net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/pci.h>
  40 #include <linux/etherdevice.h>
  41
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 struct rtnl_link {
  58         rtnl_doit_func          doit;
  59         rtnl_dumpit_func        dumpit;
  60         rtnl_calcit_func        calcit;
  61 };
  62
  63 static DEFINE_MUTEX(rtnl_mutex);
  64
  65 void rtnl_lock(void)
  66 {
  67         mutex_lock(&rtnl_mutex);
  68 }
  69 EXPORT_SYMBOL(rtnl_lock);
  70
  71 void __rtnl_unlock(void)
  72 {
  73         mutex_unlock(&rtnl_mutex);
  74 }
  75
  76 void rtnl_unlock(void)
  77 {
  78         /* This fellow will unlock it for us. */
  79         netdev_run_todo();
  80 }
  81 EXPORT_SYMBOL(rtnl_unlock);
  82
  83 int rtnl_trylock(void)
  84 {
  85         return mutex_trylock(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_trylock);
  88
  89 int rtnl_is_locked(void)
  90 {
  91         return mutex_is_locked(&rtnl_mutex);
  92 }
  93 EXPORT_SYMBOL(rtnl_is_locked);
  94
  95 #ifdef CONFIG_PROVE_LOCKING
  96 int lockdep_rtnl_is_held(void)
  97 {
  98         return lockdep_is_held(&rtnl_mutex);
  99 }
 100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105 static inline int rtm_msgindex(int msgtype)
 106 {
 107         int msgindex = msgtype - RTM_BASE;
 108
 109         /*
 110          * msgindex < 0 implies someone tried to register a netlink
 111          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112          * the message type has not been added to linux/rtnetlink.h
 113          */
 114         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116         return msgindex;
 117 }
 118
 119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120 {
 121         struct rtnl_link *tab;
 122
 123         if (protocol <= RTNL_FAMILY_MAX)
 124                 tab = rtnl_msg_handlers[protocol];
 125         else
 126                 tab = NULL;
 127
 128         if (tab == NULL || tab[msgindex].doit == NULL)
 129                 tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131         return tab ? tab[msgindex].doit : NULL;
 132 }
 133
 134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135 {
 136         struct rtnl_link *tab;
 137
 138         if (protocol <= RTNL_FAMILY_MAX)
 139                 tab = rtnl_msg_handlers[protocol];
 140         else
 141                 tab = NULL;
 142
 143         if (tab == NULL || tab[msgindex].dumpit == NULL)
 144                 tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146         return tab ? tab[msgindex].dumpit : NULL;
 147 }
 148
 149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150 {
 151         struct rtnl_link *tab;
 152
 153         if (protocol <= RTNL_FAMILY_MAX)
 154                 tab = rtnl_msg_handlers[protocol];
 155         else
 156                 tab = NULL;
 157
 158         if (tab == NULL || tab[msgindex].calcit == NULL)
 159                 tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161         return tab ? tab[msgindex].calcit : NULL;
 162 }
 163
 164 /**
 165  * __rtnl_register - Register a rtnetlink message type
 166  * @protocol: Protocol family or PF_UNSPEC
 167  * @msgtype: rtnetlink message type
 168  * @doit: Function pointer called for each request message
 169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170  * @calcit: Function pointer to calc size of dump message
 171  *
 172  * Registers the specified function pointers (at least one of them has
 173  * to be non-NULL) to be called whenever a request message for the
 174  * specified protocol family and message type is received.
 175  *
 176  * The special protocol family PF_UNSPEC may be used to define fallback
 177  * function pointers for the case when no entry for the specific protocol
 178  * family exists.
 179  *
 180  * Returns 0 on success or a negative error code.
 181  */
 182 int __rtnl_register(int protocol, int msgtype,
 183                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184                     rtnl_calcit_func calcit)
 185 {
 186         struct rtnl_link *tab;
 187         int msgindex;
 188
 189         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190         msgindex = rtm_msgindex(msgtype);
 191
 192         tab = rtnl_msg_handlers[protocol];
 193         if (tab == NULL) {
 194                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195                 if (tab == NULL)
 196                         return -ENOBUFS;
 197
 198                 rtnl_msg_handlers[protocol] = tab;
 199         }
 200
 201         if (doit)
 202                 tab[msgindex].doit = doit;
 203
 204         if (dumpit)
 205                 tab[msgindex].dumpit = dumpit;
 206
 207         if (calcit)
 208                 tab[msgindex].calcit = calcit;
 209
 210         return 0;
 211 }
 212 EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214 /**
 215  * rtnl_register - Register a rtnetlink message type
 216  *
 217  * Identical to __rtnl_register() but panics on failure. This is useful
 218  * as failure of this function is very unlikely, it can only happen due
 219  * to lack of memory when allocating the chain to store all message
 220  * handlers for a protocol. Meant for use in init functions where lack
 221  * of memory implies no sense in continuing.
 222  */
 223 void rtnl_register(int protocol, int msgtype,
 224                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225                    rtnl_calcit_func calcit)
 226 {
 227         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228                 panic("Unable to register rtnetlink message handler, "
 229                       "protocol = %d, message type = %d\n",
 230                       protocol, msgtype);
 231 }
 232 EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234 /**
 235  * rtnl_unregister - Unregister a rtnetlink message type
 236  * @protocol: Protocol family or PF_UNSPEC
 237  * @msgtype: rtnetlink message type
 238  *
 239  * Returns 0 on success or a negative error code.
 240  */
 241 int rtnl_unregister(int protocol, int msgtype)
 242 {
 243         int msgindex;
 244
 245         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246         msgindex = rtm_msgindex(msgtype);
 247
 248         if (rtnl_msg_handlers[protocol] == NULL)
 249                 return -ENOENT;
 250
 251         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254         return 0;
 255 }
 256 EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258 /**
 259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260  * @protocol : Protocol family or PF_UNSPEC
 261  *
 262  * Identical to calling rtnl_unregster() for all registered message types
 263  * of a certain protocol family.
 264  */
 265 void rtnl_unregister_all(int protocol)
 266 {
 267         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269         kfree(rtnl_msg_handlers[protocol]);
 270         rtnl_msg_handlers[protocol] = NULL;
 271 }
 272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274 static LIST_HEAD(link_ops);
 275
 276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277 {
 278         const struct rtnl_link_ops *ops;
 279
 280         list_for_each_entry(ops, &link_ops, list) {
 281                 if (!strcmp(ops->kind, kind))
 282                         return ops;
 283         }
 284         return NULL;
 285 }
 286
 287 /**
 288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289  * @ops: struct rtnl_link_ops * to register
 290  *
 291  * The caller must hold the rtnl_mutex. This function should be used
 292  * by drivers that create devices during module initialization. It
 293  * must be called before registering the devices.
 294  *
 295  * Returns 0 on success or a negative error code.
 296  */
 297 int __rtnl_link_register(struct rtnl_link_ops *ops)
 298 {
 299         if (rtnl_link_ops_get(ops->kind))
 300                 return -EEXIST;
 301
 302         if (!ops->dellink)
 303                 ops->dellink = unregister_netdevice_queue;
 304
 305         list_add_tail(&ops->list, &link_ops);
 306         return 0;
 307 }
 308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 309
 310 /**
 311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 312  * @ops: struct rtnl_link_ops * to register
 313  *
 314  * Returns 0 on success or a negative error code.
 315  */
 316 int rtnl_link_register(struct rtnl_link_ops *ops)
 317 {
 318         int err;
 319
 320         rtnl_lock();
 321         err = __rtnl_link_register(ops);
 322         rtnl_unlock();
 323         return err;
 324 }
 325 EXPORT_SYMBOL_GPL(rtnl_link_register);
 326
 327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 328 {
 329         struct net_device *dev;
 330         LIST_HEAD(list_kill);
 331
 332         for_each_netdev(net, dev) {
 333                 if (dev->rtnl_link_ops == ops)
 334                         ops->dellink(dev, &list_kill);
 335         }
 336         unregister_netdevice_many(&list_kill);
 337 }
 338
 339 /**
 340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 341  * @ops: struct rtnl_link_ops * to unregister
 342  *
 343  * The caller must hold the rtnl_mutex.
 344  */
 345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 346 {
 347         struct net *net;
 348
 349         for_each_net(net) {
 350                 __rtnl_kill_links(net, ops);
 351         }
 352         list_del(&ops->list);
 353 }
 354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 355
 356 /**
 357  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 358  * @ops: struct rtnl_link_ops * to unregister
 359  */
 360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 361 {
 362         rtnl_lock();
 363         __rtnl_link_unregister(ops);
 364         rtnl_unlock();
 365 }
 366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 367
 368 static size_t rtnl_link_get_size(const struct net_device *dev)
 369 {
 370         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 371         size_t size;
 372
 373         if (!ops)
 374                 return 0;
 375
 376         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 377                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 378
 379         if (ops->get_size)
 380                 /* IFLA_INFO_DATA + nested data */
 381                 size += nla_total_size(sizeof(struct nlattr)) +
 382                         ops->get_size(dev);
 383
 384         if (ops->get_xstats_size)
 385                 /* IFLA_INFO_XSTATS */
 386                 size += nla_total_size(ops->get_xstats_size(dev));
 387
 388         return size;
 389 }
 390
 391 static LIST_HEAD(rtnl_af_ops);
 392
 393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 394 {
 395         const struct rtnl_af_ops *ops;
 396
 397         list_for_each_entry(ops, &rtnl_af_ops, list) {
 398                 if (ops->family == family)
 399                         return ops;
 400         }
 401
 402         return NULL;
 403 }
 404
 405 /**
 406  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 407  * @ops: struct rtnl_af_ops * to register
 408  *
 409  * The caller must hold the rtnl_mutex.
 410  *
 411  * Returns 0 on success or a negative error code.
 412  */
 413 int __rtnl_af_register(struct rtnl_af_ops *ops)
 414 {
 415         list_add_tail(&ops->list, &rtnl_af_ops);
 416         return 0;
 417 }
 418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
 419
 420 /**
 421  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 422  * @ops: struct rtnl_af_ops * to register
 423  *
 424  * Returns 0 on success or a negative error code.
 425  */
 426 int rtnl_af_register(struct rtnl_af_ops *ops)
 427 {
 428         int err;
 429
 430         rtnl_lock();
 431         err = __rtnl_af_register(ops);
 432         rtnl_unlock();
 433         return err;
 434 }
 435 EXPORT_SYMBOL_GPL(rtnl_af_register);
 436
 437 /**
 438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 439  * @ops: struct rtnl_af_ops * to unregister
 440  *
 441  * The caller must hold the rtnl_mutex.
 442  */
 443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 444 {
 445         list_del(&ops->list);
 446 }
 447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 448
 449 /**
 450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 451  * @ops: struct rtnl_af_ops * to unregister
 452  */
 453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 454 {
 455         rtnl_lock();
 456         __rtnl_af_unregister(ops);
 457         rtnl_unlock();
 458 }
 459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 460
 461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 462 {
 463         struct rtnl_af_ops *af_ops;
 464         size_t size;
 465
 466         /* IFLA_AF_SPEC */
 467         size = nla_total_size(sizeof(struct nlattr));
 468
 469         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 470                 if (af_ops->get_link_af_size) {
 471                         /* AF_* + nested data */
 472                         size += nla_total_size(sizeof(struct nlattr)) +
 473                                 af_ops->get_link_af_size(dev);
 474                 }
 475         }
 476
 477         return size;
 478 }
 479
 480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 481 {
 482         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 483         struct nlattr *linkinfo, *data;
 484         int err = -EMSGSIZE;
 485
 486         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 487         if (linkinfo == NULL)
 488                 goto out;
 489
 490         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 491                 goto err_cancel_link;
 492         if (ops->fill_xstats) {
 493                 err = ops->fill_xstats(skb, dev);
 494                 if (err < 0)
 495                         goto err_cancel_link;
 496         }
 497         if (ops->fill_info) {
 498                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 499                 if (data == NULL)
 500                         goto err_cancel_link;
 501                 err = ops->fill_info(skb, dev);
 502                 if (err < 0)
 503                         goto err_cancel_data;
 504                 nla_nest_end(skb, data);
 505         }
 506
 507         nla_nest_end(skb, linkinfo);
 508         return 0;
 509
 510 err_cancel_data:
 511         nla_nest_cancel(skb, data);
 512 err_cancel_link:
 513         nla_nest_cancel(skb, linkinfo);
 514 out:
 515         return err;
 516 }
 517
 518 static const int rtm_min[RTM_NR_FAMILIES] =
 519 {
 520         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 521         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 522         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 523         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 524         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 527         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 528         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 530 };
 531
 532 static const int rta_max[RTM_NR_FAMILIES] =
 533 {
 534         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 535         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 536         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 537         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 538         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 539         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 540         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 541         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 542 };
 543
 544 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 int 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                         if (nla_put_u32(skb, i+1, metrics[i]))
 613                                 goto nla_put_failure;
 614                 }
 615         }
 616
 617         if (!valid) {
 618                 nla_nest_cancel(skb, mx);
 619                 return 0;
 620         }
 621
 622         return nla_nest_end(skb, mx);
 623
 624 nla_put_failure:
 625         nla_nest_cancel(skb, mx);
 626         return -EMSGSIZE;
 627 }
 628 EXPORT_SYMBOL(rtnetlink_put_metrics);
 629
 630 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 631                        u32 ts, u32 tsage, long expires, u32 error)
 632 {
 633         struct rta_cacheinfo ci = {
 634                 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 635                 .rta_used = dst->__use,
 636                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 637                 .rta_error = error,
 638                 .rta_id =  id,
 639                 .rta_ts = ts,
 640                 .rta_tsage = tsage,
 641         };
 642
 643         if (expires)
 644                 ci.rta_expires = jiffies_to_clock_t(expires);
 645
 646         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 647 }
 648 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 649
 650 static void set_operstate(struct net_device *dev, unsigned char transition)
 651 {
 652         unsigned char operstate = dev->operstate;
 653
 654         switch (transition) {
 655         case IF_OPER_UP:
 656                 if ((operstate == IF_OPER_DORMANT ||
 657                      operstate == IF_OPER_UNKNOWN) &&
 658                     !netif_dormant(dev))
 659                         operstate = IF_OPER_UP;
 660                 break;
 661
 662         case IF_OPER_DORMANT:
 663                 if (operstate == IF_OPER_UP ||
 664                     operstate == IF_OPER_UNKNOWN)
 665                         operstate = IF_OPER_DORMANT;
 666                 break;
 667         }
 668
 669         if (dev->operstate != operstate) {
 670                 write_lock_bh(&dev_base_lock);
 671                 dev->operstate = operstate;
 672                 write_unlock_bh(&dev_base_lock);
 673                 netdev_state_change(dev);
 674         }
 675 }
 676
 677 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 678                                            const struct ifinfomsg *ifm)
 679 {
 680         unsigned int flags = ifm->ifi_flags;
 681
 682         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 683         if (ifm->ifi_change)
 684                 flags = (flags & ifm->ifi_change) |
 685                         (dev->flags & ~ifm->ifi_change);
 686
 687         return flags;
 688 }
 689
 690 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 691                                  const struct rtnl_link_stats64 *b)
 692 {
 693         a->rx_packets = b->rx_packets;
 694         a->tx_packets = b->tx_packets;
 695         a->rx_bytes = b->rx_bytes;
 696         a->tx_bytes = b->tx_bytes;
 697         a->rx_errors = b->rx_errors;
 698         a->tx_errors = b->tx_errors;
 699         a->rx_dropped = b->rx_dropped;
 700         a->tx_dropped = b->tx_dropped;
 701
 702         a->multicast = b->multicast;
 703         a->collisions = b->collisions;
 704
 705         a->rx_length_errors = b->rx_length_errors;
 706         a->rx_over_errors = b->rx_over_errors;
 707         a->rx_crc_errors = b->rx_crc_errors;
 708         a->rx_frame_errors = b->rx_frame_errors;
 709         a->rx_fifo_errors = b->rx_fifo_errors;
 710         a->rx_missed_errors = b->rx_missed_errors;
 711
 712         a->tx_aborted_errors = b->tx_aborted_errors;
 713         a->tx_carrier_errors = b->tx_carrier_errors;
 714         a->tx_fifo_errors = b->tx_fifo_errors;
 715         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 716         a->tx_window_errors = b->tx_window_errors;
 717
 718         a->rx_compressed = b->rx_compressed;
 719         a->tx_compressed = b->tx_compressed;
 720 }
 721
 722 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 723 {
 724         memcpy(v, b, sizeof(*b));
 725 }
 726
 727 /* All VF info */
 728 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 729                                    u32 ext_filter_mask)
 730 {
 731         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 732             (ext_filter_mask & RTEXT_FILTER_VF)) {
 733                 int num_vfs = dev_num_vf(dev->dev.parent);
 734                 size_t size = nla_total_size(sizeof(struct nlattr));
 735                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 736                 size += num_vfs *
 737                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 738                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 739                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 740                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 741                 return size;
 742         } else
 743                 return 0;
 744 }
 745
 746 static size_t rtnl_port_size(const struct net_device *dev)
 747 {
 748         size_t port_size = nla_total_size(4)            /* PORT_VF */
 749                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 750                 + nla_total_size(sizeof(struct ifla_port_vsi))
 751                                                         /* PORT_VSI_TYPE */
 752                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 753                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 754                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 755                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 756         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 757         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 758                 + port_size;
 759         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 760                 + port_size;
 761
 762         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 763                 return 0;
 764         if (dev_num_vf(dev->dev.parent))
 765                 return port_self_size + vf_ports_size +
 766                         vf_port_size * dev_num_vf(dev->dev.parent);
 767         else
 768                 return port_self_size;
 769 }
 770
 771 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 772                                      u32 ext_filter_mask)
 773 {
 774         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 775                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 776                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 777                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 778                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 779                + nla_total_size(sizeof(struct rtnl_link_stats))
 780                + nla_total_size(sizeof(struct rtnl_link_stats64))
 781                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 782                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 783                + nla_total_size(4) /* IFLA_TXQLEN */
 784                + nla_total_size(4) /* IFLA_WEIGHT */
 785                + nla_total_size(4) /* IFLA_MTU */
 786                + nla_total_size(4) /* IFLA_LINK */
 787                + nla_total_size(4) /* IFLA_MASTER */
 788                + nla_total_size(4) /* IFLA_PROMISCUITY */
 789                + nla_total_size(1) /* IFLA_OPERSTATE */
 790                + nla_total_size(1) /* IFLA_LINKMODE */
 791                + nla_total_size(ext_filter_mask
 792                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 793                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 794                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 795                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 796                + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 797 }
 798
 799 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 800 {
 801         struct nlattr *vf_ports;
 802         struct nlattr *vf_port;
 803         int vf;
 804         int err;
 805
 806         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 807         if (!vf_ports)
 808                 return -EMSGSIZE;
 809
 810         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 811                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 812                 if (!vf_port)
 813                         goto nla_put_failure;
 814                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 815                         goto nla_put_failure;
 816                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 817                 if (err == -EMSGSIZE)
 818                         goto nla_put_failure;
 819                 if (err) {
 820                         nla_nest_cancel(skb, vf_port);
 821                         continue;
 822                 }
 823                 nla_nest_end(skb, vf_port);
 824         }
 825
 826         nla_nest_end(skb, vf_ports);
 827
 828         return 0;
 829
 830 nla_put_failure:
 831         nla_nest_cancel(skb, vf_ports);
 832         return -EMSGSIZE;
 833 }
 834
 835 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 836 {
 837         struct nlattr *port_self;
 838         int err;
 839
 840         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 841         if (!port_self)
 842                 return -EMSGSIZE;
 843
 844         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 845         if (err) {
 846                 nla_nest_cancel(skb, port_self);
 847                 return (err == -EMSGSIZE) ? err : 0;
 848         }
 849
 850         nla_nest_end(skb, port_self);
 851
 852         return 0;
 853 }
 854
 855 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 856 {
 857         int err;
 858
 859         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 860                 return 0;
 861
 862         err = rtnl_port_self_fill(skb, dev);
 863         if (err)
 864                 return err;
 865
 866         if (dev_num_vf(dev->dev.parent)) {
 867                 err = rtnl_vf_ports_fill(skb, dev);
 868                 if (err)
 869                         return err;
 870         }
 871
 872         return 0;
 873 }
 874
 875 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 876                             int type, u32 pid, u32 seq, u32 change,
 877                             unsigned int flags, u32 ext_filter_mask)
 878 {
 879         struct ifinfomsg *ifm;
 880         struct nlmsghdr *nlh;
 881         struct rtnl_link_stats64 temp;
 882         const struct rtnl_link_stats64 *stats;
 883         struct nlattr *attr, *af_spec;
 884         struct rtnl_af_ops *af_ops;
 885
 886         ASSERT_RTNL();
 887         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 888         if (nlh == NULL)
 889                 return -EMSGSIZE;
 890
 891         ifm = nlmsg_data(nlh);
 892         ifm->ifi_family = AF_UNSPEC;
 893         ifm->__ifi_pad = 0;
 894         ifm->ifi_type = dev->type;
 895         ifm->ifi_index = dev->ifindex;
 896         ifm->ifi_flags = dev_get_flags(dev);
 897         ifm->ifi_change = change;
 898
 899         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 900             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 901             nla_put_u8(skb, IFLA_OPERSTATE,
 902                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 903             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 904             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 905             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 906             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 907             (dev->ifindex != dev->iflink &&
 908              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 909             (dev->master &&
 910              nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 911             (dev->qdisc &&
 912              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 913             (dev->ifalias &&
 914              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 915                 goto nla_put_failure;
 916
 917         if (1) {
 918                 struct rtnl_link_ifmap map = {
 919                         .mem_start   = dev->mem_start,
 920                         .mem_end     = dev->mem_end,
 921                         .base_addr   = dev->base_addr,
 922                         .irq         = dev->irq,
 923                         .dma         = dev->dma,
 924                         .port        = dev->if_port,
 925                 };
 926                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 927                         goto nla_put_failure;
 928         }
 929
 930         if (dev->addr_len) {
 931                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 932                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 933                         goto nla_put_failure;
 934         }
 935
 936         attr = nla_reserve(skb, IFLA_STATS,
 937                         sizeof(struct rtnl_link_stats));
 938         if (attr == NULL)
 939                 goto nla_put_failure;
 940
 941         stats = dev_get_stats(dev, &temp);
 942         copy_rtnl_link_stats(nla_data(attr), stats);
 943
 944         attr = nla_reserve(skb, IFLA_STATS64,
 945                         sizeof(struct rtnl_link_stats64));
 946         if (attr == NULL)
 947                 goto nla_put_failure;
 948         copy_rtnl_link_stats64(nla_data(attr), stats);
 949
 950         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 951             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 952                 goto nla_put_failure;
 953
 954         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 955             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 956                 int i;
 957
 958                 struct nlattr *vfinfo, *vf;
 959                 int num_vfs = dev_num_vf(dev->dev.parent);
 960
 961                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 962                 if (!vfinfo)
 963                         goto nla_put_failure;
 964                 for (i = 0; i < num_vfs; i++) {
 965                         struct ifla_vf_info ivi;
 966                         struct ifla_vf_mac vf_mac;
 967                         struct ifla_vf_vlan vf_vlan;
 968                         struct ifla_vf_tx_rate vf_tx_rate;
 969                         struct ifla_vf_spoofchk vf_spoofchk;
 970
 971                         /*
 972                          * Not all SR-IOV capable drivers support the
 973                          * spoofcheck query.  Preset to -1 so the user
 974                          * space tool can detect that the driver didn't
 975                          * report anything.
 976                          */
 977                         ivi.spoofchk = -1;
 978                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 979                                 break;
 980                         vf_mac.vf =
 981                                 vf_vlan.vf =
 982                                 vf_tx_rate.vf =
 983                                 vf_spoofchk.vf = ivi.vf;
 984
 985                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 986                         vf_vlan.vlan = ivi.vlan;
 987                         vf_vlan.qos = ivi.qos;
 988                         vf_tx_rate.rate = ivi.tx_rate;
 989                         vf_spoofchk.setting = ivi.spoofchk;
 990                         vf = nla_nest_start(skb, IFLA_VF_INFO);
 991                         if (!vf) {
 992                                 nla_nest_cancel(skb, vfinfo);
 993                                 goto nla_put_failure;
 994                         }
 995                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
 996                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
 997                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
 998                                     &vf_tx_rate) ||
 999                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1000                                     &vf_spoofchk))
1001                                 goto nla_put_failure;
1002                         nla_nest_end(skb, vf);
1003                 }
1004                 nla_nest_end(skb, vfinfo);
1005         }
1006
1007         if (rtnl_port_fill(skb, dev))
1008                 goto nla_put_failure;
1009
1010         if (dev->rtnl_link_ops) {
1011                 if (rtnl_link_fill(skb, dev) < 0)
1012                         goto nla_put_failure;
1013         }
1014
1015         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1016                 goto nla_put_failure;
1017
1018         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1019                 if (af_ops->fill_link_af) {
1020                         struct nlattr *af;
1021                         int err;
1022
1023                         if (!(af = nla_nest_start(skb, af_ops->family)))
1024                                 goto nla_put_failure;
1025
1026                         err = af_ops->fill_link_af(skb, dev);
1027
1028                         /*
1029                          * Caller may return ENODATA to indicate that there
1030                          * was no data to be dumped. This is not an error, it
1031                          * means we should trim the attribute header and
1032                          * continue.
1033                          */
1034                         if (err == -ENODATA)
1035                                 nla_nest_cancel(skb, af);
1036                         else if (err < 0)
1037                                 goto nla_put_failure;
1038
1039                         nla_nest_end(skb, af);
1040                 }
1041         }
1042
1043         nla_nest_end(skb, af_spec);
1044
1045         return nlmsg_end(skb, nlh);
1046
1047 nla_put_failure:
1048         nlmsg_cancel(skb, nlh);
1049         return -EMSGSIZE;
1050 }
1051
1052 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1053 {
1054         struct net *net = sock_net(skb->sk);
1055         int h, s_h;
1056         int idx = 0, s_idx;
1057         struct net_device *dev;
1058         struct hlist_head *head;
1059         struct hlist_node *node;
1060         struct nlattr *tb[IFLA_MAX+1];
1061         u32 ext_filter_mask = 0;
1062
1063         s_h = cb->args[0];
1064         s_idx = cb->args[1];
1065
1066         rcu_read_lock();
1067         cb->seq = net->dev_base_seq;
1068
1069         if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1070                         ifla_policy) >= 0) {
1071
1072                 if (tb[IFLA_EXT_MASK])
1073                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1074         }
1075
1076         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1077                 idx = 0;
1078                 head = &net->dev_index_head[h];
1079                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1080                         if (idx < s_idx)
1081                                 goto cont;
1082                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1083                                              NETLINK_CB(cb->skb).pid,
1084                                              cb->nlh->nlmsg_seq, 0,
1085                                              NLM_F_MULTI,
1086                                              ext_filter_mask) <= 0)
1087                                 goto out;
1088
1089                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1090 cont:
1091                         idx++;
1092                 }
1093         }
1094 out:
1095         rcu_read_unlock();
1096         cb->args[1] = idx;
1097         cb->args[0] = h;
1098
1099         return skb->len;
1100 }
1101
1102 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1103         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1104         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1105         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1107         [IFLA_MTU]              = { .type = NLA_U32 },
1108         [IFLA_LINK]             = { .type = NLA_U32 },
1109         [IFLA_MASTER]           = { .type = NLA_U32 },
1110         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1111         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1112         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1113         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1114         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1115         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1116         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1117         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1118         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1119         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1120         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1121         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1122         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1123         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1124 };
1125 EXPORT_SYMBOL(ifla_policy);
1126
1127 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1128         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1129         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1130 };
1131
1132 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1133         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1134 };
1135
1136 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1137         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1138                                     .len = sizeof(struct ifla_vf_mac) },
1139         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1140                                     .len = sizeof(struct ifla_vf_vlan) },
1141         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1142                                     .len = sizeof(struct ifla_vf_tx_rate) },
1143         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1144                                     .len = sizeof(struct ifla_vf_spoofchk) },
1145 };
1146
1147 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1148         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1149         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1150                                     .len = PORT_PROFILE_MAX },
1151         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1152                                     .len = sizeof(struct ifla_port_vsi)},
1153         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1154                                       .len = PORT_UUID_MAX },
1155         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1156                                     .len = PORT_UUID_MAX },
1157         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1158         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1159 };
1160
1161 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1162 {
1163         struct net *net;
1164         /* Examine the link attributes and figure out which
1165          * network namespace we are talking about.
1166          */
1167         if (tb[IFLA_NET_NS_PID])
1168                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1169         else if (tb[IFLA_NET_NS_FD])
1170                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1171         else
1172                 net = get_net(src_net);
1173         return net;
1174 }
1175 EXPORT_SYMBOL(rtnl_link_get_net);
1176
1177 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1178 {
1179         if (dev) {
1180                 if (tb[IFLA_ADDRESS] &&
1181                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1182                         return -EINVAL;
1183
1184                 if (tb[IFLA_BROADCAST] &&
1185                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1186                         return -EINVAL;
1187         }
1188
1189         if (tb[IFLA_AF_SPEC]) {
1190                 struct nlattr *af;
1191                 int rem, err;
1192
1193                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1194                         const struct rtnl_af_ops *af_ops;
1195
1196                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1197                                 return -EAFNOSUPPORT;
1198
1199                         if (!af_ops->set_link_af)
1200                                 return -EOPNOTSUPP;
1201
1202                         if (af_ops->validate_link_af) {
1203                                 err = af_ops->validate_link_af(dev, af);
1204                                 if (err < 0)
1205                                         return err;
1206                         }
1207                 }
1208         }
1209
1210         return 0;
1211 }
1212
1213 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1214 {
1215         int rem, err = -EINVAL;
1216         struct nlattr *vf;
1217         const struct net_device_ops *ops = dev->netdev_ops;
1218
1219         nla_for_each_nested(vf, attr, rem) {
1220                 switch (nla_type(vf)) {
1221                 case IFLA_VF_MAC: {
1222                         struct ifla_vf_mac *ivm;
1223                         ivm = nla_data(vf);
1224                         err = -EOPNOTSUPP;
1225                         if (ops->ndo_set_vf_mac)
1226                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1227                                                           ivm->mac);
1228                         break;
1229                 }
1230                 case IFLA_VF_VLAN: {
1231                         struct ifla_vf_vlan *ivv;
1232                         ivv = nla_data(vf);
1233                         err = -EOPNOTSUPP;
1234                         if (ops->ndo_set_vf_vlan)
1235                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1236                                                            ivv->vlan,
1237                                                            ivv->qos);
1238                         break;
1239                 }
1240                 case IFLA_VF_TX_RATE: {
1241                         struct ifla_vf_tx_rate *ivt;
1242                         ivt = nla_data(vf);
1243                         err = -EOPNOTSUPP;
1244                         if (ops->ndo_set_vf_tx_rate)
1245                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1246                                                               ivt->rate);
1247                         break;
1248                 }
1249                 case IFLA_VF_SPOOFCHK: {
1250                         struct ifla_vf_spoofchk *ivs;
1251                         ivs = nla_data(vf);
1252                         err = -EOPNOTSUPP;
1253                         if (ops->ndo_set_vf_spoofchk)
1254                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1255                                                                ivs->setting);
1256                         break;
1257                 }
1258                 default:
1259                         err = -EINVAL;
1260                         break;
1261                 }
1262                 if (err)
1263                         break;
1264         }
1265         return err;
1266 }
1267
1268 static int do_set_master(struct net_device *dev, int ifindex)
1269 {
1270         struct net_device *master_dev;
1271         const struct net_device_ops *ops;
1272         int err;
1273
1274         if (dev->master) {
1275                 if (dev->master->ifindex == ifindex)
1276                         return 0;
1277                 ops = dev->master->netdev_ops;
1278                 if (ops->ndo_del_slave) {
1279                         err = ops->ndo_del_slave(dev->master, dev);
1280                         if (err)
1281                                 return err;
1282                 } else {
1283                         return -EOPNOTSUPP;
1284                 }
1285         }
1286
1287         if (ifindex) {
1288                 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1289                 if (!master_dev)
1290                         return -EINVAL;
1291                 ops = master_dev->netdev_ops;
1292                 if (ops->ndo_add_slave) {
1293                         err = ops->ndo_add_slave(master_dev, dev);
1294                         if (err)
1295                                 return err;
1296                 } else {
1297                         return -EOPNOTSUPP;
1298                 }
1299         }
1300         return 0;
1301 }
1302
1303 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1304                       struct nlattr **tb, char *ifname, int modified)
1305 {
1306         const struct net_device_ops *ops = dev->netdev_ops;
1307         int send_addr_notify = 0;
1308         int err;
1309
1310         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1311                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1312                 if (IS_ERR(net)) {
1313                         err = PTR_ERR(net);
1314                         goto errout;
1315                 }
1316                 err = dev_change_net_namespace(dev, net, ifname);
1317                 put_net(net);
1318                 if (err)
1319                         goto errout;
1320                 modified = 1;
1321         }
1322
1323         if (tb[IFLA_MAP]) {
1324                 struct rtnl_link_ifmap *u_map;
1325                 struct ifmap k_map;
1326
1327                 if (!ops->ndo_set_config) {
1328                         err = -EOPNOTSUPP;
1329                         goto errout;
1330                 }
1331
1332                 if (!netif_device_present(dev)) {
1333                         err = -ENODEV;
1334                         goto errout;
1335                 }
1336
1337                 u_map = nla_data(tb[IFLA_MAP]);
1338                 k_map.mem_start = (unsigned long) u_map->mem_start;
1339                 k_map.mem_end = (unsigned long) u_map->mem_end;
1340                 k_map.base_addr = (unsigned short) u_map->base_addr;
1341                 k_map.irq = (unsigned char) u_map->irq;
1342                 k_map.dma = (unsigned char) u_map->dma;
1343                 k_map.port = (unsigned char) u_map->port;
1344
1345                 err = ops->ndo_set_config(dev, &k_map);
1346                 if (err < 0)
1347                         goto errout;
1348
1349                 modified = 1;
1350         }
1351
1352         if (tb[IFLA_ADDRESS]) {
1353                 struct sockaddr *sa;
1354                 int len;
1355
1356                 if (!ops->ndo_set_mac_address) {
1357                         err = -EOPNOTSUPP;
1358                         goto errout;
1359                 }
1360
1361                 if (!netif_device_present(dev)) {
1362                         err = -ENODEV;
1363                         goto errout;
1364                 }
1365
1366                 len = sizeof(sa_family_t) + dev->addr_len;
1367                 sa = kmalloc(len, GFP_KERNEL);
1368                 if (!sa) {
1369                         err = -ENOMEM;
1370                         goto errout;
1371                 }
1372                 sa->sa_family = dev->type;
1373                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1374                        dev->addr_len);
1375                 err = ops->ndo_set_mac_address(dev, sa);
1376                 kfree(sa);
1377                 if (err)
1378                         goto errout;
1379                 send_addr_notify = 1;
1380                 modified = 1;
1381         }
1382
1383         if (tb[IFLA_MTU]) {
1384                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1385                 if (err < 0)
1386                         goto errout;
1387                 modified = 1;
1388         }
1389
1390         if (tb[IFLA_GROUP]) {
1391                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1392                 modified = 1;
1393         }
1394
1395         /*
1396          * Interface selected by interface index but interface
1397          * name provided implies that a name change has been
1398          * requested.
1399          */
1400         if (ifm->ifi_index > 0 && ifname[0]) {
1401                 err = dev_change_name(dev, ifname);
1402                 if (err < 0)
1403                         goto errout;
1404                 modified = 1;
1405         }
1406
1407         if (tb[IFLA_IFALIAS]) {
1408                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1409                                     nla_len(tb[IFLA_IFALIAS]));
1410                 if (err < 0)
1411                         goto errout;
1412                 modified = 1;
1413         }
1414
1415         if (tb[IFLA_BROADCAST]) {
1416                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1417                 send_addr_notify = 1;
1418         }
1419
1420         if (ifm->ifi_flags || ifm->ifi_change) {
1421                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1422                 if (err < 0)
1423                         goto errout;
1424         }
1425
1426         if (tb[IFLA_MASTER]) {
1427                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1428                 if (err)
1429                         goto errout;
1430                 modified = 1;
1431         }
1432
1433         if (tb[IFLA_TXQLEN])
1434                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1435
1436         if (tb[IFLA_OPERSTATE])
1437                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1438
1439         if (tb[IFLA_LINKMODE]) {
1440                 write_lock_bh(&dev_base_lock);
1441                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1442                 write_unlock_bh(&dev_base_lock);
1443         }
1444
1445         if (tb[IFLA_VFINFO_LIST]) {
1446                 struct nlattr *attr;
1447                 int rem;
1448                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1449                         if (nla_type(attr) != IFLA_VF_INFO) {
1450                                 err = -EINVAL;
1451                                 goto errout;
1452                         }
1453                         err = do_setvfinfo(dev, attr);
1454                         if (err < 0)
1455                                 goto errout;
1456                         modified = 1;
1457                 }
1458         }
1459         err = 0;
1460
1461         if (tb[IFLA_VF_PORTS]) {
1462                 struct nlattr *port[IFLA_PORT_MAX+1];
1463                 struct nlattr *attr;
1464                 int vf;
1465                 int rem;
1466
1467                 err = -EOPNOTSUPP;
1468                 if (!ops->ndo_set_vf_port)
1469                         goto errout;
1470
1471                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1472                         if (nla_type(attr) != IFLA_VF_PORT)
1473                                 continue;
1474                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1475                                 attr, ifla_port_policy);
1476                         if (err < 0)
1477                                 goto errout;
1478                         if (!port[IFLA_PORT_VF]) {
1479                                 err = -EOPNOTSUPP;
1480                                 goto errout;
1481                         }
1482                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1483                         err = ops->ndo_set_vf_port(dev, vf, port);
1484                         if (err < 0)
1485                                 goto errout;
1486                         modified = 1;
1487                 }
1488         }
1489         err = 0;
1490
1491         if (tb[IFLA_PORT_SELF]) {
1492                 struct nlattr *port[IFLA_PORT_MAX+1];
1493
1494                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1495                         tb[IFLA_PORT_SELF], ifla_port_policy);
1496                 if (err < 0)
1497                         goto errout;
1498
1499                 err = -EOPNOTSUPP;
1500                 if (ops->ndo_set_vf_port)
1501                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1502                 if (err < 0)
1503                         goto errout;
1504                 modified = 1;
1505         }
1506
1507         if (tb[IFLA_AF_SPEC]) {
1508                 struct nlattr *af;
1509                 int rem;
1510
1511                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1512                         const struct rtnl_af_ops *af_ops;
1513
1514                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1515                                 BUG();
1516
1517                         err = af_ops->set_link_af(dev, af);
1518                         if (err < 0)
1519                                 goto errout;
1520
1521                         modified = 1;
1522                 }
1523         }
1524         err = 0;
1525
1526 errout:
1527         if (err < 0 && modified && net_ratelimit())
1528                 printk(KERN_WARNING "A link change request failed with "
1529                        "some changes committed already. Interface %s may "
1530                        "have been left with an inconsistent configuration, "
1531                        "please check.\n", dev->name);
1532
1533         if (send_addr_notify)
1534                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1535
1536         return err;
1537 }
1538
1539 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1540 {
1541         struct net *net = sock_net(skb->sk);
1542         struct ifinfomsg *ifm;
1543         struct net_device *dev;
1544         int err;
1545         struct nlattr *tb[IFLA_MAX+1];
1546         char ifname[IFNAMSIZ];
1547
1548         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1549         if (err < 0)
1550                 goto errout;
1551
1552         if (tb[IFLA_IFNAME])
1553                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1554         else
1555                 ifname[0] = '\0';
1556
1557         err = -EINVAL;
1558         ifm = nlmsg_data(nlh);
1559         if (ifm->ifi_index > 0)
1560                 dev = __dev_get_by_index(net, ifm->ifi_index);
1561         else if (tb[IFLA_IFNAME])
1562                 dev = __dev_get_by_name(net, ifname);
1563         else
1564                 goto errout;
1565
1566         if (dev == NULL) {
1567                 err = -ENODEV;
1568                 goto errout;
1569         }
1570
1571         err = validate_linkmsg(dev, tb);
1572         if (err < 0)
1573                 goto errout;
1574
1575         err = do_setlink(dev, ifm, tb, ifname, 0);
1576 errout:
1577         return err;
1578 }
1579
1580 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1581 {
1582         struct net *net = sock_net(skb->sk);
1583         const struct rtnl_link_ops *ops;
1584         struct net_device *dev;
1585         struct ifinfomsg *ifm;
1586         char ifname[IFNAMSIZ];
1587         struct nlattr *tb[IFLA_MAX+1];
1588         int err;
1589         LIST_HEAD(list_kill);
1590
1591         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1592         if (err < 0)
1593                 return err;
1594
1595         if (tb[IFLA_IFNAME])
1596                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1597
1598         ifm = nlmsg_data(nlh);
1599         if (ifm->ifi_index > 0)
1600                 dev = __dev_get_by_index(net, ifm->ifi_index);
1601         else if (tb[IFLA_IFNAME])
1602                 dev = __dev_get_by_name(net, ifname);
1603         else
1604                 return -EINVAL;
1605
1606         if (!dev)
1607                 return -ENODEV;
1608
1609         ops = dev->rtnl_link_ops;
1610         if (!ops)
1611                 return -EOPNOTSUPP;
1612
1613         ops->dellink(dev, &list_kill);
1614         unregister_netdevice_many(&list_kill);
1615         list_del(&list_kill);
1616         return 0;
1617 }
1618
1619 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1620 {
1621         unsigned int old_flags;
1622         int err;
1623
1624         old_flags = dev->flags;
1625         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1626                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1627                 if (err < 0)
1628                         return err;
1629         }
1630
1631         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1632         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1633
1634         __dev_notify_flags(dev, old_flags);
1635         return 0;
1636 }
1637 EXPORT_SYMBOL(rtnl_configure_link);
1638
1639 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1640         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1641 {
1642         int err;
1643         struct net_device *dev;
1644         unsigned int num_queues = 1;
1645
1646         if (ops->get_tx_queues) {
1647                 err = ops->get_tx_queues(src_net, tb);
1648                 if (err < 0)
1649                         goto err;
1650                 num_queues = err;
1651         }
1652
1653         err = -ENOMEM;
1654         dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1655         if (!dev)
1656                 goto err;
1657
1658         dev_net_set(dev, net);
1659         dev->rtnl_link_ops = ops;
1660         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1661
1662         if (tb[IFLA_MTU])
1663                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1664         if (tb[IFLA_ADDRESS])
1665                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1666                                 nla_len(tb[IFLA_ADDRESS]));
1667         if (tb[IFLA_BROADCAST])
1668                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1669                                 nla_len(tb[IFLA_BROADCAST]));
1670         if (tb[IFLA_TXQLEN])
1671                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1672         if (tb[IFLA_OPERSTATE])
1673                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1674         if (tb[IFLA_LINKMODE])
1675                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1676         if (tb[IFLA_GROUP])
1677                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1678
1679         return dev;
1680
1681 err:
1682         return ERR_PTR(err);
1683 }
1684 EXPORT_SYMBOL(rtnl_create_link);
1685
1686 static int rtnl_group_changelink(struct net *net, int group,
1687                 struct ifinfomsg *ifm,
1688                 struct nlattr **tb)
1689 {
1690         struct net_device *dev;
1691         int err;
1692
1693         for_each_netdev(net, dev) {
1694                 if (dev->group == group) {
1695                         err = do_setlink(dev, ifm, tb, NULL, 0);
1696                         if (err < 0)
1697                                 return err;
1698                 }
1699         }
1700
1701         return 0;
1702 }
1703
1704 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1705 {
1706         struct net *net = sock_net(skb->sk);
1707         const struct rtnl_link_ops *ops;
1708         struct net_device *dev;
1709         struct ifinfomsg *ifm;
1710         char kind[MODULE_NAME_LEN];
1711         char ifname[IFNAMSIZ];
1712         struct nlattr *tb[IFLA_MAX+1];
1713         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1714         int err;
1715
1716 #ifdef CONFIG_MODULES
1717 replay:
1718 #endif
1719         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1720         if (err < 0)
1721                 return err;
1722
1723         if (tb[IFLA_IFNAME])
1724                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1725         else
1726                 ifname[0] = '\0';
1727
1728         ifm = nlmsg_data(nlh);
1729         if (ifm->ifi_index > 0)
1730                 dev = __dev_get_by_index(net, ifm->ifi_index);
1731         else {
1732                 if (ifname[0])
1733                         dev = __dev_get_by_name(net, ifname);
1734                 else
1735                         dev = NULL;
1736         }
1737
1738         err = validate_linkmsg(dev, tb);
1739         if (err < 0)
1740                 return err;
1741
1742         if (tb[IFLA_LINKINFO]) {
1743                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1744                                        tb[IFLA_LINKINFO], ifla_info_policy);
1745                 if (err < 0)
1746                         return err;
1747         } else
1748                 memset(linkinfo, 0, sizeof(linkinfo));
1749
1750         if (linkinfo[IFLA_INFO_KIND]) {
1751                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1752                 ops = rtnl_link_ops_get(kind);
1753         } else {
1754                 kind[0] = '\0';
1755                 ops = NULL;
1756         }
1757
1758         if (1) {
1759                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1760                 struct net *dest_net;
1761
1762                 if (ops) {
1763                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1764                                 err = nla_parse_nested(attr, ops->maxtype,
1765                                                        linkinfo[IFLA_INFO_DATA],
1766                                                        ops->policy);
1767                                 if (err < 0)
1768                                         return err;
1769                                 data = attr;
1770                         }
1771                         if (ops->validate) {
1772                                 err = ops->validate(tb, data);
1773                                 if (err < 0)
1774                                         return err;
1775                         }
1776                 }
1777
1778                 if (dev) {
1779                         int modified = 0;
1780
1781                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1782                                 return -EEXIST;
1783                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1784                                 return -EOPNOTSUPP;
1785
1786                         if (linkinfo[IFLA_INFO_DATA]) {
1787                                 if (!ops || ops != dev->rtnl_link_ops ||
1788                                     !ops->changelink)
1789                                         return -EOPNOTSUPP;
1790
1791                                 err = ops->changelink(dev, tb, data);
1792                                 if (err < 0)
1793                                         return err;
1794                                 modified = 1;
1795                         }
1796
1797                         return do_setlink(dev, ifm, tb, ifname, modified);
1798                 }
1799
1800                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1801                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1802                                 return rtnl_group_changelink(net,
1803                                                 nla_get_u32(tb[IFLA_GROUP]),
1804                                                 ifm, tb);
1805                         return -ENODEV;
1806                 }
1807
1808                 if (ifm->ifi_index)
1809                         return -EOPNOTSUPP;
1810                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1811                         return -EOPNOTSUPP;
1812
1813                 if (!ops) {
1814 #ifdef CONFIG_MODULES
1815                         if (kind[0]) {
1816                                 __rtnl_unlock();
1817                                 request_module("rtnl-link-%s", kind);
1818                                 rtnl_lock();
1819                                 ops = rtnl_link_ops_get(kind);
1820                                 if (ops)
1821                                         goto replay;
1822                         }
1823 #endif
1824                         return -EOPNOTSUPP;
1825                 }
1826
1827                 if (!ifname[0])
1828                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1829
1830                 dest_net = rtnl_link_get_net(net, tb);
1831                 if (IS_ERR(dest_net))
1832                         return PTR_ERR(dest_net);
1833
1834                 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1835
1836                 if (IS_ERR(dev))
1837                         err = PTR_ERR(dev);
1838                 else if (ops->newlink)
1839                         err = ops->newlink(net, dev, tb, data);
1840                 else
1841                         err = register_netdevice(dev);
1842
1843                 if (err < 0 && !IS_ERR(dev))
1844                         free_netdev(dev);
1845                 if (err < 0)
1846                         goto out;
1847
1848                 err = rtnl_configure_link(dev, ifm);
1849                 if (err < 0)
1850                         unregister_netdevice(dev);
1851 out:
1852                 put_net(dest_net);
1853                 return err;
1854         }
1855 }
1856
1857 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1858 {
1859         struct net *net = sock_net(skb->sk);
1860         struct ifinfomsg *ifm;
1861         char ifname[IFNAMSIZ];
1862         struct nlattr *tb[IFLA_MAX+1];
1863         struct net_device *dev = NULL;
1864         struct sk_buff *nskb;
1865         int err;
1866         u32 ext_filter_mask = 0;
1867
1868         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1869         if (err < 0)
1870                 return err;
1871
1872         if (tb[IFLA_IFNAME])
1873                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1874
1875         if (tb[IFLA_EXT_MASK])
1876                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1877
1878         ifm = nlmsg_data(nlh);
1879         if (ifm->ifi_index > 0)
1880                 dev = __dev_get_by_index(net, ifm->ifi_index);
1881         else if (tb[IFLA_IFNAME])
1882                 dev = __dev_get_by_name(net, ifname);
1883         else
1884                 return -EINVAL;
1885
1886         if (dev == NULL)
1887                 return -ENODEV;
1888
1889         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1890         if (nskb == NULL)
1891                 return -ENOBUFS;
1892
1893         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1894                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1895         if (err < 0) {
1896                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1897                 WARN_ON(err == -EMSGSIZE);
1898                 kfree_skb(nskb);
1899         } else
1900                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1901
1902         return err;
1903 }
1904
1905 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1906 {
1907         struct net *net = sock_net(skb->sk);
1908         struct net_device *dev;
1909         struct nlattr *tb[IFLA_MAX+1];
1910         u32 ext_filter_mask = 0;
1911         u16 min_ifinfo_dump_size = 0;
1912
1913         if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1914                         ifla_policy) >= 0) {
1915                 if (tb[IFLA_EXT_MASK])
1916                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1917         }
1918
1919         if (!ext_filter_mask)
1920                 return NLMSG_GOODSIZE;
1921         /*
1922          * traverse the list of net devices and compute the minimum
1923          * buffer size based upon the filter mask.
1924          */
1925         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1926                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1927                                              if_nlmsg_size(dev,
1928                                                            ext_filter_mask));
1929         }
1930
1931         return min_ifinfo_dump_size;
1932 }
1933
1934 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1935 {
1936         int idx;
1937         int s_idx = cb->family;
1938
1939         if (s_idx == 0)
1940                 s_idx = 1;
1941         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1942                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1943                 if (idx < s_idx || idx == PF_PACKET)
1944                         continue;
1945                 if (rtnl_msg_handlers[idx] == NULL ||
1946                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1947                         continue;
1948                 if (idx > s_idx)
1949                         memset(&cb->args[0], 0, sizeof(cb->args));
1950                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1951                         break;
1952         }
1953         cb->family = idx;
1954
1955         return skb->len;
1956 }
1957
1958 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1959 {
1960         struct net *net = dev_net(dev);
1961         struct sk_buff *skb;
1962         int err = -ENOBUFS;
1963         size_t if_info_size;
1964
1965         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1966         if (skb == NULL)
1967                 goto errout;
1968
1969         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1970         if (err < 0) {
1971                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1972                 WARN_ON(err == -EMSGSIZE);
1973                 kfree_skb(skb);
1974                 goto errout;
1975         }
1976         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1977         return;
1978 errout:
1979         if (err < 0)
1980                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1981 }
1982
1983 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1984 {
1985         struct net *net = sock_net(skb->sk);
1986         struct net_device *master = NULL;
1987         struct ndmsg *ndm;
1988         struct nlattr *tb[NDA_MAX+1];
1989         struct net_device *dev;
1990         u8 *addr;
1991         int err;
1992
1993         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1994         if (err < 0)
1995                 return err;
1996
1997         ndm = nlmsg_data(nlh);
1998         if (ndm->ndm_ifindex == 0) {
1999                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2000                 return -EINVAL;
2001         }
2002
2003         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2004         if (dev == NULL) {
2005                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2006                 return -ENODEV;
2007         }
2008
2009         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2010                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2011                 return -EINVAL;
2012         }
2013
2014         addr = nla_data(tb[NDA_LLADDR]);
2015         if (!is_valid_ether_addr(addr)) {
2016                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2017                 return -EINVAL;
2018         }
2019
2020         err = -EOPNOTSUPP;
2021
2022         /* Support fdb on master device the net/bridge default case */
2023         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2024             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2025                 master = dev->master;
2026                 err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2027                                                       nlh->nlmsg_flags);
2028                 if (err)
2029                         goto out;
2030                 else
2031                         ndm->ndm_flags &= ~NTF_MASTER;
2032         }
2033
2034         /* Embedded bridge, macvlan, and any other device support */
2035         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2036                 err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2037                                                    nlh->nlmsg_flags);
2038
2039                 if (!err)
2040                         ndm->ndm_flags &= ~NTF_SELF;
2041         }
2042 out:
2043         return err;
2044 }
2045
2046 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2047 {
2048         struct net *net = sock_net(skb->sk);
2049         struct ndmsg *ndm;
2050         struct nlattr *llattr;
2051         struct net_device *dev;
2052         int err = -EINVAL;
2053         __u8 *addr;
2054
2055         if (nlmsg_len(nlh) < sizeof(*ndm))
2056                 return -EINVAL;
2057
2058         ndm = nlmsg_data(nlh);
2059         if (ndm->ndm_ifindex == 0) {
2060                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2061                 return -EINVAL;
2062         }
2063
2064         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2065         if (dev == NULL) {
2066                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2067                 return -ENODEV;
2068         }
2069
2070         llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2071         if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2072                 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2073                 return -EINVAL;
2074         }
2075
2076         addr = nla_data(llattr);
2077         err = -EOPNOTSUPP;
2078
2079         /* Support fdb on master device the net/bridge default case */
2080         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2081             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2082                 struct net_device *master = dev->master;
2083
2084                 if (master->netdev_ops->ndo_fdb_del)
2085                         err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2086
2087                 if (err)
2088                         goto out;
2089                 else
2090                         ndm->ndm_flags &= ~NTF_MASTER;
2091         }
2092
2093         /* Embedded bridge, macvlan, and any other device support */
2094         if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2095                 err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2096
2097                 if (!err)
2098                         ndm->ndm_flags &= ~NTF_SELF;
2099         }
2100 out:
2101         return err;
2102 }
2103
2104 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2105 {
2106         int idx = 0;
2107         struct net *net = sock_net(skb->sk);
2108         struct net_device *dev;
2109
2110         rcu_read_lock();
2111         for_each_netdev_rcu(net, dev) {
2112                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2113                         struct net_device *master = dev->master;
2114                         const struct net_device_ops *ops = master->netdev_ops;
2115
2116                         if (ops->ndo_fdb_dump)
2117                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2118                 }
2119
2120                 if (dev->netdev_ops->ndo_fdb_dump)
2121                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2122         }
2123         rcu_read_unlock();
2124
2125         cb->args[0] = idx;
2126         return skb->len;
2127 }
2128
2129 /* Protected by RTNL sempahore.  */
2130 static struct rtattr **rta_buf;
2131 static int rtattr_max;
2132
2133 /* Process one rtnetlink message. */
2134
2135 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2136 {
2137         struct net *net = sock_net(skb->sk);
2138         rtnl_doit_func doit;
2139         int sz_idx, kind;
2140         int min_len;
2141         int family;
2142         int type;
2143         int err;
2144
2145         type = nlh->nlmsg_type;
2146         if (type > RTM_MAX)
2147                 return -EOPNOTSUPP;
2148
2149         type -= RTM_BASE;
2150
2151         /* All the messages must have at least 1 byte length */
2152         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2153                 return 0;
2154
2155         family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2156         sz_idx = type>>2;
2157         kind = type&3;
2158
2159         if (kind != 2 && !capable(CAP_NET_ADMIN))
2160                 return -EPERM;
2161
2162         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2163                 struct sock *rtnl;
2164                 rtnl_dumpit_func dumpit;
2165                 rtnl_calcit_func calcit;
2166                 u16 min_dump_alloc = 0;
2167
2168                 dumpit = rtnl_get_dumpit(family, type);
2169                 if (dumpit == NULL)
2170                         return -EOPNOTSUPP;
2171                 calcit = rtnl_get_calcit(family, type);
2172                 if (calcit)
2173                         min_dump_alloc = calcit(skb, nlh);
2174
2175                 __rtnl_unlock();
2176                 rtnl = net->rtnl;
2177                 {
2178                         struct netlink_dump_control c = {
2179                                 .dump           = dumpit,
2180                                 .min_dump_alloc = min_dump_alloc,
2181                         };
2182                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2183                 }
2184                 rtnl_lock();
2185                 return err;
2186         }
2187
2188         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2189
2190         min_len = rtm_min[sz_idx];
2191         if (nlh->nlmsg_len < min_len)
2192                 return -EINVAL;
2193
2194         if (nlh->nlmsg_len > min_len) {
2195                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2196                 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2197
2198                 while (RTA_OK(attr, attrlen)) {
2199                         unsigned int flavor = attr->rta_type;
2200                         if (flavor) {
2201                                 if (flavor > rta_max[sz_idx])
2202                                         return -EINVAL;
2203                                 rta_buf[flavor-1] = attr;
2204                         }
2205                         attr = RTA_NEXT(attr, attrlen);
2206                 }
2207         }
2208
2209         doit = rtnl_get_doit(family, type);
2210         if (doit == NULL)
2211                 return -EOPNOTSUPP;
2212
2213         return doit(skb, nlh, (void *)&rta_buf[0]);
2214 }
2215
2216 static void rtnetlink_rcv(struct sk_buff *skb)
2217 {
2218         rtnl_lock();
2219         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2220         rtnl_unlock();
2221 }
2222
2223 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2224 {
2225         struct net_device *dev = ptr;
2226
2227         switch (event) {
2228         case NETDEV_UP:
2229         case NETDEV_DOWN:
2230         case NETDEV_PRE_UP:
2231         case NETDEV_POST_INIT:
2232         case NETDEV_REGISTER:
2233         case NETDEV_CHANGE:
2234         case NETDEV_PRE_TYPE_CHANGE:
2235         case NETDEV_GOING_DOWN:
2236         case NETDEV_UNREGISTER:
2237         case NETDEV_UNREGISTER_BATCH:
2238         case NETDEV_RELEASE:
2239         case NETDEV_JOIN:
2240                 break;
2241         default:
2242                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2243                 break;
2244         }
2245         return NOTIFY_DONE;
2246 }
2247
2248 static struct notifier_block rtnetlink_dev_notifier = {
2249         .notifier_call  = rtnetlink_event,
2250 };
2251
2252
2253 static int __net_init rtnetlink_net_init(struct net *net)
2254 {
2255         struct sock *sk;
2256         sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2257                                    rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2258         if (!sk)
2259                 return -ENOMEM;
2260         net->rtnl = sk;
2261         return 0;
2262 }
2263
2264 static void __net_exit rtnetlink_net_exit(struct net *net)
2265 {
2266         netlink_kernel_release(net->rtnl);
2267         net->rtnl = NULL;
2268 }
2269
2270 static struct pernet_operations rtnetlink_net_ops = {
2271         .init = rtnetlink_net_init,
2272         .exit = rtnetlink_net_exit,
2273 };
2274
2275 void __init rtnetlink_init(void)
2276 {
2277         int i;
2278
2279         rtattr_max = 0;
2280         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2281                 if (rta_max[i] > rtattr_max)
2282                         rtattr_max = rta_max[i];
2283         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2284         if (!rta_buf)
2285                 panic("rtnetlink_init: cannot allocate rta_buf\n");
2286
2287         if (register_pernet_subsys(&rtnetlink_net_ops))
2288                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2289
2290         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2291         register_netdevice_notifier(&rtnetlink_dev_notifier);
2292
2293         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2294                       rtnl_dump_ifinfo, rtnl_calcit);
2295         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2296         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2297         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2298
2299         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2300         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2301
2302         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2303         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2304         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2305 }
2306