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/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[msgindex].doit;
 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[msgindex].dumpit;
 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[msgindex].calcit;
 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  * Returns 0 on success or a negative error code.
 410  */
 411 void rtnl_af_register(struct rtnl_af_ops *ops)
 412 {
 413         rtnl_lock();
 414         list_add_tail(&ops->list, &rtnl_af_ops);
 415         rtnl_unlock();
 416 }
 417 EXPORT_SYMBOL_GPL(rtnl_af_register);
 418
 419 /**
 420  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 421  * @ops: struct rtnl_af_ops * to unregister
 422  *
 423  * The caller must hold the rtnl_mutex.
 424  */
 425 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 426 {
 427         list_del(&ops->list);
 428 }
 429 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 430
 431 /**
 432  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 433  * @ops: struct rtnl_af_ops * to unregister
 434  */
 435 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 436 {
 437         rtnl_lock();
 438         __rtnl_af_unregister(ops);
 439         rtnl_unlock();
 440 }
 441 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 442
 443 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 444 {
 445         struct rtnl_af_ops *af_ops;
 446         size_t size;
 447
 448         /* IFLA_AF_SPEC */
 449         size = nla_total_size(sizeof(struct nlattr));
 450
 451         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 452                 if (af_ops->get_link_af_size) {
 453                         /* AF_* + nested data */
 454                         size += nla_total_size(sizeof(struct nlattr)) +
 455                                 af_ops->get_link_af_size(dev);
 456                 }
 457         }
 458
 459         return size;
 460 }
 461
 462 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 463 {
 464         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 465         struct nlattr *linkinfo, *data;
 466         int err = -EMSGSIZE;
 467
 468         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 469         if (linkinfo == NULL)
 470                 goto out;
 471
 472         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 473                 goto err_cancel_link;
 474         if (ops->fill_xstats) {
 475                 err = ops->fill_xstats(skb, dev);
 476                 if (err < 0)
 477                         goto err_cancel_link;
 478         }
 479         if (ops->fill_info) {
 480                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 481                 if (data == NULL) {
 482                         err = -EMSGSIZE;
 483                         goto err_cancel_link;
 484                 }
 485                 err = ops->fill_info(skb, dev);
 486                 if (err < 0)
 487                         goto err_cancel_data;
 488                 nla_nest_end(skb, data);
 489         }
 490
 491         nla_nest_end(skb, linkinfo);
 492         return 0;
 493
 494 err_cancel_data:
 495         nla_nest_cancel(skb, data);
 496 err_cancel_link:
 497         nla_nest_cancel(skb, linkinfo);
 498 out:
 499         return err;
 500 }
 501
 502 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 503 {
 504         struct sock *rtnl = net->rtnl;
 505         int err = 0;
 506
 507         NETLINK_CB(skb).dst_group = group;
 508         if (echo)
 509                 atomic_inc(&skb->users);
 510         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 511         if (echo)
 512                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 513         return err;
 514 }
 515
 516 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 517 {
 518         struct sock *rtnl = net->rtnl;
 519
 520         return nlmsg_unicast(rtnl, skb, pid);
 521 }
 522 EXPORT_SYMBOL(rtnl_unicast);
 523
 524 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 525                  struct nlmsghdr *nlh, gfp_t flags)
 526 {
 527         struct sock *rtnl = net->rtnl;
 528         int report = 0;
 529
 530         if (nlh)
 531                 report = nlmsg_report(nlh);
 532
 533         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 534 }
 535 EXPORT_SYMBOL(rtnl_notify);
 536
 537 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 538 {
 539         struct sock *rtnl = net->rtnl;
 540
 541         netlink_set_err(rtnl, 0, group, error);
 542 }
 543 EXPORT_SYMBOL(rtnl_set_sk_err);
 544
 545 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 546 {
 547         struct nlattr *mx;
 548         int i, valid = 0;
 549
 550         mx = nla_nest_start(skb, RTA_METRICS);
 551         if (mx == NULL)
 552                 return -ENOBUFS;
 553
 554         for (i = 0; i < RTAX_MAX; i++) {
 555                 if (metrics[i]) {
 556                         valid++;
 557                         if (nla_put_u32(skb, i+1, metrics[i]))
 558                                 goto nla_put_failure;
 559                 }
 560         }
 561
 562         if (!valid) {
 563                 nla_nest_cancel(skb, mx);
 564                 return 0;
 565         }
 566
 567         return nla_nest_end(skb, mx);
 568
 569 nla_put_failure:
 570         nla_nest_cancel(skb, mx);
 571         return -EMSGSIZE;
 572 }
 573 EXPORT_SYMBOL(rtnetlink_put_metrics);
 574
 575 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 576                        long expires, u32 error)
 577 {
 578         struct rta_cacheinfo ci = {
 579                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 580                 .rta_used = dst->__use,
 581                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 582                 .rta_error = error,
 583                 .rta_id =  id,
 584         };
 585
 586         if (expires) {
 587                 unsigned long clock;
 588
 589                 clock = jiffies_to_clock_t(abs(expires));
 590                 clock = min_t(unsigned long, clock, INT_MAX);
 591                 ci.rta_expires = (expires > 0) ? clock : -clock;
 592         }
 593         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 594 }
 595 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 596
 597 static void set_operstate(struct net_device *dev, unsigned char transition)
 598 {
 599         unsigned char operstate = dev->operstate;
 600
 601         switch (transition) {
 602         case IF_OPER_UP:
 603                 if ((operstate == IF_OPER_DORMANT ||
 604                      operstate == IF_OPER_UNKNOWN) &&
 605                     !netif_dormant(dev))
 606                         operstate = IF_OPER_UP;
 607                 break;
 608
 609         case IF_OPER_DORMANT:
 610                 if (operstate == IF_OPER_UP ||
 611                     operstate == IF_OPER_UNKNOWN)
 612                         operstate = IF_OPER_DORMANT;
 613                 break;
 614         }
 615
 616         if (dev->operstate != operstate) {
 617                 write_lock_bh(&dev_base_lock);
 618                 dev->operstate = operstate;
 619                 write_unlock_bh(&dev_base_lock);
 620                 netdev_state_change(dev);
 621         }
 622 }
 623
 624 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 625 {
 626         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 627                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 628 }
 629
 630 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 631                                            const struct ifinfomsg *ifm)
 632 {
 633         unsigned int flags = ifm->ifi_flags;
 634
 635         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 636         if (ifm->ifi_change)
 637                 flags = (flags & ifm->ifi_change) |
 638                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 639
 640         return flags;
 641 }
 642
 643 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 644                                  const struct rtnl_link_stats64 *b)
 645 {
 646         a->rx_packets = b->rx_packets;
 647         a->tx_packets = b->tx_packets;
 648         a->rx_bytes = b->rx_bytes;
 649         a->tx_bytes = b->tx_bytes;
 650         a->rx_errors = b->rx_errors;
 651         a->tx_errors = b->tx_errors;
 652         a->rx_dropped = b->rx_dropped;
 653         a->tx_dropped = b->tx_dropped;
 654
 655         a->multicast = b->multicast;
 656         a->collisions = b->collisions;
 657
 658         a->rx_length_errors = b->rx_length_errors;
 659         a->rx_over_errors = b->rx_over_errors;
 660         a->rx_crc_errors = b->rx_crc_errors;
 661         a->rx_frame_errors = b->rx_frame_errors;
 662         a->rx_fifo_errors = b->rx_fifo_errors;
 663         a->rx_missed_errors = b->rx_missed_errors;
 664
 665         a->tx_aborted_errors = b->tx_aborted_errors;
 666         a->tx_carrier_errors = b->tx_carrier_errors;
 667         a->tx_fifo_errors = b->tx_fifo_errors;
 668         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 669         a->tx_window_errors = b->tx_window_errors;
 670
 671         a->rx_compressed = b->rx_compressed;
 672         a->tx_compressed = b->tx_compressed;
 673 }
 674
 675 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 676 {
 677         memcpy(v, b, sizeof(*b));
 678 }
 679
 680 /* All VF info */
 681 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 682                                    u32 ext_filter_mask)
 683 {
 684         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 685             (ext_filter_mask & RTEXT_FILTER_VF)) {
 686                 int num_vfs = dev_num_vf(dev->dev.parent);
 687                 size_t size = nla_total_size(sizeof(struct nlattr));
 688                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 689                 size += num_vfs *
 690                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 691                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 692                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 693                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 694                 return size;
 695         } else
 696                 return 0;
 697 }
 698
 699 static size_t rtnl_port_size(const struct net_device *dev)
 700 {
 701         size_t port_size = nla_total_size(4)            /* PORT_VF */
 702                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 703                 + nla_total_size(sizeof(struct ifla_port_vsi))
 704                                                         /* PORT_VSI_TYPE */
 705                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 706                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 707                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 708                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 709         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 710         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 711                 + port_size;
 712         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 713                 + port_size;
 714
 715         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 716                 return 0;
 717         if (dev_num_vf(dev->dev.parent))
 718                 return port_self_size + vf_ports_size +
 719                         vf_port_size * dev_num_vf(dev->dev.parent);
 720         else
 721                 return port_self_size;
 722 }
 723
 724 static size_t rtnl_bond_slave_size(const struct net_device *dev)
 725 {
 726         struct net_device *bond;
 727         size_t slave_size =
 728                 nla_total_size(sizeof(struct nlattr)) + /* IFLA_BOND_SLAVE */
 729                 nla_total_size(1) +     /* IFLA_BOND_SLAVE_STATE */
 730                 nla_total_size(1) +     /* IFLA_BOND_SLAVE_MII_STATUS */
 731                 nla_total_size(4) +     /* IFLA_BOND_SLAVE_LINK_FAILURE_COUNT */
 732                 nla_total_size(MAX_ADDR_LEN) +  /* IFLA_BOND_SLAVE_PERM_HWADDR */
 733                 nla_total_size(2) +     /* IFLA_BOND_SLAVE_QUEUE_ID */
 734                 nla_total_size(2) +     /* IFLA_BOND_SLAVE_AD_AGGREGATOR_ID */
 735                 0;
 736
 737         if (netif_is_bond_slave((struct net_device *)dev)) {
 738                 bond = netdev_master_upper_dev_get((struct net_device *)dev);
 739                 if (bond && bond->netdev_ops->ndo_get_slave)
 740                         return slave_size;
 741         }
 742
 743         return 0;
 744 }
 745
 746 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 747                                      u32 ext_filter_mask)
 748 {
 749         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 750                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 751                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 752                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 753                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 754                + nla_total_size(sizeof(struct rtnl_link_stats))
 755                + nla_total_size(sizeof(struct rtnl_link_stats64))
 756                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 757                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 758                + nla_total_size(4) /* IFLA_TXQLEN */
 759                + nla_total_size(4) /* IFLA_WEIGHT */
 760                + nla_total_size(4) /* IFLA_MTU */
 761                + nla_total_size(4) /* IFLA_LINK */
 762                + nla_total_size(4) /* IFLA_MASTER */
 763                + nla_total_size(1) /* IFLA_CARRIER */
 764                + nla_total_size(4) /* IFLA_PROMISCUITY */
 765                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 766                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 767                + nla_total_size(1) /* IFLA_OPERSTATE */
 768                + nla_total_size(1) /* IFLA_LINKMODE */
 769                + nla_total_size(ext_filter_mask
 770                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 771                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 772                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 773                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 774                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 775                + rtnl_bond_slave_size(dev) /* IFLA_SLAVE */
 776                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 777 }
 778
 779 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 780 {
 781         struct nlattr *vf_ports;
 782         struct nlattr *vf_port;
 783         int vf;
 784         int err;
 785
 786         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 787         if (!vf_ports)
 788                 return -EMSGSIZE;
 789
 790         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 791                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 792                 if (!vf_port)
 793                         goto nla_put_failure;
 794                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 795                         goto nla_put_failure;
 796                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 797                 if (err == -EMSGSIZE)
 798                         goto nla_put_failure;
 799                 if (err) {
 800                         nla_nest_cancel(skb, vf_port);
 801                         continue;
 802                 }
 803                 nla_nest_end(skb, vf_port);
 804         }
 805
 806         nla_nest_end(skb, vf_ports);
 807
 808         return 0;
 809
 810 nla_put_failure:
 811         nla_nest_cancel(skb, vf_ports);
 812         return -EMSGSIZE;
 813 }
 814
 815 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 816 {
 817         struct nlattr *port_self;
 818         int err;
 819
 820         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 821         if (!port_self)
 822                 return -EMSGSIZE;
 823
 824         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 825         if (err) {
 826                 nla_nest_cancel(skb, port_self);
 827                 return (err == -EMSGSIZE) ? err : 0;
 828         }
 829
 830         nla_nest_end(skb, port_self);
 831
 832         return 0;
 833 }
 834
 835 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 836 {
 837         int err;
 838
 839         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 840                 return 0;
 841
 842         err = rtnl_port_self_fill(skb, dev);
 843         if (err)
 844                 return err;
 845
 846         if (dev_num_vf(dev->dev.parent)) {
 847                 err = rtnl_vf_ports_fill(skb, dev);
 848                 if (err)
 849                         return err;
 850         }
 851
 852         return 0;
 853 }
 854
 855 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 856 {
 857         int err;
 858         struct netdev_phys_port_id ppid;
 859
 860         err = dev_get_phys_port_id(dev, &ppid);
 861         if (err) {
 862                 if (err == -EOPNOTSUPP)
 863                         return 0;
 864                 return err;
 865         }
 866
 867         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 868                 return -EMSGSIZE;
 869
 870         return 0;
 871 }
 872
 873 static size_t rtnl_bond_slave_fill(struct sk_buff *skb, struct net_device *dev)
 874 {
 875         struct net_device *bond;
 876         struct nlattr *nest;
 877         int err;
 878
 879         if (!netif_is_bond_slave(dev))
 880                 return 0;
 881
 882         bond = netdev_master_upper_dev_get(dev);
 883         if (!bond || !bond->netdev_ops->ndo_get_slave)
 884                 return 0;
 885
 886         nest = nla_nest_start(skb, IFLA_BOND_SLAVE);
 887         if (!nest)
 888                 return -EMSGSIZE;
 889
 890         err = bond->netdev_ops->ndo_get_slave(dev, skb);
 891         if (err) {
 892                 nla_nest_cancel(skb, nest);
 893                 return (err == -EMSGSIZE) ? err : 0;
 894         }
 895
 896         nla_nest_end(skb, nest);
 897
 898         return 0;
 899 }
 900
 901 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 902                             int type, u32 pid, u32 seq, u32 change,
 903                             unsigned int flags, u32 ext_filter_mask)
 904 {
 905         struct ifinfomsg *ifm;
 906         struct nlmsghdr *nlh;
 907         struct rtnl_link_stats64 temp;
 908         const struct rtnl_link_stats64 *stats;
 909         struct nlattr *attr, *af_spec;
 910         struct rtnl_af_ops *af_ops;
 911         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 912
 913         ASSERT_RTNL();
 914         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 915         if (nlh == NULL)
 916                 return -EMSGSIZE;
 917
 918         ifm = nlmsg_data(nlh);
 919         ifm->ifi_family = AF_UNSPEC;
 920         ifm->__ifi_pad = 0;
 921         ifm->ifi_type = dev->type;
 922         ifm->ifi_index = dev->ifindex;
 923         ifm->ifi_flags = dev_get_flags(dev);
 924         ifm->ifi_change = change;
 925
 926         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 927             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 928             nla_put_u8(skb, IFLA_OPERSTATE,
 929                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 930             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 931             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 932             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 933             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 934             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 935 #ifdef CONFIG_RPS
 936             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 937 #endif
 938             (dev->ifindex != dev->iflink &&
 939              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 940             (upper_dev &&
 941              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
 942             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
 943             (dev->qdisc &&
 944              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 945             (dev->ifalias &&
 946              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 947                 goto nla_put_failure;
 948
 949         if (1) {
 950                 struct rtnl_link_ifmap map = {
 951                         .mem_start   = dev->mem_start,
 952                         .mem_end     = dev->mem_end,
 953                         .base_addr   = dev->base_addr,
 954                         .irq         = dev->irq,
 955                         .dma         = dev->dma,
 956                         .port        = dev->if_port,
 957                 };
 958                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 959                         goto nla_put_failure;
 960         }
 961
 962         if (dev->addr_len) {
 963                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 964                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 965                         goto nla_put_failure;
 966         }
 967
 968         if (rtnl_phys_port_id_fill(skb, dev))
 969                 goto nla_put_failure;
 970
 971         attr = nla_reserve(skb, IFLA_STATS,
 972                         sizeof(struct rtnl_link_stats));
 973         if (attr == NULL)
 974                 goto nla_put_failure;
 975
 976         stats = dev_get_stats(dev, &temp);
 977         copy_rtnl_link_stats(nla_data(attr), stats);
 978
 979         attr = nla_reserve(skb, IFLA_STATS64,
 980                         sizeof(struct rtnl_link_stats64));
 981         if (attr == NULL)
 982                 goto nla_put_failure;
 983         copy_rtnl_link_stats64(nla_data(attr), stats);
 984
 985         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 986             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 987                 goto nla_put_failure;
 988
 989         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 990             && (ext_filter_mask & RTEXT_FILTER_VF)) {
 991                 int i;
 992
 993                 struct nlattr *vfinfo, *vf;
 994                 int num_vfs = dev_num_vf(dev->dev.parent);
 995
 996                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 997                 if (!vfinfo)
 998                         goto nla_put_failure;
 999                 for (i = 0; i < num_vfs; i++) {
1000                         struct ifla_vf_info ivi;
1001                         struct ifla_vf_mac vf_mac;
1002                         struct ifla_vf_vlan vf_vlan;
1003                         struct ifla_vf_tx_rate vf_tx_rate;
1004                         struct ifla_vf_spoofchk vf_spoofchk;
1005                         struct ifla_vf_link_state vf_linkstate;
1006
1007                         /*
1008                          * Not all SR-IOV capable drivers support the
1009                          * spoofcheck query.  Preset to -1 so the user
1010                          * space tool can detect that the driver didn't
1011                          * report anything.
1012                          */
1013                         ivi.spoofchk = -1;
1014                         memset(ivi.mac, 0, sizeof(ivi.mac));
1015                         /* The default value for VF link state is "auto"
1016                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1017                          */
1018                         ivi.linkstate = 0;
1019                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1020                                 break;
1021                         vf_mac.vf =
1022                                 vf_vlan.vf =
1023                                 vf_tx_rate.vf =
1024                                 vf_spoofchk.vf =
1025                                 vf_linkstate.vf = ivi.vf;
1026
1027                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1028                         vf_vlan.vlan = ivi.vlan;
1029                         vf_vlan.qos = ivi.qos;
1030                         vf_tx_rate.rate = ivi.tx_rate;
1031                         vf_spoofchk.setting = ivi.spoofchk;
1032                         vf_linkstate.link_state = ivi.linkstate;
1033                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1034                         if (!vf) {
1035                                 nla_nest_cancel(skb, vfinfo);
1036                                 goto nla_put_failure;
1037                         }
1038                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1039                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1040                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1041                                     &vf_tx_rate) ||
1042                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1043                                     &vf_spoofchk) ||
1044                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1045                                     &vf_linkstate))
1046                                 goto nla_put_failure;
1047                         nla_nest_end(skb, vf);
1048                 }
1049                 nla_nest_end(skb, vfinfo);
1050         }
1051
1052         if (rtnl_port_fill(skb, dev))
1053                 goto nla_put_failure;
1054
1055         if (rtnl_bond_slave_fill(skb, dev))
1056                 goto nla_put_failure;
1057
1058         if (dev->rtnl_link_ops) {
1059                 if (rtnl_link_fill(skb, dev) < 0)
1060                         goto nla_put_failure;
1061         }
1062
1063         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1064                 goto nla_put_failure;
1065
1066         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1067                 if (af_ops->fill_link_af) {
1068                         struct nlattr *af;
1069                         int err;
1070
1071                         if (!(af = nla_nest_start(skb, af_ops->family)))
1072                                 goto nla_put_failure;
1073
1074                         err = af_ops->fill_link_af(skb, dev);
1075
1076                         /*
1077                          * Caller may return ENODATA to indicate that there
1078                          * was no data to be dumped. This is not an error, it
1079                          * means we should trim the attribute header and
1080                          * continue.
1081                          */
1082                         if (err == -ENODATA)
1083                                 nla_nest_cancel(skb, af);
1084                         else if (err < 0)
1085                                 goto nla_put_failure;
1086
1087                         nla_nest_end(skb, af);
1088                 }
1089         }
1090
1091         nla_nest_end(skb, af_spec);
1092
1093         return nlmsg_end(skb, nlh);
1094
1095 nla_put_failure:
1096         nlmsg_cancel(skb, nlh);
1097         return -EMSGSIZE;
1098 }
1099
1100 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1101 {
1102         struct net *net = sock_net(skb->sk);
1103         int h, s_h;
1104         int idx = 0, s_idx;
1105         struct net_device *dev;
1106         struct hlist_head *head;
1107         struct nlattr *tb[IFLA_MAX+1];
1108         u32 ext_filter_mask = 0;
1109
1110         s_h = cb->args[0];
1111         s_idx = cb->args[1];
1112
1113         rcu_read_lock();
1114         cb->seq = net->dev_base_seq;
1115
1116         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1117                         ifla_policy) >= 0) {
1118
1119                 if (tb[IFLA_EXT_MASK])
1120                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1121         }
1122
1123         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1124                 idx = 0;
1125                 head = &net->dev_index_head[h];
1126                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1127                         if (idx < s_idx)
1128                                 goto cont;
1129                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1130                                              NETLINK_CB(cb->skb).portid,
1131                                              cb->nlh->nlmsg_seq, 0,
1132                                              NLM_F_MULTI,
1133                                              ext_filter_mask) <= 0)
1134                                 goto out;
1135
1136                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1137 cont:
1138                         idx++;
1139                 }
1140         }
1141 out:
1142         rcu_read_unlock();
1143         cb->args[1] = idx;
1144         cb->args[0] = h;
1145
1146         return skb->len;
1147 }
1148
1149 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1150         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1151         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1152         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1153         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1154         [IFLA_MTU]              = { .type = NLA_U32 },
1155         [IFLA_LINK]             = { .type = NLA_U32 },
1156         [IFLA_MASTER]           = { .type = NLA_U32 },
1157         [IFLA_CARRIER]          = { .type = NLA_U8 },
1158         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1159         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1160         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1161         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1162         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1163         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1164         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1165         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1166         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1167         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1168         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1169         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1170         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1171         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1172         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1173         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1174         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1175 };
1176 EXPORT_SYMBOL(ifla_policy);
1177
1178 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1179         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1180         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1181 };
1182
1183 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1184         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1185 };
1186
1187 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1188         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1189                                     .len = sizeof(struct ifla_vf_mac) },
1190         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1191                                     .len = sizeof(struct ifla_vf_vlan) },
1192         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1193                                     .len = sizeof(struct ifla_vf_tx_rate) },
1194         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1195                                     .len = sizeof(struct ifla_vf_spoofchk) },
1196 };
1197
1198 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1199         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1200         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1201                                     .len = PORT_PROFILE_MAX },
1202         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1203                                     .len = sizeof(struct ifla_port_vsi)},
1204         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1205                                       .len = PORT_UUID_MAX },
1206         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1207                                     .len = PORT_UUID_MAX },
1208         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1209         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1210 };
1211
1212 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1213 {
1214         struct net *net;
1215         /* Examine the link attributes and figure out which
1216          * network namespace we are talking about.
1217          */
1218         if (tb[IFLA_NET_NS_PID])
1219                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1220         else if (tb[IFLA_NET_NS_FD])
1221                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1222         else
1223                 net = get_net(src_net);
1224         return net;
1225 }
1226 EXPORT_SYMBOL(rtnl_link_get_net);
1227
1228 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1229 {
1230         if (dev) {
1231                 if (tb[IFLA_ADDRESS] &&
1232                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1233                         return -EINVAL;
1234
1235                 if (tb[IFLA_BROADCAST] &&
1236                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1237                         return -EINVAL;
1238         }
1239
1240         if (tb[IFLA_AF_SPEC]) {
1241                 struct nlattr *af;
1242                 int rem, err;
1243
1244                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1245                         const struct rtnl_af_ops *af_ops;
1246
1247                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1248                                 return -EAFNOSUPPORT;
1249
1250                         if (!af_ops->set_link_af)
1251                                 return -EOPNOTSUPP;
1252
1253                         if (af_ops->validate_link_af) {
1254                                 err = af_ops->validate_link_af(dev, af);
1255                                 if (err < 0)
1256                                         return err;
1257                         }
1258                 }
1259         }
1260
1261         return 0;
1262 }
1263
1264 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1265 {
1266         int rem, err = -EINVAL;
1267         struct nlattr *vf;
1268         const struct net_device_ops *ops = dev->netdev_ops;
1269
1270         nla_for_each_nested(vf, attr, rem) {
1271                 switch (nla_type(vf)) {
1272                 case IFLA_VF_MAC: {
1273                         struct ifla_vf_mac *ivm;
1274                         ivm = nla_data(vf);
1275                         err = -EOPNOTSUPP;
1276                         if (ops->ndo_set_vf_mac)
1277                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1278                                                           ivm->mac);
1279                         break;
1280                 }
1281                 case IFLA_VF_VLAN: {
1282                         struct ifla_vf_vlan *ivv;
1283                         ivv = nla_data(vf);
1284                         err = -EOPNOTSUPP;
1285                         if (ops->ndo_set_vf_vlan)
1286                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1287                                                            ivv->vlan,
1288                                                            ivv->qos);
1289                         break;
1290                 }
1291                 case IFLA_VF_TX_RATE: {
1292                         struct ifla_vf_tx_rate *ivt;
1293                         ivt = nla_data(vf);
1294                         err = -EOPNOTSUPP;
1295                         if (ops->ndo_set_vf_tx_rate)
1296                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1297                                                               ivt->rate);
1298                         break;
1299                 }
1300                 case IFLA_VF_SPOOFCHK: {
1301                         struct ifla_vf_spoofchk *ivs;
1302                         ivs = nla_data(vf);
1303                         err = -EOPNOTSUPP;
1304                         if (ops->ndo_set_vf_spoofchk)
1305                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1306                                                                ivs->setting);
1307                         break;
1308                 }
1309                 case IFLA_VF_LINK_STATE: {
1310                         struct ifla_vf_link_state *ivl;
1311                         ivl = nla_data(vf);
1312                         err = -EOPNOTSUPP;
1313                         if (ops->ndo_set_vf_link_state)
1314                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1315                                                                  ivl->link_state);
1316                         break;
1317                 }
1318                 default:
1319                         err = -EINVAL;
1320                         break;
1321                 }
1322                 if (err)
1323                         break;
1324         }
1325         return err;
1326 }
1327
1328 static int do_set_master(struct net_device *dev, int ifindex)
1329 {
1330         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1331         const struct net_device_ops *ops;
1332         int err;
1333
1334         if (upper_dev) {
1335                 if (upper_dev->ifindex == ifindex)
1336                         return 0;
1337                 ops = upper_dev->netdev_ops;
1338                 if (ops->ndo_del_slave) {
1339                         err = ops->ndo_del_slave(upper_dev, dev);
1340                         if (err)
1341                                 return err;
1342                 } else {
1343                         return -EOPNOTSUPP;
1344                 }
1345         }
1346
1347         if (ifindex) {
1348                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1349                 if (!upper_dev)
1350                         return -EINVAL;
1351                 ops = upper_dev->netdev_ops;
1352                 if (ops->ndo_add_slave) {
1353                         err = ops->ndo_add_slave(upper_dev, dev);
1354                         if (err)
1355                                 return err;
1356                 } else {
1357                         return -EOPNOTSUPP;
1358                 }
1359         }
1360         return 0;
1361 }
1362
1363 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1364                       struct nlattr **tb, char *ifname, int modified)
1365 {
1366         const struct net_device_ops *ops = dev->netdev_ops;
1367         int err;
1368
1369         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1370                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1371                 if (IS_ERR(net)) {
1372                         err = PTR_ERR(net);
1373                         goto errout;
1374                 }
1375                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1376                         err = -EPERM;
1377                         goto errout;
1378                 }
1379                 err = dev_change_net_namespace(dev, net, ifname);
1380                 put_net(net);
1381                 if (err)
1382                         goto errout;
1383                 modified = 1;
1384         }
1385
1386         if (tb[IFLA_MAP]) {
1387                 struct rtnl_link_ifmap *u_map;
1388                 struct ifmap k_map;
1389
1390                 if (!ops->ndo_set_config) {
1391                         err = -EOPNOTSUPP;
1392                         goto errout;
1393                 }
1394
1395                 if (!netif_device_present(dev)) {
1396                         err = -ENODEV;
1397                         goto errout;
1398                 }
1399
1400                 u_map = nla_data(tb[IFLA_MAP]);
1401                 k_map.mem_start = (unsigned long) u_map->mem_start;
1402                 k_map.mem_end = (unsigned long) u_map->mem_end;
1403                 k_map.base_addr = (unsigned short) u_map->base_addr;
1404                 k_map.irq = (unsigned char) u_map->irq;
1405                 k_map.dma = (unsigned char) u_map->dma;
1406                 k_map.port = (unsigned char) u_map->port;
1407
1408                 err = ops->ndo_set_config(dev, &k_map);
1409                 if (err < 0)
1410                         goto errout;
1411
1412                 modified = 1;
1413         }
1414
1415         if (tb[IFLA_ADDRESS]) {
1416                 struct sockaddr *sa;
1417                 int len;
1418
1419                 len = sizeof(sa_family_t) + dev->addr_len;
1420                 sa = kmalloc(len, GFP_KERNEL);
1421                 if (!sa) {
1422                         err = -ENOMEM;
1423                         goto errout;
1424                 }
1425                 sa->sa_family = dev->type;
1426                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1427                        dev->addr_len);
1428                 err = dev_set_mac_address(dev, sa);
1429                 kfree(sa);
1430                 if (err)
1431                         goto errout;
1432                 modified = 1;
1433         }
1434
1435         if (tb[IFLA_MTU]) {
1436                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1437                 if (err < 0)
1438                         goto errout;
1439                 modified = 1;
1440         }
1441
1442         if (tb[IFLA_GROUP]) {
1443                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1444                 modified = 1;
1445         }
1446
1447         /*
1448          * Interface selected by interface index but interface
1449          * name provided implies that a name change has been
1450          * requested.
1451          */
1452         if (ifm->ifi_index > 0 && ifname[0]) {
1453                 err = dev_change_name(dev, ifname);
1454                 if (err < 0)
1455                         goto errout;
1456                 modified = 1;
1457         }
1458
1459         if (tb[IFLA_IFALIAS]) {
1460                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1461                                     nla_len(tb[IFLA_IFALIAS]));
1462                 if (err < 0)
1463                         goto errout;
1464                 modified = 1;
1465         }
1466
1467         if (tb[IFLA_BROADCAST]) {
1468                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1469                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1470         }
1471
1472         if (ifm->ifi_flags || ifm->ifi_change) {
1473                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1474                 if (err < 0)
1475                         goto errout;
1476         }
1477
1478         if (tb[IFLA_MASTER]) {
1479                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1480                 if (err)
1481                         goto errout;
1482                 modified = 1;
1483         }
1484
1485         if (tb[IFLA_CARRIER]) {
1486                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1487                 if (err)
1488                         goto errout;
1489                 modified = 1;
1490         }
1491
1492         if (tb[IFLA_TXQLEN])
1493                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1494
1495         if (tb[IFLA_OPERSTATE])
1496                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1497
1498         if (tb[IFLA_LINKMODE]) {
1499                 write_lock_bh(&dev_base_lock);
1500                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1501                 write_unlock_bh(&dev_base_lock);
1502         }
1503
1504         if (tb[IFLA_VFINFO_LIST]) {
1505                 struct nlattr *attr;
1506                 int rem;
1507                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1508                         if (nla_type(attr) != IFLA_VF_INFO) {
1509                                 err = -EINVAL;
1510                                 goto errout;
1511                         }
1512                         err = do_setvfinfo(dev, attr);
1513                         if (err < 0)
1514                                 goto errout;
1515                         modified = 1;
1516                 }
1517         }
1518         err = 0;
1519
1520         if (tb[IFLA_VF_PORTS]) {
1521                 struct nlattr *port[IFLA_PORT_MAX+1];
1522                 struct nlattr *attr;
1523                 int vf;
1524                 int rem;
1525
1526                 err = -EOPNOTSUPP;
1527                 if (!ops->ndo_set_vf_port)
1528                         goto errout;
1529
1530                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1531                         if (nla_type(attr) != IFLA_VF_PORT)
1532                                 continue;
1533                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1534                                 attr, ifla_port_policy);
1535                         if (err < 0)
1536                                 goto errout;
1537                         if (!port[IFLA_PORT_VF]) {
1538                                 err = -EOPNOTSUPP;
1539                                 goto errout;
1540                         }
1541                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1542                         err = ops->ndo_set_vf_port(dev, vf, port);
1543                         if (err < 0)
1544                                 goto errout;
1545                         modified = 1;
1546                 }
1547         }
1548         err = 0;
1549
1550         if (tb[IFLA_PORT_SELF]) {
1551                 struct nlattr *port[IFLA_PORT_MAX+1];
1552
1553                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1554                         tb[IFLA_PORT_SELF], ifla_port_policy);
1555                 if (err < 0)
1556                         goto errout;
1557
1558                 err = -EOPNOTSUPP;
1559                 if (ops->ndo_set_vf_port)
1560                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1561                 if (err < 0)
1562                         goto errout;
1563                 modified = 1;
1564         }
1565
1566         if (tb[IFLA_AF_SPEC]) {
1567                 struct nlattr *af;
1568                 int rem;
1569
1570                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1571                         const struct rtnl_af_ops *af_ops;
1572
1573                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1574                                 BUG();
1575
1576                         err = af_ops->set_link_af(dev, af);
1577                         if (err < 0)
1578                                 goto errout;
1579
1580                         modified = 1;
1581                 }
1582         }
1583         err = 0;
1584
1585 errout:
1586         if (err < 0 && modified)
1587                 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1588                                      dev->name);
1589
1590         return err;
1591 }
1592
1593 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1594 {
1595         struct net *net = sock_net(skb->sk);
1596         struct ifinfomsg *ifm;
1597         struct net_device *dev;
1598         int err;
1599         struct nlattr *tb[IFLA_MAX+1];
1600         char ifname[IFNAMSIZ];
1601
1602         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1603         if (err < 0)
1604                 goto errout;
1605
1606         if (tb[IFLA_IFNAME])
1607                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1608         else
1609                 ifname[0] = '\0';
1610
1611         err = -EINVAL;
1612         ifm = nlmsg_data(nlh);
1613         if (ifm->ifi_index > 0)
1614                 dev = __dev_get_by_index(net, ifm->ifi_index);
1615         else if (tb[IFLA_IFNAME])
1616                 dev = __dev_get_by_name(net, ifname);
1617         else
1618                 goto errout;
1619
1620         if (dev == NULL) {
1621                 err = -ENODEV;
1622                 goto errout;
1623         }
1624
1625         err = validate_linkmsg(dev, tb);
1626         if (err < 0)
1627                 goto errout;
1628
1629         err = do_setlink(dev, ifm, tb, ifname, 0);
1630 errout:
1631         return err;
1632 }
1633
1634 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1635 {
1636         struct net *net = sock_net(skb->sk);
1637         const struct rtnl_link_ops *ops;
1638         struct net_device *dev;
1639         struct ifinfomsg *ifm;
1640         char ifname[IFNAMSIZ];
1641         struct nlattr *tb[IFLA_MAX+1];
1642         int err;
1643         LIST_HEAD(list_kill);
1644
1645         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1646         if (err < 0)
1647                 return err;
1648
1649         if (tb[IFLA_IFNAME])
1650                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1651
1652         ifm = nlmsg_data(nlh);
1653         if (ifm->ifi_index > 0)
1654                 dev = __dev_get_by_index(net, ifm->ifi_index);
1655         else if (tb[IFLA_IFNAME])
1656                 dev = __dev_get_by_name(net, ifname);
1657         else
1658                 return -EINVAL;
1659
1660         if (!dev)
1661                 return -ENODEV;
1662
1663         ops = dev->rtnl_link_ops;
1664         if (!ops)
1665                 return -EOPNOTSUPP;
1666
1667         ops->dellink(dev, &list_kill);
1668         unregister_netdevice_many(&list_kill);
1669         list_del(&list_kill);
1670         return 0;
1671 }
1672
1673 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1674 {
1675         unsigned int old_flags;
1676         int err;
1677
1678         old_flags = dev->flags;
1679         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1680                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1681                 if (err < 0)
1682                         return err;
1683         }
1684
1685         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1686
1687         __dev_notify_flags(dev, old_flags, ~0U);
1688         return 0;
1689 }
1690 EXPORT_SYMBOL(rtnl_configure_link);
1691
1692 struct net_device *rtnl_create_link(struct net *net,
1693         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1694 {
1695         int err;
1696         struct net_device *dev;
1697         unsigned int num_tx_queues = 1;
1698         unsigned int num_rx_queues = 1;
1699
1700         if (tb[IFLA_NUM_TX_QUEUES])
1701                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1702         else if (ops->get_num_tx_queues)
1703                 num_tx_queues = ops->get_num_tx_queues();
1704
1705         if (tb[IFLA_NUM_RX_QUEUES])
1706                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1707         else if (ops->get_num_rx_queues)
1708                 num_rx_queues = ops->get_num_rx_queues();
1709
1710         err = -ENOMEM;
1711         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1712                                num_tx_queues, num_rx_queues);
1713         if (!dev)
1714                 goto err;
1715
1716         dev_net_set(dev, net);
1717         dev->rtnl_link_ops = ops;
1718         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1719
1720         if (tb[IFLA_MTU])
1721                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1722         if (tb[IFLA_ADDRESS]) {
1723                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1724                                 nla_len(tb[IFLA_ADDRESS]));
1725                 dev->addr_assign_type = NET_ADDR_SET;
1726         }
1727         if (tb[IFLA_BROADCAST])
1728                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1729                                 nla_len(tb[IFLA_BROADCAST]));
1730         if (tb[IFLA_TXQLEN])
1731                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1732         if (tb[IFLA_OPERSTATE])
1733                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1734         if (tb[IFLA_LINKMODE])
1735                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1736         if (tb[IFLA_GROUP])
1737                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1738
1739         return dev;
1740
1741 err:
1742         return ERR_PTR(err);
1743 }
1744 EXPORT_SYMBOL(rtnl_create_link);
1745
1746 static int rtnl_group_changelink(struct net *net, int group,
1747                 struct ifinfomsg *ifm,
1748                 struct nlattr **tb)
1749 {
1750         struct net_device *dev;
1751         int err;
1752
1753         for_each_netdev(net, dev) {
1754                 if (dev->group == group) {
1755                         err = do_setlink(dev, ifm, tb, NULL, 0);
1756                         if (err < 0)
1757                                 return err;
1758                 }
1759         }
1760
1761         return 0;
1762 }
1763
1764 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1765 {
1766         struct net *net = sock_net(skb->sk);
1767         const struct rtnl_link_ops *ops;
1768         struct net_device *dev;
1769         struct ifinfomsg *ifm;
1770         char kind[MODULE_NAME_LEN];
1771         char ifname[IFNAMSIZ];
1772         struct nlattr *tb[IFLA_MAX+1];
1773         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1774         int err;
1775
1776 #ifdef CONFIG_MODULES
1777 replay:
1778 #endif
1779         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1780         if (err < 0)
1781                 return err;
1782
1783         if (tb[IFLA_IFNAME])
1784                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1785         else
1786                 ifname[0] = '\0';
1787
1788         ifm = nlmsg_data(nlh);
1789         if (ifm->ifi_index > 0)
1790                 dev = __dev_get_by_index(net, ifm->ifi_index);
1791         else {
1792                 if (ifname[0])
1793                         dev = __dev_get_by_name(net, ifname);
1794                 else
1795                         dev = NULL;
1796         }
1797
1798         err = validate_linkmsg(dev, tb);
1799         if (err < 0)
1800                 return err;
1801
1802         if (tb[IFLA_LINKINFO]) {
1803                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1804                                        tb[IFLA_LINKINFO], ifla_info_policy);
1805                 if (err < 0)
1806                         return err;
1807         } else
1808                 memset(linkinfo, 0, sizeof(linkinfo));
1809
1810         if (linkinfo[IFLA_INFO_KIND]) {
1811                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1812                 ops = rtnl_link_ops_get(kind);
1813         } else {
1814                 kind[0] = '\0';
1815                 ops = NULL;
1816         }
1817
1818         if (1) {
1819                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1820                 struct net *dest_net;
1821
1822                 if (ops) {
1823                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1824                                 err = nla_parse_nested(attr, ops->maxtype,
1825                                                        linkinfo[IFLA_INFO_DATA],
1826                                                        ops->policy);
1827                                 if (err < 0)
1828                                         return err;
1829                                 data = attr;
1830                         }
1831                         if (ops->validate) {
1832                                 err = ops->validate(tb, data);
1833                                 if (err < 0)
1834                                         return err;
1835                         }
1836                 }
1837
1838                 if (dev) {
1839                         int modified = 0;
1840
1841                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1842                                 return -EEXIST;
1843                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1844                                 return -EOPNOTSUPP;
1845
1846                         if (linkinfo[IFLA_INFO_DATA]) {
1847                                 if (!ops || ops != dev->rtnl_link_ops ||
1848                                     !ops->changelink)
1849                                         return -EOPNOTSUPP;
1850
1851                                 err = ops->changelink(dev, tb, data);
1852                                 if (err < 0)
1853                                         return err;
1854                                 modified = 1;
1855                         }
1856
1857                         return do_setlink(dev, ifm, tb, ifname, modified);
1858                 }
1859
1860                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1861                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1862                                 return rtnl_group_changelink(net,
1863                                                 nla_get_u32(tb[IFLA_GROUP]),
1864                                                 ifm, tb);
1865                         return -ENODEV;
1866                 }
1867
1868                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1869                         return -EOPNOTSUPP;
1870
1871                 if (!ops) {
1872 #ifdef CONFIG_MODULES
1873                         if (kind[0]) {
1874                                 __rtnl_unlock();
1875                                 request_module("rtnl-link-%s", kind);
1876                                 rtnl_lock();
1877                                 ops = rtnl_link_ops_get(kind);
1878                                 if (ops)
1879                                         goto replay;
1880                         }
1881 #endif
1882                         return -EOPNOTSUPP;
1883                 }
1884
1885                 if (!ifname[0])
1886                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1887
1888                 dest_net = rtnl_link_get_net(net, tb);
1889                 if (IS_ERR(dest_net))
1890                         return PTR_ERR(dest_net);
1891
1892                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1893                 if (IS_ERR(dev)) {
1894                         err = PTR_ERR(dev);
1895                         goto out;
1896                 }
1897
1898                 dev->ifindex = ifm->ifi_index;
1899
1900                 if (ops->newlink)
1901                         err = ops->newlink(net, dev, tb, data);
1902                 else
1903                         err = register_netdevice(dev);
1904
1905                 if (err < 0) {
1906                         free_netdev(dev);
1907                         goto out;
1908                 }
1909
1910                 err = rtnl_configure_link(dev, ifm);
1911                 if (err < 0)
1912                         unregister_netdevice(dev);
1913 out:
1914                 put_net(dest_net);
1915                 return err;
1916         }
1917 }
1918
1919 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1920 {
1921         struct net *net = sock_net(skb->sk);
1922         struct ifinfomsg *ifm;
1923         char ifname[IFNAMSIZ];
1924         struct nlattr *tb[IFLA_MAX+1];
1925         struct net_device *dev = NULL;
1926         struct sk_buff *nskb;
1927         int err;
1928         u32 ext_filter_mask = 0;
1929
1930         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1931         if (err < 0)
1932                 return err;
1933
1934         if (tb[IFLA_IFNAME])
1935                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1936
1937         if (tb[IFLA_EXT_MASK])
1938                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1939
1940         ifm = nlmsg_data(nlh);
1941         if (ifm->ifi_index > 0)
1942                 dev = __dev_get_by_index(net, ifm->ifi_index);
1943         else if (tb[IFLA_IFNAME])
1944                 dev = __dev_get_by_name(net, ifname);
1945         else
1946                 return -EINVAL;
1947
1948         if (dev == NULL)
1949                 return -ENODEV;
1950
1951         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1952         if (nskb == NULL)
1953                 return -ENOBUFS;
1954
1955         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1956                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1957         if (err < 0) {
1958                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1959                 WARN_ON(err == -EMSGSIZE);
1960                 kfree_skb(nskb);
1961         } else
1962                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1963
1964         return err;
1965 }
1966
1967 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1968 {
1969         struct net *net = sock_net(skb->sk);
1970         struct net_device *dev;
1971         struct nlattr *tb[IFLA_MAX+1];
1972         u32 ext_filter_mask = 0;
1973         u16 min_ifinfo_dump_size = 0;
1974
1975         if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1976                         ifla_policy) >= 0) {
1977                 if (tb[IFLA_EXT_MASK])
1978                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1979         }
1980
1981         if (!ext_filter_mask)
1982                 return NLMSG_GOODSIZE;
1983         /*
1984          * traverse the list of net devices and compute the minimum
1985          * buffer size based upon the filter mask.
1986          */
1987         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1988                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1989                                              if_nlmsg_size(dev,
1990                                                            ext_filter_mask));
1991         }
1992
1993         return min_ifinfo_dump_size;
1994 }
1995
1996 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1997 {
1998         int idx;
1999         int s_idx = cb->family;
2000
2001         if (s_idx == 0)
2002                 s_idx = 1;
2003         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2004                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2005                 if (idx < s_idx || idx == PF_PACKET)
2006                         continue;
2007                 if (rtnl_msg_handlers[idx] == NULL ||
2008                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2009                         continue;
2010                 if (idx > s_idx) {
2011                         memset(&cb->args[0], 0, sizeof(cb->args));
2012                         cb->prev_seq = 0;
2013                         cb->seq = 0;
2014                 }
2015                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2016                         break;
2017         }
2018         cb->family = idx;
2019
2020         return skb->len;
2021 }
2022
2023 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2024                   gfp_t flags)
2025 {
2026         struct net *net = dev_net(dev);
2027         struct sk_buff *skb;
2028         int err = -ENOBUFS;
2029         size_t if_info_size;
2030
2031         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2032         if (skb == NULL)
2033                 goto errout;
2034
2035         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2036         if (err < 0) {
2037                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2038                 WARN_ON(err == -EMSGSIZE);
2039                 kfree_skb(skb);
2040                 goto errout;
2041         }
2042         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2043         return;
2044 errout:
2045         if (err < 0)
2046                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2047 }
2048 EXPORT_SYMBOL(rtmsg_ifinfo);
2049
2050 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2051                                    struct net_device *dev,
2052                                    u8 *addr, u32 pid, u32 seq,
2053                                    int type, unsigned int flags)
2054 {
2055         struct nlmsghdr *nlh;
2056         struct ndmsg *ndm;
2057
2058         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2059         if (!nlh)
2060                 return -EMSGSIZE;
2061
2062         ndm = nlmsg_data(nlh);
2063         ndm->ndm_family  = AF_BRIDGE;
2064         ndm->ndm_pad1    = 0;
2065         ndm->ndm_pad2    = 0;
2066         ndm->ndm_flags   = flags;
2067         ndm->ndm_type    = 0;
2068         ndm->ndm_ifindex = dev->ifindex;
2069         ndm->ndm_state   = NUD_PERMANENT;
2070
2071         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2072                 goto nla_put_failure;
2073
2074         return nlmsg_end(skb, nlh);
2075
2076 nla_put_failure:
2077         nlmsg_cancel(skb, nlh);
2078         return -EMSGSIZE;
2079 }
2080
2081 static inline size_t rtnl_fdb_nlmsg_size(void)
2082 {
2083         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2084 }
2085
2086 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2087 {
2088         struct net *net = dev_net(dev);
2089         struct sk_buff *skb;
2090         int err = -ENOBUFS;
2091
2092         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2093         if (!skb)
2094                 goto errout;
2095
2096         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2097         if (err < 0) {
2098                 kfree_skb(skb);
2099                 goto errout;
2100         }
2101
2102         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2103         return;
2104 errout:
2105         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2106 }
2107
2108 /**
2109  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2110  */
2111 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2112                      struct nlattr *tb[],
2113                      struct net_device *dev,
2114                      const unsigned char *addr,
2115                      u16 flags)
2116 {
2117         int err = -EINVAL;
2118
2119         /* If aging addresses are supported device will need to
2120          * implement its own handler for this.
2121          */
2122         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2123                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2124                 return err;
2125         }
2126
2127         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2128                 err = dev_uc_add_excl(dev, addr);
2129         else if (is_multicast_ether_addr(addr))
2130                 err = dev_mc_add_excl(dev, addr);
2131
2132         /* Only return duplicate errors if NLM_F_EXCL is set */
2133         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2134                 err = 0;
2135
2136         return err;
2137 }
2138 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2139
2140 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2141 {
2142         struct net *net = sock_net(skb->sk);
2143         struct ndmsg *ndm;
2144         struct nlattr *tb[NDA_MAX+1];
2145         struct net_device *dev;
2146         u8 *addr;
2147         int err;
2148
2149         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2150         if (err < 0)
2151                 return err;
2152
2153         ndm = nlmsg_data(nlh);
2154         if (ndm->ndm_ifindex == 0) {
2155                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2156                 return -EINVAL;
2157         }
2158
2159         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2160         if (dev == NULL) {
2161                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2162                 return -ENODEV;
2163         }
2164
2165         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2166                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2167                 return -EINVAL;
2168         }
2169
2170         addr = nla_data(tb[NDA_LLADDR]);
2171
2172         err = -EOPNOTSUPP;
2173
2174         /* Support fdb on master device the net/bridge default case */
2175         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2176             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2177                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2178                 const struct net_device_ops *ops = br_dev->netdev_ops;
2179
2180                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2181                 if (err)
2182                         goto out;
2183                 else
2184                         ndm->ndm_flags &= ~NTF_MASTER;
2185         }
2186
2187         /* Embedded bridge, macvlan, and any other device support */
2188         if ((ndm->ndm_flags & NTF_SELF)) {
2189                 if (dev->netdev_ops->ndo_fdb_add)
2190                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2191                                                            nlh->nlmsg_flags);
2192                 else
2193                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2194                                                nlh->nlmsg_flags);
2195
2196                 if (!err) {
2197                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2198                         ndm->ndm_flags &= ~NTF_SELF;
2199                 }
2200         }
2201 out:
2202         return err;
2203 }
2204
2205 /**
2206  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2207  */
2208 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2209                      struct nlattr *tb[],
2210                      struct net_device *dev,
2211                      const unsigned char *addr)
2212 {
2213         int err = -EOPNOTSUPP;
2214
2215         /* If aging addresses are supported device will need to
2216          * implement its own handler for this.
2217          */
2218         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2219                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2220                 return -EINVAL;
2221         }
2222
2223         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2224                 err = dev_uc_del(dev, addr);
2225         else if (is_multicast_ether_addr(addr))
2226                 err = dev_mc_del(dev, addr);
2227         else
2228                 err = -EINVAL;
2229
2230         return err;
2231 }
2232 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2233
2234 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2235 {
2236         struct net *net = sock_net(skb->sk);
2237         struct ndmsg *ndm;
2238         struct nlattr *tb[NDA_MAX+1];
2239         struct net_device *dev;
2240         int err = -EINVAL;
2241         __u8 *addr;
2242
2243         if (!capable(CAP_NET_ADMIN))
2244                 return -EPERM;
2245
2246         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2247         if (err < 0)
2248                 return err;
2249
2250         ndm = nlmsg_data(nlh);
2251         if (ndm->ndm_ifindex == 0) {
2252                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2253                 return -EINVAL;
2254         }
2255
2256         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2257         if (dev == NULL) {
2258                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2259                 return -ENODEV;
2260         }
2261
2262         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2263                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2264                 return -EINVAL;
2265         }
2266
2267         addr = nla_data(tb[NDA_LLADDR]);
2268
2269         err = -EOPNOTSUPP;
2270
2271         /* Support fdb on master device the net/bridge default case */
2272         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2273             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2274                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2275                 const struct net_device_ops *ops = br_dev->netdev_ops;
2276
2277                 if (ops->ndo_fdb_del)
2278                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2279
2280                 if (err)
2281                         goto out;
2282                 else
2283                         ndm->ndm_flags &= ~NTF_MASTER;
2284         }
2285
2286         /* Embedded bridge, macvlan, and any other device support */
2287         if (ndm->ndm_flags & NTF_SELF) {
2288                 if (dev->netdev_ops->ndo_fdb_del)
2289                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2290                 else
2291                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2292
2293                 if (!err) {
2294                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2295                         ndm->ndm_flags &= ~NTF_SELF;
2296                 }
2297         }
2298 out:
2299         return err;
2300 }
2301
2302 static int nlmsg_populate_fdb(struct sk_buff *skb,
2303                               struct netlink_callback *cb,
2304                               struct net_device *dev,
2305                               int *idx,
2306                               struct netdev_hw_addr_list *list)
2307 {
2308         struct netdev_hw_addr *ha;
2309         int err;
2310         u32 portid, seq;
2311
2312         portid = NETLINK_CB(cb->skb).portid;
2313         seq = cb->nlh->nlmsg_seq;
2314
2315         list_for_each_entry(ha, &list->list, list) {
2316                 if (*idx < cb->args[0])
2317                         goto skip;
2318
2319                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2320                                               portid, seq,
2321                                               RTM_NEWNEIGH, NTF_SELF);
2322                 if (err < 0)
2323                         return err;
2324 skip:
2325                 *idx += 1;
2326         }
2327         return 0;
2328 }
2329
2330 /**
2331  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2332  * @nlh: netlink message header
2333  * @dev: netdevice
2334  *
2335  * Default netdevice operation to dump the existing unicast address list.
2336  * Returns number of addresses from list put in skb.
2337  */
2338 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2339                       struct netlink_callback *cb,
2340                       struct net_device *dev,
2341                       int idx)
2342 {
2343         int err;
2344
2345         netif_addr_lock_bh(dev);
2346         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2347         if (err)
2348                 goto out;
2349         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2350 out:
2351         netif_addr_unlock_bh(dev);
2352         return idx;
2353 }
2354 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2355
2356 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2357 {
2358         int idx = 0;
2359         struct net *net = sock_net(skb->sk);
2360         struct net_device *dev;
2361
2362         rcu_read_lock();
2363         for_each_netdev_rcu(net, dev) {
2364                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2365                         struct net_device *br_dev;
2366                         const struct net_device_ops *ops;
2367
2368                         br_dev = netdev_master_upper_dev_get(dev);
2369                         ops = br_dev->netdev_ops;
2370                         if (ops->ndo_fdb_dump)
2371                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2372                 }
2373
2374                 if (dev->netdev_ops->ndo_fdb_dump)
2375                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2376                 else
2377                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2378         }
2379         rcu_read_unlock();
2380
2381         cb->args[0] = idx;
2382         return skb->len;
2383 }
2384
2385 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2386                             struct net_device *dev, u16 mode)
2387 {
2388         struct nlmsghdr *nlh;
2389         struct ifinfomsg *ifm;
2390         struct nlattr *br_afspec;
2391         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2392         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2393
2394         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2395         if (nlh == NULL)
2396                 return -EMSGSIZE;
2397
2398         ifm = nlmsg_data(nlh);
2399         ifm->ifi_family = AF_BRIDGE;
2400         ifm->__ifi_pad = 0;
2401         ifm->ifi_type = dev->type;
2402         ifm->ifi_index = dev->ifindex;
2403         ifm->ifi_flags = dev_get_flags(dev);
2404         ifm->ifi_change = 0;
2405
2406
2407         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2408             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2409             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2410             (br_dev &&
2411              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2412             (dev->addr_len &&
2413              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2414             (dev->ifindex != dev->iflink &&
2415              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2416                 goto nla_put_failure;
2417
2418         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2419         if (!br_afspec)
2420                 goto nla_put_failure;
2421
2422         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2423             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2424                 nla_nest_cancel(skb, br_afspec);
2425                 goto nla_put_failure;
2426         }
2427         nla_nest_end(skb, br_afspec);
2428
2429         return nlmsg_end(skb, nlh);
2430 nla_put_failure:
2431         nlmsg_cancel(skb, nlh);
2432         return -EMSGSIZE;
2433 }
2434 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2435
2436 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2437 {
2438         struct net *net = sock_net(skb->sk);
2439         struct net_device *dev;
2440         int idx = 0;
2441         u32 portid = NETLINK_CB(cb->skb).portid;
2442         u32 seq = cb->nlh->nlmsg_seq;
2443         struct nlattr *extfilt;
2444         u32 filter_mask = 0;
2445
2446         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2447                                   IFLA_EXT_MASK);
2448         if (extfilt)
2449                 filter_mask = nla_get_u32(extfilt);
2450
2451         rcu_read_lock();
2452         for_each_netdev_rcu(net, dev) {
2453                 const struct net_device_ops *ops = dev->netdev_ops;
2454                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2455
2456                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2457                         if (idx >= cb->args[0] &&
2458                             br_dev->netdev_ops->ndo_bridge_getlink(
2459                                     skb, portid, seq, dev, filter_mask) < 0)
2460                                 break;
2461                         idx++;
2462                 }
2463
2464                 if (ops->ndo_bridge_getlink) {
2465                         if (idx >= cb->args[0] &&
2466                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2467                                                     filter_mask) < 0)
2468                                 break;
2469                         idx++;
2470                 }
2471         }
2472         rcu_read_unlock();
2473         cb->args[0] = idx;
2474
2475         return skb->len;
2476 }
2477
2478 static inline size_t bridge_nlmsg_size(void)
2479 {
2480         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2481                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2482                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2483                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2484                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2485                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2486                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2487                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2488                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2489                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2490                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2491 }
2492
2493 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2494 {
2495         struct net *net = dev_net(dev);
2496         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2497         struct sk_buff *skb;
2498         int err = -EOPNOTSUPP;
2499
2500         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2501         if (!skb) {
2502                 err = -ENOMEM;
2503                 goto errout;
2504         }
2505
2506         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2507             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2508                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2509                 if (err < 0)
2510                         goto errout;
2511         }
2512
2513         if ((flags & BRIDGE_FLAGS_SELF) &&
2514             dev->netdev_ops->ndo_bridge_getlink) {
2515                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2516                 if (err < 0)
2517                         goto errout;
2518         }
2519
2520         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2521         return 0;
2522 errout:
2523         WARN_ON(err == -EMSGSIZE);
2524         kfree_skb(skb);
2525         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2526         return err;
2527 }
2528
2529 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2530 {
2531         struct net *net = sock_net(skb->sk);
2532         struct ifinfomsg *ifm;
2533         struct net_device *dev;
2534         struct nlattr *br_spec, *attr = NULL;
2535         int rem, err = -EOPNOTSUPP;
2536         u16 oflags, flags = 0;
2537         bool have_flags = false;
2538
2539         if (nlmsg_len(nlh) < sizeof(*ifm))
2540                 return -EINVAL;
2541
2542         ifm = nlmsg_data(nlh);
2543         if (ifm->ifi_family != AF_BRIDGE)
2544                 return -EPFNOSUPPORT;
2545
2546         dev = __dev_get_by_index(net, ifm->ifi_index);
2547         if (!dev) {
2548                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2549                 return -ENODEV;
2550         }
2551
2552         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2553         if (br_spec) {
2554                 nla_for_each_nested(attr, br_spec, rem) {
2555                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2556                                 have_flags = true;
2557                                 flags = nla_get_u16(attr);
2558                                 break;
2559                         }
2560                 }
2561         }
2562
2563         oflags = flags;
2564
2565         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2566                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2567
2568                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2569                         err = -EOPNOTSUPP;
2570                         goto out;
2571                 }
2572
2573                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2574                 if (err)
2575                         goto out;
2576
2577                 flags &= ~BRIDGE_FLAGS_MASTER;
2578         }
2579
2580         if ((flags & BRIDGE_FLAGS_SELF)) {
2581                 if (!dev->netdev_ops->ndo_bridge_setlink)
2582                         err = -EOPNOTSUPP;
2583                 else
2584                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2585
2586                 if (!err)
2587                         flags &= ~BRIDGE_FLAGS_SELF;
2588         }
2589
2590         if (have_flags)
2591                 memcpy(nla_data(attr), &flags, sizeof(flags));
2592         /* Generate event to notify upper layer of bridge change */
2593         if (!err)
2594                 err = rtnl_bridge_notify(dev, oflags);
2595 out:
2596         return err;
2597 }
2598
2599 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2600 {
2601         struct net *net = sock_net(skb->sk);
2602         struct ifinfomsg *ifm;
2603         struct net_device *dev;
2604         struct nlattr *br_spec, *attr = NULL;
2605         int rem, err = -EOPNOTSUPP;
2606         u16 oflags, flags = 0;
2607         bool have_flags = false;
2608
2609         if (nlmsg_len(nlh) < sizeof(*ifm))
2610                 return -EINVAL;
2611
2612         ifm = nlmsg_data(nlh);
2613         if (ifm->ifi_family != AF_BRIDGE)
2614                 return -EPFNOSUPPORT;
2615
2616         dev = __dev_get_by_index(net, ifm->ifi_index);
2617         if (!dev) {
2618                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2619                 return -ENODEV;
2620         }
2621
2622         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2623         if (br_spec) {
2624                 nla_for_each_nested(attr, br_spec, rem) {
2625                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2626                                 have_flags = true;
2627                                 flags = nla_get_u16(attr);
2628                                 break;
2629                         }
2630                 }
2631         }
2632
2633         oflags = flags;
2634
2635         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2636                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2637
2638                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2639                         err = -EOPNOTSUPP;
2640                         goto out;
2641                 }
2642
2643                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2644                 if (err)
2645                         goto out;
2646
2647                 flags &= ~BRIDGE_FLAGS_MASTER;
2648         }
2649
2650         if ((flags & BRIDGE_FLAGS_SELF)) {
2651                 if (!dev->netdev_ops->ndo_bridge_dellink)
2652                         err = -EOPNOTSUPP;
2653                 else
2654                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2655
2656                 if (!err)
2657                         flags &= ~BRIDGE_FLAGS_SELF;
2658         }
2659
2660         if (have_flags)
2661                 memcpy(nla_data(attr), &flags, sizeof(flags));
2662         /* Generate event to notify upper layer of bridge change */
2663         if (!err)
2664                 err = rtnl_bridge_notify(dev, oflags);
2665 out:
2666         return err;
2667 }
2668
2669 /* Process one rtnetlink message. */
2670
2671 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2672 {
2673         struct net *net = sock_net(skb->sk);
2674         rtnl_doit_func doit;
2675         int sz_idx, kind;
2676         int family;
2677         int type;
2678         int err;
2679
2680         type = nlh->nlmsg_type;
2681         if (type > RTM_MAX)
2682                 return -EOPNOTSUPP;
2683
2684         type -= RTM_BASE;
2685
2686         /* All the messages must have at least 1 byte length */
2687         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2688                 return 0;
2689
2690         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2691         sz_idx = type>>2;
2692         kind = type&3;
2693
2694         if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2695                 return -EPERM;
2696
2697         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2698                 struct sock *rtnl;
2699                 rtnl_dumpit_func dumpit;
2700                 rtnl_calcit_func calcit;
2701                 u16 min_dump_alloc = 0;
2702
2703                 dumpit = rtnl_get_dumpit(family, type);
2704                 if (dumpit == NULL)
2705                         return -EOPNOTSUPP;
2706                 calcit = rtnl_get_calcit(family, type);
2707                 if (calcit)
2708                         min_dump_alloc = calcit(skb, nlh);
2709
2710                 __rtnl_unlock();
2711                 rtnl = net->rtnl;
2712                 {
2713                         struct netlink_dump_control c = {
2714                                 .dump           = dumpit,
2715                                 .min_dump_alloc = min_dump_alloc,
2716                         };
2717                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2718                 }
2719                 rtnl_lock();
2720                 return err;
2721         }
2722
2723         doit = rtnl_get_doit(family, type);
2724         if (doit == NULL)
2725                 return -EOPNOTSUPP;
2726
2727         return doit(skb, nlh);
2728 }
2729
2730 static void rtnetlink_rcv(struct sk_buff *skb)
2731 {
2732         rtnl_lock();
2733         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2734         rtnl_unlock();
2735 }
2736
2737 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2738 {
2739         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2740
2741         switch (event) {
2742         case NETDEV_UP:
2743         case NETDEV_DOWN:
2744         case NETDEV_PRE_UP:
2745         case NETDEV_POST_INIT:
2746         case NETDEV_REGISTER:
2747         case NETDEV_CHANGE:
2748         case NETDEV_PRE_TYPE_CHANGE:
2749         case NETDEV_GOING_DOWN:
2750         case NETDEV_UNREGISTER:
2751         case NETDEV_UNREGISTER_FINAL:
2752         case NETDEV_RELEASE:
2753         case NETDEV_JOIN:
2754                 break;
2755         default:
2756                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2757                 break;
2758         }
2759         return NOTIFY_DONE;
2760 }
2761
2762 static struct notifier_block rtnetlink_dev_notifier = {
2763         .notifier_call  = rtnetlink_event,
2764 };
2765
2766
2767 static int __net_init rtnetlink_net_init(struct net *net)
2768 {
2769         struct sock *sk;
2770         struct netlink_kernel_cfg cfg = {
2771                 .groups         = RTNLGRP_MAX,
2772                 .input          = rtnetlink_rcv,
2773                 .cb_mutex       = &rtnl_mutex,
2774                 .flags          = NL_CFG_F_NONROOT_RECV,
2775         };
2776
2777         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2778         if (!sk)
2779                 return -ENOMEM;
2780         net->rtnl = sk;
2781         return 0;
2782 }
2783
2784 static void __net_exit rtnetlink_net_exit(struct net *net)
2785 {
2786         netlink_kernel_release(net->rtnl);
2787         net->rtnl = NULL;
2788 }
2789
2790 static struct pernet_operations rtnetlink_net_ops = {
2791         .init = rtnetlink_net_init,
2792         .exit = rtnetlink_net_exit,
2793 };
2794
2795 void __init rtnetlink_init(void)
2796 {
2797         if (register_pernet_subsys(&rtnetlink_net_ops))
2798                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2799
2800         register_netdevice_notifier(&rtnetlink_dev_notifier);
2801
2802         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2803                       rtnl_dump_ifinfo, rtnl_calcit);
2804         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2805         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2806         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2807
2808         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2809         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2810
2811         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2812         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2813         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2814
2815         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2816         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2817         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2818 }
2819