Pileus Git - ~andy/linux/blob - 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
  42 #include <linux/inet.h>
  43 #include <linux/netdevice.h>
  44 #include <net/ip.h>
  45 #include <net/protocol.h>
  46 #include <net/arp.h>
  47 #include <net/route.h>
  48 #include <net/udp.h>
  49 #include <net/sock.h>
  50 #include <net/pkt_sched.h>
  51 #include <net/fib_rules.h>
  52 #include <net/rtnetlink.h>
  53 #include <net/net_namespace.h>
  54
  55 struct rtnl_link {
  56         rtnl_doit_func          doit;
  57         rtnl_dumpit_func        dumpit;
  58         rtnl_calcit_func        calcit;
  59 };
  60
  61 static DEFINE_MUTEX(rtnl_mutex);
  62
  63 void rtnl_lock(void)
  64 {
  65         mutex_lock(&rtnl_mutex);
  66 }
  67 EXPORT_SYMBOL(rtnl_lock);
  68
  69 void __rtnl_unlock(void)
  70 {
  71         mutex_unlock(&rtnl_mutex);
  72 }
  73
  74 void rtnl_unlock(void)
  75 {
  76         /* This fellow will unlock it for us. */
  77         netdev_run_todo();
  78 }
  79 EXPORT_SYMBOL(rtnl_unlock);
  80
  81 int rtnl_trylock(void)
  82 {
  83         return mutex_trylock(&rtnl_mutex);
  84 }
  85 EXPORT_SYMBOL(rtnl_trylock);
  86
  87 int rtnl_is_locked(void)
  88 {
  89         return mutex_is_locked(&rtnl_mutex);
  90 }
  91 EXPORT_SYMBOL(rtnl_is_locked);
  92
  93 #ifdef CONFIG_PROVE_LOCKING
  94 int lockdep_rtnl_is_held(void)
  95 {
  96         return lockdep_is_held(&rtnl_mutex);
  97 }
  98 EXPORT_SYMBOL(lockdep_rtnl_is_held);
  99 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 100
 101 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 102
 103 static inline int rtm_msgindex(int msgtype)
 104 {
 105         int msgindex = msgtype - RTM_BASE;
 106
 107         /*
 108          * msgindex < 0 implies someone tried to register a netlink
 109          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 110          * the message type has not been added to linux/rtnetlink.h
 111          */
 112         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 113
 114         return msgindex;
 115 }
 116
 117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 118 {
 119         struct rtnl_link *tab;
 120
 121         if (protocol <= RTNL_FAMILY_MAX)
 122                 tab = rtnl_msg_handlers[protocol];
 123         else
 124                 tab = NULL;
 125
 126         if (tab == NULL || tab[msgindex].doit == NULL)
 127                 tab = rtnl_msg_handlers[PF_UNSPEC];
 128
 129         return tab ? tab[msgindex].doit : NULL;
 130 }
 131
 132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 133 {
 134         struct rtnl_link *tab;
 135
 136         if (protocol <= RTNL_FAMILY_MAX)
 137                 tab = rtnl_msg_handlers[protocol];
 138         else
 139                 tab = NULL;
 140
 141         if (tab == NULL || tab[msgindex].dumpit == NULL)
 142                 tab = rtnl_msg_handlers[PF_UNSPEC];
 143
 144         return tab ? tab[msgindex].dumpit : NULL;
 145 }
 146
 147 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 148 {
 149         struct rtnl_link *tab;
 150
 151         if (protocol <= RTNL_FAMILY_MAX)
 152                 tab = rtnl_msg_handlers[protocol];
 153         else
 154                 tab = NULL;
 155
 156         if (tab == NULL || tab[msgindex].calcit == NULL)
 157                 tab = rtnl_msg_handlers[PF_UNSPEC];
 158
 159         return tab ? tab[msgindex].calcit : NULL;
 160 }
 161
 162 /**
 163  * __rtnl_register - Register a rtnetlink message type
 164  * @protocol: Protocol family or PF_UNSPEC
 165  * @msgtype: rtnetlink message type
 166  * @doit: Function pointer called for each request message
 167  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 168  * @calcit: Function pointer to calc size of dump message
 169  *
 170  * Registers the specified function pointers (at least one of them has
 171  * to be non-NULL) to be called whenever a request message for the
 172  * specified protocol family and message type is received.
 173  *
 174  * The special protocol family PF_UNSPEC may be used to define fallback
 175  * function pointers for the case when no entry for the specific protocol
 176  * family exists.
 177  *
 178  * Returns 0 on success or a negative error code.
 179  */
 180 int __rtnl_register(int protocol, int msgtype,
 181                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 182                     rtnl_calcit_func calcit)
 183 {
 184         struct rtnl_link *tab;
 185         int msgindex;
 186
 187         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 188         msgindex = rtm_msgindex(msgtype);
 189
 190         tab = rtnl_msg_handlers[protocol];
 191         if (tab == NULL) {
 192                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 193                 if (tab == NULL)
 194                         return -ENOBUFS;
 195
 196                 rtnl_msg_handlers[protocol] = tab;
 197         }
 198
 199         if (doit)
 200                 tab[msgindex].doit = doit;
 201
 202         if (dumpit)
 203                 tab[msgindex].dumpit = dumpit;
 204
 205         if (calcit)
 206                 tab[msgindex].calcit = calcit;
 207
 208         return 0;
 209 }
 210 EXPORT_SYMBOL_GPL(__rtnl_register);
 211
 212 /**
 213  * rtnl_register - Register a rtnetlink message type
 214  *
 215  * Identical to __rtnl_register() but panics on failure. This is useful
 216  * as failure of this function is very unlikely, it can only happen due
 217  * to lack of memory when allocating the chain to store all message
 218  * handlers for a protocol. Meant for use in init functions where lack
 219  * of memory implies no sense in continuing.
 220  */
 221 void rtnl_register(int protocol, int msgtype,
 222                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 223                    rtnl_calcit_func calcit)
 224 {
 225         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 226                 panic("Unable to register rtnetlink message handler, "
 227                       "protocol = %d, message type = %d\n",
 228                       protocol, msgtype);
 229 }
 230 EXPORT_SYMBOL_GPL(rtnl_register);
 231
 232 /**
 233  * rtnl_unregister - Unregister a rtnetlink message type
 234  * @protocol: Protocol family or PF_UNSPEC
 235  * @msgtype: rtnetlink message type
 236  *
 237  * Returns 0 on success or a negative error code.
 238  */
 239 int rtnl_unregister(int protocol, int msgtype)
 240 {
 241         int msgindex;
 242
 243         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 244         msgindex = rtm_msgindex(msgtype);
 245
 246         if (rtnl_msg_handlers[protocol] == NULL)
 247                 return -ENOENT;
 248
 249         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 250         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 251
 252         return 0;
 253 }
 254 EXPORT_SYMBOL_GPL(rtnl_unregister);
 255
 256 /**
 257  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 258  * @protocol : Protocol family or PF_UNSPEC
 259  *
 260  * Identical to calling rtnl_unregster() for all registered message types
 261  * of a certain protocol family.
 262  */
 263 void rtnl_unregister_all(int protocol)
 264 {
 265         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 266
 267         kfree(rtnl_msg_handlers[protocol]);
 268         rtnl_msg_handlers[protocol] = NULL;
 269 }
 270 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 271
 272 static LIST_HEAD(link_ops);
 273
 274 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 275 {
 276         const struct rtnl_link_ops *ops;
 277
 278         list_for_each_entry(ops, &link_ops, list) {
 279                 if (!strcmp(ops->kind, kind))
 280                         return ops;
 281         }
 282         return NULL;
 283 }
 284
 285 /**
 286  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 287  * @ops: struct rtnl_link_ops * to register
 288  *
 289  * The caller must hold the rtnl_mutex. This function should be used
 290  * by drivers that create devices during module initialization. It
 291  * must be called before registering the devices.
 292  *
 293  * Returns 0 on success or a negative error code.
 294  */
 295 int __rtnl_link_register(struct rtnl_link_ops *ops)
 296 {
 297         if (rtnl_link_ops_get(ops->kind))
 298                 return -EEXIST;
 299
 300         if (!ops->dellink)
 301                 ops->dellink = unregister_netdevice_queue;
 302
 303         list_add_tail(&ops->list, &link_ops);
 304         return 0;
 305 }
 306 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 307
 308 /**
 309  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 310  * @ops: struct rtnl_link_ops * to register
 311  *
 312  * Returns 0 on success or a negative error code.
 313  */
 314 int rtnl_link_register(struct rtnl_link_ops *ops)
 315 {
 316         int err;
 317
 318         rtnl_lock();
 319         err = __rtnl_link_register(ops);
 320         rtnl_unlock();
 321         return err;
 322 }
 323 EXPORT_SYMBOL_GPL(rtnl_link_register);
 324
 325 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 326 {
 327         struct net_device *dev;
 328         LIST_HEAD(list_kill);
 329
 330         for_each_netdev(net, dev) {
 331                 if (dev->rtnl_link_ops == ops)
 332                         ops->dellink(dev, &list_kill);
 333         }
 334         unregister_netdevice_many(&list_kill);
 335 }
 336
 337 /**
 338  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 339  * @ops: struct rtnl_link_ops * to unregister
 340  *
 341  * The caller must hold the rtnl_mutex.
 342  */
 343 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 344 {
 345         struct net *net;
 346
 347         for_each_net(net) {
 348                 __rtnl_kill_links(net, ops);
 349         }
 350         list_del(&ops->list);
 351 }
 352 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 353
 354 /**
 355  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 356  * @ops: struct rtnl_link_ops * to unregister
 357  */
 358 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 359 {
 360         rtnl_lock();
 361         __rtnl_link_unregister(ops);
 362         rtnl_unlock();
 363 }
 364 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 365
 366 static size_t rtnl_link_get_size(const struct net_device *dev)
 367 {
 368         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 369         size_t size;
 370
 371         if (!ops)
 372                 return 0;
 373
 374         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 375                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 376
 377         if (ops->get_size)
 378                 /* IFLA_INFO_DATA + nested data */
 379                 size += nla_total_size(sizeof(struct nlattr)) +
 380                         ops->get_size(dev);
 381
 382         if (ops->get_xstats_size)
 383                 /* IFLA_INFO_XSTATS */
 384                 size += nla_total_size(ops->get_xstats_size(dev));
 385
 386         return size;
 387 }
 388
 389 static LIST_HEAD(rtnl_af_ops);
 390
 391 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 392 {
 393         const struct rtnl_af_ops *ops;
 394
 395         list_for_each_entry(ops, &rtnl_af_ops, list) {
 396                 if (ops->family == family)
 397                         return ops;
 398         }
 399
 400         return NULL;
 401 }
 402
 403 /**
 404  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 405  * @ops: struct rtnl_af_ops * to register
 406  *
 407  * The caller must hold the rtnl_mutex.
 408  *
 409  * Returns 0 on success or a negative error code.
 410  */
 411 int __rtnl_af_register(struct rtnl_af_ops *ops)
 412 {
 413         list_add_tail(&ops->list, &rtnl_af_ops);
 414         return 0;
 415 }
 416 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 417
 418 /**
 419  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 420  * @ops: struct rtnl_af_ops * to register
 421  *
 422  * Returns 0 on success or a negative error code.
 423  */
 424 int rtnl_af_register(struct rtnl_af_ops *ops)
 425 {
 426         int err;
 427
 428         rtnl_lock();
 429         err = __rtnl_af_register(ops);
 430         rtnl_unlock();
 431         return err;
 432 }
 433 EXPORT_SYMBOL_GPL(rtnl_af_register);
 434
 435 /**
 436  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 437  * @ops: struct rtnl_af_ops * to unregister
 438  *
 439  * The caller must hold the rtnl_mutex.
 440  */
 441 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 442 {
 443         list_del(&ops->list);
 444 }
 445 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 446
 447 /**
 448  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 449  * @ops: struct rtnl_af_ops * to unregister
 450  */
 451 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 452 {
 453         rtnl_lock();
 454         __rtnl_af_unregister(ops);
 455         rtnl_unlock();
 456 }
 457 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 458
 459 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 460 {
 461         struct rtnl_af_ops *af_ops;
 462         size_t size;
 463
 464         /* IFLA_AF_SPEC */
 465         size = nla_total_size(sizeof(struct nlattr));
 466
 467         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 468                 if (af_ops->get_link_af_size) {
 469                         /* AF_* + nested data */
 470                         size += nla_total_size(sizeof(struct nlattr)) +
 471                                 af_ops->get_link_af_size(dev);
 472                 }
 473         }
 474
 475         return size;
 476 }
 477
 478 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 479 {
 480         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 481         struct nlattr *linkinfo, *data;
 482         int err = -EMSGSIZE;
 483
 484         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 485         if (linkinfo == NULL)
 486                 goto out;
 487
 488         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 489                 goto err_cancel_link;
 490         if (ops->fill_xstats) {
 491                 err = ops->fill_xstats(skb, dev);
 492                 if (err < 0)
 493                         goto err_cancel_link;
 494         }
 495         if (ops->fill_info) {
 496                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 497                 if (data == NULL)
 498                         goto err_cancel_link;
 499                 err = ops->fill_info(skb, dev);
 500                 if (err < 0)
 501                         goto err_cancel_data;
 502                 nla_nest_end(skb, data);
 503         }
 504
 505         nla_nest_end(skb, linkinfo);
 506         return 0;
 507
 508 err_cancel_data:
 509         nla_nest_cancel(skb, data);
 510 err_cancel_link:
 511         nla_nest_cancel(skb, linkinfo);
 512 out:
 513         return err;
 514 }
 515
 516 static const int rtm_min[RTM_NR_FAMILIES] =
 517 {
 518         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 519         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 520         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 521         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 522         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 523         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 524         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 526         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 527         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 528 };
 529
 530 static const int rta_max[RTM_NR_FAMILIES] =
 531 {
 532         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 533         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 534         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 535         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 536         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 537         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 538         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 539         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 540 };
 541
 542 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 543 {
 544         struct rtattr *rta;
 545         int size = RTA_LENGTH(attrlen);
 546
 547         rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 548         rta->rta_type = attrtype;
 549         rta->rta_len = size;
 550         memcpy(RTA_DATA(rta), data, attrlen);
 551         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 552 }
 553 EXPORT_SYMBOL(__rta_fill);
 554
 555 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 556 {
 557         struct sock *rtnl = net->rtnl;
 558         int err = 0;
 559
 560         NETLINK_CB(skb).dst_group = group;
 561         if (echo)
 562                 atomic_inc(&skb->users);
 563         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 564         if (echo)
 565                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 566         return err;
 567 }
 568
 569 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 570 {
 571         struct sock *rtnl = net->rtnl;
 572
 573         return nlmsg_unicast(rtnl, skb, pid);
 574 }
 575 EXPORT_SYMBOL(rtnl_unicast);
 576
 577 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 578                  struct nlmsghdr *nlh, gfp_t flags)
 579 {
 580         struct sock *rtnl = net->rtnl;
 581         int report = 0;
 582
 583         if (nlh)
 584                 report = nlmsg_report(nlh);
 585
 586         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 587 }
 588 EXPORT_SYMBOL(rtnl_notify);
 589
 590 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 591 {
 592         struct sock *rtnl = net->rtnl;
 593
 594         netlink_set_err(rtnl, 0, group, error);
 595 }
 596 EXPORT_SYMBOL(rtnl_set_sk_err);
 597
 598 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 599 {
 600         struct nlattr *mx;
 601         int i, valid = 0;
 602
 603         mx = nla_nest_start(skb, RTA_METRICS);
 604         if (mx == NULL)
 605                 return -ENOBUFS;
 606
 607         for (i = 0; i < RTAX_MAX; i++) {
 608                 if (metrics[i]) {
 609                         valid++;
 610                         if (nla_put_u32(skb, i+1, metrics[i]))
 611                                 goto nla_put_failure;
 612                 }
 613         }
 614
 615         if (!valid) {
 616                 nla_nest_cancel(skb, mx);
 617                 return 0;
 618         }
 619
 620         return nla_nest_end(skb, mx);
 621
 622 nla_put_failure:
 623         nla_nest_cancel(skb, mx);
 624         return -EMSGSIZE;
 625 }
 626 EXPORT_SYMBOL(rtnetlink_put_metrics);
 627
 628 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 629                        u32 ts, u32 tsage, long expires, u32 error)
 630 {
 631         struct rta_cacheinfo ci = {
 632                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 633                 .rta_used = dst->__use,
 634                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 635                 .rta_error = error,
 636                 .rta_id =  id,
 637                 .rta_ts = ts,
 638                 .rta_tsage = tsage,
 639         };
 640
 641         if (expires)
 642                 ci.rta_expires = jiffies_to_clock_t(expires);
 643
 644         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 645 }
 646 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 647
 648 static void set_operstate(struct net_device *dev, unsigned char transition)
 649 {
 650         unsigned char operstate = dev->operstate;
 651
 652         switch (transition) {
 653         case IF_OPER_UP:
 654                 if ((operstate == IF_OPER_DORMANT ||
 655                      operstate == IF_OPER_UNKNOWN) &&
 656                     !netif_dormant(dev))
 657                         operstate = IF_OPER_UP;
 658                 break;
 659
 660         case IF_OPER_DORMANT:
 661                 if (operstate == IF_OPER_UP ||
 662                     operstate == IF_OPER_UNKNOWN)
 663                         operstate = IF_OPER_DORMANT;
 664                 break;
 665         }
 666
 667         if (dev->operstate != operstate) {
 668                 write_lock_bh(&dev_base_lock);
 669                 dev->operstate = operstate;
 670                 write_unlock_bh(&dev_base_lock);
 671                 netdev_state_change(dev);
 672         }
 673 }
 674
 675 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 676                                            const struct ifinfomsg *ifm)
 677 {
 678         unsigned int flags = ifm->ifi_flags;
 679
 680         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 681         if (ifm->ifi_change)
 682                 flags = (flags & ifm->ifi_change) |
 683                         (dev->flags & ~ifm->ifi_change);
 684
 685         return flags;
 686 }
 687
 688 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 689                                  const struct rtnl_link_stats64 *b)
 690 {
 691         a->rx_packets = b->rx_packets;
 692         a->tx_packets = b->tx_packets;
 693         a->rx_bytes = b->rx_bytes;
 694         a->tx_bytes = b->tx_bytes;
 695         a->rx_errors = b->rx_errors;
 696         a->tx_errors = b->tx_errors;
 697         a->rx_dropped = b->rx_dropped;
 698         a->tx_dropped = b->tx_dropped;
 699
 700         a->multicast = b->multicast;
 701         a->collisions = b->collisions;
 702
 703         a->rx_length_errors = b->rx_length_errors;
 704         a->rx_over_errors = b->rx_over_errors;
 705         a->rx_crc_errors = b->rx_crc_errors;
 706         a->rx_frame_errors = b->rx_frame_errors;
 707         a->rx_fifo_errors = b->rx_fifo_errors;
 708         a->rx_missed_errors = b->rx_missed_errors;
 709
 710         a->tx_aborted_errors = b->tx_aborted_errors;
 711         a->tx_carrier_errors = b->tx_carrier_errors;
 712         a->tx_fifo_errors = b->tx_fifo_errors;
 713         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 714         a->tx_window_errors = b->tx_window_errors;
 715
 716         a->rx_compressed = b->rx_compressed;
 717         a->tx_compressed = b->tx_compressed;
 718 }
 719
 720 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 721 {
 722         memcpy(v, b, sizeof(*b));
 723 }
 724
 725 /* All VF info */
 726 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 727                                    u32 ext_filter_mask)
 728 {
 729         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 730             (ext_filter_mask & RTEXT_FILTER_VF)) {
 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                                      u32 ext_filter_mask)
 771 {
 772         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 773                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 774                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 775                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 776                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 777                + nla_total_size(sizeof(struct rtnl_link_stats))
 778                + nla_total_size(sizeof(struct rtnl_link_stats64))
 779                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 780                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 781                + nla_total_size(4) /* IFLA_TXQLEN */
 782                + nla_total_size(4) /* IFLA_WEIGHT */
 783                + nla_total_size(4) /* IFLA_MTU */
 784                + nla_total_size(4) /* IFLA_LINK */
 785                + nla_total_size(4) /* IFLA_MASTER */
 786                + nla_total_size(4) /* IFLA_PROMISCUITY */
 787                + nla_total_size(1) /* IFLA_OPERSTATE */
 788                + nla_total_size(1) /* IFLA_LINKMODE */
 789                + nla_total_size(ext_filter_mask
 790                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 791                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 792                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 793                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 794                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 795 }
 796
 797 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 798 {
 799         struct nlattr *vf_ports;
 800         struct nlattr *vf_port;
 801         int vf;
 802         int err;
 803
 804         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 805         if (!vf_ports)
 806                 return -EMSGSIZE;
 807
 808         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 809                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 810                 if (!vf_port)
 811                         goto nla_put_failure;
 812                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 813                         goto nla_put_failure;
 814                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 815                 if (err == -EMSGSIZE)
 816                         goto nla_put_failure;
 817                 if (err) {
 818                         nla_nest_cancel(skb, vf_port);
 819                         continue;
 820                 }
 821                 nla_nest_end(skb, vf_port);
 822         }
 823
 824         nla_nest_end(skb, vf_ports);
 825
 826         return 0;
 827
 828 nla_put_failure:
 829         nla_nest_cancel(skb, vf_ports);
 830         return -EMSGSIZE;
 831 }
 832
 833 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 834 {
 835         struct nlattr *port_self;
 836         int err;
 837
 838         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 839         if (!port_self)
 840                 return -EMSGSIZE;
 841
 842         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 843         if (err) {
 844                 nla_nest_cancel(skb, port_self);
 845                 return (err == -EMSGSIZE) ? err : 0;
 846         }
 847
 848         nla_nest_end(skb, port_self);
 849
 850         return 0;
 851 }
 852
 853 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 854 {
 855         int err;
 856
 857         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 858                 return 0;
 859
 860         err = rtnl_port_self_fill(skb, dev);
 861         if (err)
 862                 return err;
 863
 864         if (dev_num_vf(dev->dev.parent)) {
 865                 err = rtnl_vf_ports_fill(skb, dev);
 866                 if (err)
 867                         return err;
 868         }
 869
 870         return 0;
 871 }
 872
 873 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 874                             int type, u32 pid, u32 seq, u32 change,
 875                             unsigned int flags, u32 ext_filter_mask)
 876 {
 877         struct ifinfomsg *ifm;
 878         struct nlmsghdr *nlh;
 879         struct rtnl_link_stats64 temp;
 880         const struct rtnl_link_stats64 *stats;
 881         struct nlattr *attr, *af_spec;
 882         struct rtnl_af_ops *af_ops;
 883
 884         ASSERT_RTNL();
 885         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 886         if (nlh == NULL)
 887                 return -EMSGSIZE;
 888
 889         ifm = nlmsg_data(nlh);
 890         ifm->ifi_family = AF_UNSPEC;
 891         ifm->__ifi_pad = 0;
 892         ifm->ifi_type = dev->type;
 893         ifm->ifi_index = dev->ifindex;
 894         ifm->ifi_flags = dev_get_flags(dev);
 895         ifm->ifi_change = change;
 896
 897         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 898             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 899             nla_put_u8(skb, IFLA_OPERSTATE,
 900                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 901             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 902             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 903             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 904             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 905             (dev->ifindex != dev->iflink &&
 906              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 907             (dev->master &&
 908              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 909             (dev->qdisc &&
 910              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 911             (dev->ifalias &&
 912              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 913                 goto nla_put_failure;
 914
 915         if (1) {
 916                 struct rtnl_link_ifmap map = {
 917                         .mem_start   = dev->mem_start,
 918                         .mem_end     = dev->mem_end,
 919                         .base_addr   = dev->base_addr,
 920                         .irq         = dev->irq,
 921                         .dma         = dev->dma,
 922                         .port        = dev->if_port,
 923                 };
 924                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 925                         goto nla_put_failure;
 926         }
 927
 928         if (dev->addr_len) {
 929                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 930                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 931                         goto nla_put_failure;
 932         }
 933
 934         attr = nla_reserve(skb, IFLA_STATS,
 935                         sizeof(struct rtnl_link_stats));
 936         if (attr == NULL)
 937                 goto nla_put_failure;
 938
 939         stats = dev_get_stats(dev, &temp);
 940         copy_rtnl_link_stats(nla_data(attr), stats);
 941
 942         attr = nla_reserve(skb, IFLA_STATS64,
 943                         sizeof(struct rtnl_link_stats64));
 944         if (attr == NULL)
 945                 goto nla_put_failure;
 946         copy_rtnl_link_stats64(nla_data(attr), stats);
 947
 948         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 949             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 950                 goto nla_put_failure;
 951
 952         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 953             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 954                 int i;
 955
 956                 struct nlattr *vfinfo, *vf;
 957                 int num_vfs = dev_num_vf(dev->dev.parent);
 958
 959                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 960                 if (!vfinfo)
 961                         goto nla_put_failure;
 962                 for (i = 0; i < num_vfs; i++) {
 963                         struct ifla_vf_info ivi;
 964                         struct ifla_vf_mac vf_mac;
 965                         struct ifla_vf_vlan vf_vlan;
 966                         struct ifla_vf_tx_rate vf_tx_rate;
 967                         struct ifla_vf_spoofchk vf_spoofchk;
 968
 969                         /*
 970                          * Not all SR-IOV capable drivers support the
 971                          * spoofcheck query.  Preset to -1 so the user
 972                          * space tool can detect that the driver didn't
 973                          * report anything.
 974                          */
 975                         ivi.spoofchk = -1;
 976                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 977                                 break;
 978                         vf_mac.vf =
 979                                 vf_vlan.vf =
 980                                 vf_tx_rate.vf =
 981                                 vf_spoofchk.vf = ivi.vf;
 982
 983                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 984                         vf_vlan.vlan = ivi.vlan;
 985                         vf_vlan.qos = ivi.qos;
 986                         vf_tx_rate.rate = ivi.tx_rate;
 987                         vf_spoofchk.setting = ivi.spoofchk;
 988                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 989                         if (!vf) {
 990                                 nla_nest_cancel(skb, vfinfo);
 991                                 goto nla_put_failure;
 992                         }
 993                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 994                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 995                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 996                                     &vf_tx_rate) ||
 997                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
 998                                     &vf_spoofchk))
 999                                 goto nla_put_failure;
1000                         nla_nest_end(skb, vf);
1001                 }
1002                 nla_nest_end(skb, vfinfo);
1003         }
1004
1005         if (rtnl_port_fill(skb, dev))
1006                 goto nla_put_failure;
1007
1008         if (dev->rtnl_link_ops) {
1009                 if (rtnl_link_fill(skb, dev) < 0)
1010                         goto nla_put_failure;
1011         }
1012
1013         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1014                 goto nla_put_failure;
1015
1016         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1017                 if (af_ops->fill_link_af) {
1018                         struct nlattr *af;
1019                         int err;
1020
1021                         if (!(af = nla_nest_start(skb, af_ops->family)))
1022                                 goto nla_put_failure;
1023
1024                         err = af_ops->fill_link_af(skb, dev);
1025
1026                         /*
1027                          * Caller may return ENODATA to indicate that there
1028                          * was no data to be dumped. This is not an error, it
1029                          * means we should trim the attribute header and
1030                          * continue.
1031                          */
1032                         if (err == -ENODATA)
1033                                 nla_nest_cancel(skb, af);
1034                         else if (err < 0)
1035                                 goto nla_put_failure;
1036
1037                         nla_nest_end(skb, af);
1038                 }
1039         }
1040
1041         nla_nest_end(skb, af_spec);
1042
1043         return nlmsg_end(skb, nlh);
1044
1045 nla_put_failure:
1046         nlmsg_cancel(skb, nlh);
1047         return -EMSGSIZE;
1048 }
1049
1050 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1051 {
1052         struct net *net = sock_net(skb->sk);
1053         int h, s_h;
1054         int idx = 0, s_idx;
1055         struct net_device *dev;
1056         struct hlist_head *head;
1057         struct hlist_node *node;
1058         struct nlattr *tb[IFLA_MAX+1];
1059         u32 ext_filter_mask = 0;
1060
1061         s_h = cb->args[0];
1062         s_idx = cb->args[1];
1063
1064         rcu_read_lock();
1065         cb->seq = net->dev_base_seq;
1066
1067         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1068                         ifla_policy) >= 0) {
1069
1070                 if (tb[IFLA_EXT_MASK])
1071                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1072         }
1073
1074         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1075                 idx = 0;
1076                 head = &net->dev_index_head[h];
1077                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1078                         if (idx < s_idx)
1079                                 goto cont;
1080                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1081                                              NETLINK_CB(cb->skb).pid,
1082                                              cb->nlh->nlmsg_seq, 0,
1083                                              NLM_F_MULTI,
1084                                              ext_filter_mask) <= 0)
1085                                 goto out;
1086
1087                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1088 cont:
1089                         idx++;
1090                 }
1091         }
1092 out:
1093         rcu_read_unlock();
1094         cb->args[1] = idx;
1095         cb->args[0] = h;
1096
1097         return skb->len;
1098 }
1099
1100 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1101         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1102         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1103         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1104         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1105         [IFLA_MTU]              = { .type = NLA_U32 },
1106         [IFLA_LINK]             = { .type = NLA_U32 },
1107         [IFLA_MASTER]           = { .type = NLA_U32 },
1108         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1109         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1110         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1111         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1112         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1113         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1114         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1115         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1116         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1117         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1118         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1119         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1120         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1121         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1122 };
1123 EXPORT_SYMBOL(ifla_policy);
1124
1125 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1126         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1127         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1128 };
1129
1130 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1131         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1132 };
1133
1134 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1135         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1136                                     .len = sizeof(struct ifla_vf_mac) },
1137         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1138                                     .len = sizeof(struct ifla_vf_vlan) },
1139         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1140                                     .len = sizeof(struct ifla_vf_tx_rate) },
1141         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1142                                     .len = sizeof(struct ifla_vf_spoofchk) },
1143 };
1144
1145 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1146         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1147         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1148                                     .len = PORT_PROFILE_MAX },
1149         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1150                                     .len = sizeof(struct ifla_port_vsi)},
1151         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1152                                       .len = PORT_UUID_MAX },
1153         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1154                                     .len = PORT_UUID_MAX },
1155         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1156         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1157 };
1158
1159 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1160 {
1161         struct net *net;
1162         /* Examine the link attributes and figure out which
1163          * network namespace we are talking about.
1164          */
1165         if (tb[IFLA_NET_NS_PID])
1166                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1167         else if (tb[IFLA_NET_NS_FD])
1168                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1169         else
1170                 net = get_net(src_net);
1171         return net;
1172 }
1173 EXPORT_SYMBOL(rtnl_link_get_net);
1174
1175 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1176 {
1177         if (dev) {
1178                 if (tb[IFLA_ADDRESS] &&
1179                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1180                         return -EINVAL;
1181
1182                 if (tb[IFLA_BROADCAST] &&
1183                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1184                         return -EINVAL;
1185         }
1186
1187         if (tb[IFLA_AF_SPEC]) {
1188                 struct nlattr *af;
1189                 int rem, err;
1190
1191                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1192                         const struct rtnl_af_ops *af_ops;
1193
1194                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1195                                 return -EAFNOSUPPORT;
1196
1197                         if (!af_ops->set_link_af)
1198                                 return -EOPNOTSUPP;
1199
1200                         if (af_ops->validate_link_af) {
1201                                 err = af_ops->validate_link_af(dev, af);
1202                                 if (err < 0)
1203                                         return err;
1204                         }
1205                 }
1206         }
1207
1208         return 0;
1209 }
1210
1211 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1212 {
1213         int rem, err = -EINVAL;
1214         struct nlattr *vf;
1215         const struct net_device_ops *ops = dev->netdev_ops;
1216
1217         nla_for_each_nested(vf, attr, rem) {
1218                 switch (nla_type(vf)) {
1219                 case IFLA_VF_MAC: {
1220                         struct ifla_vf_mac *ivm;
1221                         ivm = nla_data(vf);
1222                         err = -EOPNOTSUPP;
1223                         if (ops->ndo_set_vf_mac)
1224                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1225                                                           ivm->mac);
1226                         break;
1227                 }
1228                 case IFLA_VF_VLAN: {
1229                         struct ifla_vf_vlan *ivv;
1230                         ivv = nla_data(vf);
1231                         err = -EOPNOTSUPP;
1232                         if (ops->ndo_set_vf_vlan)
1233                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1234                                                            ivv->vlan,
1235                                                            ivv->qos);
1236                         break;
1237                 }
1238                 case IFLA_VF_TX_RATE: {
1239                         struct ifla_vf_tx_rate *ivt;
1240                         ivt = nla_data(vf);
1241                         err = -EOPNOTSUPP;
1242                         if (ops->ndo_set_vf_tx_rate)
1243                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1244                                                               ivt->rate);
1245                         break;
1246                 }
1247                 case IFLA_VF_SPOOFCHK: {
1248                         struct ifla_vf_spoofchk *ivs;
1249                         ivs = nla_data(vf);
1250                         err = -EOPNOTSUPP;
1251                         if (ops->ndo_set_vf_spoofchk)
1252                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1253                                                                ivs->setting);
1254                         break;
1255                 }
1256                 default:
1257                         err = -EINVAL;
1258                         break;
1259                 }
1260                 if (err)
1261                         break;
1262         }
1263         return err;
1264 }
1265
1266 static int do_set_master(struct net_device *dev, int ifindex)
1267 {
1268         struct net_device *master_dev;
1269         const struct net_device_ops *ops;
1270         int err;
1271
1272         if (dev->master) {
1273                 if (dev->master->ifindex == ifindex)
1274                         return 0;
1275                 ops = dev->master->netdev_ops;
1276                 if (ops->ndo_del_slave) {
1277                         err = ops->ndo_del_slave(dev->master, dev);
1278                         if (err)
1279                                 return err;
1280                 } else {
1281                         return -EOPNOTSUPP;
1282                 }
1283         }
1284
1285         if (ifindex) {
1286                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1287                 if (!master_dev)
1288                         return -EINVAL;
1289                 ops = master_dev->netdev_ops;
1290                 if (ops->ndo_add_slave) {
1291                         err = ops->ndo_add_slave(master_dev, dev);
1292                         if (err)
1293                                 return err;
1294                 } else {
1295                         return -EOPNOTSUPP;
1296                 }
1297         }
1298         return 0;
1299 }
1300
1301 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1302                       struct nlattr **tb, char *ifname, int modified)
1303 {
1304         const struct net_device_ops *ops = dev->netdev_ops;
1305         int send_addr_notify = 0;
1306         int err;
1307
1308         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1309                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1310                 if (IS_ERR(net)) {
1311                         err = PTR_ERR(net);
1312                         goto errout;
1313                 }
1314                 err = dev_change_net_namespace(dev, net, ifname);
1315                 put_net(net);
1316                 if (err)
1317                         goto errout;
1318                 modified = 1;
1319         }
1320
1321         if (tb[IFLA_MAP]) {
1322                 struct rtnl_link_ifmap *u_map;
1323                 struct ifmap k_map;
1324
1325                 if (!ops->ndo_set_config) {
1326                         err = -EOPNOTSUPP;
1327                         goto errout;
1328                 }
1329
1330                 if (!netif_device_present(dev)) {
1331                         err = -ENODEV;
1332                         goto errout;
1333                 }
1334
1335                 u_map = nla_data(tb[IFLA_MAP]);
1336                 k_map.mem_start = (unsigned long) u_map->mem_start;
1337                 k_map.mem_end = (unsigned long) u_map->mem_end;
1338                 k_map.base_addr = (unsigned short) u_map->base_addr;
1339                 k_map.irq = (unsigned char) u_map->irq;
1340                 k_map.dma = (unsigned char) u_map->dma;
1341                 k_map.port = (unsigned char) u_map->port;
1342
1343                 err = ops->ndo_set_config(dev, &k_map);
1344                 if (err < 0)
1345                         goto errout;
1346
1347                 modified = 1;
1348         }
1349
1350         if (tb[IFLA_ADDRESS]) {
1351                 struct sockaddr *sa;
1352                 int len;
1353
1354                 if (!ops->ndo_set_mac_address) {
1355                         err = -EOPNOTSUPP;
1356                         goto errout;
1357                 }
1358
1359                 if (!netif_device_present(dev)) {
1360                         err = -ENODEV;
1361                         goto errout;
1362                 }
1363
1364                 len = sizeof(sa_family_t) + dev->addr_len;
1365                 sa = kmalloc(len, GFP_KERNEL);
1366                 if (!sa) {
1367                         err = -ENOMEM;
1368                         goto errout;
1369                 }
1370                 sa->sa_family = dev->type;
1371                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1372                        dev->addr_len);
1373                 err = ops->ndo_set_mac_address(dev, sa);
1374                 kfree(sa);
1375                 if (err)
1376                         goto errout;
1377                 send_addr_notify = 1;
1378                 modified = 1;
1379         }
1380
1381         if (tb[IFLA_MTU]) {
1382                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1383                 if (err < 0)
1384                         goto errout;
1385                 modified = 1;
1386         }
1387
1388         if (tb[IFLA_GROUP]) {
1389                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1390                 modified = 1;
1391         }
1392
1393         /*
1394          * Interface selected by interface index but interface
1395          * name provided implies that a name change has been
1396          * requested.
1397          */
1398         if (ifm->ifi_index > 0 && ifname[0]) {
1399                 err = dev_change_name(dev, ifname);
1400                 if (err < 0)
1401                         goto errout;
1402                 modified = 1;
1403         }
1404
1405         if (tb[IFLA_IFALIAS]) {
1406                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1407                                     nla_len(tb[IFLA_IFALIAS]));
1408                 if (err < 0)
1409                         goto errout;
1410                 modified = 1;
1411         }
1412
1413         if (tb[IFLA_BROADCAST]) {
1414                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1415                 send_addr_notify = 1;
1416         }
1417
1418         if (ifm->ifi_flags || ifm->ifi_change) {
1419                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1420                 if (err < 0)
1421                         goto errout;
1422         }
1423
1424         if (tb[IFLA_MASTER]) {
1425                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1426                 if (err)
1427                         goto errout;
1428                 modified = 1;
1429         }
1430
1431         if (tb[IFLA_TXQLEN])
1432                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1433
1434         if (tb[IFLA_OPERSTATE])
1435                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1436
1437         if (tb[IFLA_LINKMODE]) {
1438                 write_lock_bh(&dev_base_lock);
1439                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1440                 write_unlock_bh(&dev_base_lock);
1441         }
1442
1443         if (tb[IFLA_VFINFO_LIST]) {
1444                 struct nlattr *attr;
1445                 int rem;
1446                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1447                         if (nla_type(attr) != IFLA_VF_INFO) {
1448                                 err = -EINVAL;
1449                                 goto errout;
1450                         }
1451                         err = do_setvfinfo(dev, attr);
1452                         if (err < 0)
1453                                 goto errout;
1454                         modified = 1;
1455                 }
1456         }
1457         err = 0;
1458
1459         if (tb[IFLA_VF_PORTS]) {
1460                 struct nlattr *port[IFLA_PORT_MAX+1];
1461                 struct nlattr *attr;
1462                 int vf;
1463                 int rem;
1464
1465                 err = -EOPNOTSUPP;
1466                 if (!ops->ndo_set_vf_port)
1467                         goto errout;
1468
1469                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1470                         if (nla_type(attr) != IFLA_VF_PORT)
1471                                 continue;
1472                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1473                                 attr, ifla_port_policy);
1474                         if (err < 0)
1475                                 goto errout;
1476                         if (!port[IFLA_PORT_VF]) {
1477                                 err = -EOPNOTSUPP;
1478                                 goto errout;
1479                         }
1480                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1481                         err = ops->ndo_set_vf_port(dev, vf, port);
1482                         if (err < 0)
1483                                 goto errout;
1484                         modified = 1;
1485                 }
1486         }
1487         err = 0;
1488
1489         if (tb[IFLA_PORT_SELF]) {
1490                 struct nlattr *port[IFLA_PORT_MAX+1];
1491
1492                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1493                         tb[IFLA_PORT_SELF], ifla_port_policy);
1494                 if (err < 0)
1495                         goto errout;
1496
1497                 err = -EOPNOTSUPP;
1498                 if (ops->ndo_set_vf_port)
1499                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1500                 if (err < 0)
1501                         goto errout;
1502                 modified = 1;
1503         }
1504
1505         if (tb[IFLA_AF_SPEC]) {
1506                 struct nlattr *af;
1507                 int rem;
1508
1509                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1510                         const struct rtnl_af_ops *af_ops;
1511
1512                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1513                                 BUG();
1514
1515                         err = af_ops->set_link_af(dev, af);
1516                         if (err < 0)
1517                                 goto errout;
1518
1519                         modified = 1;
1520                 }
1521         }
1522         err = 0;
1523
1524 errout:
1525         if (err < 0 && modified && net_ratelimit())
1526                 printk(KERN_WARNING "A link change request failed with "
1527                        "some changes committed already. Interface %s may "
1528                        "have been left with an inconsistent configuration, "
1529                        "please check.\n", dev->name);
1530
1531         if (send_addr_notify)
1532                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1533
1534         return err;
1535 }
1536
1537 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1538 {
1539         struct net *net = sock_net(skb->sk);
1540         struct ifinfomsg *ifm;
1541         struct net_device *dev;
1542         int err;
1543         struct nlattr *tb[IFLA_MAX+1];
1544         char ifname[IFNAMSIZ];
1545
1546         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1547         if (err < 0)
1548                 goto errout;
1549
1550         if (tb[IFLA_IFNAME])
1551                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1552         else
1553                 ifname[0] = '\0';
1554
1555         err = -EINVAL;
1556         ifm = nlmsg_data(nlh);
1557         if (ifm->ifi_index > 0)
1558                 dev = __dev_get_by_index(net, ifm->ifi_index);
1559         else if (tb[IFLA_IFNAME])
1560                 dev = __dev_get_by_name(net, ifname);
1561         else
1562                 goto errout;
1563
1564         if (dev == NULL) {
1565                 err = -ENODEV;
1566                 goto errout;
1567         }
1568
1569         err = validate_linkmsg(dev, tb);
1570         if (err < 0)
1571                 goto errout;
1572
1573         err = do_setlink(dev, ifm, tb, ifname, 0);
1574 errout:
1575         return err;
1576 }
1577
1578 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1579 {
1580         struct net *net = sock_net(skb->sk);
1581         const struct rtnl_link_ops *ops;
1582         struct net_device *dev;
1583         struct ifinfomsg *ifm;
1584         char ifname[IFNAMSIZ];
1585         struct nlattr *tb[IFLA_MAX+1];
1586         int err;
1587         LIST_HEAD(list_kill);
1588
1589         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1590         if (err < 0)
1591                 return err;
1592
1593         if (tb[IFLA_IFNAME])
1594                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1595
1596         ifm = nlmsg_data(nlh);
1597         if (ifm->ifi_index > 0)
1598                 dev = __dev_get_by_index(net, ifm->ifi_index);
1599         else if (tb[IFLA_IFNAME])
1600                 dev = __dev_get_by_name(net, ifname);
1601         else
1602                 return -EINVAL;
1603
1604         if (!dev)
1605                 return -ENODEV;
1606
1607         ops = dev->rtnl_link_ops;
1608         if (!ops)
1609                 return -EOPNOTSUPP;
1610
1611         ops->dellink(dev, &list_kill);
1612         unregister_netdevice_many(&list_kill);
1613         list_del(&list_kill);
1614         return 0;
1615 }
1616
1617 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1618 {
1619         unsigned int old_flags;
1620         int err;
1621
1622         old_flags = dev->flags;
1623         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1624                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1625                 if (err < 0)
1626                         return err;
1627         }
1628
1629         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1630         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1631
1632         __dev_notify_flags(dev, old_flags);
1633         return 0;
1634 }
1635 EXPORT_SYMBOL(rtnl_configure_link);
1636
1637 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1638         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1639 {
1640         int err;
1641         struct net_device *dev;
1642         unsigned int num_queues = 1;
1643         unsigned int real_num_queues = 1;
1644
1645         if (ops->get_tx_queues) {
1646                 err = ops->get_tx_queues(src_net, tb, &num_queues,
1647                                          &real_num_queues);
1648                 if (err)
1649                         goto err;
1650         }
1651         err = -ENOMEM;
1652         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1653         if (!dev)
1654                 goto err;
1655
1656         dev_net_set(dev, net);
1657         dev->rtnl_link_ops = ops;
1658         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1659
1660         if (tb[IFLA_MTU])
1661                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1662         if (tb[IFLA_ADDRESS])
1663                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1664                                 nla_len(tb[IFLA_ADDRESS]));
1665         if (tb[IFLA_BROADCAST])
1666                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1667                                 nla_len(tb[IFLA_BROADCAST]));
1668         if (tb[IFLA_TXQLEN])
1669                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1670         if (tb[IFLA_OPERSTATE])
1671                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1672         if (tb[IFLA_LINKMODE])
1673                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1674         if (tb[IFLA_GROUP])
1675                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1676
1677         return dev;
1678
1679 err:
1680         return ERR_PTR(err);
1681 }
1682 EXPORT_SYMBOL(rtnl_create_link);
1683
1684 static int rtnl_group_changelink(struct net *net, int group,
1685                 struct ifinfomsg *ifm,
1686                 struct nlattr **tb)
1687 {
1688         struct net_device *dev;
1689         int err;
1690
1691         for_each_netdev(net, dev) {
1692                 if (dev->group == group) {
1693                         err = do_setlink(dev, ifm, tb, NULL, 0);
1694                         if (err < 0)
1695                                 return err;
1696                 }
1697         }
1698
1699         return 0;
1700 }
1701
1702 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1703 {
1704         struct net *net = sock_net(skb->sk);
1705         const struct rtnl_link_ops *ops;
1706         struct net_device *dev;
1707         struct ifinfomsg *ifm;
1708         char kind[MODULE_NAME_LEN];
1709         char ifname[IFNAMSIZ];
1710         struct nlattr *tb[IFLA_MAX+1];
1711         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1712         int err;
1713
1714 #ifdef CONFIG_MODULES
1715 replay:
1716 #endif
1717         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1718         if (err < 0)
1719                 return err;
1720
1721         if (tb[IFLA_IFNAME])
1722                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1723         else
1724                 ifname[0] = '\0';
1725
1726         ifm = nlmsg_data(nlh);
1727         if (ifm->ifi_index > 0)
1728                 dev = __dev_get_by_index(net, ifm->ifi_index);
1729         else {
1730                 if (ifname[0])
1731                         dev = __dev_get_by_name(net, ifname);
1732                 else
1733                         dev = NULL;
1734         }
1735
1736         err = validate_linkmsg(dev, tb);
1737         if (err < 0)
1738                 return err;
1739
1740         if (tb[IFLA_LINKINFO]) {
1741                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1742                                        tb[IFLA_LINKINFO], ifla_info_policy);
1743                 if (err < 0)
1744                         return err;
1745         } else
1746                 memset(linkinfo, 0, sizeof(linkinfo));
1747
1748         if (linkinfo[IFLA_INFO_KIND]) {
1749                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1750                 ops = rtnl_link_ops_get(kind);
1751         } else {
1752                 kind[0] = '\0';
1753                 ops = NULL;
1754         }
1755
1756         if (1) {
1757                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1758                 struct net *dest_net;
1759
1760                 if (ops) {
1761                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1762                                 err = nla_parse_nested(attr, ops->maxtype,
1763                                                        linkinfo[IFLA_INFO_DATA],
1764                                                        ops->policy);
1765                                 if (err < 0)
1766                                         return err;
1767                                 data = attr;
1768                         }
1769                         if (ops->validate) {
1770                                 err = ops->validate(tb, data);
1771                                 if (err < 0)
1772                                         return err;
1773                         }
1774                 }
1775
1776                 if (dev) {
1777                         int modified = 0;
1778
1779                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1780                                 return -EEXIST;
1781                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1782                                 return -EOPNOTSUPP;
1783
1784                         if (linkinfo[IFLA_INFO_DATA]) {
1785                                 if (!ops || ops != dev->rtnl_link_ops ||
1786                                     !ops->changelink)
1787                                         return -EOPNOTSUPP;
1788
1789                                 err = ops->changelink(dev, tb, data);
1790                                 if (err < 0)
1791                                         return err;
1792                                 modified = 1;
1793                         }
1794
1795                         return do_setlink(dev, ifm, tb, ifname, modified);
1796                 }
1797
1798                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1799                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1800                                 return rtnl_group_changelink(net,
1801                                                 nla_get_u32(tb[IFLA_GROUP]),
1802                                                 ifm, tb);
1803                         return -ENODEV;
1804                 }
1805
1806                 if (ifm->ifi_index)
1807                         return -EOPNOTSUPP;
1808                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1809                         return -EOPNOTSUPP;
1810
1811                 if (!ops) {
1812 #ifdef CONFIG_MODULES
1813                         if (kind[0]) {
1814                                 __rtnl_unlock();
1815                                 request_module("rtnl-link-%s", kind);
1816                                 rtnl_lock();
1817                                 ops = rtnl_link_ops_get(kind);
1818                                 if (ops)
1819                                         goto replay;
1820                         }
1821 #endif
1822                         return -EOPNOTSUPP;
1823                 }
1824
1825                 if (!ifname[0])
1826                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1827
1828                 dest_net = rtnl_link_get_net(net, tb);
1829                 if (IS_ERR(dest_net))
1830                         return PTR_ERR(dest_net);
1831
1832                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1833
1834                 if (IS_ERR(dev))
1835                         err = PTR_ERR(dev);
1836                 else if (ops->newlink)
1837                         err = ops->newlink(net, dev, tb, data);
1838                 else
1839                         err = register_netdevice(dev);
1840
1841                 if (err < 0 && !IS_ERR(dev))
1842                         free_netdev(dev);
1843                 if (err < 0)
1844                         goto out;
1845
1846                 err = rtnl_configure_link(dev, ifm);
1847                 if (err < 0)
1848                         unregister_netdevice(dev);
1849 out:
1850                 put_net(dest_net);
1851                 return err;
1852         }
1853 }
1854
1855 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1856 {
1857         struct net *net = sock_net(skb->sk);
1858         struct ifinfomsg *ifm;
1859         char ifname[IFNAMSIZ];
1860         struct nlattr *tb[IFLA_MAX+1];
1861         struct net_device *dev = NULL;
1862         struct sk_buff *nskb;
1863         int err;
1864         u32 ext_filter_mask = 0;
1865
1866         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1867         if (err < 0)
1868                 return err;
1869
1870         if (tb[IFLA_IFNAME])
1871                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1872
1873         if (tb[IFLA_EXT_MASK])
1874                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1875
1876         ifm = nlmsg_data(nlh);
1877         if (ifm->ifi_index > 0)
1878                 dev = __dev_get_by_index(net, ifm->ifi_index);
1879         else if (tb[IFLA_IFNAME])
1880                 dev = __dev_get_by_name(net, ifname);
1881         else
1882                 return -EINVAL;
1883
1884         if (dev == NULL)
1885                 return -ENODEV;
1886
1887         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1888         if (nskb == NULL)
1889                 return -ENOBUFS;
1890
1891         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1892                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1893         if (err < 0) {
1894                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1895                 WARN_ON(err == -EMSGSIZE);
1896                 kfree_skb(nskb);
1897         } else
1898                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1899
1900         return err;
1901 }
1902
1903 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1904 {
1905         struct net *net = sock_net(skb->sk);
1906         struct net_device *dev;
1907         struct nlattr *tb[IFLA_MAX+1];
1908         u32 ext_filter_mask = 0;
1909         u16 min_ifinfo_dump_size = 0;
1910
1911         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1912                         ifla_policy) >= 0) {
1913                 if (tb[IFLA_EXT_MASK])
1914                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1915         }
1916
1917         if (!ext_filter_mask)
1918                 return NLMSG_GOODSIZE;
1919         /*
1920          * traverse the list of net devices and compute the minimum
1921          * buffer size based upon the filter mask.
1922          */
1923         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1924                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1925                                              if_nlmsg_size(dev,
1926                                                            ext_filter_mask));
1927         }
1928
1929         return min_ifinfo_dump_size;
1930 }
1931
1932 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1933 {
1934         int idx;
1935         int s_idx = cb->family;
1936
1937         if (s_idx == 0)
1938                 s_idx = 1;
1939         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1940                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1941                 if (idx < s_idx || idx == PF_PACKET)
1942                         continue;
1943                 if (rtnl_msg_handlers[idx] == NULL ||
1944                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1945                         continue;
1946                 if (idx > s_idx)
1947                         memset(&cb->args[0], 0, sizeof(cb->args));
1948                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1949                         break;
1950         }
1951         cb->family = idx;
1952
1953         return skb->len;
1954 }
1955
1956 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1957 {
1958         struct net *net = dev_net(dev);
1959         struct sk_buff *skb;
1960         int err = -ENOBUFS;
1961         size_t if_info_size;
1962
1963         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1964         if (skb == NULL)
1965                 goto errout;
1966
1967         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1968         if (err < 0) {
1969                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1970                 WARN_ON(err == -EMSGSIZE);
1971                 kfree_skb(skb);
1972                 goto errout;
1973         }
1974         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1975         return;
1976 errout:
1977         if (err < 0)
1978                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1979 }
1980
1981 /* Protected by RTNL sempahore.  */
1982 static struct rtattr **rta_buf;
1983 static int rtattr_max;
1984
1985 /* Process one rtnetlink message. */
1986
1987 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1988 {
1989         struct net *net = sock_net(skb->sk);
1990         rtnl_doit_func doit;
1991         int sz_idx, kind;
1992         int min_len;
1993         int family;
1994         int type;
1995         int err;
1996
1997         type = nlh->nlmsg_type;
1998         if (type > RTM_MAX)
1999                 return -EOPNOTSUPP;
2000
2001         type -= RTM_BASE;
2002
2003         /* All the messages must have at least 1 byte length */
2004         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2005                 return 0;
2006
2007         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2008         sz_idx = type>>2;
2009         kind = type&3;
2010
2011         if (kind != 2 && !capable(CAP_NET_ADMIN))
2012                 return -EPERM;
2013
2014         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2015                 struct sock *rtnl;
2016                 rtnl_dumpit_func dumpit;
2017                 rtnl_calcit_func calcit;
2018                 u16 min_dump_alloc = 0;
2019
2020                 dumpit = rtnl_get_dumpit(family, type);
2021                 if (dumpit == NULL)
2022                         return -EOPNOTSUPP;
2023                 calcit = rtnl_get_calcit(family, type);
2024                 if (calcit)
2025                         min_dump_alloc = calcit(skb, nlh);
2026
2027                 __rtnl_unlock();
2028                 rtnl = net->rtnl;
2029                 {
2030                         struct netlink_dump_control c = {
2031                                 .dump           = dumpit,
2032                                 .min_dump_alloc = min_dump_alloc,
2033                         };
2034                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2035                 }
2036                 rtnl_lock();
2037                 return err;
2038         }
2039
2040         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2041
2042         min_len = rtm_min[sz_idx];
2043         if (nlh->nlmsg_len < min_len)
2044                 return -EINVAL;
2045
2046         if (nlh->nlmsg_len > min_len) {
2047                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2048                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2049
2050                 while (RTA_OK(attr, attrlen)) {
2051                         unsigned flavor = attr->rta_type;
2052                         if (flavor) {
2053                                 if (flavor > rta_max[sz_idx])
2054                                         return -EINVAL;
2055                                 rta_buf[flavor-1] = attr;
2056                         }
2057                         attr = RTA_NEXT(attr, attrlen);
2058                 }
2059         }
2060
2061         doit = rtnl_get_doit(family, type);
2062         if (doit == NULL)
2063                 return -EOPNOTSUPP;
2064
2065         return doit(skb, nlh, (void *)&rta_buf[0]);
2066 }
2067
2068 static void rtnetlink_rcv(struct sk_buff *skb)
2069 {
2070         rtnl_lock();
2071         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2072         rtnl_unlock();
2073 }
2074
2075 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2076 {
2077         struct net_device *dev = ptr;
2078
2079         switch (event) {
2080         case NETDEV_UP:
2081         case NETDEV_DOWN:
2082         case NETDEV_PRE_UP:
2083         case NETDEV_POST_INIT:
2084         case NETDEV_REGISTER:
2085         case NETDEV_CHANGE:
2086         case NETDEV_PRE_TYPE_CHANGE:
2087         case NETDEV_GOING_DOWN:
2088         case NETDEV_UNREGISTER:
2089         case NETDEV_UNREGISTER_BATCH:
2090         case NETDEV_RELEASE:
2091         case NETDEV_JOIN:
2092                 break;
2093         default:
2094                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2095                 break;
2096         }
2097         return NOTIFY_DONE;
2098 }
2099
2100 static struct notifier_block rtnetlink_dev_notifier = {
2101         .notifier_call  = rtnetlink_event,
2102 };
2103
2104
2105 static int __net_init rtnetlink_net_init(struct net *net)
2106 {
2107         struct sock *sk;
2108         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2109                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2110         if (!sk)
2111                 return -ENOMEM;
2112         net->rtnl = sk;
2113         return 0;
2114 }
2115
2116 static void __net_exit rtnetlink_net_exit(struct net *net)
2117 {
2118         netlink_kernel_release(net->rtnl);
2119         net->rtnl = NULL;
2120 }
2121
2122 static struct pernet_operations rtnetlink_net_ops = {
2123         .init = rtnetlink_net_init,
2124         .exit = rtnetlink_net_exit,
2125 };
2126
2127 void __init rtnetlink_init(void)
2128 {
2129         int i;
2130
2131         rtattr_max = 0;
2132         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2133                 if (rta_max[i] > rtattr_max)
2134                         rtattr_max = rta_max[i];
2135         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2136         if (!rta_buf)
2137                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2138
2139         if (register_pernet_subsys(&rtnetlink_net_ops))
2140                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2141
2142         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2143         register_netdevice_notifier(&rtnetlink_dev_notifier);
2144
2145         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2146                       rtnl_dump_ifinfo, rtnl_calcit);
2147         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2148         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2149         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2150
2151         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2152         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2153 }
2154