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/pci.h>
  39
  40 #include <asm/uaccess.h>
  41 #include <asm/system.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/sock.h>
  51 #include <net/pkt_sched.h>
  52 #include <net/fib_rules.h>
  53 #include <net/rtnetlink.h>
  54 #include <net/net_namespace.h>
  55
  56 struct rtnl_link {
  57         rtnl_doit_func          doit;
  58         rtnl_dumpit_func        dumpit;
  59         rtnl_calcit_func        calcit;
  60 };
  61
  62 static DEFINE_MUTEX(rtnl_mutex);
  63 static u16 min_ifinfo_dump_size;
  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 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 545 {
 546         struct rtattr *rta;
 547         int size = RTA_LENGTH(attrlen);
 548
 549         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 550         rta->rta_type = attrtype;
 551         rta->rta_len = size;
 552         memcpy(RTA_DATA(rta), data, attrlen);
 553         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 554 }
 555 EXPORT_SYMBOL(__rta_fill);
 556
 557 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 558 {
 559         struct sock *rtnl = net->rtnl;
 560         int err = 0;
 561
 562         NETLINK_CB(skb).dst_group = group;
 563         if (echo)
 564                 atomic_inc(&skb->users);
 565         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 566         if (echo)
 567                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 568         return err;
 569 }
 570
 571 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 572 {
 573         struct sock *rtnl = net->rtnl;
 574
 575         return nlmsg_unicast(rtnl, skb, pid);
 576 }
 577 EXPORT_SYMBOL(rtnl_unicast);
 578
 579 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 580                  struct nlmsghdr *nlh, gfp_t flags)
 581 {
 582         struct sock *rtnl = net->rtnl;
 583         int report = 0;
 584
 585         if (nlh)
 586                 report = nlmsg_report(nlh);
 587
 588         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 589 }
 590 EXPORT_SYMBOL(rtnl_notify);
 591
 592 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 593 {
 594         struct sock *rtnl = net->rtnl;
 595
 596         netlink_set_err(rtnl, 0, group, error);
 597 }
 598 EXPORT_SYMBOL(rtnl_set_sk_err);
 599
 600 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 601 {
 602         struct nlattr *mx;
 603         int i, valid = 0;
 604
 605         mx = nla_nest_start(skb, RTA_METRICS);
 606         if (mx == NULL)
 607                 return -ENOBUFS;
 608
 609         for (i = 0; i < RTAX_MAX; i++) {
 610                 if (metrics[i]) {
 611                         valid++;
 612                         NLA_PUT_U32(skb, i+1, metrics[i]);
 613                 }
 614         }
 615
 616         if (!valid) {
 617                 nla_nest_cancel(skb, mx);
 618                 return 0;
 619         }
 620
 621         return nla_nest_end(skb, mx);
 622
 623 nla_put_failure:
 624         nla_nest_cancel(skb, mx);
 625         return -EMSGSIZE;
 626 }
 627 EXPORT_SYMBOL(rtnetlink_put_metrics);
 628
 629 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 630                        u32 ts, u32 tsage, long expires, u32 error)
 631 {
 632         struct rta_cacheinfo ci = {
 633                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 634                 .rta_used = dst->__use,
 635                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 636                 .rta_error = error,
 637                 .rta_id =  id,
 638                 .rta_ts = ts,
 639                 .rta_tsage = tsage,
 640         };
 641
 642         if (expires)
 643                 ci.rta_expires = jiffies_to_clock_t(expires);
 644
 645         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 646 }
 647 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 648
 649 static void set_operstate(struct net_device *dev, unsigned char transition)
 650 {
 651         unsigned char operstate = dev->operstate;
 652
 653         switch (transition) {
 654         case IF_OPER_UP:
 655                 if ((operstate == IF_OPER_DORMANT ||
 656                      operstate == IF_OPER_UNKNOWN) &&
 657                     !netif_dormant(dev))
 658                         operstate = IF_OPER_UP;
 659                 break;
 660
 661         case IF_OPER_DORMANT:
 662                 if (operstate == IF_OPER_UP ||
 663                     operstate == IF_OPER_UNKNOWN)
 664                         operstate = IF_OPER_DORMANT;
 665                 break;
 666         }
 667
 668         if (dev->operstate != operstate) {
 669                 write_lock_bh(&dev_base_lock);
 670                 dev->operstate = operstate;
 671                 write_unlock_bh(&dev_base_lock);
 672                 netdev_state_change(dev);
 673         }
 674 }
 675
 676 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 677                                            const struct ifinfomsg *ifm)
 678 {
 679         unsigned int flags = ifm->ifi_flags;
 680
 681         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 682         if (ifm->ifi_change)
 683                 flags = (flags & ifm->ifi_change) |
 684                         (dev->flags & ~ifm->ifi_change);
 685
 686         return flags;
 687 }
 688
 689 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 690                                  const struct rtnl_link_stats64 *b)
 691 {
 692         a->rx_packets = b->rx_packets;
 693         a->tx_packets = b->tx_packets;
 694         a->rx_bytes = b->rx_bytes;
 695         a->tx_bytes = b->tx_bytes;
 696         a->rx_errors = b->rx_errors;
 697         a->tx_errors = b->tx_errors;
 698         a->rx_dropped = b->rx_dropped;
 699         a->tx_dropped = b->tx_dropped;
 700
 701         a->multicast = b->multicast;
 702         a->collisions = b->collisions;
 703
 704         a->rx_length_errors = b->rx_length_errors;
 705         a->rx_over_errors = b->rx_over_errors;
 706         a->rx_crc_errors = b->rx_crc_errors;
 707         a->rx_frame_errors = b->rx_frame_errors;
 708         a->rx_fifo_errors = b->rx_fifo_errors;
 709         a->rx_missed_errors = b->rx_missed_errors;
 710
 711         a->tx_aborted_errors = b->tx_aborted_errors;
 712         a->tx_carrier_errors = b->tx_carrier_errors;
 713         a->tx_fifo_errors = b->tx_fifo_errors;
 714         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 715         a->tx_window_errors = b->tx_window_errors;
 716
 717         a->rx_compressed = b->rx_compressed;
 718         a->tx_compressed = b->tx_compressed;
 719 }
 720
 721 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 722 {
 723         memcpy(v, b, sizeof(*b));
 724 }
 725
 726 /* All VF info */
 727 static inline int rtnl_vfinfo_size(const struct net_device *dev)
 728 {
 729         if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
 730
 731                 int num_vfs = dev_num_vf(dev->dev.parent);
 732                 size_t size = nla_total_size(sizeof(struct nlattr));
 733                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 734                 size += num_vfs *
 735                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 736                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 737                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 738                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 739                 return size;
 740         } else
 741                 return 0;
 742 }
 743
 744 static size_t rtnl_port_size(const struct net_device *dev)
 745 {
 746         size_t port_size = nla_total_size(4)            /* PORT_VF */
 747                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 748                 + nla_total_size(sizeof(struct ifla_port_vsi))
 749                                                         /* PORT_VSI_TYPE */
 750                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 751                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 752                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 753                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 754         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 755         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 756                 + port_size;
 757         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 758                 + port_size;
 759
 760         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 761                 return 0;
 762         if (dev_num_vf(dev->dev.parent))
 763                 return port_self_size + vf_ports_size +
 764                         vf_port_size * dev_num_vf(dev->dev.parent);
 765         else
 766                 return port_self_size;
 767 }
 768
 769 static noinline size_t if_nlmsg_size(const struct net_device *dev)
 770 {
 771         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 772                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 773                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 774                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 775                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 776                + nla_total_size(sizeof(struct rtnl_link_stats))
 777                + nla_total_size(sizeof(struct rtnl_link_stats64))
 778                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 779                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 780                + nla_total_size(4) /* IFLA_TXQLEN */
 781                + nla_total_size(4) /* IFLA_WEIGHT */
 782                + nla_total_size(4) /* IFLA_MTU */
 783                + nla_total_size(4) /* IFLA_LINK */
 784                + nla_total_size(4) /* IFLA_MASTER */
 785                + nla_total_size(1) /* IFLA_OPERSTATE */
 786                + nla_total_size(1) /* IFLA_LINKMODE */
 787                + nla_total_size(4) /* IFLA_NUM_VF */
 788                + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
 789                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 790                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 791                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 792 }
 793
 794 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 795 {
 796         struct nlattr *vf_ports;
 797         struct nlattr *vf_port;
 798         int vf;
 799         int err;
 800
 801         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 802         if (!vf_ports)
 803                 return -EMSGSIZE;
 804
 805         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 806                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 807                 if (!vf_port)
 808                         goto nla_put_failure;
 809                 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
 810                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 811                 if (err == -EMSGSIZE)
 812                         goto nla_put_failure;
 813                 if (err) {
 814                         nla_nest_cancel(skb, vf_port);
 815                         continue;
 816                 }
 817                 nla_nest_end(skb, vf_port);
 818         }
 819
 820         nla_nest_end(skb, vf_ports);
 821
 822         return 0;
 823
 824 nla_put_failure:
 825         nla_nest_cancel(skb, vf_ports);
 826         return -EMSGSIZE;
 827 }
 828
 829 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 830 {
 831         struct nlattr *port_self;
 832         int err;
 833
 834         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 835         if (!port_self)
 836                 return -EMSGSIZE;
 837
 838         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 839         if (err) {
 840                 nla_nest_cancel(skb, port_self);
 841                 return (err == -EMSGSIZE) ? err : 0;
 842         }
 843
 844         nla_nest_end(skb, port_self);
 845
 846         return 0;
 847 }
 848
 849 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 850 {
 851         int err;
 852
 853         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 854                 return 0;
 855
 856         err = rtnl_port_self_fill(skb, dev);
 857         if (err)
 858                 return err;
 859
 860         if (dev_num_vf(dev->dev.parent)) {
 861                 err = rtnl_vf_ports_fill(skb, dev);
 862                 if (err)
 863                         return err;
 864         }
 865
 866         return 0;
 867 }
 868
 869 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 870                             int type, u32 pid, u32 seq, u32 change,
 871                             unsigned int flags)
 872 {
 873         struct ifinfomsg *ifm;
 874         struct nlmsghdr *nlh;
 875         struct rtnl_link_stats64 temp;
 876         const struct rtnl_link_stats64 *stats;
 877         struct nlattr *attr, *af_spec;
 878         struct rtnl_af_ops *af_ops;
 879
 880         ASSERT_RTNL();
 881         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 882         if (nlh == NULL)
 883                 return -EMSGSIZE;
 884
 885         ifm = nlmsg_data(nlh);
 886         ifm->ifi_family = AF_UNSPEC;
 887         ifm->__ifi_pad = 0;
 888         ifm->ifi_type = dev->type;
 889         ifm->ifi_index = dev->ifindex;
 890         ifm->ifi_flags = dev_get_flags(dev);
 891         ifm->ifi_change = change;
 892
 893         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 894         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 895         NLA_PUT_U8(skb, IFLA_OPERSTATE,
 896                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 897         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 898         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 899         NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
 900
 901         if (dev->ifindex != dev->iflink)
 902                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 903
 904         if (dev->master)
 905                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 906
 907         if (dev->qdisc)
 908                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
 909
 910         if (dev->ifalias)
 911                 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 912
 913         if (1) {
 914                 struct rtnl_link_ifmap map = {
 915                         .mem_start   = dev->mem_start,
 916                         .mem_end     = dev->mem_end,
 917                         .base_addr   = dev->base_addr,
 918                         .irq         = dev->irq,
 919                         .dma         = dev->dma,
 920                         .port        = dev->if_port,
 921                 };
 922                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 923         }
 924
 925         if (dev->addr_len) {
 926                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 927                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 928         }
 929
 930         attr = nla_reserve(skb, IFLA_STATS,
 931                         sizeof(struct rtnl_link_stats));
 932         if (attr == NULL)
 933                 goto nla_put_failure;
 934
 935         stats = dev_get_stats(dev, &temp);
 936         copy_rtnl_link_stats(nla_data(attr), stats);
 937
 938         attr = nla_reserve(skb, IFLA_STATS64,
 939                         sizeof(struct rtnl_link_stats64));
 940         if (attr == NULL)
 941                 goto nla_put_failure;
 942         copy_rtnl_link_stats64(nla_data(attr), stats);
 943
 944         if (dev->dev.parent)
 945                 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
 946
 947         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
 948                 int i;
 949
 950                 struct nlattr *vfinfo, *vf;
 951                 int num_vfs = dev_num_vf(dev->dev.parent);
 952
 953                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 954                 if (!vfinfo)
 955                         goto nla_put_failure;
 956                 for (i = 0; i < num_vfs; i++) {
 957                         struct ifla_vf_info ivi;
 958                         struct ifla_vf_mac vf_mac;
 959                         struct ifla_vf_vlan vf_vlan;
 960                         struct ifla_vf_tx_rate vf_tx_rate;
 961                         struct ifla_vf_spoofchk vf_spoofchk;
 962
 963                         /*
 964                          * Not all SR-IOV capable drivers support the
 965                          * spoofcheck query.  Preset to -1 so the user
 966                          * space tool can detect that the driver didn't
 967                          * report anything.
 968                          */
 969                         ivi.spoofchk = -1;
 970                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 971                                 break;
 972                         vf_mac.vf =
 973                                 vf_vlan.vf =
 974                                 vf_tx_rate.vf =
 975                                 vf_spoofchk.vf = ivi.vf;
 976
 977                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 978                         vf_vlan.vlan = ivi.vlan;
 979                         vf_vlan.qos = ivi.qos;
 980                         vf_tx_rate.rate = ivi.tx_rate;
 981                         vf_spoofchk.setting = ivi.spoofchk;
 982                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 983                         if (!vf) {
 984                                 nla_nest_cancel(skb, vfinfo);
 985                                 goto nla_put_failure;
 986                         }
 987                         NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
 988                         NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
 989                         NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 990                                 &vf_tx_rate);
 991                         NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 992                                 &vf_spoofchk);
 993                         nla_nest_end(skb, vf);
 994                 }
 995                 nla_nest_end(skb, vfinfo);
 996         }
 997
 998         if (rtnl_port_fill(skb, dev))
 999                 goto nla_put_failure;
1000
1001         if (dev->rtnl_link_ops) {
1002                 if (rtnl_link_fill(skb, dev) < 0)
1003                         goto nla_put_failure;
1004         }
1005
1006         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1007                 goto nla_put_failure;
1008
1009         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1010                 if (af_ops->fill_link_af) {
1011                         struct nlattr *af;
1012                         int err;
1013
1014                         if (!(af = nla_nest_start(skb, af_ops->family)))
1015                                 goto nla_put_failure;
1016
1017                         err = af_ops->fill_link_af(skb, dev);
1018
1019                         /*
1020                          * Caller may return ENODATA to indicate that there
1021                          * was no data to be dumped. This is not an error, it
1022                          * means we should trim the attribute header and
1023                          * continue.
1024                          */
1025                         if (err == -ENODATA)
1026                                 nla_nest_cancel(skb, af);
1027                         else if (err < 0)
1028                                 goto nla_put_failure;
1029
1030                         nla_nest_end(skb, af);
1031                 }
1032         }
1033
1034         nla_nest_end(skb, af_spec);
1035
1036         return nlmsg_end(skb, nlh);
1037
1038 nla_put_failure:
1039         nlmsg_cancel(skb, nlh);
1040         return -EMSGSIZE;
1041 }
1042
1043 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1044 {
1045         struct net *net = sock_net(skb->sk);
1046         int h, s_h;
1047         int idx = 0, s_idx;
1048         struct net_device *dev;
1049         struct hlist_head *head;
1050         struct hlist_node *node;
1051
1052         s_h = cb->args[0];
1053         s_idx = cb->args[1];
1054
1055         rcu_read_lock();
1056         cb->seq = net->dev_base_seq;
1057
1058         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1059                 idx = 0;
1060                 head = &net->dev_index_head[h];
1061                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1062                         if (idx < s_idx)
1063                                 goto cont;
1064                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1065                                              NETLINK_CB(cb->skb).pid,
1066                                              cb->nlh->nlmsg_seq, 0,
1067                                              NLM_F_MULTI) <= 0)
1068                                 goto out;
1069
1070                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1071 cont:
1072                         idx++;
1073                 }
1074         }
1075 out:
1076         rcu_read_unlock();
1077         cb->args[1] = idx;
1078         cb->args[0] = h;
1079
1080         return skb->len;
1081 }
1082
1083 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1084         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1085         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1086         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1087         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1088         [IFLA_MTU]              = { .type = NLA_U32 },
1089         [IFLA_LINK]             = { .type = NLA_U32 },
1090         [IFLA_MASTER]           = { .type = NLA_U32 },
1091         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1092         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1093         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1094         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1095         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1096         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1097         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1098         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1099         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1100         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1101         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1102         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1103 };
1104 EXPORT_SYMBOL(ifla_policy);
1105
1106 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1107         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1108         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1109 };
1110
1111 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1112         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1113 };
1114
1115 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1116         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1117                                     .len = sizeof(struct ifla_vf_mac) },
1118         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1119                                     .len = sizeof(struct ifla_vf_vlan) },
1120         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1121                                     .len = sizeof(struct ifla_vf_tx_rate) },
1122 };
1123
1124 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1125         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1126         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1127                                     .len = PORT_PROFILE_MAX },
1128         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1129                                     .len = sizeof(struct ifla_port_vsi)},
1130         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1131                                       .len = PORT_UUID_MAX },
1132         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1133                                     .len = PORT_UUID_MAX },
1134         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1135         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1136 };
1137
1138 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1139 {
1140         struct net *net;
1141         /* Examine the link attributes and figure out which
1142          * network namespace we are talking about.
1143          */
1144         if (tb[IFLA_NET_NS_PID])
1145                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1146         else if (tb[IFLA_NET_NS_FD])
1147                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1148         else
1149                 net = get_net(src_net);
1150         return net;
1151 }
1152 EXPORT_SYMBOL(rtnl_link_get_net);
1153
1154 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1155 {
1156         if (dev) {
1157                 if (tb[IFLA_ADDRESS] &&
1158                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1159                         return -EINVAL;
1160
1161                 if (tb[IFLA_BROADCAST] &&
1162                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1163                         return -EINVAL;
1164         }
1165
1166         if (tb[IFLA_AF_SPEC]) {
1167                 struct nlattr *af;
1168                 int rem, err;
1169
1170                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1171                         const struct rtnl_af_ops *af_ops;
1172
1173                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1174                                 return -EAFNOSUPPORT;
1175
1176                         if (!af_ops->set_link_af)
1177                                 return -EOPNOTSUPP;
1178
1179                         if (af_ops->validate_link_af) {
1180                                 err = af_ops->validate_link_af(dev, af);
1181                                 if (err < 0)
1182                                         return err;
1183                         }
1184                 }
1185         }
1186
1187         return 0;
1188 }
1189
1190 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1191 {
1192         int rem, err = -EINVAL;
1193         struct nlattr *vf;
1194         const struct net_device_ops *ops = dev->netdev_ops;
1195
1196         nla_for_each_nested(vf, attr, rem) {
1197                 switch (nla_type(vf)) {
1198                 case IFLA_VF_MAC: {
1199                         struct ifla_vf_mac *ivm;
1200                         ivm = nla_data(vf);
1201                         err = -EOPNOTSUPP;
1202                         if (ops->ndo_set_vf_mac)
1203                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1204                                                           ivm->mac);
1205                         break;
1206                 }
1207                 case IFLA_VF_VLAN: {
1208                         struct ifla_vf_vlan *ivv;
1209                         ivv = nla_data(vf);
1210                         err = -EOPNOTSUPP;
1211                         if (ops->ndo_set_vf_vlan)
1212                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1213                                                            ivv->vlan,
1214                                                            ivv->qos);
1215                         break;
1216                 }
1217                 case IFLA_VF_TX_RATE: {
1218                         struct ifla_vf_tx_rate *ivt;
1219                         ivt = nla_data(vf);
1220                         err = -EOPNOTSUPP;
1221                         if (ops->ndo_set_vf_tx_rate)
1222                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1223                                                               ivt->rate);
1224                         break;
1225                 }
1226                 case IFLA_VF_SPOOFCHK: {
1227                         struct ifla_vf_spoofchk *ivs;
1228                         ivs = nla_data(vf);
1229                         err = -EOPNOTSUPP;
1230                         if (ops->ndo_set_vf_spoofchk)
1231                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1232                                                                ivs->setting);
1233                         break;
1234                 }
1235                 default:
1236                         err = -EINVAL;
1237                         break;
1238                 }
1239                 if (err)
1240                         break;
1241         }
1242         return err;
1243 }
1244
1245 static int do_set_master(struct net_device *dev, int ifindex)
1246 {
1247         struct net_device *master_dev;
1248         const struct net_device_ops *ops;
1249         int err;
1250
1251         if (dev->master) {
1252                 if (dev->master->ifindex == ifindex)
1253                         return 0;
1254                 ops = dev->master->netdev_ops;
1255                 if (ops->ndo_del_slave) {
1256                         err = ops->ndo_del_slave(dev->master, dev);
1257                         if (err)
1258                                 return err;
1259                 } else {
1260                         return -EOPNOTSUPP;
1261                 }
1262         }
1263
1264         if (ifindex) {
1265                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1266                 if (!master_dev)
1267                         return -EINVAL;
1268                 ops = master_dev->netdev_ops;
1269                 if (ops->ndo_add_slave) {
1270                         err = ops->ndo_add_slave(master_dev, dev);
1271                         if (err)
1272                                 return err;
1273                 } else {
1274                         return -EOPNOTSUPP;
1275                 }
1276         }
1277         return 0;
1278 }
1279
1280 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1281                       struct nlattr **tb, char *ifname, int modified)
1282 {
1283         const struct net_device_ops *ops = dev->netdev_ops;
1284         int send_addr_notify = 0;
1285         int err;
1286
1287         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1288                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1289                 if (IS_ERR(net)) {
1290                         err = PTR_ERR(net);
1291                         goto errout;
1292                 }
1293                 err = dev_change_net_namespace(dev, net, ifname);
1294                 put_net(net);
1295                 if (err)
1296                         goto errout;
1297                 modified = 1;
1298         }
1299
1300         if (tb[IFLA_MAP]) {
1301                 struct rtnl_link_ifmap *u_map;
1302                 struct ifmap k_map;
1303
1304                 if (!ops->ndo_set_config) {
1305                         err = -EOPNOTSUPP;
1306                         goto errout;
1307                 }
1308
1309                 if (!netif_device_present(dev)) {
1310                         err = -ENODEV;
1311                         goto errout;
1312                 }
1313
1314                 u_map = nla_data(tb[IFLA_MAP]);
1315                 k_map.mem_start = (unsigned long) u_map->mem_start;
1316                 k_map.mem_end = (unsigned long) u_map->mem_end;
1317                 k_map.base_addr = (unsigned short) u_map->base_addr;
1318                 k_map.irq = (unsigned char) u_map->irq;
1319                 k_map.dma = (unsigned char) u_map->dma;
1320                 k_map.port = (unsigned char) u_map->port;
1321
1322                 err = ops->ndo_set_config(dev, &k_map);
1323                 if (err < 0)
1324                         goto errout;
1325
1326                 modified = 1;
1327         }
1328
1329         if (tb[IFLA_ADDRESS]) {
1330                 struct sockaddr *sa;
1331                 int len;
1332
1333                 if (!ops->ndo_set_mac_address) {
1334                         err = -EOPNOTSUPP;
1335                         goto errout;
1336                 }
1337
1338                 if (!netif_device_present(dev)) {
1339                         err = -ENODEV;
1340                         goto errout;
1341                 }
1342
1343                 len = sizeof(sa_family_t) + dev->addr_len;
1344                 sa = kmalloc(len, GFP_KERNEL);
1345                 if (!sa) {
1346                         err = -ENOMEM;
1347                         goto errout;
1348                 }
1349                 sa->sa_family = dev->type;
1350                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1351                        dev->addr_len);
1352                 err = ops->ndo_set_mac_address(dev, sa);
1353                 kfree(sa);
1354                 if (err)
1355                         goto errout;
1356                 send_addr_notify = 1;
1357                 modified = 1;
1358         }
1359
1360         if (tb[IFLA_MTU]) {
1361                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1362                 if (err < 0)
1363                         goto errout;
1364                 modified = 1;
1365         }
1366
1367         if (tb[IFLA_GROUP]) {
1368                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1369                 modified = 1;
1370         }
1371
1372         /*
1373          * Interface selected by interface index but interface
1374          * name provided implies that a name change has been
1375          * requested.
1376          */
1377         if (ifm->ifi_index > 0 && ifname[0]) {
1378                 err = dev_change_name(dev, ifname);
1379                 if (err < 0)
1380                         goto errout;
1381                 modified = 1;
1382         }
1383
1384         if (tb[IFLA_IFALIAS]) {
1385                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1386                                     nla_len(tb[IFLA_IFALIAS]));
1387                 if (err < 0)
1388                         goto errout;
1389                 modified = 1;
1390         }
1391
1392         if (tb[IFLA_BROADCAST]) {
1393                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1394                 send_addr_notify = 1;
1395         }
1396
1397         if (ifm->ifi_flags || ifm->ifi_change) {
1398                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1399                 if (err < 0)
1400                         goto errout;
1401         }
1402
1403         if (tb[IFLA_MASTER]) {
1404                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1405                 if (err)
1406                         goto errout;
1407                 modified = 1;
1408         }
1409
1410         if (tb[IFLA_TXQLEN])
1411                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1412
1413         if (tb[IFLA_OPERSTATE])
1414                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1415
1416         if (tb[IFLA_LINKMODE]) {
1417                 write_lock_bh(&dev_base_lock);
1418                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1419                 write_unlock_bh(&dev_base_lock);
1420         }
1421
1422         if (tb[IFLA_VFINFO_LIST]) {
1423                 struct nlattr *attr;
1424                 int rem;
1425                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1426                         if (nla_type(attr) != IFLA_VF_INFO) {
1427                                 err = -EINVAL;
1428                                 goto errout;
1429                         }
1430                         err = do_setvfinfo(dev, attr);
1431                         if (err < 0)
1432                                 goto errout;
1433                         modified = 1;
1434                 }
1435         }
1436         err = 0;
1437
1438         if (tb[IFLA_VF_PORTS]) {
1439                 struct nlattr *port[IFLA_PORT_MAX+1];
1440                 struct nlattr *attr;
1441                 int vf;
1442                 int rem;
1443
1444                 err = -EOPNOTSUPP;
1445                 if (!ops->ndo_set_vf_port)
1446                         goto errout;
1447
1448                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1449                         if (nla_type(attr) != IFLA_VF_PORT)
1450                                 continue;
1451                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1452                                 attr, ifla_port_policy);
1453                         if (err < 0)
1454                                 goto errout;
1455                         if (!port[IFLA_PORT_VF]) {
1456                                 err = -EOPNOTSUPP;
1457                                 goto errout;
1458                         }
1459                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1460                         err = ops->ndo_set_vf_port(dev, vf, port);
1461                         if (err < 0)
1462                                 goto errout;
1463                         modified = 1;
1464                 }
1465         }
1466         err = 0;
1467
1468         if (tb[IFLA_PORT_SELF]) {
1469                 struct nlattr *port[IFLA_PORT_MAX+1];
1470
1471                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1472                         tb[IFLA_PORT_SELF], ifla_port_policy);
1473                 if (err < 0)
1474                         goto errout;
1475
1476                 err = -EOPNOTSUPP;
1477                 if (ops->ndo_set_vf_port)
1478                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1479                 if (err < 0)
1480                         goto errout;
1481                 modified = 1;
1482         }
1483
1484         if (tb[IFLA_AF_SPEC]) {
1485                 struct nlattr *af;
1486                 int rem;
1487
1488                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1489                         const struct rtnl_af_ops *af_ops;
1490
1491                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1492                                 BUG();
1493
1494                         err = af_ops->set_link_af(dev, af);
1495                         if (err < 0)
1496                                 goto errout;
1497
1498                         modified = 1;
1499                 }
1500         }
1501         err = 0;
1502
1503 errout:
1504         if (err < 0 && modified && net_ratelimit())
1505                 printk(KERN_WARNING "A link change request failed with "
1506                        "some changes committed already. Interface %s may "
1507                        "have been left with an inconsistent configuration, "
1508                        "please check.\n", dev->name);
1509
1510         if (send_addr_notify)
1511                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1512         min_ifinfo_dump_size = max_t(u16, if_nlmsg_size(dev),
1513                                      min_ifinfo_dump_size);
1514
1515         return err;
1516 }
1517
1518 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1519 {
1520         struct net *net = sock_net(skb->sk);
1521         struct ifinfomsg *ifm;
1522         struct net_device *dev;
1523         int err;
1524         struct nlattr *tb[IFLA_MAX+1];
1525         char ifname[IFNAMSIZ];
1526
1527         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1528         if (err < 0)
1529                 goto errout;
1530
1531         if (tb[IFLA_IFNAME])
1532                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1533         else
1534                 ifname[0] = '\0';
1535
1536         err = -EINVAL;
1537         ifm = nlmsg_data(nlh);
1538         if (ifm->ifi_index > 0)
1539                 dev = __dev_get_by_index(net, ifm->ifi_index);
1540         else if (tb[IFLA_IFNAME])
1541                 dev = __dev_get_by_name(net, ifname);
1542         else
1543                 goto errout;
1544
1545         if (dev == NULL) {
1546                 err = -ENODEV;
1547                 goto errout;
1548         }
1549
1550         err = validate_linkmsg(dev, tb);
1551         if (err < 0)
1552                 goto errout;
1553
1554         err = do_setlink(dev, ifm, tb, ifname, 0);
1555 errout:
1556         return err;
1557 }
1558
1559 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1560 {
1561         struct net *net = sock_net(skb->sk);
1562         const struct rtnl_link_ops *ops;
1563         struct net_device *dev;
1564         struct ifinfomsg *ifm;
1565         char ifname[IFNAMSIZ];
1566         struct nlattr *tb[IFLA_MAX+1];
1567         int err;
1568         LIST_HEAD(list_kill);
1569
1570         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1571         if (err < 0)
1572                 return err;
1573
1574         if (tb[IFLA_IFNAME])
1575                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1576
1577         ifm = nlmsg_data(nlh);
1578         if (ifm->ifi_index > 0)
1579                 dev = __dev_get_by_index(net, ifm->ifi_index);
1580         else if (tb[IFLA_IFNAME])
1581                 dev = __dev_get_by_name(net, ifname);
1582         else
1583                 return -EINVAL;
1584
1585         if (!dev)
1586                 return -ENODEV;
1587
1588         ops = dev->rtnl_link_ops;
1589         if (!ops)
1590                 return -EOPNOTSUPP;
1591
1592         ops->dellink(dev, &list_kill);
1593         unregister_netdevice_many(&list_kill);
1594         list_del(&list_kill);
1595         return 0;
1596 }
1597
1598 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1599 {
1600         unsigned int old_flags;
1601         int err;
1602
1603         old_flags = dev->flags;
1604         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1605                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1606                 if (err < 0)
1607                         return err;
1608         }
1609
1610         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1611         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1612
1613         __dev_notify_flags(dev, old_flags);
1614         return 0;
1615 }
1616 EXPORT_SYMBOL(rtnl_configure_link);
1617
1618 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1619         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1620 {
1621         int err;
1622         struct net_device *dev;
1623         unsigned int num_queues = 1;
1624         unsigned int real_num_queues = 1;
1625
1626         if (ops->get_tx_queues) {
1627                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1628                                          &real_num_queues);
1629                 if (err)
1630                         goto err;
1631         }
1632         err = -ENOMEM;
1633         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1634         if (!dev)
1635                 goto err;
1636
1637         dev_net_set(dev, net);
1638         dev->rtnl_link_ops = ops;
1639         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1640
1641         if (tb[IFLA_MTU])
1642                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1643         if (tb[IFLA_ADDRESS])
1644                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1645                                 nla_len(tb[IFLA_ADDRESS]));
1646         if (tb[IFLA_BROADCAST])
1647                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1648                                 nla_len(tb[IFLA_BROADCAST]));
1649         if (tb[IFLA_TXQLEN])
1650                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1651         if (tb[IFLA_OPERSTATE])
1652                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1653         if (tb[IFLA_LINKMODE])
1654                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1655         if (tb[IFLA_GROUP])
1656                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1657
1658         return dev;
1659
1660 err:
1661         return ERR_PTR(err);
1662 }
1663 EXPORT_SYMBOL(rtnl_create_link);
1664
1665 static int rtnl_group_changelink(struct net *net, int group,
1666                 struct ifinfomsg *ifm,
1667                 struct nlattr **tb)
1668 {
1669         struct net_device *dev;
1670         int err;
1671
1672         for_each_netdev(net, dev) {
1673                 if (dev->group == group) {
1674                         err = do_setlink(dev, ifm, tb, NULL, 0);
1675                         if (err < 0)
1676                                 return err;
1677                 }
1678         }
1679
1680         return 0;
1681 }
1682
1683 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1684 {
1685         struct net *net = sock_net(skb->sk);
1686         const struct rtnl_link_ops *ops;
1687         struct net_device *dev;
1688         struct ifinfomsg *ifm;
1689         char kind[MODULE_NAME_LEN];
1690         char ifname[IFNAMSIZ];
1691         struct nlattr *tb[IFLA_MAX+1];
1692         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1693         int err;
1694
1695 #ifdef CONFIG_MODULES
1696 replay:
1697 #endif
1698         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1699         if (err < 0)
1700                 return err;
1701
1702         if (tb[IFLA_IFNAME])
1703                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1704         else
1705                 ifname[0] = '\0';
1706
1707         ifm = nlmsg_data(nlh);
1708         if (ifm->ifi_index > 0)
1709                 dev = __dev_get_by_index(net, ifm->ifi_index);
1710         else {
1711                 if (ifname[0])
1712                         dev = __dev_get_by_name(net, ifname);
1713                 else
1714                         dev = NULL;
1715         }
1716
1717         err = validate_linkmsg(dev, tb);
1718         if (err < 0)
1719                 return err;
1720
1721         if (tb[IFLA_LINKINFO]) {
1722                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1723                                        tb[IFLA_LINKINFO], ifla_info_policy);
1724                 if (err < 0)
1725                         return err;
1726         } else
1727                 memset(linkinfo, 0, sizeof(linkinfo));
1728
1729         if (linkinfo[IFLA_INFO_KIND]) {
1730                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1731                 ops = rtnl_link_ops_get(kind);
1732         } else {
1733                 kind[0] = '\0';
1734                 ops = NULL;
1735         }
1736
1737         if (1) {
1738                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1739                 struct net *dest_net;
1740
1741                 if (ops) {
1742                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1743                                 err = nla_parse_nested(attr, ops->maxtype,
1744                                                        linkinfo[IFLA_INFO_DATA],
1745                                                        ops->policy);
1746                                 if (err < 0)
1747                                         return err;
1748                                 data = attr;
1749                         }
1750                         if (ops->validate) {
1751                                 err = ops->validate(tb, data);
1752                                 if (err < 0)
1753                                         return err;
1754                         }
1755                 }
1756
1757                 if (dev) {
1758                         int modified = 0;
1759
1760                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1761                                 return -EEXIST;
1762                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1763                                 return -EOPNOTSUPP;
1764
1765                         if (linkinfo[IFLA_INFO_DATA]) {
1766                                 if (!ops || ops != dev->rtnl_link_ops ||
1767                                     !ops->changelink)
1768                                         return -EOPNOTSUPP;
1769
1770                                 err = ops->changelink(dev, tb, data);
1771                                 if (err < 0)
1772                                         return err;
1773                                 modified = 1;
1774                         }
1775
1776                         return do_setlink(dev, ifm, tb, ifname, modified);
1777                 }
1778
1779                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1780                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1781                                 return rtnl_group_changelink(net,
1782                                                 nla_get_u32(tb[IFLA_GROUP]),
1783                                                 ifm, tb);
1784                         return -ENODEV;
1785                 }
1786
1787                 if (ifm->ifi_index)
1788                         return -EOPNOTSUPP;
1789                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1790                         return -EOPNOTSUPP;
1791
1792                 if (!ops) {
1793 #ifdef CONFIG_MODULES
1794                         if (kind[0]) {
1795                                 __rtnl_unlock();
1796                                 request_module("rtnl-link-%s", kind);
1797                                 rtnl_lock();
1798                                 ops = rtnl_link_ops_get(kind);
1799                                 if (ops)
1800                                         goto replay;
1801                         }
1802 #endif
1803                         return -EOPNOTSUPP;
1804                 }
1805
1806                 if (!ifname[0])
1807                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1808
1809                 dest_net = rtnl_link_get_net(net, tb);
1810                 if (IS_ERR(dest_net))
1811                         return PTR_ERR(dest_net);
1812
1813                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1814
1815                 if (IS_ERR(dev))
1816                         err = PTR_ERR(dev);
1817                 else if (ops->newlink)
1818                         err = ops->newlink(net, dev, tb, data);
1819                 else
1820                         err = register_netdevice(dev);
1821
1822                 if (err < 0 && !IS_ERR(dev))
1823                         free_netdev(dev);
1824                 if (err < 0)
1825                         goto out;
1826
1827                 err = rtnl_configure_link(dev, ifm);
1828                 if (err < 0)
1829                         unregister_netdevice(dev);
1830 out:
1831                 put_net(dest_net);
1832                 return err;
1833         }
1834 }
1835
1836 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1837 {
1838         struct net *net = sock_net(skb->sk);
1839         struct ifinfomsg *ifm;
1840         char ifname[IFNAMSIZ];
1841         struct nlattr *tb[IFLA_MAX+1];
1842         struct net_device *dev = NULL;
1843         struct sk_buff *nskb;
1844         int err;
1845
1846         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1847         if (err < 0)
1848                 return err;
1849
1850         if (tb[IFLA_IFNAME])
1851                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1852
1853         ifm = nlmsg_data(nlh);
1854         if (ifm->ifi_index > 0)
1855                 dev = __dev_get_by_index(net, ifm->ifi_index);
1856         else if (tb[IFLA_IFNAME])
1857                 dev = __dev_get_by_name(net, ifname);
1858         else
1859                 return -EINVAL;
1860
1861         if (dev == NULL)
1862                 return -ENODEV;
1863
1864         nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1865         if (nskb == NULL)
1866                 return -ENOBUFS;
1867
1868         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1869                                nlh->nlmsg_seq, 0, 0);
1870         if (err < 0) {
1871                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1872                 WARN_ON(err == -EMSGSIZE);
1873                 kfree_skb(nskb);
1874         } else
1875                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1876
1877         return err;
1878 }
1879
1880 static u16 rtnl_calcit(struct sk_buff *skb)
1881 {
1882         return min_ifinfo_dump_size;
1883 }
1884
1885 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1886 {
1887         int idx;
1888         int s_idx = cb->family;
1889
1890         if (s_idx == 0)
1891                 s_idx = 1;
1892         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1893                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1894                 if (idx < s_idx || idx == PF_PACKET)
1895                         continue;
1896                 if (rtnl_msg_handlers[idx] == NULL ||
1897                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1898                         continue;
1899                 if (idx > s_idx)
1900                         memset(&cb->args[0], 0, sizeof(cb->args));
1901                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1902                         break;
1903         }
1904         cb->family = idx;
1905
1906         return skb->len;
1907 }
1908
1909 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1910 {
1911         struct net *net = dev_net(dev);
1912         struct sk_buff *skb;
1913         int err = -ENOBUFS;
1914         size_t if_info_size;
1915
1916         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1917         if (skb == NULL)
1918                 goto errout;
1919
1920         min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1921
1922         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1923         if (err < 0) {
1924                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1925                 WARN_ON(err == -EMSGSIZE);
1926                 kfree_skb(skb);
1927                 goto errout;
1928         }
1929         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1930         return;
1931 errout:
1932         if (err < 0)
1933                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1934 }
1935
1936 /* Protected by RTNL sempahore.  */
1937 static struct rtattr **rta_buf;
1938 static int rtattr_max;
1939
1940 /* Process one rtnetlink message. */
1941
1942 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1943 {
1944         struct net *net = sock_net(skb->sk);
1945         rtnl_doit_func doit;
1946         int sz_idx, kind;
1947         int min_len;
1948         int family;
1949         int type;
1950         int err;
1951
1952         type = nlh->nlmsg_type;
1953         if (type > RTM_MAX)
1954                 return -EOPNOTSUPP;
1955
1956         type -= RTM_BASE;
1957
1958         /* All the messages must have at least 1 byte length */
1959         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1960                 return 0;
1961
1962         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1963         sz_idx = type>>2;
1964         kind = type&3;
1965
1966         if (kind != 2 && !capable(CAP_NET_ADMIN))
1967                 return -EPERM;
1968
1969         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1970                 struct sock *rtnl;
1971                 rtnl_dumpit_func dumpit;
1972                 rtnl_calcit_func calcit;
1973                 u16 min_dump_alloc = 0;
1974
1975                 dumpit = rtnl_get_dumpit(family, type);
1976                 if (dumpit == NULL)
1977                         return -EOPNOTSUPP;
1978                 calcit = rtnl_get_calcit(family, type);
1979                 if (calcit)
1980                         min_dump_alloc = calcit(skb);
1981
1982                 __rtnl_unlock();
1983                 rtnl = net->rtnl;
1984                 {
1985                         struct netlink_dump_control c = {
1986                                 .dump           = dumpit,
1987                                 .min_dump_alloc = min_dump_alloc,
1988                         };
1989                         err = netlink_dump_start(rtnl, skb, nlh, &c);
1990                 }
1991                 rtnl_lock();
1992                 return err;
1993         }
1994
1995         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1996
1997         min_len = rtm_min[sz_idx];
1998         if (nlh->nlmsg_len < min_len)
1999                 return -EINVAL;
2000
2001         if (nlh->nlmsg_len > min_len) {
2002                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2003                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2004
2005                 while (RTA_OK(attr, attrlen)) {
2006                         unsigned flavor = attr->rta_type;
2007                         if (flavor) {
2008                                 if (flavor > rta_max[sz_idx])
2009                                         return -EINVAL;
2010                                 rta_buf[flavor-1] = attr;
2011                         }
2012                         attr = RTA_NEXT(attr, attrlen);
2013                 }
2014         }
2015
2016         doit = rtnl_get_doit(family, type);
2017         if (doit == NULL)
2018                 return -EOPNOTSUPP;
2019
2020         return doit(skb, nlh, (void *)&rta_buf[0]);
2021 }
2022
2023 static void rtnetlink_rcv(struct sk_buff *skb)
2024 {
2025         rtnl_lock();
2026         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2027         rtnl_unlock();
2028 }
2029
2030 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2031 {
2032         struct net_device *dev = ptr;
2033
2034         switch (event) {
2035         case NETDEV_UP:
2036         case NETDEV_DOWN:
2037         case NETDEV_PRE_UP:
2038         case NETDEV_POST_INIT:
2039         case NETDEV_REGISTER:
2040         case NETDEV_CHANGE:
2041         case NETDEV_PRE_TYPE_CHANGE:
2042         case NETDEV_GOING_DOWN:
2043         case NETDEV_UNREGISTER:
2044         case NETDEV_UNREGISTER_BATCH:
2045         case NETDEV_RELEASE:
2046         case NETDEV_JOIN:
2047                 break;
2048         default:
2049                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2050                 break;
2051         }
2052         return NOTIFY_DONE;
2053 }
2054
2055 static struct notifier_block rtnetlink_dev_notifier = {
2056         .notifier_call  = rtnetlink_event,
2057 };
2058
2059
2060 static int __net_init rtnetlink_net_init(struct net *net)
2061 {
2062         struct sock *sk;
2063         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2064                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2065         if (!sk)
2066                 return -ENOMEM;
2067         net->rtnl = sk;
2068         return 0;
2069 }
2070
2071 static void __net_exit rtnetlink_net_exit(struct net *net)
2072 {
2073         netlink_kernel_release(net->rtnl);
2074         net->rtnl = NULL;
2075 }
2076
2077 static struct pernet_operations rtnetlink_net_ops = {
2078         .init = rtnetlink_net_init,
2079         .exit = rtnetlink_net_exit,
2080 };
2081
2082 void __init rtnetlink_init(void)
2083 {
2084         int i;
2085
2086         rtattr_max = 0;
2087         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2088                 if (rta_max[i] > rtattr_max)
2089                         rtattr_max = rta_max[i];
2090         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2091         if (!rta_buf)
2092                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2093
2094         if (register_pernet_subsys(&rtnetlink_net_ops))
2095                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2096
2097         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2098         register_netdevice_notifier(&rtnetlink_dev_notifier);
2099
2100         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2101                       rtnl_dump_ifinfo, rtnl_calcit);
2102         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2103         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2104         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2105
2106         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2107         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2108 }
2109