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                 ci.rta_expires = jiffies_to_clock_t(expires);
 630
 631         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 632 }
 633 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 634
 635 static void set_operstate(struct net_device *dev, unsigned char transition)
 636 {
 637         unsigned char operstate = dev->operstate;
 638
 639         switch (transition) {
 640         case IF_OPER_UP:
 641                 if ((operstate == IF_OPER_DORMANT ||
 642                      operstate == IF_OPER_UNKNOWN) &&
 643                     !netif_dormant(dev))
 644                         operstate = IF_OPER_UP;
 645                 break;
 646
 647         case IF_OPER_DORMANT:
 648                 if (operstate == IF_OPER_UP ||
 649                     operstate == IF_OPER_UNKNOWN)
 650                         operstate = IF_OPER_DORMANT;
 651                 break;
 652         }
 653
 654         if (dev->operstate != operstate) {
 655                 write_lock_bh(&dev_base_lock);
 656                 dev->operstate = operstate;
 657                 write_unlock_bh(&dev_base_lock);
 658                 netdev_state_change(dev);
 659         }
 660 }
 661
 662 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 663 {
 664         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 665                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 666 }
 667
 668 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 669                                            const struct ifinfomsg *ifm)
 670 {
 671         unsigned int flags = ifm->ifi_flags;
 672
 673         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 674         if (ifm->ifi_change)
 675                 flags = (flags & ifm->ifi_change) |
 676                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 677
 678         return flags;
 679 }
 680
 681 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 682                                  const struct rtnl_link_stats64 *b)
 683 {
 684         a->rx_packets = b->rx_packets;
 685         a->tx_packets = b->tx_packets;
 686         a->rx_bytes = b->rx_bytes;
 687         a->tx_bytes = b->tx_bytes;
 688         a->rx_errors = b->rx_errors;
 689         a->tx_errors = b->tx_errors;
 690         a->rx_dropped = b->rx_dropped;
 691         a->tx_dropped = b->tx_dropped;
 692
 693         a->multicast = b->multicast;
 694         a->collisions = b->collisions;
 695
 696         a->rx_length_errors = b->rx_length_errors;
 697         a->rx_over_errors = b->rx_over_errors;
 698         a->rx_crc_errors = b->rx_crc_errors;
 699         a->rx_frame_errors = b->rx_frame_errors;
 700         a->rx_fifo_errors = b->rx_fifo_errors;
 701         a->rx_missed_errors = b->rx_missed_errors;
 702
 703         a->tx_aborted_errors = b->tx_aborted_errors;
 704         a->tx_carrier_errors = b->tx_carrier_errors;
 705         a->tx_fifo_errors = b->tx_fifo_errors;
 706         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 707         a->tx_window_errors = b->tx_window_errors;
 708
 709         a->rx_compressed = b->rx_compressed;
 710         a->tx_compressed = b->tx_compressed;
 711 }
 712
 713 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 714 {
 715         memcpy(v, b, sizeof(*b));
 716 }
 717
 718 /* All VF info */
 719 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 720                                    u32 ext_filter_mask)
 721 {
 722         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 723             (ext_filter_mask & RTEXT_FILTER_VF)) {
 724                 int num_vfs = dev_num_vf(dev->dev.parent);
 725                 size_t size = nla_total_size(sizeof(struct nlattr));
 726                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 727                 size += num_vfs *
 728                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 729                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 730                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 731                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 732                 return size;
 733         } else
 734                 return 0;
 735 }
 736
 737 static size_t rtnl_port_size(const struct net_device *dev)
 738 {
 739         size_t port_size = nla_total_size(4)            /* PORT_VF */
 740                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 741                 + nla_total_size(sizeof(struct ifla_port_vsi))
 742                                                         /* PORT_VSI_TYPE */
 743                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 744                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 745                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 746                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 747         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 748         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 749                 + port_size;
 750         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 751                 + port_size;
 752
 753         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 754                 return 0;
 755         if (dev_num_vf(dev->dev.parent))
 756                 return port_self_size + vf_ports_size +
 757                         vf_port_size * dev_num_vf(dev->dev.parent);
 758         else
 759                 return port_self_size;
 760 }
 761
 762 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 763                                      u32 ext_filter_mask)
 764 {
 765         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 766                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 767                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 768                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 769                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 770                + nla_total_size(sizeof(struct rtnl_link_stats))
 771                + nla_total_size(sizeof(struct rtnl_link_stats64))
 772                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 773                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 774                + nla_total_size(4) /* IFLA_TXQLEN */
 775                + nla_total_size(4) /* IFLA_WEIGHT */
 776                + nla_total_size(4) /* IFLA_MTU */
 777                + nla_total_size(4) /* IFLA_LINK */
 778                + nla_total_size(4) /* IFLA_MASTER */
 779                + nla_total_size(4) /* IFLA_PROMISCUITY */
 780                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 781                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 782                + nla_total_size(1) /* IFLA_OPERSTATE */
 783                + nla_total_size(1) /* IFLA_LINKMODE */
 784                + nla_total_size(ext_filter_mask
 785                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 786                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 787                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 788                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 789                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 790 }
 791
 792 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 793 {
 794         struct nlattr *vf_ports;
 795         struct nlattr *vf_port;
 796         int vf;
 797         int err;
 798
 799         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 800         if (!vf_ports)
 801                 return -EMSGSIZE;
 802
 803         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 804                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 805                 if (!vf_port)
 806                         goto nla_put_failure;
 807                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 808                         goto nla_put_failure;
 809                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 810                 if (err == -EMSGSIZE)
 811                         goto nla_put_failure;
 812                 if (err) {
 813                         nla_nest_cancel(skb, vf_port);
 814                         continue;
 815                 }
 816                 nla_nest_end(skb, vf_port);
 817         }
 818
 819         nla_nest_end(skb, vf_ports);
 820
 821         return 0;
 822
 823 nla_put_failure:
 824         nla_nest_cancel(skb, vf_ports);
 825         return -EMSGSIZE;
 826 }
 827
 828 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 829 {
 830         struct nlattr *port_self;
 831         int err;
 832
 833         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 834         if (!port_self)
 835                 return -EMSGSIZE;
 836
 837         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 838         if (err) {
 839                 nla_nest_cancel(skb, port_self);
 840                 return (err == -EMSGSIZE) ? err : 0;
 841         }
 842
 843         nla_nest_end(skb, port_self);
 844
 845         return 0;
 846 }
 847
 848 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 849 {
 850         int err;
 851
 852         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 853                 return 0;
 854
 855         err = rtnl_port_self_fill(skb, dev);
 856         if (err)
 857                 return err;
 858
 859         if (dev_num_vf(dev->dev.parent)) {
 860                 err = rtnl_vf_ports_fill(skb, dev);
 861                 if (err)
 862                         return err;
 863         }
 864
 865         return 0;
 866 }
 867
 868 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 869                             int type, u32 pid, u32 seq, u32 change,
 870                             unsigned int flags, u32 ext_filter_mask)
 871 {
 872         struct ifinfomsg *ifm;
 873         struct nlmsghdr *nlh;
 874         struct rtnl_link_stats64 temp;
 875         const struct rtnl_link_stats64 *stats;
 876         struct nlattr *attr, *af_spec;
 877         struct rtnl_af_ops *af_ops;
 878
 879         ASSERT_RTNL();
 880         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 881         if (nlh == NULL)
 882                 return -EMSGSIZE;
 883
 884         ifm = nlmsg_data(nlh);
 885         ifm->ifi_family = AF_UNSPEC;
 886         ifm->__ifi_pad = 0;
 887         ifm->ifi_type = dev->type;
 888         ifm->ifi_index = dev->ifindex;
 889         ifm->ifi_flags = dev_get_flags(dev);
 890         ifm->ifi_change = change;
 891
 892         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 893             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 894             nla_put_u8(skb, IFLA_OPERSTATE,
 895                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 896             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 897             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 898             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 899             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 900             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 901 #ifdef CONFIG_RPS
 902             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 903 #endif
 904             (dev->ifindex != dev->iflink &&
 905              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 906             (dev->master &&
 907              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 908             (dev->qdisc &&
 909              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 910             (dev->ifalias &&
 911              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 912                 goto nla_put_failure;
 913
 914         if (1) {
 915                 struct rtnl_link_ifmap map = {
 916                         .mem_start   = dev->mem_start,
 917                         .mem_end     = dev->mem_end,
 918                         .base_addr   = dev->base_addr,
 919                         .irq         = dev->irq,
 920                         .dma         = dev->dma,
 921                         .port        = dev->if_port,
 922                 };
 923                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 924                         goto nla_put_failure;
 925         }
 926
 927         if (dev->addr_len) {
 928                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 929                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 930                         goto nla_put_failure;
 931         }
 932
 933         attr = nla_reserve(skb, IFLA_STATS,
 934                         sizeof(struct rtnl_link_stats));
 935         if (attr == NULL)
 936                 goto nla_put_failure;
 937
 938         stats = dev_get_stats(dev, &temp);
 939         copy_rtnl_link_stats(nla_data(attr), stats);
 940
 941         attr = nla_reserve(skb, IFLA_STATS64,
 942                         sizeof(struct rtnl_link_stats64));
 943         if (attr == NULL)
 944                 goto nla_put_failure;
 945         copy_rtnl_link_stats64(nla_data(attr), stats);
 946
 947         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 948             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 949                 goto nla_put_failure;
 950
 951         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 952             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 953                 int i;
 954
 955                 struct nlattr *vfinfo, *vf;
 956                 int num_vfs = dev_num_vf(dev->dev.parent);
 957
 958                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 959                 if (!vfinfo)
 960                         goto nla_put_failure;
 961                 for (i = 0; i < num_vfs; i++) {
 962                         struct ifla_vf_info ivi;
 963                         struct ifla_vf_mac vf_mac;
 964                         struct ifla_vf_vlan vf_vlan;
 965                         struct ifla_vf_tx_rate vf_tx_rate;
 966                         struct ifla_vf_spoofchk vf_spoofchk;
 967
 968                         /*
 969                          * Not all SR-IOV capable drivers support the
 970                          * spoofcheck query.  Preset to -1 so the user
 971                          * space tool can detect that the driver didn't
 972                          * report anything.
 973                          */
 974                         ivi.spoofchk = -1;
 975                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 976                                 break;
 977                         vf_mac.vf =
 978                                 vf_vlan.vf =
 979                                 vf_tx_rate.vf =
 980                                 vf_spoofchk.vf = ivi.vf;
 981
 982                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 983                         vf_vlan.vlan = ivi.vlan;
 984                         vf_vlan.qos = ivi.qos;
 985                         vf_tx_rate.rate = ivi.tx_rate;
 986                         vf_spoofchk.setting = ivi.spoofchk;
 987                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 988                         if (!vf) {
 989                                 nla_nest_cancel(skb, vfinfo);
 990                                 goto nla_put_failure;
 991                         }
 992                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 993                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 994                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 995                                     &vf_tx_rate) ||
 996                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 997                                     &vf_spoofchk))
 998                                 goto nla_put_failure;
 999                         nla_nest_end(skb, vf);
1000                 }
1001                 nla_nest_end(skb, vfinfo);
1002         }
1003
1004         if (rtnl_port_fill(skb, dev))
1005                 goto nla_put_failure;
1006
1007         if (dev->rtnl_link_ops) {
1008                 if (rtnl_link_fill(skb, dev) < 0)
1009                         goto nla_put_failure;
1010         }
1011
1012         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1013                 goto nla_put_failure;
1014
1015         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1016                 if (af_ops->fill_link_af) {
1017                         struct nlattr *af;
1018                         int err;
1019
1020                         if (!(af = nla_nest_start(skb, af_ops->family)))
1021                                 goto nla_put_failure;
1022
1023                         err = af_ops->fill_link_af(skb, dev);
1024
1025                         /*
1026                          * Caller may return ENODATA to indicate that there
1027                          * was no data to be dumped. This is not an error, it
1028                          * means we should trim the attribute header and
1029                          * continue.
1030                          */
1031                         if (err == -ENODATA)
1032                                 nla_nest_cancel(skb, af);
1033                         else if (err < 0)
1034                                 goto nla_put_failure;
1035
1036                         nla_nest_end(skb, af);
1037                 }
1038         }
1039
1040         nla_nest_end(skb, af_spec);
1041
1042         return nlmsg_end(skb, nlh);
1043
1044 nla_put_failure:
1045         nlmsg_cancel(skb, nlh);
1046         return -EMSGSIZE;
1047 }
1048
1049 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1050 {
1051         struct net *net = sock_net(skb->sk);
1052         int h, s_h;
1053         int idx = 0, s_idx;
1054         struct net_device *dev;
1055         struct hlist_head *head;
1056         struct hlist_node *node;
1057         struct nlattr *tb[IFLA_MAX+1];
1058         u32 ext_filter_mask = 0;
1059
1060         s_h = cb->args[0];
1061         s_idx = cb->args[1];
1062
1063         rcu_read_lock();
1064         cb->seq = net->dev_base_seq;
1065
1066         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1067                         ifla_policy) >= 0) {
1068
1069                 if (tb[IFLA_EXT_MASK])
1070                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1071         }
1072
1073         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1074                 idx = 0;
1075                 head = &net->dev_index_head[h];
1076                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1077                         if (idx < s_idx)
1078                                 goto cont;
1079                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1080                                              NETLINK_CB(cb->skb).pid,
1081                                              cb->nlh->nlmsg_seq, 0,
1082                                              NLM_F_MULTI,
1083                                              ext_filter_mask) <= 0)
1084                                 goto out;
1085
1086                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1087 cont:
1088                         idx++;
1089                 }
1090         }
1091 out:
1092         rcu_read_unlock();
1093         cb->args[1] = idx;
1094         cb->args[0] = h;
1095
1096         return skb->len;
1097 }
1098
1099 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1100         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1101         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1102         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1103         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1104         [IFLA_MTU]              = { .type = NLA_U32 },
1105         [IFLA_LINK]             = { .type = NLA_U32 },
1106         [IFLA_MASTER]           = { .type = NLA_U32 },
1107         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1108         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1109         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1110         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1111         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1112         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1113         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1114         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1115         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1116         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1117         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1118         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1119         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1120         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1121         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1122         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1123 };
1124 EXPORT_SYMBOL(ifla_policy);
1125
1126 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1127         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1128         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1129 };
1130
1131 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1132         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1133 };
1134
1135 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1136         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1137                                     .len = sizeof(struct ifla_vf_mac) },
1138         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1139                                     .len = sizeof(struct ifla_vf_vlan) },
1140         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1141                                     .len = sizeof(struct ifla_vf_tx_rate) },
1142         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1143                                     .len = sizeof(struct ifla_vf_spoofchk) },
1144 };
1145
1146 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1147         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1148         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1149                                     .len = PORT_PROFILE_MAX },
1150         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1151                                     .len = sizeof(struct ifla_port_vsi)},
1152         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1153                                       .len = PORT_UUID_MAX },
1154         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1155                                     .len = PORT_UUID_MAX },
1156         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1157         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1158 };
1159
1160 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1161 {
1162         struct net *net;
1163         /* Examine the link attributes and figure out which
1164          * network namespace we are talking about.
1165          */
1166         if (tb[IFLA_NET_NS_PID])
1167                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1168         else if (tb[IFLA_NET_NS_FD])
1169                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1170         else
1171                 net = get_net(src_net);
1172         return net;
1173 }
1174 EXPORT_SYMBOL(rtnl_link_get_net);
1175
1176 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1177 {
1178         if (dev) {
1179                 if (tb[IFLA_ADDRESS] &&
1180                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1181                         return -EINVAL;
1182
1183                 if (tb[IFLA_BROADCAST] &&
1184                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1185                         return -EINVAL;
1186         }
1187
1188         if (tb[IFLA_AF_SPEC]) {
1189                 struct nlattr *af;
1190                 int rem, err;
1191
1192                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1193                         const struct rtnl_af_ops *af_ops;
1194
1195                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1196                                 return -EAFNOSUPPORT;
1197
1198                         if (!af_ops->set_link_af)
1199                                 return -EOPNOTSUPP;
1200
1201                         if (af_ops->validate_link_af) {
1202                                 err = af_ops->validate_link_af(dev, af);
1203                                 if (err < 0)
1204                                         return err;
1205                         }
1206                 }
1207         }
1208
1209         return 0;
1210 }
1211
1212 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1213 {
1214         int rem, err = -EINVAL;
1215         struct nlattr *vf;
1216         const struct net_device_ops *ops = dev->netdev_ops;
1217
1218         nla_for_each_nested(vf, attr, rem) {
1219                 switch (nla_type(vf)) {
1220                 case IFLA_VF_MAC: {
1221                         struct ifla_vf_mac *ivm;
1222                         ivm = nla_data(vf);
1223                         err = -EOPNOTSUPP;
1224                         if (ops->ndo_set_vf_mac)
1225                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1226                                                           ivm->mac);
1227                         break;
1228                 }
1229                 case IFLA_VF_VLAN: {
1230                         struct ifla_vf_vlan *ivv;
1231                         ivv = nla_data(vf);
1232                         err = -EOPNOTSUPP;
1233                         if (ops->ndo_set_vf_vlan)
1234                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1235                                                            ivv->vlan,
1236                                                            ivv->qos);
1237                         break;
1238                 }
1239                 case IFLA_VF_TX_RATE: {
1240                         struct ifla_vf_tx_rate *ivt;
1241                         ivt = nla_data(vf);
1242                         err = -EOPNOTSUPP;
1243                         if (ops->ndo_set_vf_tx_rate)
1244                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1245                                                               ivt->rate);
1246                         break;
1247                 }
1248                 case IFLA_VF_SPOOFCHK: {
1249                         struct ifla_vf_spoofchk *ivs;
1250                         ivs = nla_data(vf);
1251                         err = -EOPNOTSUPP;
1252                         if (ops->ndo_set_vf_spoofchk)
1253                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1254                                                                ivs->setting);
1255                         break;
1256                 }
1257                 default:
1258                         err = -EINVAL;
1259                         break;
1260                 }
1261                 if (err)
1262                         break;
1263         }
1264         return err;
1265 }
1266
1267 static int do_set_master(struct net_device *dev, int ifindex)
1268 {
1269         struct net_device *master_dev;
1270         const struct net_device_ops *ops;
1271         int err;
1272
1273         if (dev->master) {
1274                 if (dev->master->ifindex == ifindex)
1275                         return 0;
1276                 ops = dev->master->netdev_ops;
1277                 if (ops->ndo_del_slave) {
1278                         err = ops->ndo_del_slave(dev->master, dev);
1279                         if (err)
1280                                 return err;
1281                 } else {
1282                         return -EOPNOTSUPP;
1283                 }
1284         }
1285
1286         if (ifindex) {
1287                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1288                 if (!master_dev)
1289                         return -EINVAL;
1290                 ops = master_dev->netdev_ops;
1291                 if (ops->ndo_add_slave) {
1292                         err = ops->ndo_add_slave(master_dev, dev);
1293                         if (err)
1294                                 return err;
1295                 } else {
1296                         return -EOPNOTSUPP;
1297                 }
1298         }
1299         return 0;
1300 }
1301
1302 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1303                       struct nlattr **tb, char *ifname, int modified)
1304 {
1305         const struct net_device_ops *ops = dev->netdev_ops;
1306         int send_addr_notify = 0;
1307         int err;
1308
1309         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1310                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1311                 if (IS_ERR(net)) {
1312                         err = PTR_ERR(net);
1313                         goto errout;
1314                 }
1315                 err = dev_change_net_namespace(dev, net, ifname);
1316                 put_net(net);
1317                 if (err)
1318                         goto errout;
1319                 modified = 1;
1320         }
1321
1322         if (tb[IFLA_MAP]) {
1323                 struct rtnl_link_ifmap *u_map;
1324                 struct ifmap k_map;
1325
1326                 if (!ops->ndo_set_config) {
1327                         err = -EOPNOTSUPP;
1328                         goto errout;
1329                 }
1330
1331                 if (!netif_device_present(dev)) {
1332                         err = -ENODEV;
1333                         goto errout;
1334                 }
1335
1336                 u_map = nla_data(tb[IFLA_MAP]);
1337                 k_map.mem_start = (unsigned long) u_map->mem_start;
1338                 k_map.mem_end = (unsigned long) u_map->mem_end;
1339                 k_map.base_addr = (unsigned short) u_map->base_addr;
1340                 k_map.irq = (unsigned char) u_map->irq;
1341                 k_map.dma = (unsigned char) u_map->dma;
1342                 k_map.port = (unsigned char) u_map->port;
1343
1344                 err = ops->ndo_set_config(dev, &k_map);
1345                 if (err < 0)
1346                         goto errout;
1347
1348                 modified = 1;
1349         }
1350
1351         if (tb[IFLA_ADDRESS]) {
1352                 struct sockaddr *sa;
1353                 int len;
1354
1355                 if (!ops->ndo_set_mac_address) {
1356                         err = -EOPNOTSUPP;
1357                         goto errout;
1358                 }
1359
1360                 if (!netif_device_present(dev)) {
1361                         err = -ENODEV;
1362                         goto errout;
1363                 }
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 = ops->ndo_set_mac_address(dev, sa);
1375                 kfree(sa);
1376                 if (err)
1377                         goto errout;
1378                 send_addr_notify = 1;
1379                 modified = 1;
1380         }
1381
1382         if (tb[IFLA_MTU]) {
1383                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1384                 if (err < 0)
1385                         goto errout;
1386                 modified = 1;
1387         }
1388
1389         if (tb[IFLA_GROUP]) {
1390                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1391                 modified = 1;
1392         }
1393
1394         /*
1395          * Interface selected by interface index but interface
1396          * name provided implies that a name change has been
1397          * requested.
1398          */
1399         if (ifm->ifi_index > 0 && ifname[0]) {
1400                 err = dev_change_name(dev, ifname);
1401                 if (err < 0)
1402                         goto errout;
1403                 modified = 1;
1404         }
1405
1406         if (tb[IFLA_IFALIAS]) {
1407                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1408                                     nla_len(tb[IFLA_IFALIAS]));
1409                 if (err < 0)
1410                         goto errout;
1411                 modified = 1;
1412         }
1413
1414         if (tb[IFLA_BROADCAST]) {
1415                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1416                 send_addr_notify = 1;
1417         }
1418
1419         if (ifm->ifi_flags || ifm->ifi_change) {
1420                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1421                 if (err < 0)
1422                         goto errout;
1423         }
1424
1425         if (tb[IFLA_MASTER]) {
1426                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1427                 if (err)
1428                         goto errout;
1429                 modified = 1;
1430         }
1431
1432         if (tb[IFLA_TXQLEN])
1433                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1434
1435         if (tb[IFLA_OPERSTATE])
1436                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1437
1438         if (tb[IFLA_LINKMODE]) {
1439                 write_lock_bh(&dev_base_lock);
1440                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1441                 write_unlock_bh(&dev_base_lock);
1442         }
1443
1444         if (tb[IFLA_VFINFO_LIST]) {
1445                 struct nlattr *attr;
1446                 int rem;
1447                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1448                         if (nla_type(attr) != IFLA_VF_INFO) {
1449                                 err = -EINVAL;
1450                                 goto errout;
1451                         }
1452                         err = do_setvfinfo(dev, attr);
1453                         if (err < 0)
1454                                 goto errout;
1455                         modified = 1;
1456                 }
1457         }
1458         err = 0;
1459
1460         if (tb[IFLA_VF_PORTS]) {
1461                 struct nlattr *port[IFLA_PORT_MAX+1];
1462                 struct nlattr *attr;
1463                 int vf;
1464                 int rem;
1465
1466                 err = -EOPNOTSUPP;
1467                 if (!ops->ndo_set_vf_port)
1468                         goto errout;
1469
1470                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1471                         if (nla_type(attr) != IFLA_VF_PORT)
1472                                 continue;
1473                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1474                                 attr, ifla_port_policy);
1475                         if (err < 0)
1476                                 goto errout;
1477                         if (!port[IFLA_PORT_VF]) {
1478                                 err = -EOPNOTSUPP;
1479                                 goto errout;
1480                         }
1481                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1482                         err = ops->ndo_set_vf_port(dev, vf, port);
1483                         if (err < 0)
1484                                 goto errout;
1485                         modified = 1;
1486                 }
1487         }
1488         err = 0;
1489
1490         if (tb[IFLA_PORT_SELF]) {
1491                 struct nlattr *port[IFLA_PORT_MAX+1];
1492
1493                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1494                         tb[IFLA_PORT_SELF], ifla_port_policy);
1495                 if (err < 0)
1496                         goto errout;
1497
1498                 err = -EOPNOTSUPP;
1499                 if (ops->ndo_set_vf_port)
1500                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1501                 if (err < 0)
1502                         goto errout;
1503                 modified = 1;
1504         }
1505
1506         if (tb[IFLA_AF_SPEC]) {
1507                 struct nlattr *af;
1508                 int rem;
1509
1510                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1511                         const struct rtnl_af_ops *af_ops;
1512
1513                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1514                                 BUG();
1515
1516                         err = af_ops->set_link_af(dev, af);
1517                         if (err < 0)
1518                                 goto errout;
1519
1520                         modified = 1;
1521                 }
1522         }
1523         err = 0;
1524
1525 errout:
1526         if (err < 0 && modified)
1527                 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",
1528                                      dev->name);
1529
1530         if (send_addr_notify)
1531                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1532
1533         return err;
1534 }
1535
1536 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1537 {
1538         struct net *net = sock_net(skb->sk);
1539         struct ifinfomsg *ifm;
1540         struct net_device *dev;
1541         int err;
1542         struct nlattr *tb[IFLA_MAX+1];
1543         char ifname[IFNAMSIZ];
1544
1545         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1546         if (err < 0)
1547                 goto errout;
1548
1549         if (tb[IFLA_IFNAME])
1550                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1551         else
1552                 ifname[0] = '\0';
1553
1554         err = -EINVAL;
1555         ifm = nlmsg_data(nlh);
1556         if (ifm->ifi_index > 0)
1557                 dev = __dev_get_by_index(net, ifm->ifi_index);
1558         else if (tb[IFLA_IFNAME])
1559                 dev = __dev_get_by_name(net, ifname);
1560         else
1561                 goto errout;
1562
1563         if (dev == NULL) {
1564                 err = -ENODEV;
1565                 goto errout;
1566         }
1567
1568         err = validate_linkmsg(dev, tb);
1569         if (err < 0)
1570                 goto errout;
1571
1572         err = do_setlink(dev, ifm, tb, ifname, 0);
1573 errout:
1574         return err;
1575 }
1576
1577 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1578 {
1579         struct net *net = sock_net(skb->sk);
1580         const struct rtnl_link_ops *ops;
1581         struct net_device *dev;
1582         struct ifinfomsg *ifm;
1583         char ifname[IFNAMSIZ];
1584         struct nlattr *tb[IFLA_MAX+1];
1585         int err;
1586         LIST_HEAD(list_kill);
1587
1588         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1589         if (err < 0)
1590                 return err;
1591
1592         if (tb[IFLA_IFNAME])
1593                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1594
1595         ifm = nlmsg_data(nlh);
1596         if (ifm->ifi_index > 0)
1597                 dev = __dev_get_by_index(net, ifm->ifi_index);
1598         else if (tb[IFLA_IFNAME])
1599                 dev = __dev_get_by_name(net, ifname);
1600         else
1601                 return -EINVAL;
1602
1603         if (!dev)
1604                 return -ENODEV;
1605
1606         ops = dev->rtnl_link_ops;
1607         if (!ops)
1608                 return -EOPNOTSUPP;
1609
1610         ops->dellink(dev, &list_kill);
1611         unregister_netdevice_many(&list_kill);
1612         list_del(&list_kill);
1613         return 0;
1614 }
1615
1616 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1617 {
1618         unsigned int old_flags;
1619         int err;
1620
1621         old_flags = dev->flags;
1622         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1623                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1624                 if (err < 0)
1625                         return err;
1626         }
1627
1628         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1629         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1630
1631         __dev_notify_flags(dev, old_flags);
1632         return 0;
1633 }
1634 EXPORT_SYMBOL(rtnl_configure_link);
1635
1636 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1637         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1638 {
1639         int err;
1640         struct net_device *dev;
1641         unsigned int num_tx_queues = 1;
1642         unsigned int num_rx_queues = 1;
1643
1644         if (tb[IFLA_NUM_TX_QUEUES])
1645                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1646         else if (ops->get_num_tx_queues)
1647                 num_tx_queues = ops->get_num_tx_queues();
1648
1649         if (tb[IFLA_NUM_RX_QUEUES])
1650                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1651         else if (ops->get_num_rx_queues)
1652                 num_rx_queues = ops->get_num_rx_queues();
1653
1654         err = -ENOMEM;
1655         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1656                                num_tx_queues, num_rx_queues);
1657         if (!dev)
1658                 goto err;
1659
1660         dev_net_set(dev, net);
1661         dev->rtnl_link_ops = ops;
1662         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1663
1664         if (tb[IFLA_MTU])
1665                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1666         if (tb[IFLA_ADDRESS])
1667                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1668                                 nla_len(tb[IFLA_ADDRESS]));
1669         if (tb[IFLA_BROADCAST])
1670                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1671                                 nla_len(tb[IFLA_BROADCAST]));
1672         if (tb[IFLA_TXQLEN])
1673                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1674         if (tb[IFLA_OPERSTATE])
1675                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1676         if (tb[IFLA_LINKMODE])
1677                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1678         if (tb[IFLA_GROUP])
1679                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1680
1681         return dev;
1682
1683 err:
1684         return ERR_PTR(err);
1685 }
1686 EXPORT_SYMBOL(rtnl_create_link);
1687
1688 static int rtnl_group_changelink(struct net *net, int group,
1689                 struct ifinfomsg *ifm,
1690                 struct nlattr **tb)
1691 {
1692         struct net_device *dev;
1693         int err;
1694
1695         for_each_netdev(net, dev) {
1696                 if (dev->group == group) {
1697                         err = do_setlink(dev, ifm, tb, NULL, 0);
1698                         if (err < 0)
1699                                 return err;
1700                 }
1701         }
1702
1703         return 0;
1704 }
1705
1706 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1707 {
1708         struct net *net = sock_net(skb->sk);
1709         const struct rtnl_link_ops *ops;
1710         struct net_device *dev;
1711         struct ifinfomsg *ifm;
1712         char kind[MODULE_NAME_LEN];
1713         char ifname[IFNAMSIZ];
1714         struct nlattr *tb[IFLA_MAX+1];
1715         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1716         int err;
1717
1718 #ifdef CONFIG_MODULES
1719 replay:
1720 #endif
1721         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1722         if (err < 0)
1723                 return err;
1724
1725         if (tb[IFLA_IFNAME])
1726                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1727         else
1728                 ifname[0] = '\0';
1729
1730         ifm = nlmsg_data(nlh);
1731         if (ifm->ifi_index > 0)
1732                 dev = __dev_get_by_index(net, ifm->ifi_index);
1733         else {
1734                 if (ifname[0])
1735                         dev = __dev_get_by_name(net, ifname);
1736                 else
1737                         dev = NULL;
1738         }
1739
1740         err = validate_linkmsg(dev, tb);
1741         if (err < 0)
1742                 return err;
1743
1744         if (tb[IFLA_LINKINFO]) {
1745                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1746                                        tb[IFLA_LINKINFO], ifla_info_policy);
1747                 if (err < 0)
1748                         return err;
1749         } else
1750                 memset(linkinfo, 0, sizeof(linkinfo));
1751
1752         if (linkinfo[IFLA_INFO_KIND]) {
1753                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1754                 ops = rtnl_link_ops_get(kind);
1755         } else {
1756                 kind[0] = '\0';
1757                 ops = NULL;
1758         }
1759
1760         if (1) {
1761                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1762                 struct net *dest_net;
1763
1764                 if (ops) {
1765                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1766                                 err = nla_parse_nested(attr, ops->maxtype,
1767                                                        linkinfo[IFLA_INFO_DATA],
1768                                                        ops->policy);
1769                                 if (err < 0)
1770                                         return err;
1771                                 data = attr;
1772                         }
1773                         if (ops->validate) {
1774                                 err = ops->validate(tb, data);
1775                                 if (err < 0)
1776                                         return err;
1777                         }
1778                 }
1779
1780                 if (dev) {
1781                         int modified = 0;
1782
1783                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1784                                 return -EEXIST;
1785                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1786                                 return -EOPNOTSUPP;
1787
1788                         if (linkinfo[IFLA_INFO_DATA]) {
1789                                 if (!ops || ops != dev->rtnl_link_ops ||
1790                                     !ops->changelink)
1791                                         return -EOPNOTSUPP;
1792
1793                                 err = ops->changelink(dev, tb, data);
1794                                 if (err < 0)
1795                                         return err;
1796                                 modified = 1;
1797                         }
1798
1799                         return do_setlink(dev, ifm, tb, ifname, modified);
1800                 }
1801
1802                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1803                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1804                                 return rtnl_group_changelink(net,
1805                                                 nla_get_u32(tb[IFLA_GROUP]),
1806                                                 ifm, tb);
1807                         return -ENODEV;
1808                 }
1809
1810                 if (ifm->ifi_index)
1811                         return -EOPNOTSUPP;
1812                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1813                         return -EOPNOTSUPP;
1814
1815                 if (!ops) {
1816 #ifdef CONFIG_MODULES
1817                         if (kind[0]) {
1818                                 __rtnl_unlock();
1819                                 request_module("rtnl-link-%s", kind);
1820                                 rtnl_lock();
1821                                 ops = rtnl_link_ops_get(kind);
1822                                 if (ops)
1823                                         goto replay;
1824                         }
1825 #endif
1826                         return -EOPNOTSUPP;
1827                 }
1828
1829                 if (!ifname[0])
1830                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1831
1832                 dest_net = rtnl_link_get_net(net, tb);
1833                 if (IS_ERR(dest_net))
1834                         return PTR_ERR(dest_net);
1835
1836                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1837
1838                 if (IS_ERR(dev))
1839                         err = PTR_ERR(dev);
1840                 else if (ops->newlink)
1841                         err = ops->newlink(net, dev, tb, data);
1842                 else
1843                         err = register_netdevice(dev);
1844
1845                 if (err < 0 && !IS_ERR(dev))
1846                         free_netdev(dev);
1847                 if (err < 0)
1848                         goto out;
1849
1850                 err = rtnl_configure_link(dev, ifm);
1851                 if (err < 0)
1852                         unregister_netdevice(dev);
1853 out:
1854                 put_net(dest_net);
1855                 return err;
1856         }
1857 }
1858
1859 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1860 {
1861         struct net *net = sock_net(skb->sk);
1862         struct ifinfomsg *ifm;
1863         char ifname[IFNAMSIZ];
1864         struct nlattr *tb[IFLA_MAX+1];
1865         struct net_device *dev = NULL;
1866         struct sk_buff *nskb;
1867         int err;
1868         u32 ext_filter_mask = 0;
1869
1870         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1871         if (err < 0)
1872                 return err;
1873
1874         if (tb[IFLA_IFNAME])
1875                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1876
1877         if (tb[IFLA_EXT_MASK])
1878                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1879
1880         ifm = nlmsg_data(nlh);
1881         if (ifm->ifi_index > 0)
1882                 dev = __dev_get_by_index(net, ifm->ifi_index);
1883         else if (tb[IFLA_IFNAME])
1884                 dev = __dev_get_by_name(net, ifname);
1885         else
1886                 return -EINVAL;
1887
1888         if (dev == NULL)
1889                 return -ENODEV;
1890
1891         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1892         if (nskb == NULL)
1893                 return -ENOBUFS;
1894
1895         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1896                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1897         if (err < 0) {
1898                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1899                 WARN_ON(err == -EMSGSIZE);
1900                 kfree_skb(nskb);
1901         } else
1902                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1903
1904         return err;
1905 }
1906
1907 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1908 {
1909         struct net *net = sock_net(skb->sk);
1910         struct net_device *dev;
1911         struct nlattr *tb[IFLA_MAX+1];
1912         u32 ext_filter_mask = 0;
1913         u16 min_ifinfo_dump_size = 0;
1914
1915         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1916                         ifla_policy) >= 0) {
1917                 if (tb[IFLA_EXT_MASK])
1918                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1919         }
1920
1921         if (!ext_filter_mask)
1922                 return NLMSG_GOODSIZE;
1923         /*
1924          * traverse the list of net devices and compute the minimum
1925          * buffer size based upon the filter mask.
1926          */
1927         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1928                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1929                                              if_nlmsg_size(dev,
1930                                                            ext_filter_mask));
1931         }
1932
1933         return min_ifinfo_dump_size;
1934 }
1935
1936 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1937 {
1938         int idx;
1939         int s_idx = cb->family;
1940
1941         if (s_idx == 0)
1942                 s_idx = 1;
1943         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1944                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1945                 if (idx < s_idx || idx == PF_PACKET)
1946                         continue;
1947                 if (rtnl_msg_handlers[idx] == NULL ||
1948                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1949                         continue;
1950                 if (idx > s_idx)
1951                         memset(&cb->args[0], 0, sizeof(cb->args));
1952                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1953                         break;
1954         }
1955         cb->family = idx;
1956
1957         return skb->len;
1958 }
1959
1960 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1961 {
1962         struct net *net = dev_net(dev);
1963         struct sk_buff *skb;
1964         int err = -ENOBUFS;
1965         size_t if_info_size;
1966
1967         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1968         if (skb == NULL)
1969                 goto errout;
1970
1971         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1972         if (err < 0) {
1973                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1974                 WARN_ON(err == -EMSGSIZE);
1975                 kfree_skb(skb);
1976                 goto errout;
1977         }
1978         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1979         return;
1980 errout:
1981         if (err < 0)
1982                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1983 }
1984
1985 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1986                                    struct net_device *dev,
1987                                    u8 *addr, u32 pid, u32 seq,
1988                                    int type, unsigned int flags)
1989 {
1990         struct nlmsghdr *nlh;
1991         struct ndmsg *ndm;
1992
1993         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1994         if (!nlh)
1995                 return -EMSGSIZE;
1996
1997         ndm = nlmsg_data(nlh);
1998         ndm->ndm_family  = AF_BRIDGE;
1999         ndm->ndm_pad1    = 0;
2000         ndm->ndm_pad2    = 0;
2001         ndm->ndm_flags   = flags;
2002         ndm->ndm_type    = 0;
2003         ndm->ndm_ifindex = dev->ifindex;
2004         ndm->ndm_state   = NUD_PERMANENT;
2005
2006         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2007                 goto nla_put_failure;
2008
2009         return nlmsg_end(skb, nlh);
2010
2011 nla_put_failure:
2012         nlmsg_cancel(skb, nlh);
2013         return -EMSGSIZE;
2014 }
2015
2016 static inline size_t rtnl_fdb_nlmsg_size(void)
2017 {
2018         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2019 }
2020
2021 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2022 {
2023         struct net *net = dev_net(dev);
2024         struct sk_buff *skb;
2025         int err = -ENOBUFS;
2026
2027         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2028         if (!skb)
2029                 goto errout;
2030
2031         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2032         if (err < 0) {
2033                 kfree_skb(skb);
2034                 goto errout;
2035         }
2036
2037         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2038         return;
2039 errout:
2040         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2041 }
2042
2043 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2044 {
2045         struct net *net = sock_net(skb->sk);
2046         struct net_device *master = NULL;
2047         struct ndmsg *ndm;
2048         struct nlattr *tb[NDA_MAX+1];
2049         struct net_device *dev;
2050         u8 *addr;
2051         int err;
2052
2053         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2054         if (err < 0)
2055                 return err;
2056
2057         ndm = nlmsg_data(nlh);
2058         if (ndm->ndm_ifindex == 0) {
2059                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2060                 return -EINVAL;
2061         }
2062
2063         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2064         if (dev == NULL) {
2065                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2066                 return -ENODEV;
2067         }
2068
2069         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2070                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2071                 return -EINVAL;
2072         }
2073
2074         addr = nla_data(tb[NDA_LLADDR]);
2075         if (!is_valid_ether_addr(addr)) {
2076                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2077                 return -EINVAL;
2078         }
2079
2080         err = -EOPNOTSUPP;
2081
2082         /* Support fdb on master device the net/bridge default case */
2083         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2084             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2085                 master = dev->master;
2086                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2087                                                       nlh->nlmsg_flags);
2088                 if (err)
2089                         goto out;
2090                 else
2091                         ndm->ndm_flags &= ~NTF_MASTER;
2092         }
2093
2094         /* Embedded bridge, macvlan, and any other device support */
2095         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2096                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2097                                                    nlh->nlmsg_flags);
2098
2099                 if (!err) {
2100                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2101                         ndm->ndm_flags &= ~NTF_SELF;
2102                 }
2103         }
2104 out:
2105         return err;
2106 }
2107
2108 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2109 {
2110         struct net *net = sock_net(skb->sk);
2111         struct ndmsg *ndm;
2112         struct nlattr *llattr;
2113         struct net_device *dev;
2114         int err = -EINVAL;
2115         __u8 *addr;
2116
2117         if (nlmsg_len(nlh) < sizeof(*ndm))
2118                 return -EINVAL;
2119
2120         ndm = nlmsg_data(nlh);
2121         if (ndm->ndm_ifindex == 0) {
2122                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2123                 return -EINVAL;
2124         }
2125
2126         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2127         if (dev == NULL) {
2128                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2129                 return -ENODEV;
2130         }
2131
2132         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2133         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2134                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2135                 return -EINVAL;
2136         }
2137
2138         addr = nla_data(llattr);
2139         err = -EOPNOTSUPP;
2140
2141         /* Support fdb on master device the net/bridge default case */
2142         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2143             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2144                 struct net_device *master = dev->master;
2145
2146                 if (master->netdev_ops->ndo_fdb_del)
2147                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2148
2149                 if (err)
2150                         goto out;
2151                 else
2152                         ndm->ndm_flags &= ~NTF_MASTER;
2153         }
2154
2155         /* Embedded bridge, macvlan, and any other device support */
2156         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2157                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2158
2159                 if (!err) {
2160                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2161                         ndm->ndm_flags &= ~NTF_SELF;
2162                 }
2163         }
2164 out:
2165         return err;
2166 }
2167
2168 static int nlmsg_populate_fdb(struct sk_buff *skb,
2169                               struct netlink_callback *cb,
2170                               struct net_device *dev,
2171                               int *idx,
2172                               struct netdev_hw_addr_list *list)
2173 {
2174         struct netdev_hw_addr *ha;
2175         int err;
2176         u32 pid, seq;
2177
2178         pid = NETLINK_CB(cb->skb).pid;
2179         seq = cb->nlh->nlmsg_seq;
2180
2181         list_for_each_entry(ha, &list->list, list) {
2182                 if (*idx < cb->args[0])
2183                         goto skip;
2184
2185                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2186                                               pid, seq, 0, NTF_SELF);
2187                 if (err < 0)
2188                         return err;
2189 skip:
2190                 *idx += 1;
2191         }
2192         return 0;
2193 }
2194
2195 /**
2196  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2197  * @nlh: netlink message header
2198  * @dev: netdevice
2199  *
2200  * Default netdevice operation to dump the existing unicast address list.
2201  * Returns zero on success.
2202  */
2203 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2204                       struct netlink_callback *cb,
2205                       struct net_device *dev,
2206                       int idx)
2207 {
2208         int err;
2209
2210         netif_addr_lock_bh(dev);
2211         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2212         if (err)
2213                 goto out;
2214         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2215 out:
2216         netif_addr_unlock_bh(dev);
2217         return idx;
2218 }
2219 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2220
2221 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2222 {
2223         int idx = 0;
2224         struct net *net = sock_net(skb->sk);
2225         struct net_device *dev;
2226
2227         rcu_read_lock();
2228         for_each_netdev_rcu(net, dev) {
2229                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2230                         struct net_device *master = dev->master;
2231                         const struct net_device_ops *ops = master->netdev_ops;
2232
2233                         if (ops->ndo_fdb_dump)
2234                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2235                 }
2236
2237                 if (dev->netdev_ops->ndo_fdb_dump)
2238                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2239         }
2240         rcu_read_unlock();
2241
2242         cb->args[0] = idx;
2243         return skb->len;
2244 }
2245
2246 /* Protected by RTNL sempahore.  */
2247 static struct rtattr **rta_buf;
2248 static int rtattr_max;
2249
2250 /* Process one rtnetlink message. */
2251
2252 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2253 {
2254         struct net *net = sock_net(skb->sk);
2255         rtnl_doit_func doit;
2256         int sz_idx, kind;
2257         int min_len;
2258         int family;
2259         int type;
2260         int err;
2261
2262         type = nlh->nlmsg_type;
2263         if (type > RTM_MAX)
2264                 return -EOPNOTSUPP;
2265
2266         type -= RTM_BASE;
2267
2268         /* All the messages must have at least 1 byte length */
2269         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2270                 return 0;
2271
2272         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2273         sz_idx = type>>2;
2274         kind = type&3;
2275
2276         if (kind != 2 && !capable(CAP_NET_ADMIN))
2277                 return -EPERM;
2278
2279         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2280                 struct sock *rtnl;
2281                 rtnl_dumpit_func dumpit;
2282                 rtnl_calcit_func calcit;
2283                 u16 min_dump_alloc = 0;
2284
2285                 dumpit = rtnl_get_dumpit(family, type);
2286                 if (dumpit == NULL)
2287                         return -EOPNOTSUPP;
2288                 calcit = rtnl_get_calcit(family, type);
2289                 if (calcit)
2290                         min_dump_alloc = calcit(skb, nlh);
2291
2292                 __rtnl_unlock();
2293                 rtnl = net->rtnl;
2294                 {
2295                         struct netlink_dump_control c = {
2296                                 .dump           = dumpit,
2297                                 .min_dump_alloc = min_dump_alloc,
2298                         };
2299                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2300                 }
2301                 rtnl_lock();
2302                 return err;
2303         }
2304
2305         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2306
2307         min_len = rtm_min[sz_idx];
2308         if (nlh->nlmsg_len < min_len)
2309                 return -EINVAL;
2310
2311         if (nlh->nlmsg_len > min_len) {
2312                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2313                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2314
2315                 while (RTA_OK(attr, attrlen)) {
2316                         unsigned int flavor = attr->rta_type;
2317                         if (flavor) {
2318                                 if (flavor > rta_max[sz_idx])
2319                                         return -EINVAL;
2320                                 rta_buf[flavor-1] = attr;
2321                         }
2322                         attr = RTA_NEXT(attr, attrlen);
2323                 }
2324         }
2325
2326         doit = rtnl_get_doit(family, type);
2327         if (doit == NULL)
2328                 return -EOPNOTSUPP;
2329
2330         return doit(skb, nlh, (void *)&rta_buf[0]);
2331 }
2332
2333 static void rtnetlink_rcv(struct sk_buff *skb)
2334 {
2335         rtnl_lock();
2336         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2337         rtnl_unlock();
2338 }
2339
2340 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2341 {
2342         struct net_device *dev = ptr;
2343
2344         switch (event) {
2345         case NETDEV_UP:
2346         case NETDEV_DOWN:
2347         case NETDEV_PRE_UP:
2348         case NETDEV_POST_INIT:
2349         case NETDEV_REGISTER:
2350         case NETDEV_CHANGE:
2351         case NETDEV_PRE_TYPE_CHANGE:
2352         case NETDEV_GOING_DOWN:
2353         case NETDEV_UNREGISTER:
2354         case NETDEV_UNREGISTER_BATCH:
2355         case NETDEV_RELEASE:
2356         case NETDEV_JOIN:
2357                 break;
2358         default:
2359                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2360                 break;
2361         }
2362         return NOTIFY_DONE;
2363 }
2364
2365 static struct notifier_block rtnetlink_dev_notifier = {
2366         .notifier_call  = rtnetlink_event,
2367 };
2368
2369
2370 static int __net_init rtnetlink_net_init(struct net *net)
2371 {
2372         struct sock *sk;
2373         struct netlink_kernel_cfg cfg = {
2374                 .groups         = RTNLGRP_MAX,
2375                 .input          = rtnetlink_rcv,
2376                 .cb_mutex       = &rtnl_mutex,
2377         };
2378
2379         sk = netlink_kernel_create(net, NETLINK_ROUTE, THIS_MODULE, &cfg);
2380         if (!sk)
2381                 return -ENOMEM;
2382         net->rtnl = sk;
2383         return 0;
2384 }
2385
2386 static void __net_exit rtnetlink_net_exit(struct net *net)
2387 {
2388         netlink_kernel_release(net->rtnl);
2389         net->rtnl = NULL;
2390 }
2391
2392 static struct pernet_operations rtnetlink_net_ops = {
2393         .init = rtnetlink_net_init,
2394         .exit = rtnetlink_net_exit,
2395 };
2396
2397 void __init rtnetlink_init(void)
2398 {
2399         int i;
2400
2401         rtattr_max = 0;
2402         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2403                 if (rta_max[i] > rtattr_max)
2404                         rtattr_max = rta_max[i];
2405         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2406         if (!rta_buf)
2407                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2408
2409         if (register_pernet_subsys(&rtnetlink_net_ops))
2410                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2411
2412         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2413         register_netdevice_notifier(&rtnetlink_dev_notifier);
2414
2415         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2416                       rtnl_dump_ifinfo, rtnl_calcit);
2417         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2418         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2419         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2420
2421         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2422         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2423
2424         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2425         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2426         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2427 }
2428