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                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 983                                 break;
 984                         vf_mac.vf =
 985                                 vf_vlan.vf =
 986                                 vf_tx_rate.vf =
 987                                 vf_spoofchk.vf = ivi.vf;
 988
 989                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 990                         vf_vlan.vlan = ivi.vlan;
 991                         vf_vlan.qos = ivi.qos;
 992                         vf_tx_rate.rate = ivi.tx_rate;
 993                         vf_spoofchk.setting = ivi.spoofchk;
 994                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 995                         if (!vf) {
 996                                 nla_nest_cancel(skb, vfinfo);
 997                                 goto nla_put_failure;
 998                         }
 999                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1000                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1001                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1002                                     &vf_tx_rate) ||
1003                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1004                                     &vf_spoofchk))
1005                                 goto nla_put_failure;
1006                         nla_nest_end(skb, vf);
1007                 }
1008                 nla_nest_end(skb, vfinfo);
1009         }
1010
1011         if (rtnl_port_fill(skb, dev))
1012                 goto nla_put_failure;
1013
1014         if (dev->rtnl_link_ops) {
1015                 if (rtnl_link_fill(skb, dev) < 0)
1016                         goto nla_put_failure;
1017         }
1018
1019         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1020                 goto nla_put_failure;
1021
1022         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1023                 if (af_ops->fill_link_af) {
1024                         struct nlattr *af;
1025                         int err;
1026
1027                         if (!(af = nla_nest_start(skb, af_ops->family)))
1028                                 goto nla_put_failure;
1029
1030                         err = af_ops->fill_link_af(skb, dev);
1031
1032                         /*
1033                          * Caller may return ENODATA to indicate that there
1034                          * was no data to be dumped. This is not an error, it
1035                          * means we should trim the attribute header and
1036                          * continue.
1037                          */
1038                         if (err == -ENODATA)
1039                                 nla_nest_cancel(skb, af);
1040                         else if (err < 0)
1041                                 goto nla_put_failure;
1042
1043                         nla_nest_end(skb, af);
1044                 }
1045         }
1046
1047         nla_nest_end(skb, af_spec);
1048
1049         return nlmsg_end(skb, nlh);
1050
1051 nla_put_failure:
1052         nlmsg_cancel(skb, nlh);
1053         return -EMSGSIZE;
1054 }
1055
1056 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1057 {
1058         struct net *net = sock_net(skb->sk);
1059         int h, s_h;
1060         int idx = 0, s_idx;
1061         struct net_device *dev;
1062         struct hlist_head *head;
1063         struct nlattr *tb[IFLA_MAX+1];
1064         u32 ext_filter_mask = 0;
1065
1066         s_h = cb->args[0];
1067         s_idx = cb->args[1];
1068
1069         rcu_read_lock();
1070         cb->seq = net->dev_base_seq;
1071
1072         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1073                         ifla_policy) >= 0) {
1074
1075                 if (tb[IFLA_EXT_MASK])
1076                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1077         }
1078
1079         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1080                 idx = 0;
1081                 head = &net->dev_index_head[h];
1082                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1083                         if (idx < s_idx)
1084                                 goto cont;
1085                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1086                                              NETLINK_CB(cb->skb).portid,
1087                                              cb->nlh->nlmsg_seq, 0,
1088                                              NLM_F_MULTI,
1089                                              ext_filter_mask) <= 0)
1090                                 goto out;
1091
1092                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1093 cont:
1094                         idx++;
1095                 }
1096         }
1097 out:
1098         rcu_read_unlock();
1099         cb->args[1] = idx;
1100         cb->args[0] = h;
1101
1102         return skb->len;
1103 }
1104
1105 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1106         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1107         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1108         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1109         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1110         [IFLA_MTU]              = { .type = NLA_U32 },
1111         [IFLA_LINK]             = { .type = NLA_U32 },
1112         [IFLA_MASTER]           = { .type = NLA_U32 },
1113         [IFLA_CARRIER]          = { .type = NLA_U8 },
1114         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1115         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1116         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1117         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1118         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1119         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1120         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1121         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1122         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1123         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1124         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1125         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1126         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1127         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1128         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1129         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1130 };
1131 EXPORT_SYMBOL(ifla_policy);
1132
1133 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1134         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1135         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1136 };
1137
1138 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1139         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1140 };
1141
1142 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1143         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1144                                     .len = sizeof(struct ifla_vf_mac) },
1145         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1146                                     .len = sizeof(struct ifla_vf_vlan) },
1147         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1148                                     .len = sizeof(struct ifla_vf_tx_rate) },
1149         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1150                                     .len = sizeof(struct ifla_vf_spoofchk) },
1151 };
1152
1153 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1154         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1155         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1156                                     .len = PORT_PROFILE_MAX },
1157         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1158                                     .len = sizeof(struct ifla_port_vsi)},
1159         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1160                                       .len = PORT_UUID_MAX },
1161         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1162                                     .len = PORT_UUID_MAX },
1163         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1164         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1165 };
1166
1167 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1168 {
1169         struct net *net;
1170         /* Examine the link attributes and figure out which
1171          * network namespace we are talking about.
1172          */
1173         if (tb[IFLA_NET_NS_PID])
1174                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1175         else if (tb[IFLA_NET_NS_FD])
1176                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1177         else
1178                 net = get_net(src_net);
1179         return net;
1180 }
1181 EXPORT_SYMBOL(rtnl_link_get_net);
1182
1183 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1184 {
1185         if (dev) {
1186                 if (tb[IFLA_ADDRESS] &&
1187                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1188                         return -EINVAL;
1189
1190                 if (tb[IFLA_BROADCAST] &&
1191                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1192                         return -EINVAL;
1193         }
1194
1195         if (tb[IFLA_AF_SPEC]) {
1196                 struct nlattr *af;
1197                 int rem, err;
1198
1199                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1200                         const struct rtnl_af_ops *af_ops;
1201
1202                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1203                                 return -EAFNOSUPPORT;
1204
1205                         if (!af_ops->set_link_af)
1206                                 return -EOPNOTSUPP;
1207
1208                         if (af_ops->validate_link_af) {
1209                                 err = af_ops->validate_link_af(dev, af);
1210                                 if (err < 0)
1211                                         return err;
1212                         }
1213                 }
1214         }
1215
1216         return 0;
1217 }
1218
1219 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1220 {
1221         int rem, err = -EINVAL;
1222         struct nlattr *vf;
1223         const struct net_device_ops *ops = dev->netdev_ops;
1224
1225         nla_for_each_nested(vf, attr, rem) {
1226                 switch (nla_type(vf)) {
1227                 case IFLA_VF_MAC: {
1228                         struct ifla_vf_mac *ivm;
1229                         ivm = nla_data(vf);
1230                         err = -EOPNOTSUPP;
1231                         if (ops->ndo_set_vf_mac)
1232                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1233                                                           ivm->mac);
1234                         break;
1235                 }
1236                 case IFLA_VF_VLAN: {
1237                         struct ifla_vf_vlan *ivv;
1238                         ivv = nla_data(vf);
1239                         err = -EOPNOTSUPP;
1240                         if (ops->ndo_set_vf_vlan)
1241                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1242                                                            ivv->vlan,
1243                                                            ivv->qos);
1244                         break;
1245                 }
1246                 case IFLA_VF_TX_RATE: {
1247                         struct ifla_vf_tx_rate *ivt;
1248                         ivt = nla_data(vf);
1249                         err = -EOPNOTSUPP;
1250                         if (ops->ndo_set_vf_tx_rate)
1251                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1252                                                               ivt->rate);
1253                         break;
1254                 }
1255                 case IFLA_VF_SPOOFCHK: {
1256                         struct ifla_vf_spoofchk *ivs;
1257                         ivs = nla_data(vf);
1258                         err = -EOPNOTSUPP;
1259                         if (ops->ndo_set_vf_spoofchk)
1260                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1261                                                                ivs->setting);
1262                         break;
1263                 }
1264                 default:
1265                         err = -EINVAL;
1266                         break;
1267                 }
1268                 if (err)
1269                         break;
1270         }
1271         return err;
1272 }
1273
1274 static int do_set_master(struct net_device *dev, int ifindex)
1275 {
1276         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1277         const struct net_device_ops *ops;
1278         int err;
1279
1280         if (upper_dev) {
1281                 if (upper_dev->ifindex == ifindex)
1282                         return 0;
1283                 ops = upper_dev->netdev_ops;
1284                 if (ops->ndo_del_slave) {
1285                         err = ops->ndo_del_slave(upper_dev, dev);
1286                         if (err)
1287                                 return err;
1288                 } else {
1289                         return -EOPNOTSUPP;
1290                 }
1291         }
1292
1293         if (ifindex) {
1294                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1295                 if (!upper_dev)
1296                         return -EINVAL;
1297                 ops = upper_dev->netdev_ops;
1298                 if (ops->ndo_add_slave) {
1299                         err = ops->ndo_add_slave(upper_dev, dev);
1300                         if (err)
1301                                 return err;
1302                 } else {
1303                         return -EOPNOTSUPP;
1304                 }
1305         }
1306         return 0;
1307 }
1308
1309 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1310                       struct nlattr **tb, char *ifname, int modified)
1311 {
1312         const struct net_device_ops *ops = dev->netdev_ops;
1313         int err;
1314
1315         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1316                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1317                 if (IS_ERR(net)) {
1318                         err = PTR_ERR(net);
1319                         goto errout;
1320                 }
1321                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1322                         err = -EPERM;
1323                         goto errout;
1324                 }
1325                 err = dev_change_net_namespace(dev, net, ifname);
1326                 put_net(net);
1327                 if (err)
1328                         goto errout;
1329                 modified = 1;
1330         }
1331
1332         if (tb[IFLA_MAP]) {
1333                 struct rtnl_link_ifmap *u_map;
1334                 struct ifmap k_map;
1335
1336                 if (!ops->ndo_set_config) {
1337                         err = -EOPNOTSUPP;
1338                         goto errout;
1339                 }
1340
1341                 if (!netif_device_present(dev)) {
1342                         err = -ENODEV;
1343                         goto errout;
1344                 }
1345
1346                 u_map = nla_data(tb[IFLA_MAP]);
1347                 k_map.mem_start = (unsigned long) u_map->mem_start;
1348                 k_map.mem_end = (unsigned long) u_map->mem_end;
1349                 k_map.base_addr = (unsigned short) u_map->base_addr;
1350                 k_map.irq = (unsigned char) u_map->irq;
1351                 k_map.dma = (unsigned char) u_map->dma;
1352                 k_map.port = (unsigned char) u_map->port;
1353
1354                 err = ops->ndo_set_config(dev, &k_map);
1355                 if (err < 0)
1356                         goto errout;
1357
1358                 modified = 1;
1359         }
1360
1361         if (tb[IFLA_ADDRESS]) {
1362                 struct sockaddr *sa;
1363                 int len;
1364
1365                 len = sizeof(sa_family_t) + dev->addr_len;
1366                 sa = kmalloc(len, GFP_KERNEL);
1367                 if (!sa) {
1368                         err = -ENOMEM;
1369                         goto errout;
1370                 }
1371                 sa->sa_family = dev->type;
1372                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1373                        dev->addr_len);
1374                 err = dev_set_mac_address(dev, sa);
1375                 kfree(sa);
1376                 if (err)
1377                         goto errout;
1378                 modified = 1;
1379         }
1380
1381         if (tb[IFLA_MTU]) {
1382                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1383                 if (err < 0)
1384                         goto errout;
1385                 modified = 1;
1386         }
1387
1388         if (tb[IFLA_GROUP]) {
1389                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1390                 modified = 1;
1391         }
1392
1393         /*
1394          * Interface selected by interface index but interface
1395          * name provided implies that a name change has been
1396          * requested.
1397          */
1398         if (ifm->ifi_index > 0 && ifname[0]) {
1399                 err = dev_change_name(dev, ifname);
1400                 if (err < 0)
1401                         goto errout;
1402                 modified = 1;
1403         }
1404
1405         if (tb[IFLA_IFALIAS]) {
1406                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1407                                     nla_len(tb[IFLA_IFALIAS]));
1408                 if (err < 0)
1409                         goto errout;
1410                 modified = 1;
1411         }
1412
1413         if (tb[IFLA_BROADCAST]) {
1414                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1415                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1416         }
1417
1418         if (ifm->ifi_flags || ifm->ifi_change) {
1419                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1420                 if (err < 0)
1421                         goto errout;
1422         }
1423
1424         if (tb[IFLA_MASTER]) {
1425                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1426                 if (err)
1427                         goto errout;
1428                 modified = 1;
1429         }
1430
1431         if (tb[IFLA_CARRIER]) {
1432                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1433                 if (err)
1434                         goto errout;
1435                 modified = 1;
1436         }
1437
1438         if (tb[IFLA_TXQLEN])
1439                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1440
1441         if (tb[IFLA_OPERSTATE])
1442                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1443
1444         if (tb[IFLA_LINKMODE]) {
1445                 write_lock_bh(&dev_base_lock);
1446                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1447                 write_unlock_bh(&dev_base_lock);
1448         }
1449
1450         if (tb[IFLA_VFINFO_LIST]) {
1451                 struct nlattr *attr;
1452                 int rem;
1453                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1454                         if (nla_type(attr) != IFLA_VF_INFO) {
1455                                 err = -EINVAL;
1456                                 goto errout;
1457                         }
1458                         err = do_setvfinfo(dev, attr);
1459                         if (err < 0)
1460                                 goto errout;
1461                         modified = 1;
1462                 }
1463         }
1464         err = 0;
1465
1466         if (tb[IFLA_VF_PORTS]) {
1467                 struct nlattr *port[IFLA_PORT_MAX+1];
1468                 struct nlattr *attr;
1469                 int vf;
1470                 int rem;
1471
1472                 err = -EOPNOTSUPP;
1473                 if (!ops->ndo_set_vf_port)
1474                         goto errout;
1475
1476                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1477                         if (nla_type(attr) != IFLA_VF_PORT)
1478                                 continue;
1479                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1480                                 attr, ifla_port_policy);
1481                         if (err < 0)
1482                                 goto errout;
1483                         if (!port[IFLA_PORT_VF]) {
1484                                 err = -EOPNOTSUPP;
1485                                 goto errout;
1486                         }
1487                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1488                         err = ops->ndo_set_vf_port(dev, vf, port);
1489                         if (err < 0)
1490                                 goto errout;
1491                         modified = 1;
1492                 }
1493         }
1494         err = 0;
1495
1496         if (tb[IFLA_PORT_SELF]) {
1497                 struct nlattr *port[IFLA_PORT_MAX+1];
1498
1499                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1500                         tb[IFLA_PORT_SELF], ifla_port_policy);
1501                 if (err < 0)
1502                         goto errout;
1503
1504                 err = -EOPNOTSUPP;
1505                 if (ops->ndo_set_vf_port)
1506                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1507                 if (err < 0)
1508                         goto errout;
1509                 modified = 1;
1510         }
1511
1512         if (tb[IFLA_AF_SPEC]) {
1513                 struct nlattr *af;
1514                 int rem;
1515
1516                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1517                         const struct rtnl_af_ops *af_ops;
1518
1519                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1520                                 BUG();
1521
1522                         err = af_ops->set_link_af(dev, af);
1523                         if (err < 0)
1524                                 goto errout;
1525
1526                         modified = 1;
1527                 }
1528         }
1529         err = 0;
1530
1531 errout:
1532         if (err < 0 && modified)
1533                 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",
1534                                      dev->name);
1535
1536         return err;
1537 }
1538
1539 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1540 {
1541         struct net *net = sock_net(skb->sk);
1542         struct ifinfomsg *ifm;
1543         struct net_device *dev;
1544         int err;
1545         struct nlattr *tb[IFLA_MAX+1];
1546         char ifname[IFNAMSIZ];
1547
1548         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1549         if (err < 0)
1550                 goto errout;
1551
1552         if (tb[IFLA_IFNAME])
1553                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1554         else
1555                 ifname[0] = '\0';
1556
1557         err = -EINVAL;
1558         ifm = nlmsg_data(nlh);
1559         if (ifm->ifi_index > 0)
1560                 dev = __dev_get_by_index(net, ifm->ifi_index);
1561         else if (tb[IFLA_IFNAME])
1562                 dev = __dev_get_by_name(net, ifname);
1563         else
1564                 goto errout;
1565
1566         if (dev == NULL) {
1567                 err = -ENODEV;
1568                 goto errout;
1569         }
1570
1571         err = validate_linkmsg(dev, tb);
1572         if (err < 0)
1573                 goto errout;
1574
1575         err = do_setlink(dev, ifm, tb, ifname, 0);
1576 errout:
1577         return err;
1578 }
1579
1580 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1581 {
1582         struct net *net = sock_net(skb->sk);
1583         const struct rtnl_link_ops *ops;
1584         struct net_device *dev;
1585         struct ifinfomsg *ifm;
1586         char ifname[IFNAMSIZ];
1587         struct nlattr *tb[IFLA_MAX+1];
1588         int err;
1589         LIST_HEAD(list_kill);
1590
1591         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1592         if (err < 0)
1593                 return err;
1594
1595         if (tb[IFLA_IFNAME])
1596                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1597
1598         ifm = nlmsg_data(nlh);
1599         if (ifm->ifi_index > 0)
1600                 dev = __dev_get_by_index(net, ifm->ifi_index);
1601         else if (tb[IFLA_IFNAME])
1602                 dev = __dev_get_by_name(net, ifname);
1603         else
1604                 return -EINVAL;
1605
1606         if (!dev)
1607                 return -ENODEV;
1608
1609         ops = dev->rtnl_link_ops;
1610         if (!ops)
1611                 return -EOPNOTSUPP;
1612
1613         ops->dellink(dev, &list_kill);
1614         unregister_netdevice_many(&list_kill);
1615         list_del(&list_kill);
1616         return 0;
1617 }
1618
1619 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1620 {
1621         unsigned int old_flags;
1622         int err;
1623
1624         old_flags = dev->flags;
1625         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1626                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1627                 if (err < 0)
1628                         return err;
1629         }
1630
1631         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1632         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1633
1634         __dev_notify_flags(dev, old_flags);
1635         return 0;
1636 }
1637 EXPORT_SYMBOL(rtnl_configure_link);
1638
1639 struct net_device *rtnl_create_link(struct net *net,
1640         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1641 {
1642         int err;
1643         struct net_device *dev;
1644         unsigned int num_tx_queues = 1;
1645         unsigned int num_rx_queues = 1;
1646
1647         if (tb[IFLA_NUM_TX_QUEUES])
1648                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1649         else if (ops->get_num_tx_queues)
1650                 num_tx_queues = ops->get_num_tx_queues();
1651
1652         if (tb[IFLA_NUM_RX_QUEUES])
1653                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1654         else if (ops->get_num_rx_queues)
1655                 num_rx_queues = ops->get_num_rx_queues();
1656
1657         err = -ENOMEM;
1658         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1659                                num_tx_queues, num_rx_queues);
1660         if (!dev)
1661                 goto err;
1662
1663         dev_net_set(dev, net);
1664         dev->rtnl_link_ops = ops;
1665         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1666
1667         if (tb[IFLA_MTU])
1668                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1669         if (tb[IFLA_ADDRESS]) {
1670                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1671                                 nla_len(tb[IFLA_ADDRESS]));
1672                 dev->addr_assign_type = NET_ADDR_SET;
1673         }
1674         if (tb[IFLA_BROADCAST])
1675                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1676                                 nla_len(tb[IFLA_BROADCAST]));
1677         if (tb[IFLA_TXQLEN])
1678                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1679         if (tb[IFLA_OPERSTATE])
1680                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1681         if (tb[IFLA_LINKMODE])
1682                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1683         if (tb[IFLA_GROUP])
1684                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1685
1686         return dev;
1687
1688 err:
1689         return ERR_PTR(err);
1690 }
1691 EXPORT_SYMBOL(rtnl_create_link);
1692
1693 static int rtnl_group_changelink(struct net *net, int group,
1694                 struct ifinfomsg *ifm,
1695                 struct nlattr **tb)
1696 {
1697         struct net_device *dev;
1698         int err;
1699
1700         for_each_netdev(net, dev) {
1701                 if (dev->group == group) {
1702                         err = do_setlink(dev, ifm, tb, NULL, 0);
1703                         if (err < 0)
1704                                 return err;
1705                 }
1706         }
1707
1708         return 0;
1709 }
1710
1711 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1712 {
1713         struct net *net = sock_net(skb->sk);
1714         const struct rtnl_link_ops *ops;
1715         struct net_device *dev;
1716         struct ifinfomsg *ifm;
1717         char kind[MODULE_NAME_LEN];
1718         char ifname[IFNAMSIZ];
1719         struct nlattr *tb[IFLA_MAX+1];
1720         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1721         int err;
1722
1723 #ifdef CONFIG_MODULES
1724 replay:
1725 #endif
1726         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1727         if (err < 0)
1728                 return err;
1729
1730         if (tb[IFLA_IFNAME])
1731                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1732         else
1733                 ifname[0] = '\0';
1734
1735         ifm = nlmsg_data(nlh);
1736         if (ifm->ifi_index > 0)
1737                 dev = __dev_get_by_index(net, ifm->ifi_index);
1738         else {
1739                 if (ifname[0])
1740                         dev = __dev_get_by_name(net, ifname);
1741                 else
1742                         dev = NULL;
1743         }
1744
1745         err = validate_linkmsg(dev, tb);
1746         if (err < 0)
1747                 return err;
1748
1749         if (tb[IFLA_LINKINFO]) {
1750                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1751                                        tb[IFLA_LINKINFO], ifla_info_policy);
1752                 if (err < 0)
1753                         return err;
1754         } else
1755                 memset(linkinfo, 0, sizeof(linkinfo));
1756
1757         if (linkinfo[IFLA_INFO_KIND]) {
1758                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1759                 ops = rtnl_link_ops_get(kind);
1760         } else {
1761                 kind[0] = '\0';
1762                 ops = NULL;
1763         }
1764
1765         if (1) {
1766                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1767                 struct net *dest_net;
1768
1769                 if (ops) {
1770                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1771                                 err = nla_parse_nested(attr, ops->maxtype,
1772                                                        linkinfo[IFLA_INFO_DATA],
1773                                                        ops->policy);
1774                                 if (err < 0)
1775                                         return err;
1776                                 data = attr;
1777                         }
1778                         if (ops->validate) {
1779                                 err = ops->validate(tb, data);
1780                                 if (err < 0)
1781                                         return err;
1782                         }
1783                 }
1784
1785                 if (dev) {
1786                         int modified = 0;
1787
1788                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1789                                 return -EEXIST;
1790                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1791                                 return -EOPNOTSUPP;
1792
1793                         if (linkinfo[IFLA_INFO_DATA]) {
1794                                 if (!ops || ops != dev->rtnl_link_ops ||
1795                                     !ops->changelink)
1796                                         return -EOPNOTSUPP;
1797
1798                                 err = ops->changelink(dev, tb, data);
1799                                 if (err < 0)
1800                                         return err;
1801                                 modified = 1;
1802                         }
1803
1804                         return do_setlink(dev, ifm, tb, ifname, modified);
1805                 }
1806
1807                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1808                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1809                                 return rtnl_group_changelink(net,
1810                                                 nla_get_u32(tb[IFLA_GROUP]),
1811                                                 ifm, tb);
1812                         return -ENODEV;
1813                 }
1814
1815                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1816                         return -EOPNOTSUPP;
1817
1818                 if (!ops) {
1819 #ifdef CONFIG_MODULES
1820                         if (kind[0]) {
1821                                 __rtnl_unlock();
1822                                 request_module("rtnl-link-%s", kind);
1823                                 rtnl_lock();
1824                                 ops = rtnl_link_ops_get(kind);
1825                                 if (ops)
1826                                         goto replay;
1827                         }
1828 #endif
1829                         return -EOPNOTSUPP;
1830                 }
1831
1832                 if (!ifname[0])
1833                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1834
1835                 dest_net = rtnl_link_get_net(net, tb);
1836                 if (IS_ERR(dest_net))
1837                         return PTR_ERR(dest_net);
1838
1839                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1840                 if (IS_ERR(dev)) {
1841                         err = PTR_ERR(dev);
1842                         goto out;
1843                 }
1844
1845                 dev->ifindex = ifm->ifi_index;
1846
1847                 if (ops->newlink)
1848                         err = ops->newlink(net, dev, tb, data);
1849                 else
1850                         err = register_netdevice(dev);
1851
1852                 if (err < 0 && !IS_ERR(dev))
1853                         free_netdev(dev);
1854                 if (err < 0)
1855                         goto out;
1856
1857                 err = rtnl_configure_link(dev, ifm);
1858                 if (err < 0)
1859                         unregister_netdevice(dev);
1860 out:
1861                 put_net(dest_net);
1862                 return err;
1863         }
1864 }
1865
1866 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1867 {
1868         struct net *net = sock_net(skb->sk);
1869         struct ifinfomsg *ifm;
1870         char ifname[IFNAMSIZ];
1871         struct nlattr *tb[IFLA_MAX+1];
1872         struct net_device *dev = NULL;
1873         struct sk_buff *nskb;
1874         int err;
1875         u32 ext_filter_mask = 0;
1876
1877         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1878         if (err < 0)
1879                 return err;
1880
1881         if (tb[IFLA_IFNAME])
1882                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1883
1884         if (tb[IFLA_EXT_MASK])
1885                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1886
1887         ifm = nlmsg_data(nlh);
1888         if (ifm->ifi_index > 0)
1889                 dev = __dev_get_by_index(net, ifm->ifi_index);
1890         else if (tb[IFLA_IFNAME])
1891                 dev = __dev_get_by_name(net, ifname);
1892         else
1893                 return -EINVAL;
1894
1895         if (dev == NULL)
1896                 return -ENODEV;
1897
1898         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1899         if (nskb == NULL)
1900                 return -ENOBUFS;
1901
1902         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1903                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1904         if (err < 0) {
1905                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1906                 WARN_ON(err == -EMSGSIZE);
1907                 kfree_skb(nskb);
1908         } else
1909                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1910
1911         return err;
1912 }
1913
1914 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1915 {
1916         struct net *net = sock_net(skb->sk);
1917         struct net_device *dev;
1918         struct nlattr *tb[IFLA_MAX+1];
1919         u32 ext_filter_mask = 0;
1920         u16 min_ifinfo_dump_size = 0;
1921
1922         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1923                         ifla_policy) >= 0) {
1924                 if (tb[IFLA_EXT_MASK])
1925                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1926         }
1927
1928         if (!ext_filter_mask)
1929                 return NLMSG_GOODSIZE;
1930         /*
1931          * traverse the list of net devices and compute the minimum
1932          * buffer size based upon the filter mask.
1933          */
1934         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1935                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1936                                              if_nlmsg_size(dev,
1937                                                            ext_filter_mask));
1938         }
1939
1940         return min_ifinfo_dump_size;
1941 }
1942
1943 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1944 {
1945         int idx;
1946         int s_idx = cb->family;
1947
1948         if (s_idx == 0)
1949                 s_idx = 1;
1950         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1951                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1952                 if (idx < s_idx || idx == PF_PACKET)
1953                         continue;
1954                 if (rtnl_msg_handlers[idx] == NULL ||
1955                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1956                         continue;
1957                 if (idx > s_idx)
1958                         memset(&cb->args[0], 0, sizeof(cb->args));
1959                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1960                         break;
1961         }
1962         cb->family = idx;
1963
1964         return skb->len;
1965 }
1966
1967 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1968 {
1969         struct net *net = dev_net(dev);
1970         struct sk_buff *skb;
1971         int err = -ENOBUFS;
1972         size_t if_info_size;
1973
1974         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1975         if (skb == NULL)
1976                 goto errout;
1977
1978         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1979         if (err < 0) {
1980                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1981                 WARN_ON(err == -EMSGSIZE);
1982                 kfree_skb(skb);
1983                 goto errout;
1984         }
1985         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1986         return;
1987 errout:
1988         if (err < 0)
1989                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1990 }
1991 EXPORT_SYMBOL(rtmsg_ifinfo);
1992
1993 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1994                                    struct net_device *dev,
1995                                    u8 *addr, u32 pid, u32 seq,
1996                                    int type, unsigned int flags)
1997 {
1998         struct nlmsghdr *nlh;
1999         struct ndmsg *ndm;
2000
2001         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2002         if (!nlh)
2003                 return -EMSGSIZE;
2004
2005         ndm = nlmsg_data(nlh);
2006         ndm->ndm_family  = AF_BRIDGE;
2007         ndm->ndm_pad1    = 0;
2008         ndm->ndm_pad2    = 0;
2009         ndm->ndm_flags   = flags;
2010         ndm->ndm_type    = 0;
2011         ndm->ndm_ifindex = dev->ifindex;
2012         ndm->ndm_state   = NUD_PERMANENT;
2013
2014         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2015                 goto nla_put_failure;
2016
2017         return nlmsg_end(skb, nlh);
2018
2019 nla_put_failure:
2020         nlmsg_cancel(skb, nlh);
2021         return -EMSGSIZE;
2022 }
2023
2024 static inline size_t rtnl_fdb_nlmsg_size(void)
2025 {
2026         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2027 }
2028
2029 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2030 {
2031         struct net *net = dev_net(dev);
2032         struct sk_buff *skb;
2033         int err = -ENOBUFS;
2034
2035         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2036         if (!skb)
2037                 goto errout;
2038
2039         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2040         if (err < 0) {
2041                 kfree_skb(skb);
2042                 goto errout;
2043         }
2044
2045         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2046         return;
2047 errout:
2048         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2049 }
2050
2051 /**
2052  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2053  */
2054 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2055                      struct nlattr *tb[],
2056                      struct net_device *dev,
2057                      const unsigned char *addr,
2058                      u16 flags)
2059 {
2060         int err = -EINVAL;
2061
2062         /* If aging addresses are supported device will need to
2063          * implement its own handler for this.
2064          */
2065         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2066                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2067                 return err;
2068         }
2069
2070         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2071                 err = dev_uc_add_excl(dev, addr);
2072         else if (is_multicast_ether_addr(addr))
2073                 err = dev_mc_add_excl(dev, addr);
2074
2075         /* Only return duplicate errors if NLM_F_EXCL is set */
2076         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2077                 err = 0;
2078
2079         return err;
2080 }
2081 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2082
2083 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2084 {
2085         struct net *net = sock_net(skb->sk);
2086         struct ndmsg *ndm;
2087         struct nlattr *tb[NDA_MAX+1];
2088         struct net_device *dev;
2089         u8 *addr;
2090         int err;
2091
2092         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2093         if (err < 0)
2094                 return err;
2095
2096         ndm = nlmsg_data(nlh);
2097         if (ndm->ndm_ifindex == 0) {
2098                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2099                 return -EINVAL;
2100         }
2101
2102         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2103         if (dev == NULL) {
2104                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2105                 return -ENODEV;
2106         }
2107
2108         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2109                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2110                 return -EINVAL;
2111         }
2112
2113         addr = nla_data(tb[NDA_LLADDR]);
2114         if (!is_valid_ether_addr(addr)) {
2115                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2116                 return -EINVAL;
2117         }
2118
2119         err = -EOPNOTSUPP;
2120
2121         /* Support fdb on master device the net/bridge default case */
2122         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2123             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2124                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2125                 const struct net_device_ops *ops = br_dev->netdev_ops;
2126
2127                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2128                 if (err)
2129                         goto out;
2130                 else
2131                         ndm->ndm_flags &= ~NTF_MASTER;
2132         }
2133
2134         /* Embedded bridge, macvlan, and any other device support */
2135         if ((ndm->ndm_flags & NTF_SELF)) {
2136                 if (dev->netdev_ops->ndo_fdb_add)
2137                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2138                                                            nlh->nlmsg_flags);
2139                 else
2140                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2141                                                nlh->nlmsg_flags);
2142
2143                 if (!err) {
2144                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2145                         ndm->ndm_flags &= ~NTF_SELF;
2146                 }
2147         }
2148 out:
2149         return err;
2150 }
2151
2152 /**
2153  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2154  */
2155 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2156                      struct nlattr *tb[],
2157                      struct net_device *dev,
2158                      const unsigned char *addr)
2159 {
2160         int err = -EOPNOTSUPP;
2161
2162         /* If aging addresses are supported device will need to
2163          * implement its own handler for this.
2164          */
2165         if (ndm->ndm_state & NUD_PERMANENT) {
2166                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2167                 return -EINVAL;
2168         }
2169
2170         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2171                 err = dev_uc_del(dev, addr);
2172         else if (is_multicast_ether_addr(addr))
2173                 err = dev_mc_del(dev, addr);
2174         else
2175                 err = -EINVAL;
2176
2177         return err;
2178 }
2179 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2180
2181 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2182 {
2183         struct net *net = sock_net(skb->sk);
2184         struct ndmsg *ndm;
2185         struct nlattr *tb[NDA_MAX+1];
2186         struct net_device *dev;
2187         int err = -EINVAL;
2188         __u8 *addr;
2189
2190         if (!capable(CAP_NET_ADMIN))
2191                 return -EPERM;
2192
2193         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2194         if (err < 0)
2195                 return err;
2196
2197         ndm = nlmsg_data(nlh);
2198         if (ndm->ndm_ifindex == 0) {
2199                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2200                 return -EINVAL;
2201         }
2202
2203         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2204         if (dev == NULL) {
2205                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2206                 return -ENODEV;
2207         }
2208
2209         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2210                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2211                 return -EINVAL;
2212         }
2213
2214         addr = nla_data(tb[NDA_LLADDR]);
2215         if (!is_valid_ether_addr(addr)) {
2216                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2217                 return -EINVAL;
2218         }
2219
2220         err = -EOPNOTSUPP;
2221
2222         /* Support fdb on master device the net/bridge default case */
2223         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2224             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2225                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2226                 const struct net_device_ops *ops = br_dev->netdev_ops;
2227
2228                 if (ops->ndo_fdb_del)
2229                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2230
2231                 if (err)
2232                         goto out;
2233                 else
2234                         ndm->ndm_flags &= ~NTF_MASTER;
2235         }
2236
2237         /* Embedded bridge, macvlan, and any other device support */
2238         if (ndm->ndm_flags & NTF_SELF) {
2239                 if (dev->netdev_ops->ndo_fdb_del)
2240                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2241                 else
2242                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2243
2244                 if (!err) {
2245                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2246                         ndm->ndm_flags &= ~NTF_SELF;
2247                 }
2248         }
2249 out:
2250         return err;
2251 }
2252
2253 static int nlmsg_populate_fdb(struct sk_buff *skb,
2254                               struct netlink_callback *cb,
2255                               struct net_device *dev,
2256                               int *idx,
2257                               struct netdev_hw_addr_list *list)
2258 {
2259         struct netdev_hw_addr *ha;
2260         int err;
2261         u32 portid, seq;
2262
2263         portid = NETLINK_CB(cb->skb).portid;
2264         seq = cb->nlh->nlmsg_seq;
2265
2266         list_for_each_entry(ha, &list->list, list) {
2267                 if (*idx < cb->args[0])
2268                         goto skip;
2269
2270                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2271                                               portid, seq,
2272                                               RTM_NEWNEIGH, NTF_SELF);
2273                 if (err < 0)
2274                         return err;
2275 skip:
2276                 *idx += 1;
2277         }
2278         return 0;
2279 }
2280
2281 /**
2282  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2283  * @nlh: netlink message header
2284  * @dev: netdevice
2285  *
2286  * Default netdevice operation to dump the existing unicast address list.
2287  * Returns zero on success.
2288  */
2289 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2290                       struct netlink_callback *cb,
2291                       struct net_device *dev,
2292                       int idx)
2293 {
2294         int err;
2295
2296         netif_addr_lock_bh(dev);
2297         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2298         if (err)
2299                 goto out;
2300         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2301 out:
2302         netif_addr_unlock_bh(dev);
2303         return idx;
2304 }
2305 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2306
2307 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2308 {
2309         int idx = 0;
2310         struct net *net = sock_net(skb->sk);
2311         struct net_device *dev;
2312
2313         rcu_read_lock();
2314         for_each_netdev_rcu(net, dev) {
2315                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2316                         struct net_device *br_dev;
2317                         const struct net_device_ops *ops;
2318
2319                         br_dev = netdev_master_upper_dev_get(dev);
2320                         ops = br_dev->netdev_ops;
2321                         if (ops->ndo_fdb_dump)
2322                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2323                 }
2324
2325                 if (dev->netdev_ops->ndo_fdb_dump)
2326                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2327                 else
2328                         ndo_dflt_fdb_dump(skb, cb, dev, idx);
2329         }
2330         rcu_read_unlock();
2331
2332         cb->args[0] = idx;
2333         return skb->len;
2334 }
2335
2336 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2337                             struct net_device *dev, u16 mode)
2338 {
2339         struct nlmsghdr *nlh;
2340         struct ifinfomsg *ifm;
2341         struct nlattr *br_afspec;
2342         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2343         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2344
2345         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2346         if (nlh == NULL)
2347                 return -EMSGSIZE;
2348
2349         ifm = nlmsg_data(nlh);
2350         ifm->ifi_family = AF_BRIDGE;
2351         ifm->__ifi_pad = 0;
2352         ifm->ifi_type = dev->type;
2353         ifm->ifi_index = dev->ifindex;
2354         ifm->ifi_flags = dev_get_flags(dev);
2355         ifm->ifi_change = 0;
2356
2357
2358         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2359             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2360             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2361             (br_dev &&
2362              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2363             (dev->addr_len &&
2364              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2365             (dev->ifindex != dev->iflink &&
2366              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2367                 goto nla_put_failure;
2368
2369         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2370         if (!br_afspec)
2371                 goto nla_put_failure;
2372
2373         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2374             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2375                 nla_nest_cancel(skb, br_afspec);
2376                 goto nla_put_failure;
2377         }
2378         nla_nest_end(skb, br_afspec);
2379
2380         return nlmsg_end(skb, nlh);
2381 nla_put_failure:
2382         nlmsg_cancel(skb, nlh);
2383         return -EMSGSIZE;
2384 }
2385 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2386
2387 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2388 {
2389         struct net *net = sock_net(skb->sk);
2390         struct net_device *dev;
2391         int idx = 0;
2392         u32 portid = NETLINK_CB(cb->skb).portid;
2393         u32 seq = cb->nlh->nlmsg_seq;
2394         struct nlattr *extfilt;
2395         u32 filter_mask = 0;
2396
2397         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2398                                   IFLA_EXT_MASK);
2399         if (extfilt)
2400                 filter_mask = nla_get_u32(extfilt);
2401
2402         rcu_read_lock();
2403         for_each_netdev_rcu(net, dev) {
2404                 const struct net_device_ops *ops = dev->netdev_ops;
2405                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2406
2407                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2408                         if (idx >= cb->args[0] &&
2409                             br_dev->netdev_ops->ndo_bridge_getlink(
2410                                     skb, portid, seq, dev, filter_mask) < 0)
2411                                 break;
2412                         idx++;
2413                 }
2414
2415                 if (ops->ndo_bridge_getlink) {
2416                         if (idx >= cb->args[0] &&
2417                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2418                                                     filter_mask) < 0)
2419                                 break;
2420                         idx++;
2421                 }
2422         }
2423         rcu_read_unlock();
2424         cb->args[0] = idx;
2425
2426         return skb->len;
2427 }
2428
2429 static inline size_t bridge_nlmsg_size(void)
2430 {
2431         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2432                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2433                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2434                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2435                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2436                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2437                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2438                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2439                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2440                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2441                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2442 }
2443
2444 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2445 {
2446         struct net *net = dev_net(dev);
2447         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2448         struct sk_buff *skb;
2449         int err = -EOPNOTSUPP;
2450
2451         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2452         if (!skb) {
2453                 err = -ENOMEM;
2454                 goto errout;
2455         }
2456
2457         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2458             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2459                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2460                 if (err < 0)
2461                         goto errout;
2462         }
2463
2464         if ((flags & BRIDGE_FLAGS_SELF) &&
2465             dev->netdev_ops->ndo_bridge_getlink) {
2466                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2467                 if (err < 0)
2468                         goto errout;
2469         }
2470
2471         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2472         return 0;
2473 errout:
2474         WARN_ON(err == -EMSGSIZE);
2475         kfree_skb(skb);
2476         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2477         return err;
2478 }
2479
2480 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2481                                void *arg)
2482 {
2483         struct net *net = sock_net(skb->sk);
2484         struct ifinfomsg *ifm;
2485         struct net_device *dev;
2486         struct nlattr *br_spec, *attr = NULL;
2487         int rem, err = -EOPNOTSUPP;
2488         u16 oflags, flags = 0;
2489         bool have_flags = false;
2490
2491         if (nlmsg_len(nlh) < sizeof(*ifm))
2492                 return -EINVAL;
2493
2494         ifm = nlmsg_data(nlh);
2495         if (ifm->ifi_family != AF_BRIDGE)
2496                 return -EPFNOSUPPORT;
2497
2498         dev = __dev_get_by_index(net, ifm->ifi_index);
2499         if (!dev) {
2500                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2501                 return -ENODEV;
2502         }
2503
2504         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2505         if (br_spec) {
2506                 nla_for_each_nested(attr, br_spec, rem) {
2507                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2508                                 have_flags = true;
2509                                 flags = nla_get_u16(attr);
2510                                 break;
2511                         }
2512                 }
2513         }
2514
2515         oflags = flags;
2516
2517         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2518                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2519
2520                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2521                         err = -EOPNOTSUPP;
2522                         goto out;
2523                 }
2524
2525                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2526                 if (err)
2527                         goto out;
2528
2529                 flags &= ~BRIDGE_FLAGS_MASTER;
2530         }
2531
2532         if ((flags & BRIDGE_FLAGS_SELF)) {
2533                 if (!dev->netdev_ops->ndo_bridge_setlink)
2534                         err = -EOPNOTSUPP;
2535                 else
2536                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2537
2538                 if (!err)
2539                         flags &= ~BRIDGE_FLAGS_SELF;
2540         }
2541
2542         if (have_flags)
2543                 memcpy(nla_data(attr), &flags, sizeof(flags));
2544         /* Generate event to notify upper layer of bridge change */
2545         if (!err)
2546                 err = rtnl_bridge_notify(dev, oflags);
2547 out:
2548         return err;
2549 }
2550
2551 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2552                                void *arg)
2553 {
2554         struct net *net = sock_net(skb->sk);
2555         struct ifinfomsg *ifm;
2556         struct net_device *dev;
2557         struct nlattr *br_spec, *attr = NULL;
2558         int rem, err = -EOPNOTSUPP;
2559         u16 oflags, flags = 0;
2560         bool have_flags = false;
2561
2562         if (nlmsg_len(nlh) < sizeof(*ifm))
2563                 return -EINVAL;
2564
2565         ifm = nlmsg_data(nlh);
2566         if (ifm->ifi_family != AF_BRIDGE)
2567                 return -EPFNOSUPPORT;
2568
2569         dev = __dev_get_by_index(net, ifm->ifi_index);
2570         if (!dev) {
2571                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2572                 return -ENODEV;
2573         }
2574
2575         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2576         if (br_spec) {
2577                 nla_for_each_nested(attr, br_spec, rem) {
2578                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2579                                 have_flags = true;
2580                                 flags = nla_get_u16(attr);
2581                                 break;
2582                         }
2583                 }
2584         }
2585
2586         oflags = flags;
2587
2588         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2589                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2590
2591                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2592                         err = -EOPNOTSUPP;
2593                         goto out;
2594                 }
2595
2596                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2597                 if (err)
2598                         goto out;
2599
2600                 flags &= ~BRIDGE_FLAGS_MASTER;
2601         }
2602
2603         if ((flags & BRIDGE_FLAGS_SELF)) {
2604                 if (!dev->netdev_ops->ndo_bridge_dellink)
2605                         err = -EOPNOTSUPP;
2606                 else
2607                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2608
2609                 if (!err)
2610                         flags &= ~BRIDGE_FLAGS_SELF;
2611         }
2612
2613         if (have_flags)
2614                 memcpy(nla_data(attr), &flags, sizeof(flags));
2615         /* Generate event to notify upper layer of bridge change */
2616         if (!err)
2617                 err = rtnl_bridge_notify(dev, oflags);
2618 out:
2619         return err;
2620 }
2621
2622 /* Protected by RTNL sempahore.  */
2623 static struct rtattr **rta_buf;
2624 static int rtattr_max;
2625
2626 /* Process one rtnetlink message. */
2627
2628 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2629 {
2630         struct net *net = sock_net(skb->sk);
2631         rtnl_doit_func doit;
2632         int sz_idx, kind;
2633         int min_len;
2634         int family;
2635         int type;
2636         int err;
2637
2638         type = nlh->nlmsg_type;
2639         if (type > RTM_MAX)
2640                 return -EOPNOTSUPP;
2641
2642         type -= RTM_BASE;
2643
2644         /* All the messages must have at least 1 byte length */
2645         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2646                 return 0;
2647
2648         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2649         sz_idx = type>>2;
2650         kind = type&3;
2651
2652         if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2653                 return -EPERM;
2654
2655         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2656                 struct sock *rtnl;
2657                 rtnl_dumpit_func dumpit;
2658                 rtnl_calcit_func calcit;
2659                 u16 min_dump_alloc = 0;
2660
2661                 dumpit = rtnl_get_dumpit(family, type);
2662                 if (dumpit == NULL)
2663                         return -EOPNOTSUPP;
2664                 calcit = rtnl_get_calcit(family, type);
2665                 if (calcit)
2666                         min_dump_alloc = calcit(skb, nlh);
2667
2668                 __rtnl_unlock();
2669                 rtnl = net->rtnl;
2670                 {
2671                         struct netlink_dump_control c = {
2672                                 .dump           = dumpit,
2673                                 .min_dump_alloc = min_dump_alloc,
2674                         };
2675                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2676                 }
2677                 rtnl_lock();
2678                 return err;
2679         }
2680
2681         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2682
2683         min_len = rtm_min[sz_idx];
2684         if (nlh->nlmsg_len < min_len)
2685                 return -EINVAL;
2686
2687         if (nlh->nlmsg_len > min_len) {
2688                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2689                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2690
2691                 while (RTA_OK(attr, attrlen)) {
2692                         unsigned int flavor = attr->rta_type;
2693                         if (flavor) {
2694                                 if (flavor > rta_max[sz_idx])
2695                                         return -EINVAL;
2696                                 rta_buf[flavor-1] = attr;
2697                         }
2698                         attr = RTA_NEXT(attr, attrlen);
2699                 }
2700         }
2701
2702         doit = rtnl_get_doit(family, type);
2703         if (doit == NULL)
2704                 return -EOPNOTSUPP;
2705
2706         return doit(skb, nlh, (void *)&rta_buf[0]);
2707 }
2708
2709 static void rtnetlink_rcv(struct sk_buff *skb)
2710 {
2711         rtnl_lock();
2712         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2713         rtnl_unlock();
2714 }
2715
2716 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2717 {
2718         struct net_device *dev = ptr;
2719
2720         switch (event) {
2721         case NETDEV_UP:
2722         case NETDEV_DOWN:
2723         case NETDEV_PRE_UP:
2724         case NETDEV_POST_INIT:
2725         case NETDEV_REGISTER:
2726         case NETDEV_CHANGE:
2727         case NETDEV_PRE_TYPE_CHANGE:
2728         case NETDEV_GOING_DOWN:
2729         case NETDEV_UNREGISTER:
2730         case NETDEV_UNREGISTER_FINAL:
2731         case NETDEV_RELEASE:
2732         case NETDEV_JOIN:
2733                 break;
2734         default:
2735                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2736                 break;
2737         }
2738         return NOTIFY_DONE;
2739 }
2740
2741 static struct notifier_block rtnetlink_dev_notifier = {
2742         .notifier_call  = rtnetlink_event,
2743 };
2744
2745
2746 static int __net_init rtnetlink_net_init(struct net *net)
2747 {
2748         struct sock *sk;
2749         struct netlink_kernel_cfg cfg = {
2750                 .groups         = RTNLGRP_MAX,
2751                 .input          = rtnetlink_rcv,
2752                 .cb_mutex       = &rtnl_mutex,
2753                 .flags          = NL_CFG_F_NONROOT_RECV,
2754         };
2755
2756         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2757         if (!sk)
2758                 return -ENOMEM;
2759         net->rtnl = sk;
2760         return 0;
2761 }
2762
2763 static void __net_exit rtnetlink_net_exit(struct net *net)
2764 {
2765         netlink_kernel_release(net->rtnl);
2766         net->rtnl = NULL;
2767 }
2768
2769 static struct pernet_operations rtnetlink_net_ops = {
2770         .init = rtnetlink_net_init,
2771         .exit = rtnetlink_net_exit,
2772 };
2773
2774 void __init rtnetlink_init(void)
2775 {
2776         int i;
2777
2778         rtattr_max = 0;
2779         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2780                 if (rta_max[i] > rtattr_max)
2781                         rtattr_max = rta_max[i];
2782         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2783         if (!rta_buf)
2784                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2785
2786         if (register_pernet_subsys(&rtnetlink_net_ops))
2787                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2788
2789         register_netdevice_notifier(&rtnetlink_dev_notifier);
2790
2791         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2792                       rtnl_dump_ifinfo, rtnl_calcit);
2793         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2794         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2795         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2796
2797         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2798         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2799
2800         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2801         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2802         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2803
2804         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2805         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2806         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2807 }
2808