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