net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/pci.h>
  40 #include <linux/etherdevice.h>
  41
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 struct rtnl_link {
  58         rtnl_doit_func          doit;
  59         rtnl_dumpit_func        dumpit;
  60         rtnl_calcit_func        calcit;
  61 };
  62
  63 static DEFINE_MUTEX(rtnl_mutex);
  64
  65 void rtnl_lock(void)
  66 {
  67         mutex_lock(&rtnl_mutex);
  68 }
  69 EXPORT_SYMBOL(rtnl_lock);
  70
  71 void __rtnl_unlock(void)
  72 {
  73         mutex_unlock(&rtnl_mutex);
  74 }
  75
  76 void rtnl_unlock(void)
  77 {
  78         /* This fellow will unlock it for us. */
  79         netdev_run_todo();
  80 }
  81 EXPORT_SYMBOL(rtnl_unlock);
  82
  83 int rtnl_trylock(void)
  84 {
  85         return mutex_trylock(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_trylock);
  88
  89 int rtnl_is_locked(void)
  90 {
  91         return mutex_is_locked(&rtnl_mutex);
  92 }
  93 EXPORT_SYMBOL(rtnl_is_locked);
  94
  95 #ifdef CONFIG_PROVE_LOCKING
  96 int lockdep_rtnl_is_held(void)
  97 {
  98         return lockdep_is_held(&rtnl_mutex);
  99 }
 100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105 static inline int rtm_msgindex(int msgtype)
 106 {
 107         int msgindex = msgtype - RTM_BASE;
 108
 109         /*
 110          * msgindex < 0 implies someone tried to register a netlink
 111          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112          * the message type has not been added to linux/rtnetlink.h
 113          */
 114         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116         return msgindex;
 117 }
 118
 119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120 {
 121         struct rtnl_link *tab;
 122
 123         if (protocol <= RTNL_FAMILY_MAX)
 124                 tab = rtnl_msg_handlers[protocol];
 125         else
 126                 tab = NULL;
 127
 128         if (tab == NULL || tab[msgindex].doit == NULL)
 129                 tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131         return tab ? tab[msgindex].doit : NULL;
 132 }
 133
 134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135 {
 136         struct rtnl_link *tab;
 137
 138         if (protocol <= RTNL_FAMILY_MAX)
 139                 tab = rtnl_msg_handlers[protocol];
 140         else
 141                 tab = NULL;
 142
 143         if (tab == NULL || tab[msgindex].dumpit == NULL)
 144                 tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146         return tab ? tab[msgindex].dumpit : NULL;
 147 }
 148
 149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150 {
 151         struct rtnl_link *tab;
 152
 153         if (protocol <= RTNL_FAMILY_MAX)
 154                 tab = rtnl_msg_handlers[protocol];
 155         else
 156                 tab = NULL;
 157
 158         if (tab == NULL || tab[msgindex].calcit == NULL)
 159                 tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161         return tab ? tab[msgindex].calcit : NULL;
 162 }
 163
 164 /**
 165  * __rtnl_register - Register a rtnetlink message type
 166  * @protocol: Protocol family or PF_UNSPEC
 167  * @msgtype: rtnetlink message type
 168  * @doit: Function pointer called for each request message
 169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170  * @calcit: Function pointer to calc size of dump message
 171  *
 172  * Registers the specified function pointers (at least one of them has
 173  * to be non-NULL) to be called whenever a request message for the
 174  * specified protocol family and message type is received.
 175  *
 176  * The special protocol family PF_UNSPEC may be used to define fallback
 177  * function pointers for the case when no entry for the specific protocol
 178  * family exists.
 179  *
 180  * Returns 0 on success or a negative error code.
 181  */
 182 int __rtnl_register(int protocol, int msgtype,
 183                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184                     rtnl_calcit_func calcit)
 185 {
 186         struct rtnl_link *tab;
 187         int msgindex;
 188
 189         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190         msgindex = rtm_msgindex(msgtype);
 191
 192         tab = rtnl_msg_handlers[protocol];
 193         if (tab == NULL) {
 194                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195                 if (tab == NULL)
 196                         return -ENOBUFS;
 197
 198                 rtnl_msg_handlers[protocol] = tab;
 199         }
 200
 201         if (doit)
 202                 tab[msgindex].doit = doit;
 203
 204         if (dumpit)
 205                 tab[msgindex].dumpit = dumpit;
 206
 207         if (calcit)
 208                 tab[msgindex].calcit = calcit;
 209
 210         return 0;
 211 }
 212 EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214 /**
 215  * rtnl_register - Register a rtnetlink message type
 216  *
 217  * Identical to __rtnl_register() but panics on failure. This is useful
 218  * as failure of this function is very unlikely, it can only happen due
 219  * to lack of memory when allocating the chain to store all message
 220  * handlers for a protocol. Meant for use in init functions where lack
 221  * of memory implies no sense in continuing.
 222  */
 223 void rtnl_register(int protocol, int msgtype,
 224                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225                    rtnl_calcit_func calcit)
 226 {
 227         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228                 panic("Unable to register rtnetlink message handler, "
 229                       "protocol = %d, message type = %d\n",
 230                       protocol, msgtype);
 231 }
 232 EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234 /**
 235  * rtnl_unregister - Unregister a rtnetlink message type
 236  * @protocol: Protocol family or PF_UNSPEC
 237  * @msgtype: rtnetlink message type
 238  *
 239  * Returns 0 on success or a negative error code.
 240  */
 241 int rtnl_unregister(int protocol, int msgtype)
 242 {
 243         int msgindex;
 244
 245         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246         msgindex = rtm_msgindex(msgtype);
 247
 248         if (rtnl_msg_handlers[protocol] == NULL)
 249                 return -ENOENT;
 250
 251         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254         return 0;
 255 }
 256 EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258 /**
 259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260  * @protocol : Protocol family or PF_UNSPEC
 261  *
 262  * Identical to calling rtnl_unregster() for all registered message types
 263  * of a certain protocol family.
 264  */
 265 void rtnl_unregister_all(int protocol)
 266 {
 267         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269         kfree(rtnl_msg_handlers[protocol]);
 270         rtnl_msg_handlers[protocol] = NULL;
 271 }
 272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274 static LIST_HEAD(link_ops);
 275
 276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277 {
 278         const struct rtnl_link_ops *ops;
 279
 280         list_for_each_entry(ops, &link_ops, list) {
 281                 if (!strcmp(ops->kind, kind))
 282                         return ops;
 283         }
 284         return NULL;
 285 }
 286
 287 /**
 288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289  * @ops: struct rtnl_link_ops * to register
 290  *
 291  * The caller must hold the rtnl_mutex. This function should be used
 292  * by drivers that create devices during module initialization. It
 293  * must be called before registering the devices.
 294  *
 295  * Returns 0 on success or a negative error code.
 296  */
 297 int __rtnl_link_register(struct rtnl_link_ops *ops)
 298 {
 299         if (rtnl_link_ops_get(ops->kind))
 300                 return -EEXIST;
 301
 302         if (!ops->dellink)
 303                 ops->dellink = unregister_netdevice_queue;
 304
 305         list_add_tail(&ops->list, &link_ops);
 306         return 0;
 307 }
 308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 309
 310 /**
 311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 312  * @ops: struct rtnl_link_ops * to register
 313  *
 314  * Returns 0 on success or a negative error code.
 315  */
 316 int rtnl_link_register(struct rtnl_link_ops *ops)
 317 {
 318         int err;
 319
 320         rtnl_lock();
 321         err = __rtnl_link_register(ops);
 322         rtnl_unlock();
 323         return err;
 324 }
 325 EXPORT_SYMBOL_GPL(rtnl_link_register);
 326
 327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 328 {
 329         struct net_device *dev;
 330         LIST_HEAD(list_kill);
 331
 332         for_each_netdev(net, dev) {
 333                 if (dev->rtnl_link_ops == ops)
 334                         ops->dellink(dev, &list_kill);
 335         }
 336         unregister_netdevice_many(&list_kill);
 337 }
 338
 339 /**
 340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 341  * @ops: struct rtnl_link_ops * to unregister
 342  *
 343  * The caller must hold the rtnl_mutex.
 344  */
 345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 346 {
 347         struct net *net;
 348
 349         for_each_net(net) {
 350                 __rtnl_kill_links(net, ops);
 351         }
 352         list_del(&ops->list);
 353 }
 354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 355
 356 /**
 357  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 358  * @ops: struct rtnl_link_ops * to unregister
 359  */
 360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 361 {
 362         rtnl_lock();
 363         __rtnl_link_unregister(ops);
 364         rtnl_unlock();
 365 }
 366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 367
 368 static size_t rtnl_link_get_size(const struct net_device *dev)
 369 {
 370         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 371         size_t size;
 372
 373         if (!ops)
 374                 return 0;
 375
 376         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 377                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 378
 379         if (ops->get_size)
 380                 /* IFLA_INFO_DATA + nested data */
 381                 size += nla_total_size(sizeof(struct nlattr)) +
 382                         ops->get_size(dev);
 383
 384         if (ops->get_xstats_size)
 385                 /* IFLA_INFO_XSTATS */
 386                 size += nla_total_size(ops->get_xstats_size(dev));
 387
 388         return size;
 389 }
 390
 391 static LIST_HEAD(rtnl_af_ops);
 392
 393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 394 {
 395         const struct rtnl_af_ops *ops;
 396
 397         list_for_each_entry(ops, &rtnl_af_ops, list) {
 398                 if (ops->family == family)
 399                         return ops;
 400         }
 401
 402         return NULL;
 403 }
 404
 405 /**
 406  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 407  * @ops: struct rtnl_af_ops * to register
 408  *
 409  * The caller must hold the rtnl_mutex.
 410  *
 411  * Returns 0 on success or a negative error code.
 412  */
 413 int __rtnl_af_register(struct rtnl_af_ops *ops)
 414 {
 415         list_add_tail(&ops->list, &rtnl_af_ops);
 416         return 0;
 417 }
 418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 419
 420 /**
 421  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 422  * @ops: struct rtnl_af_ops * to register
 423  *
 424  * Returns 0 on success or a negative error code.
 425  */
 426 int rtnl_af_register(struct rtnl_af_ops *ops)
 427 {
 428         int err;
 429
 430         rtnl_lock();
 431         err = __rtnl_af_register(ops);
 432         rtnl_unlock();
 433         return err;
 434 }
 435 EXPORT_SYMBOL_GPL(rtnl_af_register);
 436
 437 /**
 438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 439  * @ops: struct rtnl_af_ops * to unregister
 440  *
 441  * The caller must hold the rtnl_mutex.
 442  */
 443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 444 {
 445         list_del(&ops->list);
 446 }
 447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 448
 449 /**
 450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 451  * @ops: struct rtnl_af_ops * to unregister
 452  */
 453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 454 {
 455         rtnl_lock();
 456         __rtnl_af_unregister(ops);
 457         rtnl_unlock();
 458 }
 459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 460
 461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 462 {
 463         struct rtnl_af_ops *af_ops;
 464         size_t size;
 465
 466         /* IFLA_AF_SPEC */
 467         size = nla_total_size(sizeof(struct nlattr));
 468
 469         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 470                 if (af_ops->get_link_af_size) {
 471                         /* AF_* + nested data */
 472                         size += nla_total_size(sizeof(struct nlattr)) +
 473                                 af_ops->get_link_af_size(dev);
 474                 }
 475         }
 476
 477         return size;
 478 }
 479
 480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 481 {
 482         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 483         struct nlattr *linkinfo, *data;
 484         int err = -EMSGSIZE;
 485
 486         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 487         if (linkinfo == NULL)
 488                 goto out;
 489
 490         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 491                 goto err_cancel_link;
 492         if (ops->fill_xstats) {
 493                 err = ops->fill_xstats(skb, dev);
 494                 if (err < 0)
 495                         goto err_cancel_link;
 496         }
 497         if (ops->fill_info) {
 498                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 499                 if (data == NULL)
 500                         goto err_cancel_link;
 501                 err = ops->fill_info(skb, dev);
 502                 if (err < 0)
 503                         goto err_cancel_data;
 504                 nla_nest_end(skb, data);
 505         }
 506
 507         nla_nest_end(skb, linkinfo);
 508         return 0;
 509
 510 err_cancel_data:
 511         nla_nest_cancel(skb, data);
 512 err_cancel_link:
 513         nla_nest_cancel(skb, linkinfo);
 514 out:
 515         return err;
 516 }
 517
 518 static const int rtm_min[RTM_NR_FAMILIES] =
 519 {
 520         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 521         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 522         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 523         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 524         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 527         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 528         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 530 };
 531
 532 static const int rta_max[RTM_NR_FAMILIES] =
 533 {
 534         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 535         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 536         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 537         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 538         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 539         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 540         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 541         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 542 };
 543
 544 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 545 {
 546         struct sock *rtnl = net->rtnl;
 547         int err = 0;
 548
 549         NETLINK_CB(skb).dst_group = group;
 550         if (echo)
 551                 atomic_inc(&skb->users);
 552         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 553         if (echo)
 554                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 555         return err;
 556 }
 557
 558 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 559 {
 560         struct sock *rtnl = net->rtnl;
 561
 562         return nlmsg_unicast(rtnl, skb, pid);
 563 }
 564 EXPORT_SYMBOL(rtnl_unicast);
 565
 566 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 567                  struct nlmsghdr *nlh, gfp_t flags)
 568 {
 569         struct sock *rtnl = net->rtnl;
 570         int report = 0;
 571
 572         if (nlh)
 573                 report = nlmsg_report(nlh);
 574
 575         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 576 }
 577 EXPORT_SYMBOL(rtnl_notify);
 578
 579 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 580 {
 581         struct sock *rtnl = net->rtnl;
 582
 583         netlink_set_err(rtnl, 0, group, error);
 584 }
 585 EXPORT_SYMBOL(rtnl_set_sk_err);
 586
 587 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 588 {
 589         struct nlattr *mx;
 590         int i, valid = 0;
 591
 592         mx = nla_nest_start(skb, RTA_METRICS);
 593         if (mx == NULL)
 594                 return -ENOBUFS;
 595
 596         for (i = 0; i < RTAX_MAX; i++) {
 597                 if (metrics[i]) {
 598                         valid++;
 599                         if (nla_put_u32(skb, i+1, metrics[i]))
 600                                 goto nla_put_failure;
 601                 }
 602         }
 603
 604         if (!valid) {
 605                 nla_nest_cancel(skb, mx);
 606                 return 0;
 607         }
 608
 609         return nla_nest_end(skb, mx);
 610
 611 nla_put_failure:
 612         nla_nest_cancel(skb, mx);
 613         return -EMSGSIZE;
 614 }
 615 EXPORT_SYMBOL(rtnetlink_put_metrics);
 616
 617 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 618                        long expires, u32 error)
 619 {
 620         struct rta_cacheinfo ci = {
 621                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 622                 .rta_used = dst->__use,
 623                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 624                 .rta_error = error,
 625                 .rta_id =  id,
 626         };
 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(4) /* IFLA_PROMISCUITY */
 784                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 785                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 786                + nla_total_size(1) /* IFLA_OPERSTATE */
 787                + nla_total_size(1) /* IFLA_LINKMODE */
 788                + nla_total_size(ext_filter_mask
 789                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 790                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 791                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 792                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 793                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 794 }
 795
 796 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 797 {
 798         struct nlattr *vf_ports;
 799         struct nlattr *vf_port;
 800         int vf;
 801         int err;
 802
 803         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 804         if (!vf_ports)
 805                 return -EMSGSIZE;
 806
 807         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 808                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 809                 if (!vf_port)
 810                         goto nla_put_failure;
 811                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 812                         goto nla_put_failure;
 813                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 814                 if (err == -EMSGSIZE)
 815                         goto nla_put_failure;
 816                 if (err) {
 817                         nla_nest_cancel(skb, vf_port);
 818                         continue;
 819                 }
 820                 nla_nest_end(skb, vf_port);
 821         }
 822
 823         nla_nest_end(skb, vf_ports);
 824
 825         return 0;
 826
 827 nla_put_failure:
 828         nla_nest_cancel(skb, vf_ports);
 829         return -EMSGSIZE;
 830 }
 831
 832 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 833 {
 834         struct nlattr *port_self;
 835         int err;
 836
 837         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 838         if (!port_self)
 839                 return -EMSGSIZE;
 840
 841         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 842         if (err) {
 843                 nla_nest_cancel(skb, port_self);
 844                 return (err == -EMSGSIZE) ? err : 0;
 845         }
 846
 847         nla_nest_end(skb, port_self);
 848
 849         return 0;
 850 }
 851
 852 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 853 {
 854         int err;
 855
 856         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 857                 return 0;
 858
 859         err = rtnl_port_self_fill(skb, dev);
 860         if (err)
 861                 return err;
 862
 863         if (dev_num_vf(dev->dev.parent)) {
 864                 err = rtnl_vf_ports_fill(skb, dev);
 865                 if (err)
 866                         return err;
 867         }
 868
 869         return 0;
 870 }
 871
 872 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 873                             int type, u32 pid, u32 seq, u32 change,
 874                             unsigned int flags, u32 ext_filter_mask)
 875 {
 876         struct ifinfomsg *ifm;
 877         struct nlmsghdr *nlh;
 878         struct rtnl_link_stats64 temp;
 879         const struct rtnl_link_stats64 *stats;
 880         struct nlattr *attr, *af_spec;
 881         struct rtnl_af_ops *af_ops;
 882
 883         ASSERT_RTNL();
 884         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 885         if (nlh == NULL)
 886                 return -EMSGSIZE;
 887
 888         ifm = nlmsg_data(nlh);
 889         ifm->ifi_family = AF_UNSPEC;
 890         ifm->__ifi_pad = 0;
 891         ifm->ifi_type = dev->type;
 892         ifm->ifi_index = dev->ifindex;
 893         ifm->ifi_flags = dev_get_flags(dev);
 894         ifm->ifi_change = change;
 895
 896         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 897             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 898             nla_put_u8(skb, IFLA_OPERSTATE,
 899                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 900             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 901             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 902             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 903             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 904             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 905 #ifdef CONFIG_RPS
 906             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 907 #endif
 908             (dev->ifindex != dev->iflink &&
 909              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 910             (dev->master &&
 911              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 912             (dev->qdisc &&
 913              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 914             (dev->ifalias &&
 915              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 916                 goto nla_put_failure;
 917
 918         if (1) {
 919                 struct rtnl_link_ifmap map = {
 920                         .mem_start   = dev->mem_start,
 921                         .mem_end     = dev->mem_end,
 922                         .base_addr   = dev->base_addr,
 923                         .irq         = dev->irq,
 924                         .dma         = dev->dma,
 925                         .port        = dev->if_port,
 926                 };
 927                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 928                         goto nla_put_failure;
 929         }
 930
 931         if (dev->addr_len) {
 932                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 933                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 934                         goto nla_put_failure;
 935         }
 936
 937         attr = nla_reserve(skb, IFLA_STATS,
 938                         sizeof(struct rtnl_link_stats));
 939         if (attr == NULL)
 940                 goto nla_put_failure;
 941
 942         stats = dev_get_stats(dev, &temp);
 943         copy_rtnl_link_stats(nla_data(attr), stats);
 944
 945         attr = nla_reserve(skb, IFLA_STATS64,
 946                         sizeof(struct rtnl_link_stats64));
 947         if (attr == NULL)
 948                 goto nla_put_failure;
 949         copy_rtnl_link_stats64(nla_data(attr), stats);
 950
 951         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 952             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 953                 goto nla_put_failure;
 954
 955         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 956             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 957                 int i;
 958
 959                 struct nlattr *vfinfo, *vf;
 960                 int num_vfs = dev_num_vf(dev->dev.parent);
 961
 962                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 963                 if (!vfinfo)
 964                         goto nla_put_failure;
 965                 for (i = 0; i < num_vfs; i++) {
 966                         struct ifla_vf_info ivi;
 967                         struct ifla_vf_mac vf_mac;
 968                         struct ifla_vf_vlan vf_vlan;
 969                         struct ifla_vf_tx_rate vf_tx_rate;
 970                         struct ifla_vf_spoofchk vf_spoofchk;
 971
 972                         /*
 973                          * Not all SR-IOV capable drivers support the
 974                          * spoofcheck query.  Preset to -1 so the user
 975                          * space tool can detect that the driver didn't
 976                          * report anything.
 977                          */
 978                         ivi.spoofchk = -1;
 979                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 980                                 break;
 981                         vf_mac.vf =
 982                                 vf_vlan.vf =
 983                                 vf_tx_rate.vf =
 984                                 vf_spoofchk.vf = ivi.vf;
 985
 986                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 987                         vf_vlan.vlan = ivi.vlan;
 988                         vf_vlan.qos = ivi.qos;
 989                         vf_tx_rate.rate = ivi.tx_rate;
 990                         vf_spoofchk.setting = ivi.spoofchk;
 991                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 992                         if (!vf) {
 993                                 nla_nest_cancel(skb, vfinfo);
 994                                 goto nla_put_failure;
 995                         }
 996                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 997                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 998                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 999                                     &vf_tx_rate) ||
1000                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1001                                     &vf_spoofchk))
1002                                 goto nla_put_failure;
1003                         nla_nest_end(skb, vf);
1004                 }
1005                 nla_nest_end(skb, vfinfo);
1006         }
1007
1008         if (rtnl_port_fill(skb, dev))
1009                 goto nla_put_failure;
1010
1011         if (dev->rtnl_link_ops) {
1012                 if (rtnl_link_fill(skb, dev) < 0)
1013                         goto nla_put_failure;
1014         }
1015
1016         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1017                 goto nla_put_failure;
1018
1019         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1020                 if (af_ops->fill_link_af) {
1021                         struct nlattr *af;
1022                         int err;
1023
1024                         if (!(af = nla_nest_start(skb, af_ops->family)))
1025                                 goto nla_put_failure;
1026
1027                         err = af_ops->fill_link_af(skb, dev);
1028
1029                         /*
1030                          * Caller may return ENODATA to indicate that there
1031                          * was no data to be dumped. This is not an error, it
1032                          * means we should trim the attribute header and
1033                          * continue.
1034                          */
1035                         if (err == -ENODATA)
1036                                 nla_nest_cancel(skb, af);
1037                         else if (err < 0)
1038                                 goto nla_put_failure;
1039
1040                         nla_nest_end(skb, af);
1041                 }
1042         }
1043
1044         nla_nest_end(skb, af_spec);
1045
1046         return nlmsg_end(skb, nlh);
1047
1048 nla_put_failure:
1049         nlmsg_cancel(skb, nlh);
1050         return -EMSGSIZE;
1051 }
1052
1053 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1054 {
1055         struct net *net = sock_net(skb->sk);
1056         int h, s_h;
1057         int idx = 0, s_idx;
1058         struct net_device *dev;
1059         struct hlist_head *head;
1060         struct hlist_node *node;
1061         struct nlattr *tb[IFLA_MAX+1];
1062         u32 ext_filter_mask = 0;
1063
1064         s_h = cb->args[0];
1065         s_idx = cb->args[1];
1066
1067         rcu_read_lock();
1068         cb->seq = net->dev_base_seq;
1069
1070         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1071                         ifla_policy) >= 0) {
1072
1073                 if (tb[IFLA_EXT_MASK])
1074                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1075         }
1076
1077         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1078                 idx = 0;
1079                 head = &net->dev_index_head[h];
1080                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1081                         if (idx < s_idx)
1082                                 goto cont;
1083                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1084                                              NETLINK_CB(cb->skb).pid,
1085                                              cb->nlh->nlmsg_seq, 0,
1086                                              NLM_F_MULTI,
1087                                              ext_filter_mask) <= 0)
1088                                 goto out;
1089
1090                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1091 cont:
1092                         idx++;
1093                 }
1094         }
1095 out:
1096         rcu_read_unlock();
1097         cb->args[1] = idx;
1098         cb->args[0] = h;
1099
1100         return skb->len;
1101 }
1102
1103 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1104         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1105         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1107         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1108         [IFLA_MTU]              = { .type = NLA_U32 },
1109         [IFLA_LINK]             = { .type = NLA_U32 },
1110         [IFLA_MASTER]           = { .type = NLA_U32 },
1111         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1112         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1113         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1114         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1115         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1116         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1117         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1118         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1119         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1120         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1121         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1122         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1123         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1124         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1125         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1126         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1127 };
1128 EXPORT_SYMBOL(ifla_policy);
1129
1130 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1131         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1132         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1133 };
1134
1135 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1136         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1137 };
1138
1139 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1140         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1141                                     .len = sizeof(struct ifla_vf_mac) },
1142         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1143                                     .len = sizeof(struct ifla_vf_vlan) },
1144         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1145                                     .len = sizeof(struct ifla_vf_tx_rate) },
1146         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1147                                     .len = sizeof(struct ifla_vf_spoofchk) },
1148 };
1149
1150 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1151         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1152         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1153                                     .len = PORT_PROFILE_MAX },
1154         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1155                                     .len = sizeof(struct ifla_port_vsi)},
1156         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1157                                       .len = PORT_UUID_MAX },
1158         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1159                                     .len = PORT_UUID_MAX },
1160         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1161         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1162 };
1163
1164 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1165 {
1166         struct net *net;
1167         /* Examine the link attributes and figure out which
1168          * network namespace we are talking about.
1169          */
1170         if (tb[IFLA_NET_NS_PID])
1171                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1172         else if (tb[IFLA_NET_NS_FD])
1173                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1174         else
1175                 net = get_net(src_net);
1176         return net;
1177 }
1178 EXPORT_SYMBOL(rtnl_link_get_net);
1179
1180 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1181 {
1182         if (dev) {
1183                 if (tb[IFLA_ADDRESS] &&
1184                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1185                         return -EINVAL;
1186
1187                 if (tb[IFLA_BROADCAST] &&
1188                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1189                         return -EINVAL;
1190         }
1191
1192         if (tb[IFLA_AF_SPEC]) {
1193                 struct nlattr *af;
1194                 int rem, err;
1195
1196                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1197                         const struct rtnl_af_ops *af_ops;
1198
1199                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1200                                 return -EAFNOSUPPORT;
1201
1202                         if (!af_ops->set_link_af)
1203                                 return -EOPNOTSUPP;
1204
1205                         if (af_ops->validate_link_af) {
1206                                 err = af_ops->validate_link_af(dev, af);
1207                                 if (err < 0)
1208                                         return err;
1209                         }
1210                 }
1211         }
1212
1213         return 0;
1214 }
1215
1216 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1217 {
1218         int rem, err = -EINVAL;
1219         struct nlattr *vf;
1220         const struct net_device_ops *ops = dev->netdev_ops;
1221
1222         nla_for_each_nested(vf, attr, rem) {
1223                 switch (nla_type(vf)) {
1224                 case IFLA_VF_MAC: {
1225                         struct ifla_vf_mac *ivm;
1226                         ivm = nla_data(vf);
1227                         err = -EOPNOTSUPP;
1228                         if (ops->ndo_set_vf_mac)
1229                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1230                                                           ivm->mac);
1231                         break;
1232                 }
1233                 case IFLA_VF_VLAN: {
1234                         struct ifla_vf_vlan *ivv;
1235                         ivv = nla_data(vf);
1236                         err = -EOPNOTSUPP;
1237                         if (ops->ndo_set_vf_vlan)
1238                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1239                                                            ivv->vlan,
1240                                                            ivv->qos);
1241                         break;
1242                 }
1243                 case IFLA_VF_TX_RATE: {
1244                         struct ifla_vf_tx_rate *ivt;
1245                         ivt = nla_data(vf);
1246                         err = -EOPNOTSUPP;
1247                         if (ops->ndo_set_vf_tx_rate)
1248                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1249                                                               ivt->rate);
1250                         break;
1251                 }
1252                 case IFLA_VF_SPOOFCHK: {
1253                         struct ifla_vf_spoofchk *ivs;
1254                         ivs = nla_data(vf);
1255                         err = -EOPNOTSUPP;
1256                         if (ops->ndo_set_vf_spoofchk)
1257                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1258                                                                ivs->setting);
1259                         break;
1260                 }
1261                 default:
1262                         err = -EINVAL;
1263                         break;
1264                 }
1265                 if (err)
1266                         break;
1267         }
1268         return err;
1269 }
1270
1271 static int do_set_master(struct net_device *dev, int ifindex)
1272 {
1273         struct net_device *master_dev;
1274         const struct net_device_ops *ops;
1275         int err;
1276
1277         if (dev->master) {
1278                 if (dev->master->ifindex == ifindex)
1279                         return 0;
1280                 ops = dev->master->netdev_ops;
1281                 if (ops->ndo_del_slave) {
1282                         err = ops->ndo_del_slave(dev->master, dev);
1283                         if (err)
1284                                 return err;
1285                 } else {
1286                         return -EOPNOTSUPP;
1287                 }
1288         }
1289
1290         if (ifindex) {
1291                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1292                 if (!master_dev)
1293                         return -EINVAL;
1294                 ops = master_dev->netdev_ops;
1295                 if (ops->ndo_add_slave) {
1296                         err = ops->ndo_add_slave(master_dev, dev);
1297                         if (err)
1298                                 return err;
1299                 } else {
1300                         return -EOPNOTSUPP;
1301                 }
1302         }
1303         return 0;
1304 }
1305
1306 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1307                       struct nlattr **tb, char *ifname, int modified)
1308 {
1309         const struct net_device_ops *ops = dev->netdev_ops;
1310         int send_addr_notify = 0;
1311         int err;
1312
1313         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1314                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1315                 if (IS_ERR(net)) {
1316                         err = PTR_ERR(net);
1317                         goto errout;
1318                 }
1319                 err = dev_change_net_namespace(dev, net, ifname);
1320                 put_net(net);
1321                 if (err)
1322                         goto errout;
1323                 modified = 1;
1324         }
1325
1326         if (tb[IFLA_MAP]) {
1327                 struct rtnl_link_ifmap *u_map;
1328                 struct ifmap k_map;
1329
1330                 if (!ops->ndo_set_config) {
1331                         err = -EOPNOTSUPP;
1332                         goto errout;
1333                 }
1334
1335                 if (!netif_device_present(dev)) {
1336                         err = -ENODEV;
1337                         goto errout;
1338                 }
1339
1340                 u_map = nla_data(tb[IFLA_MAP]);
1341                 k_map.mem_start = (unsigned long) u_map->mem_start;
1342                 k_map.mem_end = (unsigned long) u_map->mem_end;
1343                 k_map.base_addr = (unsigned short) u_map->base_addr;
1344                 k_map.irq = (unsigned char) u_map->irq;
1345                 k_map.dma = (unsigned char) u_map->dma;
1346                 k_map.port = (unsigned char) u_map->port;
1347
1348                 err = ops->ndo_set_config(dev, &k_map);
1349                 if (err < 0)
1350                         goto errout;
1351
1352                 modified = 1;
1353         }
1354
1355         if (tb[IFLA_ADDRESS]) {
1356                 struct sockaddr *sa;
1357                 int len;
1358
1359                 if (!ops->ndo_set_mac_address) {
1360                         err = -EOPNOTSUPP;
1361                         goto errout;
1362                 }
1363
1364                 if (!netif_device_present(dev)) {
1365                         err = -ENODEV;
1366                         goto errout;
1367                 }
1368
1369                 len = sizeof(sa_family_t) + dev->addr_len;
1370                 sa = kmalloc(len, GFP_KERNEL);
1371                 if (!sa) {
1372                         err = -ENOMEM;
1373                         goto errout;
1374                 }
1375                 sa->sa_family = dev->type;
1376                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1377                        dev->addr_len);
1378                 err = ops->ndo_set_mac_address(dev, sa);
1379                 kfree(sa);
1380                 if (err)
1381                         goto errout;
1382                 send_addr_notify = 1;
1383                 modified = 1;
1384         }
1385
1386         if (tb[IFLA_MTU]) {
1387                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1388                 if (err < 0)
1389                         goto errout;
1390                 modified = 1;
1391         }
1392
1393         if (tb[IFLA_GROUP]) {
1394                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1395                 modified = 1;
1396         }
1397
1398         /*
1399          * Interface selected by interface index but interface
1400          * name provided implies that a name change has been
1401          * requested.
1402          */
1403         if (ifm->ifi_index > 0 && ifname[0]) {
1404                 err = dev_change_name(dev, ifname);
1405                 if (err < 0)
1406                         goto errout;
1407                 modified = 1;
1408         }
1409
1410         if (tb[IFLA_IFALIAS]) {
1411                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1412                                     nla_len(tb[IFLA_IFALIAS]));
1413                 if (err < 0)
1414                         goto errout;
1415                 modified = 1;
1416         }
1417
1418         if (tb[IFLA_BROADCAST]) {
1419                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1420                 send_addr_notify = 1;
1421         }
1422
1423         if (ifm->ifi_flags || ifm->ifi_change) {
1424                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1425                 if (err < 0)
1426                         goto errout;
1427         }
1428
1429         if (tb[IFLA_MASTER]) {
1430                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1431                 if (err)
1432                         goto errout;
1433                 modified = 1;
1434         }
1435
1436         if (tb[IFLA_TXQLEN])
1437                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1438
1439         if (tb[IFLA_OPERSTATE])
1440                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1441
1442         if (tb[IFLA_LINKMODE]) {
1443                 write_lock_bh(&dev_base_lock);
1444                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1445                 write_unlock_bh(&dev_base_lock);
1446         }
1447
1448         if (tb[IFLA_VFINFO_LIST]) {
1449                 struct nlattr *attr;
1450                 int rem;
1451                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1452                         if (nla_type(attr) != IFLA_VF_INFO) {
1453                                 err = -EINVAL;
1454                                 goto errout;
1455                         }
1456                         err = do_setvfinfo(dev, attr);
1457                         if (err < 0)
1458                                 goto errout;
1459                         modified = 1;
1460                 }
1461         }
1462         err = 0;
1463
1464         if (tb[IFLA_VF_PORTS]) {
1465                 struct nlattr *port[IFLA_PORT_MAX+1];
1466                 struct nlattr *attr;
1467                 int vf;
1468                 int rem;
1469
1470                 err = -EOPNOTSUPP;
1471                 if (!ops->ndo_set_vf_port)
1472                         goto errout;
1473
1474                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1475                         if (nla_type(attr) != IFLA_VF_PORT)
1476                                 continue;
1477                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1478                                 attr, ifla_port_policy);
1479                         if (err < 0)
1480                                 goto errout;
1481                         if (!port[IFLA_PORT_VF]) {
1482                                 err = -EOPNOTSUPP;
1483                                 goto errout;
1484                         }
1485                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1486                         err = ops->ndo_set_vf_port(dev, vf, port);
1487                         if (err < 0)
1488                                 goto errout;
1489                         modified = 1;
1490                 }
1491         }
1492         err = 0;
1493
1494         if (tb[IFLA_PORT_SELF]) {
1495                 struct nlattr *port[IFLA_PORT_MAX+1];
1496
1497                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1498                         tb[IFLA_PORT_SELF], ifla_port_policy);
1499                 if (err < 0)
1500                         goto errout;
1501
1502                 err = -EOPNOTSUPP;
1503                 if (ops->ndo_set_vf_port)
1504                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1505                 if (err < 0)
1506                         goto errout;
1507                 modified = 1;
1508         }
1509
1510         if (tb[IFLA_AF_SPEC]) {
1511                 struct nlattr *af;
1512                 int rem;
1513
1514                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1515                         const struct rtnl_af_ops *af_ops;
1516
1517                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1518                                 BUG();
1519
1520                         err = af_ops->set_link_af(dev, af);
1521                         if (err < 0)
1522                                 goto errout;
1523
1524                         modified = 1;
1525                 }
1526         }
1527         err = 0;
1528
1529 errout:
1530         if (err < 0 && modified)
1531                 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",
1532                                      dev->name);
1533
1534         if (send_addr_notify)
1535                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1536
1537         return err;
1538 }
1539
1540 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1541 {
1542         struct net *net = sock_net(skb->sk);
1543         struct ifinfomsg *ifm;
1544         struct net_device *dev;
1545         int err;
1546         struct nlattr *tb[IFLA_MAX+1];
1547         char ifname[IFNAMSIZ];
1548
1549         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1550         if (err < 0)
1551                 goto errout;
1552
1553         if (tb[IFLA_IFNAME])
1554                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1555         else
1556                 ifname[0] = '\0';
1557
1558         err = -EINVAL;
1559         ifm = nlmsg_data(nlh);
1560         if (ifm->ifi_index > 0)
1561                 dev = __dev_get_by_index(net, ifm->ifi_index);
1562         else if (tb[IFLA_IFNAME])
1563                 dev = __dev_get_by_name(net, ifname);
1564         else
1565                 goto errout;
1566
1567         if (dev == NULL) {
1568                 err = -ENODEV;
1569                 goto errout;
1570         }
1571
1572         err = validate_linkmsg(dev, tb);
1573         if (err < 0)
1574                 goto errout;
1575
1576         err = do_setlink(dev, ifm, tb, ifname, 0);
1577 errout:
1578         return err;
1579 }
1580
1581 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1582 {
1583         struct net *net = sock_net(skb->sk);
1584         const struct rtnl_link_ops *ops;
1585         struct net_device *dev;
1586         struct ifinfomsg *ifm;
1587         char ifname[IFNAMSIZ];
1588         struct nlattr *tb[IFLA_MAX+1];
1589         int err;
1590         LIST_HEAD(list_kill);
1591
1592         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1593         if (err < 0)
1594                 return err;
1595
1596         if (tb[IFLA_IFNAME])
1597                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1598
1599         ifm = nlmsg_data(nlh);
1600         if (ifm->ifi_index > 0)
1601                 dev = __dev_get_by_index(net, ifm->ifi_index);
1602         else if (tb[IFLA_IFNAME])
1603                 dev = __dev_get_by_name(net, ifname);
1604         else
1605                 return -EINVAL;
1606
1607         if (!dev)
1608                 return -ENODEV;
1609
1610         ops = dev->rtnl_link_ops;
1611         if (!ops)
1612                 return -EOPNOTSUPP;
1613
1614         ops->dellink(dev, &list_kill);
1615         unregister_netdevice_many(&list_kill);
1616         list_del(&list_kill);
1617         return 0;
1618 }
1619
1620 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1621 {
1622         unsigned int old_flags;
1623         int err;
1624
1625         old_flags = dev->flags;
1626         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1627                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1628                 if (err < 0)
1629                         return err;
1630         }
1631
1632         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1633         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1634
1635         __dev_notify_flags(dev, old_flags);
1636         return 0;
1637 }
1638 EXPORT_SYMBOL(rtnl_configure_link);
1639
1640 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1641         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1642 {
1643         int err;
1644         struct net_device *dev;
1645         unsigned int num_tx_queues = 1;
1646         unsigned int num_rx_queues = 1;
1647
1648         if (tb[IFLA_NUM_TX_QUEUES])
1649                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1650         else if (ops->get_num_tx_queues)
1651                 num_tx_queues = ops->get_num_tx_queues();
1652
1653         if (tb[IFLA_NUM_RX_QUEUES])
1654                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1655         else if (ops->get_num_rx_queues)
1656                 num_rx_queues = ops->get_num_rx_queues();
1657
1658         err = -ENOMEM;
1659         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1660                                num_tx_queues, num_rx_queues);
1661         if (!dev)
1662                 goto err;
1663
1664         dev_net_set(dev, net);
1665         dev->rtnl_link_ops = ops;
1666         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1667
1668         if (tb[IFLA_MTU])
1669                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1670         if (tb[IFLA_ADDRESS])
1671                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1672                                 nla_len(tb[IFLA_ADDRESS]));
1673         if (tb[IFLA_BROADCAST])
1674                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1675                                 nla_len(tb[IFLA_BROADCAST]));
1676         if (tb[IFLA_TXQLEN])
1677                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1678         if (tb[IFLA_OPERSTATE])
1679                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1680         if (tb[IFLA_LINKMODE])
1681                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1682         if (tb[IFLA_GROUP])
1683                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1684
1685         return dev;
1686
1687 err:
1688         return ERR_PTR(err);
1689 }
1690 EXPORT_SYMBOL(rtnl_create_link);
1691
1692 static int rtnl_group_changelink(struct net *net, int group,
1693                 struct ifinfomsg *ifm,
1694                 struct nlattr **tb)
1695 {
1696         struct net_device *dev;
1697         int err;
1698
1699         for_each_netdev(net, dev) {
1700                 if (dev->group == group) {
1701                         err = do_setlink(dev, ifm, tb, NULL, 0);
1702                         if (err < 0)
1703                                 return err;
1704                 }
1705         }
1706
1707         return 0;
1708 }
1709
1710 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1711 {
1712         struct net *net = sock_net(skb->sk);
1713         const struct rtnl_link_ops *ops;
1714         struct net_device *dev;
1715         struct ifinfomsg *ifm;
1716         char kind[MODULE_NAME_LEN];
1717         char ifname[IFNAMSIZ];
1718         struct nlattr *tb[IFLA_MAX+1];
1719         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1720         int err;
1721
1722 #ifdef CONFIG_MODULES
1723 replay:
1724 #endif
1725         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1726         if (err < 0)
1727                 return err;
1728
1729         if (tb[IFLA_IFNAME])
1730                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1731         else
1732                 ifname[0] = '\0';
1733
1734         ifm = nlmsg_data(nlh);
1735         if (ifm->ifi_index > 0)
1736                 dev = __dev_get_by_index(net, ifm->ifi_index);
1737         else {
1738                 if (ifname[0])
1739                         dev = __dev_get_by_name(net, ifname);
1740                 else
1741                         dev = NULL;
1742         }
1743
1744         err = validate_linkmsg(dev, tb);
1745         if (err < 0)
1746                 return err;
1747
1748         if (tb[IFLA_LINKINFO]) {
1749                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1750                                        tb[IFLA_LINKINFO], ifla_info_policy);
1751                 if (err < 0)
1752                         return err;
1753         } else
1754                 memset(linkinfo, 0, sizeof(linkinfo));
1755
1756         if (linkinfo[IFLA_INFO_KIND]) {
1757                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1758                 ops = rtnl_link_ops_get(kind);
1759         } else {
1760                 kind[0] = '\0';
1761                 ops = NULL;
1762         }
1763
1764         if (1) {
1765                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1766                 struct net *dest_net;
1767
1768                 if (ops) {
1769                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1770                                 err = nla_parse_nested(attr, ops->maxtype,
1771                                                        linkinfo[IFLA_INFO_DATA],
1772                                                        ops->policy);
1773                                 if (err < 0)
1774                                         return err;
1775                                 data = attr;
1776                         }
1777                         if (ops->validate) {
1778                                 err = ops->validate(tb, data);
1779                                 if (err < 0)
1780                                         return err;
1781                         }
1782                 }
1783
1784                 if (dev) {
1785                         int modified = 0;
1786
1787                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1788                                 return -EEXIST;
1789                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1790                                 return -EOPNOTSUPP;
1791
1792                         if (linkinfo[IFLA_INFO_DATA]) {
1793                                 if (!ops || ops != dev->rtnl_link_ops ||
1794                                     !ops->changelink)
1795                                         return -EOPNOTSUPP;
1796
1797                                 err = ops->changelink(dev, tb, data);
1798                                 if (err < 0)
1799                                         return err;
1800                                 modified = 1;
1801                         }
1802
1803                         return do_setlink(dev, ifm, tb, ifname, modified);
1804                 }
1805
1806                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1807                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1808                                 return rtnl_group_changelink(net,
1809                                                 nla_get_u32(tb[IFLA_GROUP]),
1810                                                 ifm, tb);
1811                         return -ENODEV;
1812                 }
1813
1814                 if (ifm->ifi_index)
1815                         return -EOPNOTSUPP;
1816                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1817                         return -EOPNOTSUPP;
1818
1819                 if (!ops) {
1820 #ifdef CONFIG_MODULES
1821                         if (kind[0]) {
1822                                 __rtnl_unlock();
1823                                 request_module("rtnl-link-%s", kind);
1824                                 rtnl_lock();
1825                                 ops = rtnl_link_ops_get(kind);
1826                                 if (ops)
1827                                         goto replay;
1828                         }
1829 #endif
1830                         return -EOPNOTSUPP;
1831                 }
1832
1833                 if (!ifname[0])
1834                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1835
1836                 dest_net = rtnl_link_get_net(net, tb);
1837                 if (IS_ERR(dest_net))
1838                         return PTR_ERR(dest_net);
1839
1840                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1841
1842                 if (IS_ERR(dev))
1843                         err = PTR_ERR(dev);
1844                 else if (ops->newlink)
1845                         err = ops->newlink(net, dev, tb, data);
1846                 else
1847                         err = register_netdevice(dev);
1848
1849                 if (err < 0 && !IS_ERR(dev))
1850                         free_netdev(dev);
1851                 if (err < 0)
1852                         goto out;
1853
1854                 err = rtnl_configure_link(dev, ifm);
1855                 if (err < 0)
1856                         unregister_netdevice(dev);
1857 out:
1858                 put_net(dest_net);
1859                 return err;
1860         }
1861 }
1862
1863 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1864 {
1865         struct net *net = sock_net(skb->sk);
1866         struct ifinfomsg *ifm;
1867         char ifname[IFNAMSIZ];
1868         struct nlattr *tb[IFLA_MAX+1];
1869         struct net_device *dev = NULL;
1870         struct sk_buff *nskb;
1871         int err;
1872         u32 ext_filter_mask = 0;
1873
1874         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1875         if (err < 0)
1876                 return err;
1877
1878         if (tb[IFLA_IFNAME])
1879                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1880
1881         if (tb[IFLA_EXT_MASK])
1882                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1883
1884         ifm = nlmsg_data(nlh);
1885         if (ifm->ifi_index > 0)
1886                 dev = __dev_get_by_index(net, ifm->ifi_index);
1887         else if (tb[IFLA_IFNAME])
1888                 dev = __dev_get_by_name(net, ifname);
1889         else
1890                 return -EINVAL;
1891
1892         if (dev == NULL)
1893                 return -ENODEV;
1894
1895         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1896         if (nskb == NULL)
1897                 return -ENOBUFS;
1898
1899         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1900                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1901         if (err < 0) {
1902                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1903                 WARN_ON(err == -EMSGSIZE);
1904                 kfree_skb(nskb);
1905         } else
1906                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1907
1908         return err;
1909 }
1910
1911 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1912 {
1913         struct net *net = sock_net(skb->sk);
1914         struct net_device *dev;
1915         struct nlattr *tb[IFLA_MAX+1];
1916         u32 ext_filter_mask = 0;
1917         u16 min_ifinfo_dump_size = 0;
1918
1919         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1920                         ifla_policy) >= 0) {
1921                 if (tb[IFLA_EXT_MASK])
1922                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1923         }
1924
1925         if (!ext_filter_mask)
1926                 return NLMSG_GOODSIZE;
1927         /*
1928          * traverse the list of net devices and compute the minimum
1929          * buffer size based upon the filter mask.
1930          */
1931         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1932                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1933                                              if_nlmsg_size(dev,
1934                                                            ext_filter_mask));
1935         }
1936
1937         return min_ifinfo_dump_size;
1938 }
1939
1940 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1941 {
1942         int idx;
1943         int s_idx = cb->family;
1944
1945         if (s_idx == 0)
1946                 s_idx = 1;
1947         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1948                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1949                 if (idx < s_idx || idx == PF_PACKET)
1950                         continue;
1951                 if (rtnl_msg_handlers[idx] == NULL ||
1952                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1953                         continue;
1954                 if (idx > s_idx)
1955                         memset(&cb->args[0], 0, sizeof(cb->args));
1956                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1957                         break;
1958         }
1959         cb->family = idx;
1960
1961         return skb->len;
1962 }
1963
1964 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1965 {
1966         struct net *net = dev_net(dev);
1967         struct sk_buff *skb;
1968         int err = -ENOBUFS;
1969         size_t if_info_size;
1970
1971         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1972         if (skb == NULL)
1973                 goto errout;
1974
1975         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1976         if (err < 0) {
1977                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1978                 WARN_ON(err == -EMSGSIZE);
1979                 kfree_skb(skb);
1980                 goto errout;
1981         }
1982         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1983         return;
1984 errout:
1985         if (err < 0)
1986                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1987 }
1988
1989 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1990                                    struct net_device *dev,
1991                                    u8 *addr, u32 pid, u32 seq,
1992                                    int type, unsigned int flags)
1993 {
1994         struct nlmsghdr *nlh;
1995         struct ndmsg *ndm;
1996
1997         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1998         if (!nlh)
1999                 return -EMSGSIZE;
2000
2001         ndm = nlmsg_data(nlh);
2002         ndm->ndm_family  = AF_BRIDGE;
2003         ndm->ndm_pad1    = 0;
2004         ndm->ndm_pad2    = 0;
2005         ndm->ndm_flags   = flags;
2006         ndm->ndm_type    = 0;
2007         ndm->ndm_ifindex = dev->ifindex;
2008         ndm->ndm_state   = NUD_PERMANENT;
2009
2010         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2011                 goto nla_put_failure;
2012
2013         return nlmsg_end(skb, nlh);
2014
2015 nla_put_failure:
2016         nlmsg_cancel(skb, nlh);
2017         return -EMSGSIZE;
2018 }
2019
2020 static inline size_t rtnl_fdb_nlmsg_size(void)
2021 {
2022         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2023 }
2024
2025 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2026 {
2027         struct net *net = dev_net(dev);
2028         struct sk_buff *skb;
2029         int err = -ENOBUFS;
2030
2031         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2032         if (!skb)
2033                 goto errout;
2034
2035         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2036         if (err < 0) {
2037                 kfree_skb(skb);
2038                 goto errout;
2039         }
2040
2041         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2042         return;
2043 errout:
2044         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2045 }
2046
2047 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2048 {
2049         struct net *net = sock_net(skb->sk);
2050         struct net_device *master = NULL;
2051         struct ndmsg *ndm;
2052         struct nlattr *tb[NDA_MAX+1];
2053         struct net_device *dev;
2054         u8 *addr;
2055         int err;
2056
2057         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2058         if (err < 0)
2059                 return err;
2060
2061         ndm = nlmsg_data(nlh);
2062         if (ndm->ndm_ifindex == 0) {
2063                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2064                 return -EINVAL;
2065         }
2066
2067         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2068         if (dev == NULL) {
2069                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2070                 return -ENODEV;
2071         }
2072
2073         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2074                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2075                 return -EINVAL;
2076         }
2077
2078         addr = nla_data(tb[NDA_LLADDR]);
2079         if (!is_valid_ether_addr(addr)) {
2080                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2081                 return -EINVAL;
2082         }
2083
2084         err = -EOPNOTSUPP;
2085
2086         /* Support fdb on master device the net/bridge default case */
2087         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2088             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2089                 master = dev->master;
2090                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2091                                                       nlh->nlmsg_flags);
2092                 if (err)
2093                         goto out;
2094                 else
2095                         ndm->ndm_flags &= ~NTF_MASTER;
2096         }
2097
2098         /* Embedded bridge, macvlan, and any other device support */
2099         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2100                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2101                                                    nlh->nlmsg_flags);
2102
2103                 if (!err) {
2104                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2105                         ndm->ndm_flags &= ~NTF_SELF;
2106                 }
2107         }
2108 out:
2109         return err;
2110 }
2111
2112 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2113 {
2114         struct net *net = sock_net(skb->sk);
2115         struct ndmsg *ndm;
2116         struct nlattr *llattr;
2117         struct net_device *dev;
2118         int err = -EINVAL;
2119         __u8 *addr;
2120
2121         if (nlmsg_len(nlh) < sizeof(*ndm))
2122                 return -EINVAL;
2123
2124         ndm = nlmsg_data(nlh);
2125         if (ndm->ndm_ifindex == 0) {
2126                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2127                 return -EINVAL;
2128         }
2129
2130         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2131         if (dev == NULL) {
2132                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2133                 return -ENODEV;
2134         }
2135
2136         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2137         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2138                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2139                 return -EINVAL;
2140         }
2141
2142         addr = nla_data(llattr);
2143         err = -EOPNOTSUPP;
2144
2145         /* Support fdb on master device the net/bridge default case */
2146         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2147             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2148                 struct net_device *master = dev->master;
2149
2150                 if (master->netdev_ops->ndo_fdb_del)
2151                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2152
2153                 if (err)
2154                         goto out;
2155                 else
2156                         ndm->ndm_flags &= ~NTF_MASTER;
2157         }
2158
2159         /* Embedded bridge, macvlan, and any other device support */
2160         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2161                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2162
2163                 if (!err) {
2164                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2165                         ndm->ndm_flags &= ~NTF_SELF;
2166                 }
2167         }
2168 out:
2169         return err;
2170 }
2171
2172 static int nlmsg_populate_fdb(struct sk_buff *skb,
2173                               struct netlink_callback *cb,
2174                               struct net_device *dev,
2175                               int *idx,
2176                               struct netdev_hw_addr_list *list)
2177 {
2178         struct netdev_hw_addr *ha;
2179         int err;
2180         u32 pid, seq;
2181
2182         pid = NETLINK_CB(cb->skb).pid;
2183         seq = cb->nlh->nlmsg_seq;
2184
2185         list_for_each_entry(ha, &list->list, list) {
2186                 if (*idx < cb->args[0])
2187                         goto skip;
2188
2189                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2190                                               pid, seq, 0, NTF_SELF);
2191                 if (err < 0)
2192                         return err;
2193 skip:
2194                 *idx += 1;
2195         }
2196         return 0;
2197 }
2198
2199 /**
2200  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2201  * @nlh: netlink message header
2202  * @dev: netdevice
2203  *
2204  * Default netdevice operation to dump the existing unicast address list.
2205  * Returns zero on success.
2206  */
2207 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2208                       struct netlink_callback *cb,
2209                       struct net_device *dev,
2210                       int idx)
2211 {
2212         int err;
2213
2214         netif_addr_lock_bh(dev);
2215         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2216         if (err)
2217                 goto out;
2218         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2219 out:
2220         netif_addr_unlock_bh(dev);
2221         return idx;
2222 }
2223 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2224
2225 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2226 {
2227         int idx = 0;
2228         struct net *net = sock_net(skb->sk);
2229         struct net_device *dev;
2230
2231         rcu_read_lock();
2232         for_each_netdev_rcu(net, dev) {
2233                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2234                         struct net_device *master = dev->master;
2235                         const struct net_device_ops *ops = master->netdev_ops;
2236
2237                         if (ops->ndo_fdb_dump)
2238                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2239                 }
2240
2241                 if (dev->netdev_ops->ndo_fdb_dump)
2242                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2243         }
2244         rcu_read_unlock();
2245
2246         cb->args[0] = idx;
2247         return skb->len;
2248 }
2249
2250 /* Protected by RTNL sempahore.  */
2251 static struct rtattr **rta_buf;
2252 static int rtattr_max;
2253
2254 /* Process one rtnetlink message. */
2255
2256 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2257 {
2258         struct net *net = sock_net(skb->sk);
2259         rtnl_doit_func doit;
2260         int sz_idx, kind;
2261         int min_len;
2262         int family;
2263         int type;
2264         int err;
2265
2266         type = nlh->nlmsg_type;
2267         if (type > RTM_MAX)
2268                 return -EOPNOTSUPP;
2269
2270         type -= RTM_BASE;
2271
2272         /* All the messages must have at least 1 byte length */
2273         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2274                 return 0;
2275
2276         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2277         sz_idx = type>>2;
2278         kind = type&3;
2279
2280         if (kind != 2 && !capable(CAP_NET_ADMIN))
2281                 return -EPERM;
2282
2283         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2284                 struct sock *rtnl;
2285                 rtnl_dumpit_func dumpit;
2286                 rtnl_calcit_func calcit;
2287                 u16 min_dump_alloc = 0;
2288
2289                 dumpit = rtnl_get_dumpit(family, type);
2290                 if (dumpit == NULL)
2291                         return -EOPNOTSUPP;
2292                 calcit = rtnl_get_calcit(family, type);
2293                 if (calcit)
2294                         min_dump_alloc = calcit(skb, nlh);
2295
2296                 __rtnl_unlock();
2297                 rtnl = net->rtnl;
2298                 {
2299                         struct netlink_dump_control c = {
2300                                 .dump           = dumpit,
2301                                 .min_dump_alloc = min_dump_alloc,
2302                         };
2303                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2304                 }
2305                 rtnl_lock();
2306                 return err;
2307         }
2308
2309         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2310
2311         min_len = rtm_min[sz_idx];
2312         if (nlh->nlmsg_len < min_len)
2313                 return -EINVAL;
2314
2315         if (nlh->nlmsg_len > min_len) {
2316                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2317                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2318
2319                 while (RTA_OK(attr, attrlen)) {
2320                         unsigned int flavor = attr->rta_type;
2321                         if (flavor) {
2322                                 if (flavor > rta_max[sz_idx])
2323                                         return -EINVAL;
2324                                 rta_buf[flavor-1] = attr;
2325                         }
2326                         attr = RTA_NEXT(attr, attrlen);
2327                 }
2328         }
2329
2330         doit = rtnl_get_doit(family, type);
2331         if (doit == NULL)
2332                 return -EOPNOTSUPP;
2333
2334         return doit(skb, nlh, (void *)&rta_buf[0]);
2335 }
2336
2337 static void rtnetlink_rcv(struct sk_buff *skb)
2338 {
2339         rtnl_lock();
2340         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2341         rtnl_unlock();
2342 }
2343
2344 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2345 {
2346         struct net_device *dev = ptr;
2347
2348         switch (event) {
2349         case NETDEV_UP:
2350         case NETDEV_DOWN:
2351         case NETDEV_PRE_UP:
2352         case NETDEV_POST_INIT:
2353         case NETDEV_REGISTER:
2354         case NETDEV_CHANGE:
2355         case NETDEV_PRE_TYPE_CHANGE:
2356         case NETDEV_GOING_DOWN:
2357         case NETDEV_UNREGISTER:
2358         case NETDEV_UNREGISTER_BATCH:
2359         case NETDEV_RELEASE:
2360         case NETDEV_JOIN:
2361                 break;
2362         default:
2363                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2364                 break;
2365         }
2366         return NOTIFY_DONE;
2367 }
2368
2369 static struct notifier_block rtnetlink_dev_notifier = {
2370         .notifier_call  = rtnetlink_event,
2371 };
2372
2373
2374 static int __net_init rtnetlink_net_init(struct net *net)
2375 {
2376         struct sock *sk;
2377         struct netlink_kernel_cfg cfg = {
2378                 .groups         = RTNLGRP_MAX,
2379                 .input          = rtnetlink_rcv,
2380                 .cb_mutex       = &rtnl_mutex,
2381         };
2382
2383         sk = netlink_kernel_create(net, NETLINK_ROUTE, THIS_MODULE, &cfg);
2384         if (!sk)
2385                 return -ENOMEM;
2386         net->rtnl = sk;
2387         return 0;
2388 }
2389
2390 static void __net_exit rtnetlink_net_exit(struct net *net)
2391 {
2392         netlink_kernel_release(net->rtnl);
2393         net->rtnl = NULL;
2394 }
2395
2396 static struct pernet_operations rtnetlink_net_ops = {
2397         .init = rtnetlink_net_init,
2398         .exit = rtnetlink_net_exit,
2399 };
2400
2401 void __init rtnetlink_init(void)
2402 {
2403         int i;
2404
2405         rtattr_max = 0;
2406         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2407                 if (rta_max[i] > rtattr_max)
2408                         rtattr_max = rta_max[i];
2409         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2410         if (!rta_buf)
2411                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2412
2413         if (register_pernet_subsys(&rtnetlink_net_ops))
2414                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2415
2416         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2417         register_netdevice_notifier(&rtnetlink_dev_notifier);
2418
2419         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2420                       rtnl_dump_ifinfo, rtnl_calcit);
2421         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2422         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2423         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2424
2425         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2426         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2427
2428         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2429         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2430         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2431 }
2432