -EINVAL;
2107         }
2108
2109         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2110         if (dev == NULL) {
2111                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2112                 return -ENODEV;
2113         }
2114
2115         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2116         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2117                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2118                 return -EINVAL;
2119         }
2120
2121         addr = nla_data(llattr);
2122         err = -EOPNOTSUPP;
2123
2124         /* Support fdb on master device the net/bridge default case */
2125         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2126             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2127                 struct net_device *master = dev->master;
2128
2129                 if (master->netdev_ops->ndo_fdb_del)
2130                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2131
2132                 if (err)
2133                         goto out;
2134                 else
2135                         ndm->ndm_flags &= ~NTF_MASTER;
2136         }
2137
2138         /* Embedded bridge, macvlan, and any other device support */
2139         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2140                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2141
2142                 if (!err) {
2143                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2144                         ndm->ndm_flags &= ~NTF_SELF;
2145                 }
2146         }
2147 out:
2148         return err;
2149 }
2150
2151 static int nlmsg_populate_fdb(struct sk_buff *skb,
2152                               struct netlink_callback *cb,
2153                               struct net_device *dev,
2154                               int *idx,
2155                               struct netdev_hw_addr_list *list)
2156 {
2157         struct netdev_hw_addr *ha;
2158         int err;
2159         u32 pid, seq;
2160
2161         pid = NETLINK_CB(cb->skb).pid;
2162         seq = cb->nlh->nlmsg_seq;
2163
2164         list_for_each_entry(ha, &list->list, list) {
2165                 if (*idx < cb->args[0])
2166                         goto skip;
2167
2168                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2169                                               pid, seq, 0, NTF_SELF);
2170                 if (err < 0)
2171                         return err;
2172 skip:
2173                 *idx += 1;
2174         }
2175         return 0;
2176 }
2177
2178 /**
2179  * ndo_dflt_fdb_dump: default netdevice operation to dump an FDB table.
2180  * @nlh: netlink message header
2181  * @dev: netdevice
2182  *
2183  * Default netdevice operation to dump the existing unicast address list.
2184  * Returns zero on success.
2185  */
2186 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2187                       struct netlink_callback *cb,
2188                       struct net_device *dev,
2189                       int idx)
2190 {
2191         int err;
2192
2193         netif_addr_lock_bh(dev);
2194         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2195         if (err)
2196                 goto out;
2197         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2198 out:
2199         netif_addr_unlock_bh(dev);
2200         return idx;
2201 }
2202 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2203
2204 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2205 {
2206         int idx = 0;
2207         struct net *net = sock_net(skb->sk);
2208         struct net_device *dev;
2209
2210         rcu_read_lock();
2211         for_each_netdev_rcu(net, dev) {
2212                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2213                         struct net_device *master = dev->master;
2214                         const struct net_device_ops *ops = master->netdev_ops;
2215
2216                         if (ops->ndo_fdb_dump)
2217                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2218                 }
2219
2220                 if (dev->netdev_ops->ndo_fdb_dump)
2221                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2222         }
2223         rcu_read_unlock();
2224
2225         cb->args[0] = idx;
2226         return skb->len;
2227 }
2228
2229 /* Protected by RTNL sempahore.  */
2230 static struct rtattr **rta_buf;
2231 static int rtattr_max;
2232
2233 /* Process one rtnetlink message. */
2234
2235 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2236 {
2237         struct net *net = sock_net(skb->sk);
2238         rtnl_doit_func doit;
2239         int sz_idx, kind;
2240         int min_len;
2241         int family;
2242         int type;
2243         int err;
2244
2245         type = nlh->nlmsg_type;
2246         if (type > RTM_MAX)
2247                 return -EOPNOTSUPP;
2248
2249         type -= RTM_BASE;
2250
2251         /* All the messages must have at least 1 byte length */
2252         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2253                 return 0;
2254
2255         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2256         sz_idx = type>>2;
2257         kind = type&3;
2258
2259         if (kind != 2 && !capable(CAP_NET_ADMIN))
2260                 return -EPERM;
2261
2262         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2263                 struct sock *rtnl;
2264                 rtnl_dumpit_func dumpit;
2265                 rtnl_calcit_func calcit;
2266                 u16 min_dump_alloc = 0;
2267
2268                 dumpit = rtnl_get_dumpit(family, type);
2269                 if (dumpit == NULL)
2270                         return -EOPNOTSUPP;
2271                 calcit = rtnl_get_calcit(family, type);
2272                 if (calcit)
2273                         min_dump_alloc = calcit(skb, nlh);
2274
2275                 __rtnl_unlock();
2276                 rtnl = net->rtnl;
2277                 {
2278                         struct netlink_dump_control c = {
2279                                 .dump           = dumpit,
2280                                 .min_dump_alloc = min_dump_alloc,
2281                         };
2282                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2283                 }
2284                 rtnl_lock();
2285                 return err;
2286         }
2287
2288         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2289
2290         min_len = rtm_min[sz_idx];
2291         if (nlh->nlmsg_len < min_len)
2292                 return -EINVAL;
2293
2294         if (nlh->nlmsg_len > min_len) {
2295                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2296                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2297
2298                 while (RTA_OK(attr, attrlen)) {
2299                         unsigned int flavor = attr->rta_type;
2300                         if (flavor) {
2301                                 if (flavor > rta_max[sz_idx])
2302                                         return -EINVAL;
2303                                 rta_buf[flavor-1] = attr;
2304                         }
2305                         attr = RTA_NEXT(attr, attrlen);
2306                 }
2307         }
2308
2309         doit = rtnl_get_doit(family, type);
2310         if (doit == NULL)
2311                 return -EOPNOTSUPP;
2312
2313         return doit(skb, nlh, (void *)&rta_buf[0]);
2314 }
2315
2316 static void rtnetlink_rcv(struct sk_buff *skb)
2317 {
2318         rtnl_lock();
2319         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2320         rtnl_unlock();
2321 }
2322
2323 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2324 {
2325         struct net_device *dev = ptr;
2326
2327         switch (event) {
2328         case NETDEV_UP:
2329         case NETDEV_DOWN:
2330         case NETDEV_PRE_UP:
2331         case NETDEV_POST_INIT:
2332         case NETDEV_REGISTER:
2333         case NETDEV_CHANGE:
2334         case NETDEV_PRE_TYPE_CHANGE:
2335         case NETDEV_GOING_DOWN:
2336         case NETDEV_UNREGISTER:
2337         case NETDEV_UNREGISTER_BATCH:
2338         case NETDEV_RELEASE:
2339         case NETDEV_JOIN:
2340                 break;
2341         default:
2342                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2343                 break;
2344         }
2345         return NOTIFY_DONE;
2346 }
2347
2348 static struct notifier_block rtnetlink_dev_notifier = {
2349         .notifier_call  = rtnetlink_event,
2350 };
2351
2352
2353 static int __net_init rtnetlink_net_init(struct net *net)
2354 {
2355         struct sock *sk;
2356         struct netlink_kernel_cfg cfg = {
2357                 .groups         = RTNLGRP_MAX,
2358                 .input          = rtnetlink_rcv,
2359                 .cb_mutex       = &rtnl_mutex,
2360         };
2361
2362         sk = netlink_kernel_create(net, NETLINK_ROUTE, THIS_MODULE, &cfg);
2363         if (!sk)
2364                 return -ENOMEM;
2365         net->rtnl = sk;
2366         return 0;
2367 }
2368
2369 static void __net_exit rtnetlink_net_exit(struct net *net)
2370 {
2371         netlink_kernel_release(net->rtnl);
2372         net->rtnl = NULL;
2373 }
2374
2375 static struct pernet_operations rtnetlink_net_ops = {
2376         .init = rtnetlink_net_init,
2377         .exit = rtnetlink_net_exit,
2378 };
2379
2380 void __init rtnetlink_init(void)
2381 {
2382         int i;
2383
2384         rtattr_max = 0;
2385         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2386                 if (rta_max[i] > rtattr_max)
2387                         rtattr_max = rta_max[i];
2388         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2389         if (!rta_buf)
2390                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2391
2392         if (register_pernet_subsys(&rtnetlink_net_ops))
2393                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2394
2395         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2396         register_netdevice_notifier(&rtnetlink_dev_notifier);
2397
2398         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2399                       rtnl_dump_ifinfo, rtnl_calcit);
2400         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2401         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2402         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2403
2404         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2405         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2406
2407         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2408         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2409         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2410 }
2411