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[msgindex].doit;
 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[msgindex].dumpit;
 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[msgindex].calcit;
 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                        long expires, u32 error)
 619 {
 620         struct rta_cacheinfo ci = {
 621                 .rta_lastuse = jiffies_delta_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         };
 627
 628         if (expires) {
 629                 unsigned long clock;
 630
 631                 clock = jiffies_to_clock_t(abs(expires));
 632                 clock = min_t(unsigned long, clock, INT_MAX);
 633                 ci.rta_expires = (expires > 0) ? clock : -clock;
 634         }
 635         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 636 }
 637 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 638
 639 static void set_operstate(struct net_device *dev, unsigned char transition)
 640 {
 641         unsigned char operstate = dev->operstate;
 642
 643         switch (transition) {
 644         case IF_OPER_UP:
 645                 if ((operstate == IF_OPER_DORMANT ||
 646                      operstate == IF_OPER_UNKNOWN) &&
 647                     !netif_dormant(dev))
 648                         operstate = IF_OPER_UP;
 649                 break;
 650
 651         case IF_OPER_DORMANT:
 652                 if (operstate == IF_OPER_UP ||
 653                     operstate == IF_OPER_UNKNOWN)
 654                         operstate = IF_OPER_DORMANT;
 655                 break;
 656         }
 657
 658         if (dev->operstate != operstate) {
 659                 write_lock_bh(&dev_base_lock);
 660                 dev->operstate = operstate;
 661                 write_unlock_bh(&dev_base_lock);
 662                 netdev_state_change(dev);
 663         }
 664 }
 665
 666 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 667 {
 668         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 669                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 670 }
 671
 672 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 673                                            const struct ifinfomsg *ifm)
 674 {
 675         unsigned int flags = ifm->ifi_flags;
 676
 677         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 678         if (ifm->ifi_change)
 679                 flags = (flags & ifm->ifi_change) |
 680                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 681
 682         return flags;
 683 }
 684
 685 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 686                                  const struct rtnl_link_stats64 *b)
 687 {
 688         a->rx_packets = b->rx_packets;
 689         a->tx_packets = b->tx_packets;
 690         a->rx_bytes = b->rx_bytes;
 691         a->tx_bytes = b->tx_bytes;
 692         a->rx_errors = b->rx_errors;
 693         a->tx_errors = b->tx_errors;
 694         a->rx_dropped = b->rx_dropped;
 695         a->tx_dropped = b->tx_dropped;
 696
 697         a->multicast = b->multicast;
 698         a->collisions = b->collisions;
 699
 700         a->rx_length_errors = b->rx_length_errors;
 701         a->rx_over_errors = b->rx_over_errors;
 702         a->rx_crc_errors = b->rx_crc_errors;
 703         a->rx_frame_errors = b->rx_frame_errors;
 704         a->rx_fifo_errors = b->rx_fifo_errors;
 705         a->rx_missed_errors = b->rx_missed_errors;
 706
 707         a->tx_aborted_errors = b->tx_aborted_errors;
 708         a->tx_carrier_errors = b->tx_carrier_errors;
 709         a->tx_fifo_errors = b->tx_fifo_errors;
 710         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 711         a->tx_window_errors = b->tx_window_errors;
 712
 713         a->rx_compressed = b->rx_compressed;
 714         a->tx_compressed = b->tx_compressed;
 715 }
 716
 717 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 718 {
 719         memcpy(v, b, sizeof(*b));
 720 }
 721
 722 /* All VF info */
 723 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 724                                    u32 ext_filter_mask)
 725 {
 726         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 727             (ext_filter_mask & RTEXT_FILTER_VF)) {
 728                 int num_vfs = dev_num_vf(dev->dev.parent);
 729                 size_t size = nla_total_size(sizeof(struct nlattr));
 730                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 731                 size += num_vfs *
 732                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 733                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 734                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 735                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 736                 return size;
 737         } else
 738                 return 0;
 739 }
 740
 741 static size_t rtnl_port_size(const struct net_device *dev)
 742 {
 743         size_t port_size = nla_total_size(4)            /* PORT_VF */
 744                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 745                 + nla_total_size(sizeof(struct ifla_port_vsi))
 746                                                         /* PORT_VSI_TYPE */
 747                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 748                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 749                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 750                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 751         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 752         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 753                 + port_size;
 754         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 755                 + port_size;
 756
 757         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 758                 return 0;
 759         if (dev_num_vf(dev->dev.parent))
 760                 return port_self_size + vf_ports_size +
 761                         vf_port_size * dev_num_vf(dev->dev.parent);
 762         else
 763                 return port_self_size;
 764 }
 765
 766 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 767                                      u32 ext_filter_mask)
 768 {
 769         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 770                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 771                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 772                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 773                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 774                + nla_total_size(sizeof(struct rtnl_link_stats))
 775                + nla_total_size(sizeof(struct rtnl_link_stats64))
 776                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 777                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 778                + nla_total_size(4) /* IFLA_TXQLEN */
 779                + nla_total_size(4) /* IFLA_WEIGHT */
 780                + nla_total_size(4) /* IFLA_MTU */
 781                + nla_total_size(4) /* IFLA_LINK */
 782                + nla_total_size(4) /* IFLA_MASTER */
 783                + nla_total_size(1) /* IFLA_CARRIER */
 784                + nla_total_size(4) /* IFLA_PROMISCUITY */
 785                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 786                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 787                + nla_total_size(1) /* IFLA_OPERSTATE */
 788                + nla_total_size(1) /* IFLA_LINKMODE */
 789                + nla_total_size(ext_filter_mask
 790                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 791                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 792                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 793                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 794                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 795 }
 796
 797 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 798 {
 799         struct nlattr *vf_ports;
 800         struct nlattr *vf_port;
 801         int vf;
 802         int err;
 803
 804         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 805         if (!vf_ports)
 806                 return -EMSGSIZE;
 807
 808         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 809                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 810                 if (!vf_port)
 811                         goto nla_put_failure;
 812                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 813                         goto nla_put_failure;
 814                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 815                 if (err == -EMSGSIZE)
 816                         goto nla_put_failure;
 817                 if (err) {
 818                         nla_nest_cancel(skb, vf_port);
 819                         continue;
 820                 }
 821                 nla_nest_end(skb, vf_port);
 822         }
 823
 824         nla_nest_end(skb, vf_ports);
 825
 826         return 0;
 827
 828 nla_put_failure:
 829         nla_nest_cancel(skb, vf_ports);
 830         return -EMSGSIZE;
 831 }
 832
 833 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 834 {
 835         struct nlattr *port_self;
 836         int err;
 837
 838         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 839         if (!port_self)
 840                 return -EMSGSIZE;
 841
 842         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 843         if (err) {
 844                 nla_nest_cancel(skb, port_self);
 845                 return (err == -EMSGSIZE) ? err : 0;
 846         }
 847
 848         nla_nest_end(skb, port_self);
 849
 850         return 0;
 851 }
 852
 853 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 854 {
 855         int err;
 856
 857         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 858                 return 0;
 859
 860         err = rtnl_port_self_fill(skb, dev);
 861         if (err)
 862                 return err;
 863
 864         if (dev_num_vf(dev->dev.parent)) {
 865                 err = rtnl_vf_ports_fill(skb, dev);
 866                 if (err)
 867                         return err;
 868         }
 869
 870         return 0;
 871 }
 872
 873 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 874                             int type, u32 pid, u32 seq, u32 change,
 875                             unsigned int flags, u32 ext_filter_mask)
 876 {
 877         struct ifinfomsg *ifm;
 878         struct nlmsghdr *nlh;
 879         struct rtnl_link_stats64 temp;
 880         const struct rtnl_link_stats64 *stats;
 881         struct nlattr *attr, *af_spec;
 882         struct rtnl_af_ops *af_ops;
 883         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 884
 885         ASSERT_RTNL();
 886         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 887         if (nlh == NULL)
 888                 return -EMSGSIZE;
 889
 890         ifm = nlmsg_data(nlh);
 891         ifm->ifi_family = AF_UNSPEC;
 892         ifm->__ifi_pad = 0;
 893         ifm->ifi_type = dev->type;
 894         ifm->ifi_index = dev->ifindex;
 895         ifm->ifi_flags = dev_get_flags(dev);
 896         ifm->ifi_change = change;
 897
 898         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 899             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 900             nla_put_u8(skb, IFLA_OPERSTATE,
 901                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 902             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 903             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 904             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 905             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 906             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 907 #ifdef CONFIG_RPS
 908             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 909 #endif
 910             (dev->ifindex != dev->iflink &&
 911              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 912             (upper_dev &&
 913              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
 914             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
 915             (dev->qdisc &&
 916              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 917             (dev->ifalias &&
 918              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 919                 goto nla_put_failure;
 920
 921         if (1) {
 922                 struct rtnl_link_ifmap map = {
 923                         .mem_start   = dev->mem_start,
 924                         .mem_end     = dev->mem_end,
 925                         .base_addr   = dev->base_addr,
 926                         .irq         = dev->irq,
 927                         .dma         = dev->dma,
 928                         .port        = dev->if_port,
 929                 };
 930                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 931                         goto nla_put_failure;
 932         }
 933
 934         if (dev->addr_len) {
 935                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 936                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 937                         goto nla_put_failure;
 938         }
 939
 940         attr = nla_reserve(skb, IFLA_STATS,
 941                         sizeof(struct rtnl_link_stats));
 942         if (attr == NULL)
 943                 goto nla_put_failure;
 944
 945         stats = dev_get_stats(dev, &temp);
 946         copy_rtnl_link_stats(nla_data(attr), stats);
 947
 948         attr = nla_reserve(skb, IFLA_STATS64,
 949                         sizeof(struct rtnl_link_stats64));
 950         if (attr == NULL)
 951                 goto nla_put_failure;
 952         copy_rtnl_link_stats64(nla_data(attr), stats);
 953
 954         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 955             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 956                 goto nla_put_failure;
 957
 958         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 959             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 960                 int i;
 961
 962                 struct nlattr *vfinfo, *vf;
 963                 int num_vfs = dev_num_vf(dev->dev.parent);
 964
 965                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 966                 if (!vfinfo)
 967                         goto nla_put_failure;
 968                 for (i = 0; i < num_vfs; i++) {
 969                         struct ifla_vf_info ivi;
 970                         struct ifla_vf_mac vf_mac;
 971                         struct ifla_vf_vlan vf_vlan;
 972                         struct ifla_vf_tx_rate vf_tx_rate;
 973                         struct ifla_vf_spoofchk vf_spoofchk;
 974
 975                         /*
 976                          * Not all SR-IOV capable drivers support the
 977                          * spoofcheck query.  Preset to -1 so the user
 978                          * space tool can detect that the driver didn't
 979                          * report anything.
 980                          */
 981                         ivi.spoofchk = -1;
 982                         memset(ivi.mac, 0, sizeof(ivi.mac));
 983                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 984                                 break;
 985                         vf_mac.vf =
 986                                 vf_vlan.vf =
 987                                 vf_tx_rate.vf =
 988                                 vf_spoofchk.vf = ivi.vf;
 989
 990                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 991                         vf_vlan.vlan = ivi.vlan;
 992                         vf_vlan.qos = ivi.qos;
 993                         vf_tx_rate.rate = ivi.tx_rate;
 994                         vf_spoofchk.setting = ivi.spoofchk;
 995                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 996                         if (!vf) {
 997                                 nla_nest_cancel(skb, vfinfo);
 998                                 goto nla_put_failure;
 999                         }
1000                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1001                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1002                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1003                                     &vf_tx_rate) ||
1004                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1005                                     &vf_spoofchk))
1006                                 goto nla_put_failure;
1007                         nla_nest_end(skb, vf);
1008                 }
1009                 nla_nest_end(skb, vfinfo);
1010         }
1011
1012         if (rtnl_port_fill(skb, dev))
1013                 goto nla_put_failure;
1014
1015         if (dev->rtnl_link_ops) {
1016                 if (rtnl_link_fill(skb, dev) < 0)
1017                         goto nla_put_failure;
1018         }
1019
1020         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1021                 goto nla_put_failure;
1022
1023         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1024                 if (af_ops->fill_link_af) {
1025                         struct nlattr *af;
1026                         int err;
1027
1028                         if (!(af = nla_nest_start(skb, af_ops->family)))
1029                                 goto nla_put_failure;
1030
1031                         err = af_ops->fill_link_af(skb, dev);
1032
1033                         /*
1034                          * Caller may return ENODATA to indicate that there
1035                          * was no data to be dumped. This is not an error, it
1036                          * means we should trim the attribute header and
1037                          * continue.
1038                          */
1039                         if (err == -ENODATA)
1040                                 nla_nest_cancel(skb, af);
1041                         else if (err < 0)
1042                                 goto nla_put_failure;
1043
1044                         nla_nest_end(skb, af);
1045                 }
1046         }
1047
1048         nla_nest_end(skb, af_spec);
1049
1050         return nlmsg_end(skb, nlh);
1051
1052 nla_put_failure:
1053         nlmsg_cancel(skb, nlh);
1054         return -EMSGSIZE;
1055 }
1056
1057 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1058 {
1059         struct net *net = sock_net(skb->sk);
1060         int h, s_h;
1061         int idx = 0, s_idx;
1062         struct net_device *dev;
1063         struct hlist_head *head;
1064         struct nlattr *tb[IFLA_MAX+1];
1065         u32 ext_filter_mask = 0;
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         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1074                         ifla_policy) >= 0) {
1075
1076                 if (tb[IFLA_EXT_MASK])
1077                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1078         }
1079
1080         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1081                 idx = 0;
1082                 head = &net->dev_index_head[h];
1083                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1084                         if (idx < s_idx)
1085                                 goto cont;
1086                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1087                                              NETLINK_CB(cb->skb).portid,
1088                                              cb->nlh->nlmsg_seq, 0,
1089                                              NLM_F_MULTI,
1090                                              ext_filter_mask) <= 0)
1091                                 goto out;
1092
1093                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1094 cont:
1095                         idx++;
1096                 }
1097         }
1098 out:
1099         rcu_read_unlock();
1100         cb->args[1] = idx;
1101         cb->args[0] = h;
1102
1103         return skb->len;
1104 }
1105
1106 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1107         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1108         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1109         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1110         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1111         [IFLA_MTU]              = { .type = NLA_U32 },
1112         [IFLA_LINK]             = { .type = NLA_U32 },
1113         [IFLA_MASTER]           = { .type = NLA_U32 },
1114         [IFLA_CARRIER]          = { .type = NLA_U8 },
1115         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1116         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1117         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1118         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1119         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1120         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1121         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1122         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1123         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1124         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1125         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1126         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1127         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1128         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1129         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1130         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1131 };
1132 EXPORT_SYMBOL(ifla_policy);
1133
1134 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1135         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1136         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1137 };
1138
1139 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1140         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1141 };
1142
1143 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1144         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1145                                     .len = sizeof(struct ifla_vf_mac) },
1146         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1147                                     .len = sizeof(struct ifla_vf_vlan) },
1148         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1149                                     .len = sizeof(struct ifla_vf_tx_rate) },
1150         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1151                                     .len = sizeof(struct ifla_vf_spoofchk) },
1152 };
1153
1154 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1155         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1156         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1157                                     .len = PORT_PROFILE_MAX },
1158         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1159                                     .len = sizeof(struct ifla_port_vsi)},
1160         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1161                                       .len = PORT_UUID_MAX },
1162         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1163                                     .len = PORT_UUID_MAX },
1164         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1165         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1166 };
1167
1168 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1169 {
1170         struct net *net;
1171         /* Examine the link attributes and figure out which
1172          * network namespace we are talking about.
1173          */
1174         if (tb[IFLA_NET_NS_PID])
1175                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1176         else if (tb[IFLA_NET_NS_FD])
1177                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1178         else
1179                 net = get_net(src_net);
1180         return net;
1181 }
1182 EXPORT_SYMBOL(rtnl_link_get_net);
1183
1184 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1185 {
1186         if (dev) {
1187                 if (tb[IFLA_ADDRESS] &&
1188                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1189                         return -EINVAL;
1190
1191                 if (tb[IFLA_BROADCAST] &&
1192                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1193                         return -EINVAL;
1194         }
1195
1196         if (tb[IFLA_AF_SPEC]) {
1197                 struct nlattr *af;
1198                 int rem, err;
1199
1200                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1201                         const struct rtnl_af_ops *af_ops;
1202
1203                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1204                                 return -EAFNOSUPPORT;
1205
1206                         if (!af_ops->set_link_af)
1207                                 return -EOPNOTSUPP;
1208
1209                         if (af_ops->validate_link_af) {
1210                                 err = af_ops->validate_link_af(dev, af);
1211                                 if (err < 0)
1212                                         return err;
1213                         }
1214                 }
1215         }
1216
1217         return 0;
1218 }
1219
1220 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1221 {
1222         int rem, err = -EINVAL;
1223         struct nlattr *vf;
1224         const struct net_device_ops *ops = dev->netdev_ops;
1225
1226         nla_for_each_nested(vf, attr, rem) {
1227                 switch (nla_type(vf)) {
1228                 case IFLA_VF_MAC: {
1229                         struct ifla_vf_mac *ivm;
1230                         ivm = nla_data(vf);
1231                         err = -EOPNOTSUPP;
1232                         if (ops->ndo_set_vf_mac)
1233                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1234                                                           ivm->mac);
1235                         break;
1236                 }
1237                 case IFLA_VF_VLAN: {
1238                         struct ifla_vf_vlan *ivv;
1239                         ivv = nla_data(vf);
1240                         err = -EOPNOTSUPP;
1241                         if (ops->ndo_set_vf_vlan)
1242                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1243                                                            ivv->vlan,
1244                                                            ivv->qos);
1245                         break;
1246                 }
1247                 case IFLA_VF_TX_RATE: {
1248                         struct ifla_vf_tx_rate *ivt;
1249                         ivt = nla_data(vf);
1250                         err = -EOPNOTSUPP;
1251                         if (ops->ndo_set_vf_tx_rate)
1252                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1253                                                               ivt->rate);
1254                         break;
1255                 }
1256                 case IFLA_VF_SPOOFCHK: {
1257                         struct ifla_vf_spoofchk *ivs;
1258                         ivs = nla_data(vf);
1259                         err = -EOPNOTSUPP;
1260                         if (ops->ndo_set_vf_spoofchk)
1261                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1262                                                                ivs->setting);
1263                         break;
1264                 }
1265                 default:
1266                         err = -EINVAL;
1267                         break;
1268                 }
1269                 if (err)
1270                         break;
1271         }
1272         return err;
1273 }
1274
1275 static int do_set_master(struct net_device *dev, int ifindex)
1276 {
1277         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1278         const struct net_device_ops *ops;
1279         int err;
1280
1281         if (upper_dev) {
1282                 if (upper_dev->ifindex == ifindex)
1283                         return 0;
1284                 ops = upper_dev->netdev_ops;
1285                 if (ops->ndo_del_slave) {
1286                         err = ops->ndo_del_slave(upper_dev, dev);
1287                         if (err)
1288                                 return err;
1289                 } else {
1290                         return -EOPNOTSUPP;
1291                 }
1292         }
1293
1294         if (ifindex) {
1295                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1296                 if (!upper_dev)
1297                         return -EINVAL;
1298                 ops = upper_dev->netdev_ops;
1299                 if (ops->ndo_add_slave) {
1300                         err = ops->ndo_add_slave(upper_dev, dev);
1301                         if (err)
1302                                 return err;
1303                 } else {
1304                         return -EOPNOTSUPP;
1305                 }
1306         }
1307         return 0;
1308 }
1309
1310 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1311                       struct nlattr **tb, char *ifname, int modified)
1312 {
1313         const struct net_device_ops *ops = dev->netdev_ops;
1314         int err;
1315
1316         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1317                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1318                 if (IS_ERR(net)) {
1319                         err = PTR_ERR(net);
1320                         goto errout;
1321                 }
1322                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1323                         err = -EPERM;
1324                         goto errout;
1325                 }
1326                 err = dev_change_net_namespace(dev, net, ifname);
1327                 put_net(net);
1328                 if (err)
1329                         goto errout;
1330                 modified = 1;
1331         }
1332
1333         if (tb[IFLA_MAP]) {
1334                 struct rtnl_link_ifmap *u_map;
1335                 struct ifmap k_map;
1336
1337                 if (!ops->ndo_set_config) {
1338                         err = -EOPNOTSUPP;
1339                         goto errout;
1340                 }
1341
1342                 if (!netif_device_present(dev)) {
1343                         err = -ENODEV;
1344                         goto errout;
1345                 }
1346
1347                 u_map = nla_data(tb[IFLA_MAP]);
1348                 k_map.mem_start = (unsigned long) u_map->mem_start;
1349                 k_map.mem_end = (unsigned long) u_map->mem_end;
1350                 k_map.base_addr = (unsigned short) u_map->base_addr;
1351                 k_map.irq = (unsigned char) u_map->irq;
1352                 k_map.dma = (unsigned char) u_map->dma;
1353                 k_map.port = (unsigned char) u_map->port;
1354
1355                 err = ops->ndo_set_config(dev, &k_map);
1356                 if (err < 0)
1357                         goto errout;
1358
1359                 modified = 1;
1360         }
1361
1362         if (tb[IFLA_ADDRESS]) {
1363                 struct sockaddr *sa;
1364                 int len;
1365
1366                 len = sizeof(sa_family_t) + dev->addr_len;
1367                 sa = kmalloc(len, GFP_KERNEL);
1368                 if (!sa) {
1369                         err = -ENOMEM;
1370                         goto errout;
1371                 }
1372                 sa->sa_family = dev->type;
1373                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1374                        dev->addr_len);
1375                 err = dev_set_mac_address(dev, sa);
1376                 kfree(sa);
1377                 if (err)
1378                         goto errout;
1379                 modified = 1;
1380         }
1381
1382         if (tb[IFLA_MTU]) {
1383                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1384                 if (err < 0)
1385                         goto errout;
1386                 modified = 1;
1387         }
1388
1389         if (tb[IFLA_GROUP]) {
1390                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1391                 modified = 1;
1392         }
1393
1394         /*
1395          * Interface selected by interface index but interface
1396          * name provided implies that a name change has been
1397          * requested.
1398          */
1399         if (ifm->ifi_index > 0 && ifname[0]) {
1400                 err = dev_change_name(dev, ifname);
1401                 if (err < 0)
1402                         goto errout;
1403                 modified = 1;
1404         }
1405
1406         if (tb[IFLA_IFALIAS]) {
1407                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1408                                     nla_len(tb[IFLA_IFALIAS]));
1409                 if (err < 0)
1410                         goto errout;
1411                 modified = 1;
1412         }
1413
1414         if (tb[IFLA_BROADCAST]) {
1415                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1416                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1417         }
1418
1419         if (ifm->ifi_flags || ifm->ifi_change) {
1420                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1421                 if (err < 0)
1422                         goto errout;
1423         }
1424
1425         if (tb[IFLA_MASTER]) {
1426                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1427                 if (err)
1428                         goto errout;
1429                 modified = 1;
1430         }
1431
1432         if (tb[IFLA_CARRIER]) {
1433                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1434                 if (err)
1435                         goto errout;
1436                 modified = 1;
1437         }
1438
1439         if (tb[IFLA_TXQLEN])
1440                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1441
1442         if (tb[IFLA_OPERSTATE])
1443                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1444
1445         if (tb[IFLA_LINKMODE]) {
1446                 write_lock_bh(&dev_base_lock);
1447                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1448                 write_unlock_bh(&dev_base_lock);
1449         }
1450
1451         if (tb[IFLA_VFINFO_LIST]) {
1452                 struct nlattr *attr;
1453                 int rem;
1454                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1455                         if (nla_type(attr) != IFLA_VF_INFO) {
1456                                 err = -EINVAL;
1457                                 goto errout;
1458                         }
1459                         err = do_setvfinfo(dev, attr);
1460                         if (err < 0)
1461                                 goto errout;
1462                         modified = 1;
1463                 }
1464         }
1465         err = 0;
1466
1467         if (tb[IFLA_VF_PORTS]) {
1468                 struct nlattr *port[IFLA_PORT_MAX+1];
1469                 struct nlattr *attr;
1470                 int vf;
1471                 int rem;
1472
1473                 err = -EOPNOTSUPP;
1474                 if (!ops->ndo_set_vf_port)
1475                         goto errout;
1476
1477                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1478                         if (nla_type(attr) != IFLA_VF_PORT)
1479                                 continue;
1480                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1481                                 attr, ifla_port_policy);
1482                         if (err < 0)
1483                                 goto errout;
1484                         if (!port[IFLA_PORT_VF]) {
1485                                 err = -EOPNOTSUPP;
1486                                 goto errout;
1487                         }
1488                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1489                         err = ops->ndo_set_vf_port(dev, vf, port);
1490                         if (err < 0)
1491                                 goto errout;
1492                         modified = 1;
1493                 }
1494         }
1495         err = 0;
1496
1497         if (tb[IFLA_PORT_SELF]) {
1498                 struct nlattr *port[IFLA_PORT_MAX+1];
1499
1500                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1501                         tb[IFLA_PORT_SELF], ifla_port_policy);
1502                 if (err < 0)
1503                         goto errout;
1504
1505                 err = -EOPNOTSUPP;
1506                 if (ops->ndo_set_vf_port)
1507                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1508                 if (err < 0)
1509                         goto errout;
1510                 modified = 1;
1511         }
1512
1513         if (tb[IFLA_AF_SPEC]) {
1514                 struct nlattr *af;
1515                 int rem;
1516
1517                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1518                         const struct rtnl_af_ops *af_ops;
1519
1520                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1521                                 BUG();
1522
1523                         err = af_ops->set_link_af(dev, af);
1524                         if (err < 0)
1525                                 goto errout;
1526
1527                         modified = 1;
1528                 }
1529         }
1530         err = 0;
1531
1532 errout:
1533         if (err < 0 && modified)
1534                 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",
1535                                      dev->name);
1536
1537         return err;
1538 }
1539
1540 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1541 {
1542         struct net *net = sock_net(skb->sk);
1543         struct ifinfomsg *ifm;
1544         struct net_device *dev;
1545         int err;
1546         struct nlattr *tb[IFLA_MAX+1];
1547         char ifname[IFNAMSIZ];
1548
1549         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1550         if (err < 0)
1551                 goto errout;
1552
1553         if (tb[IFLA_IFNAME])
1554                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1555         else
1556                 ifname[0] = '\0';
1557
1558         err = -EINVAL;
1559         ifm = nlmsg_data(nlh);
1560         if (ifm->ifi_index > 0)
1561                 dev = __dev_get_by_index(net, ifm->ifi_index);
1562         else if (tb[IFLA_IFNAME])
1563                 dev = __dev_get_by_name(net, ifname);
1564         else
1565                 goto errout;
1566
1567         if (dev == NULL) {
1568                 err = -ENODEV;
1569                 goto errout;
1570         }
1571
1572         err = validate_linkmsg(dev, tb);
1573         if (err < 0)
1574                 goto errout;
1575
1576         err = do_setlink(dev, ifm, tb, ifname, 0);
1577 errout:
1578         return err;
1579 }
1580
1581 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1582 {
1583         struct net *net = sock_net(skb->sk);
1584         const struct rtnl_link_ops *ops;
1585         struct net_device *dev;
1586         struct ifinfomsg *ifm;
1587         char ifname[IFNAMSIZ];
1588         struct nlattr *tb[IFLA_MAX+1];
1589         int err;
1590         LIST_HEAD(list_kill);
1591
1592         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1593         if (err < 0)
1594                 return err;
1595
1596         if (tb[IFLA_IFNAME])
1597                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1598
1599         ifm = nlmsg_data(nlh);
1600         if (ifm->ifi_index > 0)
1601                 dev = __dev_get_by_index(net, ifm->ifi_index);
1602         else if (tb[IFLA_IFNAME])
1603                 dev = __dev_get_by_name(net, ifname);
1604         else
1605                 return -EINVAL;
1606
1607         if (!dev)
1608                 return -ENODEV;
1609
1610         ops = dev->rtnl_link_ops;
1611         if (!ops)
1612                 return -EOPNOTSUPP;
1613
1614         ops->dellink(dev, &list_kill);
1615         unregister_netdevice_many(&list_kill);
1616         list_del(&list_kill);
1617         return 0;
1618 }
1619
1620 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1621 {
1622         unsigned int old_flags;
1623         int err;
1624
1625         old_flags = dev->flags;
1626         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1627                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1628                 if (err < 0)
1629                         return err;
1630         }
1631
1632         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1633         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1634
1635         __dev_notify_flags(dev, old_flags);
1636         return 0;
1637 }
1638 EXPORT_SYMBOL(rtnl_configure_link);
1639
1640 struct net_device *rtnl_create_link(struct net *net,
1641         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1642 {
1643         int err;
1644         struct net_device *dev;
1645         unsigned int num_tx_queues = 1;
1646         unsigned int num_rx_queues = 1;
1647
1648         if (tb[IFLA_NUM_TX_QUEUES])
1649                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1650         else if (ops->get_num_tx_queues)
1651                 num_tx_queues = ops->get_num_tx_queues();
1652
1653         if (tb[IFLA_NUM_RX_QUEUES])
1654                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1655         else if (ops->get_num_rx_queues)
1656                 num_rx_queues = ops->get_num_rx_queues();
1657
1658         err = -ENOMEM;
1659         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1660                                num_tx_queues, num_rx_queues);
1661         if (!dev)
1662                 goto err;
1663
1664         dev_net_set(dev, net);
1665         dev->rtnl_link_ops = ops;
1666         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1667
1668         if (tb[IFLA_MTU])
1669                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1670         if (tb[IFLA_ADDRESS]) {
1671                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1672                                 nla_len(tb[IFLA_ADDRESS]));
1673                 dev->addr_assign_type = NET_ADDR_SET;
1674         }
1675         if (tb[IFLA_BROADCAST])
1676                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1677                                 nla_len(tb[IFLA_BROADCAST]));
1678         if (tb[IFLA_TXQLEN])
1679                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1680         if (tb[IFLA_OPERSTATE])
1681                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1682         if (tb[IFLA_LINKMODE])
1683                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1684         if (tb[IFLA_GROUP])
1685                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1686
1687         return dev;
1688
1689 err:
1690         return ERR_PTR(err);
1691 }
1692 EXPORT_SYMBOL(rtnl_create_link);
1693
1694 static int rtnl_group_changelink(struct net *net, int group,
1695                 struct ifinfomsg *ifm,
1696                 struct nlattr **tb)
1697 {
1698         struct net_device *dev;
1699         int err;
1700
1701         for_each_netdev(net, dev) {
1702                 if (dev->group == group) {
1703                         err = do_setlink(dev, ifm, tb, NULL, 0);
1704                         if (err < 0)
1705                                 return err;
1706                 }
1707         }
1708
1709         return 0;
1710 }
1711
1712 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1713 {
1714         struct net *net = sock_net(skb->sk);
1715         const struct rtnl_link_ops *ops;
1716         struct net_device *dev;
1717         struct ifinfomsg *ifm;
1718         char kind[MODULE_NAME_LEN];
1719         char ifname[IFNAMSIZ];
1720         struct nlattr *tb[IFLA_MAX+1];
1721         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1722         int err;
1723
1724 #ifdef CONFIG_MODULES
1725 replay:
1726 #endif
1727         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1728         if (err < 0)
1729                 return err;
1730
1731         if (tb[IFLA_IFNAME])
1732                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1733         else
1734                 ifname[0] = '\0';
1735
1736         ifm = nlmsg_data(nlh);
1737         if (ifm->ifi_index > 0)
1738                 dev = __dev_get_by_index(net, ifm->ifi_index);
1739         else {
1740                 if (ifname[0])
1741                         dev = __dev_get_by_name(net, ifname);
1742                 else
1743                         dev = NULL;
1744         }
1745
1746         err = validate_linkmsg(dev, tb);
1747         if (err < 0)
1748                 return err;
1749
1750         if (tb[IFLA_LINKINFO]) {
1751                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1752                                        tb[IFLA_LINKINFO], ifla_info_policy);
1753                 if (err < 0)
1754                         return err;
1755         } else
1756                 memset(linkinfo, 0, sizeof(linkinfo));
1757
1758         if (linkinfo[IFLA_INFO_KIND]) {
1759                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1760                 ops = rtnl_link_ops_get(kind);
1761         } else {
1762                 kind[0] = '\0';
1763                 ops = NULL;
1764         }
1765
1766         if (1) {
1767                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1768                 struct net *dest_net;
1769
1770                 if (ops) {
1771                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1772                                 err = nla_parse_nested(attr, ops->maxtype,
1773                                                        linkinfo[IFLA_INFO_DATA],
1774                                                        ops->policy);
1775                                 if (err < 0)
1776                                         return err;
1777                                 data = attr;
1778                         }
1779                         if (ops->validate) {
1780                                 err = ops->validate(tb, data);
1781                                 if (err < 0)
1782                                         return err;
1783                         }
1784                 }
1785
1786                 if (dev) {
1787                         int modified = 0;
1788
1789                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1790                                 return -EEXIST;
1791                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1792                                 return -EOPNOTSUPP;
1793
1794                         if (linkinfo[IFLA_INFO_DATA]) {
1795                                 if (!ops || ops != dev->rtnl_link_ops ||
1796                                     !ops->changelink)
1797                                         return -EOPNOTSUPP;
1798
1799                                 err = ops->changelink(dev, tb, data);
1800                                 if (err < 0)
1801                                         return err;
1802                                 modified = 1;
1803                         }
1804
1805                         return do_setlink(dev, ifm, tb, ifname, modified);
1806                 }
1807
1808                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1809                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1810                                 return rtnl_group_changelink(net,
1811                                                 nla_get_u32(tb[IFLA_GROUP]),
1812                                                 ifm, tb);
1813                         return -ENODEV;
1814                 }
1815
1816                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1817                         return -EOPNOTSUPP;
1818
1819                 if (!ops) {
1820 #ifdef CONFIG_MODULES
1821                         if (kind[0]) {
1822                                 __rtnl_unlock();
1823                                 request_module("rtnl-link-%s", kind);
1824                                 rtnl_lock();
1825                                 ops = rtnl_link_ops_get(kind);
1826                                 if (ops)
1827                                         goto replay;
1828                         }
1829 #endif
1830                         return -EOPNOTSUPP;
1831                 }
1832
1833                 if (!ifname[0])
1834                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1835
1836                 dest_net = rtnl_link_get_net(net, tb);
1837                 if (IS_ERR(dest_net))
1838                         return PTR_ERR(dest_net);
1839
1840                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1841                 if (IS_ERR(dev)) {
1842                         err = PTR_ERR(dev);
1843                         goto out;
1844                 }
1845
1846                 dev->ifindex = ifm->ifi_index;
1847
1848                 if (ops->newlink)
1849                         err = ops->newlink(net, dev, tb, data);
1850                 else
1851                         err = register_netdevice(dev);
1852
1853                 if (err < 0 && !IS_ERR(dev))
1854                         free_netdev(dev);
1855                 if (err < 0)
1856                         goto out;
1857
1858                 err = rtnl_configure_link(dev, ifm);
1859                 if (err < 0)
1860                         unregister_netdevice(dev);
1861 out:
1862                 put_net(dest_net);
1863                 return err;
1864         }
1865 }
1866
1867 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1868 {
1869         struct net *net = sock_net(skb->sk);
1870         struct ifinfomsg *ifm;
1871         char ifname[IFNAMSIZ];
1872         struct nlattr *tb[IFLA_MAX+1];
1873         struct net_device *dev = NULL;
1874         struct sk_buff *nskb;
1875         int err;
1876         u32 ext_filter_mask = 0;
1877
1878         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1879         if (err < 0)
1880                 return err;
1881
1882         if (tb[IFLA_IFNAME])
1883                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1884
1885         if (tb[IFLA_EXT_MASK])
1886                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1887
1888         ifm = nlmsg_data(nlh);
1889         if (ifm->ifi_index > 0)
1890                 dev = __dev_get_by_index(net, ifm->ifi_index);
1891         else if (tb[IFLA_IFNAME])
1892                 dev = __dev_get_by_name(net, ifname);
1893         else
1894                 return -EINVAL;
1895
1896         if (dev == NULL)
1897                 return -ENODEV;
1898
1899         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1900         if (nskb == NULL)
1901                 return -ENOBUFS;
1902
1903         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1904                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1905         if (err < 0) {
1906                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1907                 WARN_ON(err == -EMSGSIZE);
1908                 kfree_skb(nskb);
1909         } else
1910                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1911
1912         return err;
1913 }
1914
1915 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1916 {
1917         struct net *net = sock_net(skb->sk);
1918         struct net_device *dev;
1919         struct nlattr *tb[IFLA_MAX+1];
1920         u32 ext_filter_mask = 0;
1921         u16 min_ifinfo_dump_size = 0;
1922
1923         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1924                         ifla_policy) >= 0) {
1925                 if (tb[IFLA_EXT_MASK])
1926                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1927         }
1928
1929         if (!ext_filter_mask)
1930                 return NLMSG_GOODSIZE;
1931         /*
1932          * traverse the list of net devices and compute the minimum
1933          * buffer size based upon the filter mask.
1934          */
1935         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1936                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1937                                              if_nlmsg_size(dev,
1938                                                            ext_filter_mask));
1939         }
1940
1941         return min_ifinfo_dump_size;
1942 }
1943
1944 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1945 {
1946         int idx;
1947         int s_idx = cb->family;
1948
1949         if (s_idx == 0)
1950                 s_idx = 1;
1951         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1952                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1953                 if (idx < s_idx || idx == PF_PACKET)
1954                         continue;
1955                 if (rtnl_msg_handlers[idx] == NULL ||
1956                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1957                         continue;
1958                 if (idx > s_idx)
1959                         memset(&cb->args[0], 0, sizeof(cb->args));
1960                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1961                         break;
1962         }
1963         cb->family = idx;
1964
1965         return skb->len;
1966 }
1967
1968 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1969 {
1970         struct net *net = dev_net(dev);
1971         struct sk_buff *skb;
1972         int err = -ENOBUFS;
1973         size_t if_info_size;
1974
1975         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1976         if (skb == NULL)
1977                 goto errout;
1978
1979         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1980         if (err < 0) {
1981                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1982                 WARN_ON(err == -EMSGSIZE);
1983                 kfree_skb(skb);
1984                 goto errout;
1985         }
1986         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1987         return;
1988 errout:
1989         if (err < 0)
1990                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1991 }
1992 EXPORT_SYMBOL(rtmsg_ifinfo);
1993
1994 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1995                                    struct net_device *dev,
1996                                    u8 *addr, u32 pid, u32 seq,
1997                                    int type, unsigned int flags)
1998 {
1999         struct nlmsghdr *nlh;
2000         struct ndmsg *ndm;
2001
2002         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2003         if (!nlh)
2004                 return -EMSGSIZE;
2005
2006         ndm = nlmsg_data(nlh);
2007         ndm->ndm_family  = AF_BRIDGE;
2008         ndm->ndm_pad1    = 0;
2009         ndm->ndm_pad2    = 0;
2010         ndm->ndm_flags   = flags;
2011         ndm->ndm_type    = 0;
2012         ndm->ndm_ifindex = dev->ifindex;
2013         ndm->ndm_state   = NUD_PERMANENT;
2014
2015         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2016                 goto nla_put_failure;
2017
2018         return nlmsg_end(skb, nlh);
2019
2020 nla_put_failure:
2021         nlmsg_cancel(skb, nlh);
2022         return -EMSGSIZE;
2023 }
2024
2025 static inline size_t rtnl_fdb_nlmsg_size(void)
2026 {
2027         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2028 }
2029
2030 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2031 {
2032         struct net *net = dev_net(dev);
2033         struct sk_buff *skb;
2034         int err = -ENOBUFS;
2035
2036         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2037         if (!skb)
2038                 goto errout;
2039
2040         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2041         if (err < 0) {
2042                 kfree_skb(skb);
2043                 goto errout;
2044         }
2045
2046         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2047         return;
2048 errout:
2049         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2050 }
2051
2052 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2053 {
2054         struct net *net = sock_net(skb->sk);
2055         struct ndmsg *ndm;
2056         struct nlattr *tb[NDA_MAX+1];
2057         struct net_device *dev;
2058         u8 *addr;
2059         int err;
2060
2061         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2062         if (err < 0)
2063                 return err;
2064
2065         ndm = nlmsg_data(nlh);
2066         if (ndm->ndm_ifindex == 0) {
2067                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2068                 return -EINVAL;
2069         }
2070
2071         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2072         if (dev == NULL) {
2073                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2074                 return -ENODEV;
2075         }
2076
2077         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2078                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2079                 return -EINVAL;
2080         }
2081
2082         addr = nla_data(tb[NDA_LLADDR]);
2083         if (!is_valid_ether_addr(addr)) {
2084                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2085                 return -EINVAL;
2086         }
2087
2088         err = -EOPNOTSUPP;
2089
2090         /* Support fdb on master device the net/bridge default case */
2091         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2092             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2093                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2094                 const struct net_device_ops *ops = br_dev->netdev_ops;
2095
2096                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2097                 if (err)
2098                         goto out;
2099                 else
2100                         ndm->ndm_flags &= ~NTF_MASTER;
2101         }
2102
2103         /* Embedded bridge, macvlan, and any other device support */
2104         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2105                 err = dev->netdev_ops->ndo_fdb_add(ndm, tb,
2106                                                    dev, addr,
2107                                                    nlh->nlmsg_flags);
2108
2109                 if (!err) {
2110                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2111                         ndm->ndm_flags &= ~NTF_SELF;
2112                 }
2113         }
2114 out:
2115         return err;
2116 }
2117
2118 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2119 {
2120         struct net *net = sock_net(skb->sk);
2121         struct ndmsg *ndm;
2122         struct nlattr *tb[NDA_MAX+1];
2123         struct net_device *dev;
2124         int err = -EINVAL;
2125         __u8 *addr;
2126
2127         if (!capable(CAP_NET_ADMIN))
2128                 return -EPERM;
2129
2130         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2131         if (err < 0)
2132                 return err;
2133
2134         ndm = nlmsg_data(nlh);
2135         if (ndm->ndm_ifindex == 0) {
2136                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2137                 return -EINVAL;
2138         }
2139
2140         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2141         if (dev == NULL) {
2142                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2143                 return -ENODEV;
2144         }
2145
2146         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2147                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2148                 return -EINVAL;
2149         }
2150
2151         addr = nla_data(tb[NDA_LLADDR]);
2152         if (!is_valid_ether_addr(addr)) {
2153                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2154                 return -EINVAL;
2155         }
2156
2157         err = -EOPNOTSUPP;
2158
2159         /* Support fdb on master device the net/bridge default case */
2160         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2161             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2162                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2163                 const struct net_device_ops *ops = br_dev->netdev_ops;
2164
2165                 if (ops->ndo_fdb_del)
2166                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2167
2168                 if (err)
2169                         goto out;
2170                 else
2171                         ndm->ndm_flags &= ~NTF_MASTER;
2172         }
2173
2174         /* Embedded bridge, macvlan, and any other device support */
2175         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2176                 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2177
2178                 if (!err) {
2179                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2180                         ndm->ndm_flags &= ~NTF_SELF;
2181                 }
2182         }
2183 out:
2184         return err;
2185 }
2186
2187 static int nlmsg_populate_fdb(struct sk_buff *skb,
2188                               struct netlink_callback *cb,
2189                               struct net_device *dev,
2190                               int *idx,
2191                               struct netdev_hw_addr_list *list)
2192 {
2193         struct netdev_hw_addr *ha;
2194         int err;
2195         u32 portid, seq;
2196
2197         portid = NETLINK_CB(cb->skb).portid;
2198         seq = cb->nlh->nlmsg_seq;
2199
2200         list_for_each_entry(ha, &list->list, list) {
2201                 if (*idx < cb->args[0])
2202                         goto skip;
2203
2204                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2205                                               portid, seq,
2206                                               RTM_NEWNEIGH, NTF_SELF);
2207                 if (err < 0)
2208                         return err;
2209 skip:
2210                 *idx += 1;
2211         }
2212         return 0;
2213 }
2214
2215 /**
2216  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2217  * @nlh: netlink message header
2218  * @dev: netdevice
2219  *
2220  * Default netdevice operation to dump the existing unicast address list.
2221  * Returns zero on success.
2222  */
2223 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2224                       struct netlink_callback *cb,
2225                       struct net_device *dev,
2226                       int idx)
2227 {
2228         int err;
2229
2230         netif_addr_lock_bh(dev);
2231         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2232         if (err)
2233                 goto out;
2234         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2235 out:
2236         netif_addr_unlock_bh(dev);
2237         return idx;
2238 }
2239 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2240
2241 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2242 {
2243         int idx = 0;
2244         struct net *net = sock_net(skb->sk);
2245         struct net_device *dev;
2246
2247         rcu_read_lock();
2248         for_each_netdev_rcu(net, dev) {
2249                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2250                         struct net_device *br_dev;
2251                         const struct net_device_ops *ops;
2252
2253                         br_dev = netdev_master_upper_dev_get(dev);
2254                         ops = br_dev->netdev_ops;
2255                         if (ops->ndo_fdb_dump)
2256                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2257                 }
2258
2259                 if (dev->netdev_ops->ndo_fdb_dump)
2260                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2261         }
2262         rcu_read_unlock();
2263
2264         cb->args[0] = idx;
2265         return skb->len;
2266 }
2267
2268 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2269                             struct net_device *dev, u16 mode)
2270 {
2271         struct nlmsghdr *nlh;
2272         struct ifinfomsg *ifm;
2273         struct nlattr *br_afspec;
2274         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2275         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2276
2277         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2278         if (nlh == NULL)
2279                 return -EMSGSIZE;
2280
2281         ifm = nlmsg_data(nlh);
2282         ifm->ifi_family = AF_BRIDGE;
2283         ifm->__ifi_pad = 0;
2284         ifm->ifi_type = dev->type;
2285         ifm->ifi_index = dev->ifindex;
2286         ifm->ifi_flags = dev_get_flags(dev);
2287         ifm->ifi_change = 0;
2288
2289
2290         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2291             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2292             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2293             (br_dev &&
2294              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2295             (dev->addr_len &&
2296              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2297             (dev->ifindex != dev->iflink &&
2298              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2299                 goto nla_put_failure;
2300
2301         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2302         if (!br_afspec)
2303                 goto nla_put_failure;
2304
2305         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2306             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2307                 nla_nest_cancel(skb, br_afspec);
2308                 goto nla_put_failure;
2309         }
2310         nla_nest_end(skb, br_afspec);
2311
2312         return nlmsg_end(skb, nlh);
2313 nla_put_failure:
2314         nlmsg_cancel(skb, nlh);
2315         return -EMSGSIZE;
2316 }
2317 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2318
2319 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2320 {
2321         struct net *net = sock_net(skb->sk);
2322         struct net_device *dev;
2323         int idx = 0;
2324         u32 portid = NETLINK_CB(cb->skb).portid;
2325         u32 seq = cb->nlh->nlmsg_seq;
2326         struct nlattr *extfilt;
2327         u32 filter_mask = 0;
2328
2329         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2330                                   IFLA_EXT_MASK);
2331         if (extfilt)
2332                 filter_mask = nla_get_u32(extfilt);
2333
2334         rcu_read_lock();
2335         for_each_netdev_rcu(net, dev) {
2336                 const struct net_device_ops *ops = dev->netdev_ops;
2337                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2338
2339                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2340                         if (idx >= cb->args[0] &&
2341                             br_dev->netdev_ops->ndo_bridge_getlink(
2342                                     skb, portid, seq, dev, filter_mask) < 0)
2343                                 break;
2344                         idx++;
2345                 }
2346
2347                 if (ops->ndo_bridge_getlink) {
2348                         if (idx >= cb->args[0] &&
2349                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2350                                                     filter_mask) < 0)
2351                                 break;
2352                         idx++;
2353                 }
2354         }
2355         rcu_read_unlock();
2356         cb->args[0] = idx;
2357
2358         return skb->len;
2359 }
2360
2361 static inline size_t bridge_nlmsg_size(void)
2362 {
2363         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2364                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2365                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2366                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2367                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2368                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2369                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2370                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2371                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2372                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2373                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2374 }
2375
2376 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2377 {
2378         struct net *net = dev_net(dev);
2379         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2380         struct sk_buff *skb;
2381         int err = -EOPNOTSUPP;
2382
2383         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2384         if (!skb) {
2385                 err = -ENOMEM;
2386                 goto errout;
2387         }
2388
2389         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2390             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2391                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2392                 if (err < 0)
2393                         goto errout;
2394         }
2395
2396         if ((flags & BRIDGE_FLAGS_SELF) &&
2397             dev->netdev_ops->ndo_bridge_getlink) {
2398                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2399                 if (err < 0)
2400                         goto errout;
2401         }
2402
2403         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2404         return 0;
2405 errout:
2406         WARN_ON(err == -EMSGSIZE);
2407         kfree_skb(skb);
2408         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2409         return err;
2410 }
2411
2412 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2413                                void *arg)
2414 {
2415         struct net *net = sock_net(skb->sk);
2416         struct ifinfomsg *ifm;
2417         struct net_device *dev;
2418         struct nlattr *br_spec, *attr = NULL;
2419         int rem, err = -EOPNOTSUPP;
2420         u16 oflags, flags = 0;
2421         bool have_flags = false;
2422
2423         if (nlmsg_len(nlh) < sizeof(*ifm))
2424                 return -EINVAL;
2425
2426         ifm = nlmsg_data(nlh);
2427         if (ifm->ifi_family != AF_BRIDGE)
2428                 return -EPFNOSUPPORT;
2429
2430         dev = __dev_get_by_index(net, ifm->ifi_index);
2431         if (!dev) {
2432                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2433                 return -ENODEV;
2434         }
2435
2436         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2437         if (br_spec) {
2438                 nla_for_each_nested(attr, br_spec, rem) {
2439                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2440                                 have_flags = true;
2441                                 flags = nla_get_u16(attr);
2442                                 break;
2443                         }
2444                 }
2445         }
2446
2447         oflags = flags;
2448
2449         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2450                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2451
2452                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2453                         err = -EOPNOTSUPP;
2454                         goto out;
2455                 }
2456
2457                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2458                 if (err)
2459                         goto out;
2460
2461                 flags &= ~BRIDGE_FLAGS_MASTER;
2462         }
2463
2464         if ((flags & BRIDGE_FLAGS_SELF)) {
2465                 if (!dev->netdev_ops->ndo_bridge_setlink)
2466                         err = -EOPNOTSUPP;
2467                 else
2468                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2469
2470                 if (!err)
2471                         flags &= ~BRIDGE_FLAGS_SELF;
2472         }
2473
2474         if (have_flags)
2475                 memcpy(nla_data(attr), &flags, sizeof(flags));
2476         /* Generate event to notify upper layer of bridge change */
2477         if (!err)
2478                 err = rtnl_bridge_notify(dev, oflags);
2479 out:
2480         return err;
2481 }
2482
2483 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2484                                void *arg)
2485 {
2486         struct net *net = sock_net(skb->sk);
2487         struct ifinfomsg *ifm;
2488         struct net_device *dev;
2489         struct nlattr *br_spec, *attr = NULL;
2490         int rem, err = -EOPNOTSUPP;
2491         u16 oflags, flags = 0;
2492         bool have_flags = false;
2493
2494         if (nlmsg_len(nlh) < sizeof(*ifm))
2495                 return -EINVAL;
2496
2497         ifm = nlmsg_data(nlh);
2498         if (ifm->ifi_family != AF_BRIDGE)
2499                 return -EPFNOSUPPORT;
2500
2501         dev = __dev_get_by_index(net, ifm->ifi_index);
2502         if (!dev) {
2503                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2504                 return -ENODEV;
2505         }
2506
2507         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2508         if (br_spec) {
2509                 nla_for_each_nested(attr, br_spec, rem) {
2510                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2511                                 have_flags = true;
2512                                 flags = nla_get_u16(attr);
2513                                 break;
2514                         }
2515                 }
2516         }
2517
2518         oflags = flags;
2519
2520         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2521                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2522
2523                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2524                         err = -EOPNOTSUPP;
2525                         goto out;
2526                 }
2527
2528                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2529                 if (err)
2530                         goto out;
2531
2532                 flags &= ~BRIDGE_FLAGS_MASTER;
2533         }
2534
2535         if ((flags & BRIDGE_FLAGS_SELF)) {
2536                 if (!dev->netdev_ops->ndo_bridge_dellink)
2537                         err = -EOPNOTSUPP;
2538                 else
2539                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2540
2541                 if (!err)
2542                         flags &= ~BRIDGE_FLAGS_SELF;
2543         }
2544
2545         if (have_flags)
2546                 memcpy(nla_data(attr), &flags, sizeof(flags));
2547         /* Generate event to notify upper layer of bridge change */
2548         if (!err)
2549                 err = rtnl_bridge_notify(dev, oflags);
2550 out:
2551         return err;
2552 }
2553
2554 /* Protected by RTNL sempahore.  */
2555 static struct rtattr **rta_buf;
2556 static int rtattr_max;
2557
2558 /* Process one rtnetlink message. */
2559
2560 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2561 {
2562         struct net *net = sock_net(skb->sk);
2563         rtnl_doit_func doit;
2564         int sz_idx, kind;
2565         int min_len;
2566         int family;
2567         int type;
2568         int err;
2569
2570         type = nlh->nlmsg_type;
2571         if (type > RTM_MAX)
2572                 return -EOPNOTSUPP;
2573
2574         type -= RTM_BASE;
2575
2576         /* All the messages must have at least 1 byte length */
2577         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2578                 return 0;
2579
2580         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2581         sz_idx = type>>2;
2582         kind = type&3;
2583
2584         if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2585                 return -EPERM;
2586
2587         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2588                 struct sock *rtnl;
2589                 rtnl_dumpit_func dumpit;
2590                 rtnl_calcit_func calcit;
2591                 u16 min_dump_alloc = 0;
2592
2593                 dumpit = rtnl_get_dumpit(family, type);
2594                 if (dumpit == NULL)
2595                         return -EOPNOTSUPP;
2596                 calcit = rtnl_get_calcit(family, type);
2597                 if (calcit)
2598                         min_dump_alloc = calcit(skb, nlh);
2599
2600                 __rtnl_unlock();
2601                 rtnl = net->rtnl;
2602                 {
2603                         struct netlink_dump_control c = {
2604                                 .dump           = dumpit,
2605                                 .min_dump_alloc = min_dump_alloc,
2606                         };
2607                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2608                 }
2609                 rtnl_lock();
2610                 return err;
2611         }
2612
2613         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2614
2615         min_len = rtm_min[sz_idx];
2616         if (nlh->nlmsg_len < min_len)
2617                 return -EINVAL;
2618
2619         if (nlh->nlmsg_len > min_len) {
2620                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2621                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2622
2623                 while (RTA_OK(attr, attrlen)) {
2624                         unsigned int flavor = attr->rta_type & NLA_TYPE_MASK;
2625                         if (flavor) {
2626                                 if (flavor > rta_max[sz_idx])
2627                                         return -EINVAL;
2628                                 rta_buf[flavor-1] = attr;
2629                         }
2630                         attr = RTA_NEXT(attr, attrlen);
2631                 }
2632         }
2633
2634         doit = rtnl_get_doit(family, type);
2635         if (doit == NULL)
2636                 return -EOPNOTSUPP;
2637
2638         return doit(skb, nlh, (void *)&rta_buf[0]);
2639 }
2640
2641 static void rtnetlink_rcv(struct sk_buff *skb)
2642 {
2643         rtnl_lock();
2644         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2645         rtnl_unlock();
2646 }
2647
2648 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2649 {
2650         struct net_device *dev = ptr;
2651
2652         switch (event) {
2653         case NETDEV_UP:
2654         case NETDEV_DOWN:
2655         case NETDEV_PRE_UP:
2656         case NETDEV_POST_INIT:
2657         case NETDEV_REGISTER:
2658         case NETDEV_CHANGE:
2659         case NETDEV_PRE_TYPE_CHANGE:
2660         case NETDEV_GOING_DOWN:
2661         case NETDEV_UNREGISTER:
2662         case NETDEV_UNREGISTER_FINAL:
2663         case NETDEV_RELEASE:
2664         case NETDEV_JOIN:
2665                 break;
2666         default:
2667                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2668                 break;
2669         }
2670         return NOTIFY_DONE;
2671 }
2672
2673 static struct notifier_block rtnetlink_dev_notifier = {
2674         .notifier_call  = rtnetlink_event,
2675 };
2676
2677
2678 static int __net_init rtnetlink_net_init(struct net *net)
2679 {
2680         struct sock *sk;
2681         struct netlink_kernel_cfg cfg = {
2682                 .groups         = RTNLGRP_MAX,
2683                 .input          = rtnetlink_rcv,
2684                 .cb_mutex       = &rtnl_mutex,
2685                 .flags          = NL_CFG_F_NONROOT_RECV,
2686         };
2687
2688         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2689         if (!sk)
2690                 return -ENOMEM;
2691         net->rtnl = sk;
2692         return 0;
2693 }
2694
2695 static void __net_exit rtnetlink_net_exit(struct net *net)
2696 {
2697         netlink_kernel_release(net->rtnl);
2698         net->rtnl = NULL;
2699 }
2700
2701 static struct pernet_operations rtnetlink_net_ops = {
2702         .init = rtnetlink_net_init,
2703         .exit = rtnetlink_net_exit,
2704 };
2705
2706 void __init rtnetlink_init(void)
2707 {
2708         int i;
2709
2710         rtattr_max = 0;
2711         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2712                 if (rta_max[i] > rtattr_max)
2713                         rtattr_max = rta_max[i];
2714         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2715         if (!rta_buf)
2716                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2717
2718         if (register_pernet_subsys(&rtnetlink_net_ops))
2719                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2720
2721         register_netdevice_notifier(&rtnetlink_dev_notifier);
2722
2723         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2724                       rtnl_dump_ifinfo, rtnl_calcit);
2725         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2726         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2727         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2728
2729         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2730         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2731
2732         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2733         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2734         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2735
2736         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2737         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2738         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2739 }
2740