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_slave_info_data_size(const struct net_device *dev)
 369 {
 370         struct net_device *master_dev;
 371         const struct rtnl_link_ops *ops;
 372
 373         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 374         if (!master_dev)
 375                 return 0;
 376         ops = master_dev->rtnl_link_ops;
 377         if (!ops || !ops->get_slave_size)
 378                 return 0;
 379         /* IFLA_INFO_SLAVE_DATA + nested data */
 380         return nla_total_size(sizeof(struct nlattr)) +
 381                ops->get_slave_size(master_dev, dev);
 382 }
 383
 384 static size_t rtnl_link_get_size(const struct net_device *dev)
 385 {
 386         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 387         size_t size;
 388
 389         if (!ops)
 390                 return 0;
 391
 392         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 393                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 394
 395         if (ops->get_size)
 396                 /* IFLA_INFO_DATA + nested data */
 397                 size += nla_total_size(sizeof(struct nlattr)) +
 398                         ops->get_size(dev);
 399
 400         if (ops->get_xstats_size)
 401                 /* IFLA_INFO_XSTATS */
 402                 size += nla_total_size(ops->get_xstats_size(dev));
 403
 404         size += rtnl_link_get_slave_info_data_size(dev);
 405
 406         return size;
 407 }
 408
 409 static LIST_HEAD(rtnl_af_ops);
 410
 411 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 412 {
 413         const struct rtnl_af_ops *ops;
 414
 415         list_for_each_entry(ops, &rtnl_af_ops, list) {
 416                 if (ops->family == family)
 417                         return ops;
 418         }
 419
 420         return NULL;
 421 }
 422
 423 /**
 424  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 425  * @ops: struct rtnl_af_ops * to register
 426  *
 427  * Returns 0 on success or a negative error code.
 428  */
 429 void rtnl_af_register(struct rtnl_af_ops *ops)
 430 {
 431         rtnl_lock();
 432         list_add_tail(&ops->list, &rtnl_af_ops);
 433         rtnl_unlock();
 434 }
 435 EXPORT_SYMBOL_GPL(rtnl_af_register);
 436
 437 /**
 438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 439  * @ops: struct rtnl_af_ops * to unregister
 440  *
 441  * The caller must hold the rtnl_mutex.
 442  */
 443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 444 {
 445         list_del(&ops->list);
 446 }
 447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 448
 449 /**
 450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 451  * @ops: struct rtnl_af_ops * to unregister
 452  */
 453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 454 {
 455         rtnl_lock();
 456         __rtnl_af_unregister(ops);
 457         rtnl_unlock();
 458 }
 459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 460
 461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 462 {
 463         struct rtnl_af_ops *af_ops;
 464         size_t size;
 465
 466         /* IFLA_AF_SPEC */
 467         size = nla_total_size(sizeof(struct nlattr));
 468
 469         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 470                 if (af_ops->get_link_af_size) {
 471                         /* AF_* + nested data */
 472                         size += nla_total_size(sizeof(struct nlattr)) +
 473                                 af_ops->get_link_af_size(dev);
 474                 }
 475         }
 476
 477         return size;
 478 }
 479
 480 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 481 {
 482         struct net_device *master_dev;
 483
 484         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 485         if (master_dev && master_dev->rtnl_link_ops)
 486                 return true;
 487         return false;
 488 }
 489
 490 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 491                                      const struct net_device *dev)
 492 {
 493         struct net_device *master_dev;
 494         const struct rtnl_link_ops *ops;
 495         struct nlattr *slave_data;
 496         int err;
 497
 498         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 499         if (!master_dev)
 500                 return 0;
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops)
 503                 return 0;
 504         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 505                 return -EMSGSIZE;
 506         if (ops->fill_slave_info) {
 507                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 508                 if (!slave_data)
 509                         return -EMSGSIZE;
 510                 err = ops->fill_slave_info(skb, master_dev, dev);
 511                 if (err < 0)
 512                         goto err_cancel_slave_data;
 513                 nla_nest_end(skb, slave_data);
 514         }
 515         return 0;
 516
 517 err_cancel_slave_data:
 518         nla_nest_cancel(skb, slave_data);
 519         return err;
 520 }
 521
 522 static int rtnl_link_info_fill(struct sk_buff *skb,
 523                                const struct net_device *dev)
 524 {
 525         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 526         struct nlattr *data;
 527         int err;
 528
 529         if (!ops)
 530                 return 0;
 531         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 532                 return -EMSGSIZE;
 533         if (ops->fill_xstats) {
 534                 err = ops->fill_xstats(skb, dev);
 535                 if (err < 0)
 536                         return err;
 537         }
 538         if (ops->fill_info) {
 539                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 540                 if (data == NULL)
 541                         return -EMSGSIZE;
 542                 err = ops->fill_info(skb, dev);
 543                 if (err < 0)
 544                         goto err_cancel_data;
 545                 nla_nest_end(skb, data);
 546         }
 547         return 0;
 548
 549 err_cancel_data:
 550         nla_nest_cancel(skb, data);
 551         return err;
 552 }
 553
 554 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 555 {
 556         struct nlattr *linkinfo;
 557         int err = -EMSGSIZE;
 558
 559         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 560         if (linkinfo == NULL)
 561                 goto out;
 562
 563         err = rtnl_link_info_fill(skb, dev);
 564         if (err < 0)
 565                 goto err_cancel_link;
 566
 567         err = rtnl_link_slave_info_fill(skb, dev);
 568         if (err < 0)
 569                 goto err_cancel_link;
 570
 571         nla_nest_end(skb, linkinfo);
 572         return 0;
 573
 574 err_cancel_link:
 575         nla_nest_cancel(skb, linkinfo);
 576 out:
 577         return err;
 578 }
 579
 580 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 581 {
 582         struct sock *rtnl = net->rtnl;
 583         int err = 0;
 584
 585         NETLINK_CB(skb).dst_group = group;
 586         if (echo)
 587                 atomic_inc(&skb->users);
 588         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 589         if (echo)
 590                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 591         return err;
 592 }
 593
 594 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 595 {
 596         struct sock *rtnl = net->rtnl;
 597
 598         return nlmsg_unicast(rtnl, skb, pid);
 599 }
 600 EXPORT_SYMBOL(rtnl_unicast);
 601
 602 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 603                  struct nlmsghdr *nlh, gfp_t flags)
 604 {
 605         struct sock *rtnl = net->rtnl;
 606         int report = 0;
 607
 608         if (nlh)
 609                 report = nlmsg_report(nlh);
 610
 611         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 612 }
 613 EXPORT_SYMBOL(rtnl_notify);
 614
 615 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 616 {
 617         struct sock *rtnl = net->rtnl;
 618
 619         netlink_set_err(rtnl, 0, group, error);
 620 }
 621 EXPORT_SYMBOL(rtnl_set_sk_err);
 622
 623 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 624 {
 625         struct nlattr *mx;
 626         int i, valid = 0;
 627
 628         mx = nla_nest_start(skb, RTA_METRICS);
 629         if (mx == NULL)
 630                 return -ENOBUFS;
 631
 632         for (i = 0; i < RTAX_MAX; i++) {
 633                 if (metrics[i]) {
 634                         valid++;
 635                         if (nla_put_u32(skb, i+1, metrics[i]))
 636                                 goto nla_put_failure;
 637                 }
 638         }
 639
 640         if (!valid) {
 641                 nla_nest_cancel(skb, mx);
 642                 return 0;
 643         }
 644
 645         return nla_nest_end(skb, mx);
 646
 647 nla_put_failure:
 648         nla_nest_cancel(skb, mx);
 649         return -EMSGSIZE;
 650 }
 651 EXPORT_SYMBOL(rtnetlink_put_metrics);
 652
 653 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 654                        long expires, u32 error)
 655 {
 656         struct rta_cacheinfo ci = {
 657                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 658                 .rta_used = dst->__use,
 659                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 660                 .rta_error = error,
 661                 .rta_id =  id,
 662         };
 663
 664         if (expires) {
 665                 unsigned long clock;
 666
 667                 clock = jiffies_to_clock_t(abs(expires));
 668                 clock = min_t(unsigned long, clock, INT_MAX);
 669                 ci.rta_expires = (expires > 0) ? clock : -clock;
 670         }
 671         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 672 }
 673 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 674
 675 static void set_operstate(struct net_device *dev, unsigned char transition)
 676 {
 677         unsigned char operstate = dev->operstate;
 678
 679         switch (transition) {
 680         case IF_OPER_UP:
 681                 if ((operstate == IF_OPER_DORMANT ||
 682                      operstate == IF_OPER_UNKNOWN) &&
 683                     !netif_dormant(dev))
 684                         operstate = IF_OPER_UP;
 685                 break;
 686
 687         case IF_OPER_DORMANT:
 688                 if (operstate == IF_OPER_UP ||
 689                     operstate == IF_OPER_UNKNOWN)
 690                         operstate = IF_OPER_DORMANT;
 691                 break;
 692         }
 693
 694         if (dev->operstate != operstate) {
 695                 write_lock_bh(&dev_base_lock);
 696                 dev->operstate = operstate;
 697                 write_unlock_bh(&dev_base_lock);
 698                 netdev_state_change(dev);
 699         }
 700 }
 701
 702 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 703 {
 704         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 705                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 706 }
 707
 708 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 709                                            const struct ifinfomsg *ifm)
 710 {
 711         unsigned int flags = ifm->ifi_flags;
 712
 713         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 714         if (ifm->ifi_change)
 715                 flags = (flags & ifm->ifi_change) |
 716                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 717
 718         return flags;
 719 }
 720
 721 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 722                                  const struct rtnl_link_stats64 *b)
 723 {
 724         a->rx_packets = b->rx_packets;
 725         a->tx_packets = b->tx_packets;
 726         a->rx_bytes = b->rx_bytes;
 727         a->tx_bytes = b->tx_bytes;
 728         a->rx_errors = b->rx_errors;
 729         a->tx_errors = b->tx_errors;
 730         a->rx_dropped = b->rx_dropped;
 731         a->tx_dropped = b->tx_dropped;
 732
 733         a->multicast = b->multicast;
 734         a->collisions = b->collisions;
 735
 736         a->rx_length_errors = b->rx_length_errors;
 737         a->rx_over_errors = b->rx_over_errors;
 738         a->rx_crc_errors = b->rx_crc_errors;
 739         a->rx_frame_errors = b->rx_frame_errors;
 740         a->rx_fifo_errors = b->rx_fifo_errors;
 741         a->rx_missed_errors = b->rx_missed_errors;
 742
 743         a->tx_aborted_errors = b->tx_aborted_errors;
 744         a->tx_carrier_errors = b->tx_carrier_errors;
 745         a->tx_fifo_errors = b->tx_fifo_errors;
 746         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 747         a->tx_window_errors = b->tx_window_errors;
 748
 749         a->rx_compressed = b->rx_compressed;
 750         a->tx_compressed = b->tx_compressed;
 751 }
 752
 753 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 754 {
 755         memcpy(v, b, sizeof(*b));
 756 }
 757
 758 /* All VF info */
 759 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 760                                    u32 ext_filter_mask)
 761 {
 762         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 763             (ext_filter_mask & RTEXT_FILTER_VF)) {
 764                 int num_vfs = dev_num_vf(dev->dev.parent);
 765                 size_t size = nla_total_size(sizeof(struct nlattr));
 766                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 767                 size += num_vfs *
 768                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 769                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 770                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 771                          nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 772                 return size;
 773         } else
 774                 return 0;
 775 }
 776
 777 static size_t rtnl_port_size(const struct net_device *dev)
 778 {
 779         size_t port_size = nla_total_size(4)            /* PORT_VF */
 780                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 781                 + nla_total_size(sizeof(struct ifla_port_vsi))
 782                                                         /* PORT_VSI_TYPE */
 783                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 784                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 785                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 786                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 787         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 788         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 789                 + port_size;
 790         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 791                 + port_size;
 792
 793         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 794                 return 0;
 795         if (dev_num_vf(dev->dev.parent))
 796                 return port_self_size + vf_ports_size +
 797                         vf_port_size * dev_num_vf(dev->dev.parent);
 798         else
 799                 return port_self_size;
 800 }
 801
 802 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 803                                      u32 ext_filter_mask)
 804 {
 805         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 806                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 807                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 808                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 809                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 810                + nla_total_size(sizeof(struct rtnl_link_stats))
 811                + nla_total_size(sizeof(struct rtnl_link_stats64))
 812                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 813                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 814                + nla_total_size(4) /* IFLA_TXQLEN */
 815                + nla_total_size(4) /* IFLA_WEIGHT */
 816                + nla_total_size(4) /* IFLA_MTU */
 817                + nla_total_size(4) /* IFLA_LINK */
 818                + nla_total_size(4) /* IFLA_MASTER */
 819                + nla_total_size(1) /* IFLA_CARRIER */
 820                + nla_total_size(4) /* IFLA_PROMISCUITY */
 821                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 822                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 823                + nla_total_size(1) /* IFLA_OPERSTATE */
 824                + nla_total_size(1) /* IFLA_LINKMODE */
 825                + nla_total_size(ext_filter_mask
 826                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 827                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 828                + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 829                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 830                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 831                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 832 }
 833
 834 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 835 {
 836         struct nlattr *vf_ports;
 837         struct nlattr *vf_port;
 838         int vf;
 839         int err;
 840
 841         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 842         if (!vf_ports)
 843                 return -EMSGSIZE;
 844
 845         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 846                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 847                 if (!vf_port)
 848                         goto nla_put_failure;
 849                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 850                         goto nla_put_failure;
 851                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 852                 if (err == -EMSGSIZE)
 853                         goto nla_put_failure;
 854                 if (err) {
 855                         nla_nest_cancel(skb, vf_port);
 856                         continue;
 857                 }
 858                 nla_nest_end(skb, vf_port);
 859         }
 860
 861         nla_nest_end(skb, vf_ports);
 862
 863         return 0;
 864
 865 nla_put_failure:
 866         nla_nest_cancel(skb, vf_ports);
 867         return -EMSGSIZE;
 868 }
 869
 870 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 871 {
 872         struct nlattr *port_self;
 873         int err;
 874
 875         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 876         if (!port_self)
 877                 return -EMSGSIZE;
 878
 879         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 880         if (err) {
 881                 nla_nest_cancel(skb, port_self);
 882                 return (err == -EMSGSIZE) ? err : 0;
 883         }
 884
 885         nla_nest_end(skb, port_self);
 886
 887         return 0;
 888 }
 889
 890 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 891 {
 892         int err;
 893
 894         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 895                 return 0;
 896
 897         err = rtnl_port_self_fill(skb, dev);
 898         if (err)
 899                 return err;
 900
 901         if (dev_num_vf(dev->dev.parent)) {
 902                 err = rtnl_vf_ports_fill(skb, dev);
 903                 if (err)
 904                         return err;
 905         }
 906
 907         return 0;
 908 }
 909
 910 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 911 {
 912         int err;
 913         struct netdev_phys_port_id ppid;
 914
 915         err = dev_get_phys_port_id(dev, &ppid);
 916         if (err) {
 917                 if (err == -EOPNOTSUPP)
 918                         return 0;
 919                 return err;
 920         }
 921
 922         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 923                 return -EMSGSIZE;
 924
 925         return 0;
 926 }
 927
 928 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 929                             int type, u32 pid, u32 seq, u32 change,
 930                             unsigned int flags, u32 ext_filter_mask)
 931 {
 932         struct ifinfomsg *ifm;
 933         struct nlmsghdr *nlh;
 934         struct rtnl_link_stats64 temp;
 935         const struct rtnl_link_stats64 *stats;
 936         struct nlattr *attr, *af_spec;
 937         struct rtnl_af_ops *af_ops;
 938         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 939
 940         ASSERT_RTNL();
 941         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 942         if (nlh == NULL)
 943                 return -EMSGSIZE;
 944
 945         ifm = nlmsg_data(nlh);
 946         ifm->ifi_family = AF_UNSPEC;
 947         ifm->__ifi_pad = 0;
 948         ifm->ifi_type = dev->type;
 949         ifm->ifi_index = dev->ifindex;
 950         ifm->ifi_flags = dev_get_flags(dev);
 951         ifm->ifi_change = change;
 952
 953         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 954             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 955             nla_put_u8(skb, IFLA_OPERSTATE,
 956                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 957             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 958             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 959             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 960             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 961             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
 962 #ifdef CONFIG_RPS
 963             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
 964 #endif
 965             (dev->ifindex != dev->iflink &&
 966              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 967             (upper_dev &&
 968              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
 969             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
 970             (dev->qdisc &&
 971              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 972             (dev->ifalias &&
 973              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 974                 goto nla_put_failure;
 975
 976         if (1) {
 977                 struct rtnl_link_ifmap map = {
 978                         .mem_start   = dev->mem_start,
 979                         .mem_end     = dev->mem_end,
 980                         .base_addr   = dev->base_addr,
 981                         .irq         = dev->irq,
 982                         .dma         = dev->dma,
 983                         .port        = dev->if_port,
 984                 };
 985                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 986                         goto nla_put_failure;
 987         }
 988
 989         if (dev->addr_len) {
 990                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 991                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 992                         goto nla_put_failure;
 993         }
 994
 995         if (rtnl_phys_port_id_fill(skb, dev))
 996                 goto nla_put_failure;
 997
 998         attr = nla_reserve(skb, IFLA_STATS,
 999                         sizeof(struct rtnl_link_stats));
1000         if (attr == NULL)
1001                 goto nla_put_failure;
1002
1003         stats = dev_get_stats(dev, &temp);
1004         copy_rtnl_link_stats(nla_data(attr), stats);
1005
1006         attr = nla_reserve(skb, IFLA_STATS64,
1007                         sizeof(struct rtnl_link_stats64));
1008         if (attr == NULL)
1009                 goto nla_put_failure;
1010         copy_rtnl_link_stats64(nla_data(attr), stats);
1011
1012         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1013             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1014                 goto nla_put_failure;
1015
1016         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1017             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1018                 int i;
1019
1020                 struct nlattr *vfinfo, *vf;
1021                 int num_vfs = dev_num_vf(dev->dev.parent);
1022
1023                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1024                 if (!vfinfo)
1025                         goto nla_put_failure;
1026                 for (i = 0; i < num_vfs; i++) {
1027                         struct ifla_vf_info ivi;
1028                         struct ifla_vf_mac vf_mac;
1029                         struct ifla_vf_vlan vf_vlan;
1030                         struct ifla_vf_tx_rate vf_tx_rate;
1031                         struct ifla_vf_spoofchk vf_spoofchk;
1032                         struct ifla_vf_link_state vf_linkstate;
1033
1034                         /*
1035                          * Not all SR-IOV capable drivers support the
1036                          * spoofcheck query.  Preset to -1 so the user
1037                          * space tool can detect that the driver didn't
1038                          * report anything.
1039                          */
1040                         ivi.spoofchk = -1;
1041                         memset(ivi.mac, 0, sizeof(ivi.mac));
1042                         /* The default value for VF link state is "auto"
1043                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1044                          */
1045                         ivi.linkstate = 0;
1046                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1047                                 break;
1048                         vf_mac.vf =
1049                                 vf_vlan.vf =
1050                                 vf_tx_rate.vf =
1051                                 vf_spoofchk.vf =
1052                                 vf_linkstate.vf = ivi.vf;
1053
1054                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1055                         vf_vlan.vlan = ivi.vlan;
1056                         vf_vlan.qos = ivi.qos;
1057                         vf_tx_rate.rate = ivi.tx_rate;
1058                         vf_spoofchk.setting = ivi.spoofchk;
1059                         vf_linkstate.link_state = ivi.linkstate;
1060                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1061                         if (!vf) {
1062                                 nla_nest_cancel(skb, vfinfo);
1063                                 goto nla_put_failure;
1064                         }
1065                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1066                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1067                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1068                                     &vf_tx_rate) ||
1069                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1070                                     &vf_spoofchk) ||
1071                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1072                                     &vf_linkstate))
1073                                 goto nla_put_failure;
1074                         nla_nest_end(skb, vf);
1075                 }
1076                 nla_nest_end(skb, vfinfo);
1077         }
1078
1079         if (rtnl_port_fill(skb, dev))
1080                 goto nla_put_failure;
1081
1082         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1083                 if (rtnl_link_fill(skb, dev) < 0)
1084                         goto nla_put_failure;
1085         }
1086
1087         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1088                 goto nla_put_failure;
1089
1090         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1091                 if (af_ops->fill_link_af) {
1092                         struct nlattr *af;
1093                         int err;
1094
1095                         if (!(af = nla_nest_start(skb, af_ops->family)))
1096                                 goto nla_put_failure;
1097
1098                         err = af_ops->fill_link_af(skb, dev);
1099
1100                         /*
1101                          * Caller may return ENODATA to indicate that there
1102                          * was no data to be dumped. This is not an error, it
1103                          * means we should trim the attribute header and
1104                          * continue.
1105                          */
1106                         if (err == -ENODATA)
1107                                 nla_nest_cancel(skb, af);
1108                         else if (err < 0)
1109                                 goto nla_put_failure;
1110
1111                         nla_nest_end(skb, af);
1112                 }
1113         }
1114
1115         nla_nest_end(skb, af_spec);
1116
1117         return nlmsg_end(skb, nlh);
1118
1119 nla_put_failure:
1120         nlmsg_cancel(skb, nlh);
1121         return -EMSGSIZE;
1122 }
1123
1124 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1125 {
1126         struct net *net = sock_net(skb->sk);
1127         int h, s_h;
1128         int idx = 0, s_idx;
1129         struct net_device *dev;
1130         struct hlist_head *head;
1131         struct nlattr *tb[IFLA_MAX+1];
1132         u32 ext_filter_mask = 0;
1133         int err;
1134
1135         s_h = cb->args[0];
1136         s_idx = cb->args[1];
1137
1138         rcu_read_lock();
1139         cb->seq = net->dev_base_seq;
1140
1141         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1142                         ifla_policy) >= 0) {
1143
1144                 if (tb[IFLA_EXT_MASK])
1145                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1146         }
1147
1148         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1149                 idx = 0;
1150                 head = &net->dev_index_head[h];
1151                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1152                         if (idx < s_idx)
1153                                 goto cont;
1154                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1155                                                NETLINK_CB(cb->skb).portid,
1156                                                cb->nlh->nlmsg_seq, 0,
1157                                                NLM_F_MULTI,
1158                                                ext_filter_mask);
1159                         /* If we ran out of room on the first message,
1160                          * we're in trouble
1161                          */
1162                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1163
1164                         if (err <= 0)
1165                                 goto out;
1166
1167                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1168 cont:
1169                         idx++;
1170                 }
1171         }
1172 out:
1173         rcu_read_unlock();
1174         cb->args[1] = idx;
1175         cb->args[0] = h;
1176
1177         return skb->len;
1178 }
1179
1180 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1181         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1182         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1183         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1184         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1185         [IFLA_MTU]              = { .type = NLA_U32 },
1186         [IFLA_LINK]             = { .type = NLA_U32 },
1187         [IFLA_MASTER]           = { .type = NLA_U32 },
1188         [IFLA_CARRIER]          = { .type = NLA_U8 },
1189         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1190         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1191         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1192         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1193         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1194         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1195         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1196         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1197         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1198         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1199         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1200         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1201         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1202         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1203         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1204         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1205         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1206 };
1207 EXPORT_SYMBOL(ifla_policy);
1208
1209 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1210         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1211         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1212         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1213         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1214 };
1215
1216 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1217         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1218 };
1219
1220 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1221         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1222                                     .len = sizeof(struct ifla_vf_mac) },
1223         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1224                                     .len = sizeof(struct ifla_vf_vlan) },
1225         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1226                                     .len = sizeof(struct ifla_vf_tx_rate) },
1227         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1228                                     .len = sizeof(struct ifla_vf_spoofchk) },
1229 };
1230
1231 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1232         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1233         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1234                                     .len = PORT_PROFILE_MAX },
1235         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1236                                     .len = sizeof(struct ifla_port_vsi)},
1237         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1238                                       .len = PORT_UUID_MAX },
1239         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1240                                     .len = PORT_UUID_MAX },
1241         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1242         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1243 };
1244
1245 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1246 {
1247         struct net *net;
1248         /* Examine the link attributes and figure out which
1249          * network namespace we are talking about.
1250          */
1251         if (tb[IFLA_NET_NS_PID])
1252                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1253         else if (tb[IFLA_NET_NS_FD])
1254                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1255         else
1256                 net = get_net(src_net);
1257         return net;
1258 }
1259 EXPORT_SYMBOL(rtnl_link_get_net);
1260
1261 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1262 {
1263         if (dev) {
1264                 if (tb[IFLA_ADDRESS] &&
1265                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1266                         return -EINVAL;
1267
1268                 if (tb[IFLA_BROADCAST] &&
1269                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1270                         return -EINVAL;
1271         }
1272
1273         if (tb[IFLA_AF_SPEC]) {
1274                 struct nlattr *af;
1275                 int rem, err;
1276
1277                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1278                         const struct rtnl_af_ops *af_ops;
1279
1280                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1281                                 return -EAFNOSUPPORT;
1282
1283                         if (!af_ops->set_link_af)
1284                                 return -EOPNOTSUPP;
1285
1286                         if (af_ops->validate_link_af) {
1287                                 err = af_ops->validate_link_af(dev, af);
1288                                 if (err < 0)
1289                                         return err;
1290                         }
1291                 }
1292         }
1293
1294         return 0;
1295 }
1296
1297 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1298 {
1299         int rem, err = -EINVAL;
1300         struct nlattr *vf;
1301         const struct net_device_ops *ops = dev->netdev_ops;
1302
1303         nla_for_each_nested(vf, attr, rem) {
1304                 switch (nla_type(vf)) {
1305                 case IFLA_VF_MAC: {
1306                         struct ifla_vf_mac *ivm;
1307                         ivm = nla_data(vf);
1308                         err = -EOPNOTSUPP;
1309                         if (ops->ndo_set_vf_mac)
1310                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1311                                                           ivm->mac);
1312                         break;
1313                 }
1314                 case IFLA_VF_VLAN: {
1315                         struct ifla_vf_vlan *ivv;
1316                         ivv = nla_data(vf);
1317                         err = -EOPNOTSUPP;
1318                         if (ops->ndo_set_vf_vlan)
1319                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1320                                                            ivv->vlan,
1321                                                            ivv->qos);
1322                         break;
1323                 }
1324                 case IFLA_VF_TX_RATE: {
1325                         struct ifla_vf_tx_rate *ivt;
1326                         ivt = nla_data(vf);
1327                         err = -EOPNOTSUPP;
1328                         if (ops->ndo_set_vf_tx_rate)
1329                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1330                                                               ivt->rate);
1331                         break;
1332                 }
1333                 case IFLA_VF_SPOOFCHK: {
1334                         struct ifla_vf_spoofchk *ivs;
1335                         ivs = nla_data(vf);
1336                         err = -EOPNOTSUPP;
1337                         if (ops->ndo_set_vf_spoofchk)
1338                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1339                                                                ivs->setting);
1340                         break;
1341                 }
1342                 case IFLA_VF_LINK_STATE: {
1343                         struct ifla_vf_link_state *ivl;
1344                         ivl = nla_data(vf);
1345                         err = -EOPNOTSUPP;
1346                         if (ops->ndo_set_vf_link_state)
1347                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1348                                                                  ivl->link_state);
1349                         break;
1350                 }
1351                 default:
1352                         err = -EINVAL;
1353                         break;
1354                 }
1355                 if (err)
1356                         break;
1357         }
1358         return err;
1359 }
1360
1361 static int do_set_master(struct net_device *dev, int ifindex)
1362 {
1363         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1364         const struct net_device_ops *ops;
1365         int err;
1366
1367         if (upper_dev) {
1368                 if (upper_dev->ifindex == ifindex)
1369                         return 0;
1370                 ops = upper_dev->netdev_ops;
1371                 if (ops->ndo_del_slave) {
1372                         err = ops->ndo_del_slave(upper_dev, dev);
1373                         if (err)
1374                                 return err;
1375                 } else {
1376                         return -EOPNOTSUPP;
1377                 }
1378         }
1379
1380         if (ifindex) {
1381                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1382                 if (!upper_dev)
1383                         return -EINVAL;
1384                 ops = upper_dev->netdev_ops;
1385                 if (ops->ndo_add_slave) {
1386                         err = ops->ndo_add_slave(upper_dev, dev);
1387                         if (err)
1388                                 return err;
1389                 } else {
1390                         return -EOPNOTSUPP;
1391                 }
1392         }
1393         return 0;
1394 }
1395
1396 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1397                       struct nlattr **tb, char *ifname, int modified)
1398 {
1399         const struct net_device_ops *ops = dev->netdev_ops;
1400         int err;
1401
1402         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1403                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1404                 if (IS_ERR(net)) {
1405                         err = PTR_ERR(net);
1406                         goto errout;
1407                 }
1408                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1409                         err = -EPERM;
1410                         goto errout;
1411                 }
1412                 err = dev_change_net_namespace(dev, net, ifname);
1413                 put_net(net);
1414                 if (err)
1415                         goto errout;
1416                 modified = 1;
1417         }
1418
1419         if (tb[IFLA_MAP]) {
1420                 struct rtnl_link_ifmap *u_map;
1421                 struct ifmap k_map;
1422
1423                 if (!ops->ndo_set_config) {
1424                         err = -EOPNOTSUPP;
1425                         goto errout;
1426                 }
1427
1428                 if (!netif_device_present(dev)) {
1429                         err = -ENODEV;
1430                         goto errout;
1431                 }
1432
1433                 u_map = nla_data(tb[IFLA_MAP]);
1434                 k_map.mem_start = (unsigned long) u_map->mem_start;
1435                 k_map.mem_end = (unsigned long) u_map->mem_end;
1436                 k_map.base_addr = (unsigned short) u_map->base_addr;
1437                 k_map.irq = (unsigned char) u_map->irq;
1438                 k_map.dma = (unsigned char) u_map->dma;
1439                 k_map.port = (unsigned char) u_map->port;
1440
1441                 err = ops->ndo_set_config(dev, &k_map);
1442                 if (err < 0)
1443                         goto errout;
1444
1445                 modified = 1;
1446         }
1447
1448         if (tb[IFLA_ADDRESS]) {
1449                 struct sockaddr *sa;
1450                 int len;
1451
1452                 len = sizeof(sa_family_t) + dev->addr_len;
1453                 sa = kmalloc(len, GFP_KERNEL);
1454                 if (!sa) {
1455                         err = -ENOMEM;
1456                         goto errout;
1457                 }
1458                 sa->sa_family = dev->type;
1459                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1460                        dev->addr_len);
1461                 err = dev_set_mac_address(dev, sa);
1462                 kfree(sa);
1463                 if (err)
1464                         goto errout;
1465                 modified = 1;
1466         }
1467
1468         if (tb[IFLA_MTU]) {
1469                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1470                 if (err < 0)
1471                         goto errout;
1472                 modified = 1;
1473         }
1474
1475         if (tb[IFLA_GROUP]) {
1476                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1477                 modified = 1;
1478         }
1479
1480         /*
1481          * Interface selected by interface index but interface
1482          * name provided implies that a name change has been
1483          * requested.
1484          */
1485         if (ifm->ifi_index > 0 && ifname[0]) {
1486                 err = dev_change_name(dev, ifname);
1487                 if (err < 0)
1488                         goto errout;
1489                 modified = 1;
1490         }
1491
1492         if (tb[IFLA_IFALIAS]) {
1493                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1494                                     nla_len(tb[IFLA_IFALIAS]));
1495                 if (err < 0)
1496                         goto errout;
1497                 modified = 1;
1498         }
1499
1500         if (tb[IFLA_BROADCAST]) {
1501                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1502                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1503         }
1504
1505         if (ifm->ifi_flags || ifm->ifi_change) {
1506                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1507                 if (err < 0)
1508                         goto errout;
1509         }
1510
1511         if (tb[IFLA_MASTER]) {
1512                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1513                 if (err)
1514                         goto errout;
1515                 modified = 1;
1516         }
1517
1518         if (tb[IFLA_CARRIER]) {
1519                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1520                 if (err)
1521                         goto errout;
1522                 modified = 1;
1523         }
1524
1525         if (tb[IFLA_TXQLEN])
1526                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1527
1528         if (tb[IFLA_OPERSTATE])
1529                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1530
1531         if (tb[IFLA_LINKMODE]) {
1532                 write_lock_bh(&dev_base_lock);
1533                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1534                 write_unlock_bh(&dev_base_lock);
1535         }
1536
1537         if (tb[IFLA_VFINFO_LIST]) {
1538                 struct nlattr *attr;
1539                 int rem;
1540                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1541                         if (nla_type(attr) != IFLA_VF_INFO) {
1542                                 err = -EINVAL;
1543                                 goto errout;
1544                         }
1545                         err = do_setvfinfo(dev, attr);
1546                         if (err < 0)
1547                                 goto errout;
1548                         modified = 1;
1549                 }
1550         }
1551         err = 0;
1552
1553         if (tb[IFLA_VF_PORTS]) {
1554                 struct nlattr *port[IFLA_PORT_MAX+1];
1555                 struct nlattr *attr;
1556                 int vf;
1557                 int rem;
1558
1559                 err = -EOPNOTSUPP;
1560                 if (!ops->ndo_set_vf_port)
1561                         goto errout;
1562
1563                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1564                         if (nla_type(attr) != IFLA_VF_PORT)
1565                                 continue;
1566                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1567                                 attr, ifla_port_policy);
1568                         if (err < 0)
1569                                 goto errout;
1570                         if (!port[IFLA_PORT_VF]) {
1571                                 err = -EOPNOTSUPP;
1572                                 goto errout;
1573                         }
1574                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1575                         err = ops->ndo_set_vf_port(dev, vf, port);
1576                         if (err < 0)
1577                                 goto errout;
1578                         modified = 1;
1579                 }
1580         }
1581         err = 0;
1582
1583         if (tb[IFLA_PORT_SELF]) {
1584                 struct nlattr *port[IFLA_PORT_MAX+1];
1585
1586                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1587                         tb[IFLA_PORT_SELF], ifla_port_policy);
1588                 if (err < 0)
1589                         goto errout;
1590
1591                 err = -EOPNOTSUPP;
1592                 if (ops->ndo_set_vf_port)
1593                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1594                 if (err < 0)
1595                         goto errout;
1596                 modified = 1;
1597         }
1598
1599         if (tb[IFLA_AF_SPEC]) {
1600                 struct nlattr *af;
1601                 int rem;
1602
1603                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1604                         const struct rtnl_af_ops *af_ops;
1605
1606                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1607                                 BUG();
1608
1609                         err = af_ops->set_link_af(dev, af);
1610                         if (err < 0)
1611                                 goto errout;
1612
1613                         modified = 1;
1614                 }
1615         }
1616         err = 0;
1617
1618 errout:
1619         if (err < 0 && modified)
1620                 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",
1621                                      dev->name);
1622
1623         return err;
1624 }
1625
1626 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1627 {
1628         struct net *net = sock_net(skb->sk);
1629         struct ifinfomsg *ifm;
1630         struct net_device *dev;
1631         int err;
1632         struct nlattr *tb[IFLA_MAX+1];
1633         char ifname[IFNAMSIZ];
1634
1635         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1636         if (err < 0)
1637                 goto errout;
1638
1639         if (tb[IFLA_IFNAME])
1640                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1641         else
1642                 ifname[0] = '\0';
1643
1644         err = -EINVAL;
1645         ifm = nlmsg_data(nlh);
1646         if (ifm->ifi_index > 0)
1647                 dev = __dev_get_by_index(net, ifm->ifi_index);
1648         else if (tb[IFLA_IFNAME])
1649                 dev = __dev_get_by_name(net, ifname);
1650         else
1651                 goto errout;
1652
1653         if (dev == NULL) {
1654                 err = -ENODEV;
1655                 goto errout;
1656         }
1657
1658         err = validate_linkmsg(dev, tb);
1659         if (err < 0)
1660                 goto errout;
1661
1662         err = do_setlink(dev, ifm, tb, ifname, 0);
1663 errout:
1664         return err;
1665 }
1666
1667 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1668 {
1669         struct net *net = sock_net(skb->sk);
1670         const struct rtnl_link_ops *ops;
1671         struct net_device *dev;
1672         struct ifinfomsg *ifm;
1673         char ifname[IFNAMSIZ];
1674         struct nlattr *tb[IFLA_MAX+1];
1675         int err;
1676         LIST_HEAD(list_kill);
1677
1678         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1679         if (err < 0)
1680                 return err;
1681
1682         if (tb[IFLA_IFNAME])
1683                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1684
1685         ifm = nlmsg_data(nlh);
1686         if (ifm->ifi_index > 0)
1687                 dev = __dev_get_by_index(net, ifm->ifi_index);
1688         else if (tb[IFLA_IFNAME])
1689                 dev = __dev_get_by_name(net, ifname);
1690         else
1691                 return -EINVAL;
1692
1693         if (!dev)
1694                 return -ENODEV;
1695
1696         ops = dev->rtnl_link_ops;
1697         if (!ops)
1698                 return -EOPNOTSUPP;
1699
1700         ops->dellink(dev, &list_kill);
1701         unregister_netdevice_many(&list_kill);
1702         list_del(&list_kill);
1703         return 0;
1704 }
1705
1706 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1707 {
1708         unsigned int old_flags;
1709         int err;
1710
1711         old_flags = dev->flags;
1712         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1713                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1714                 if (err < 0)
1715                         return err;
1716         }
1717
1718         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1719
1720         __dev_notify_flags(dev, old_flags, ~0U);
1721         return 0;
1722 }
1723 EXPORT_SYMBOL(rtnl_configure_link);
1724
1725 struct net_device *rtnl_create_link(struct net *net,
1726         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1727 {
1728         int err;
1729         struct net_device *dev;
1730         unsigned int num_tx_queues = 1;
1731         unsigned int num_rx_queues = 1;
1732
1733         if (tb[IFLA_NUM_TX_QUEUES])
1734                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1735         else if (ops->get_num_tx_queues)
1736                 num_tx_queues = ops->get_num_tx_queues();
1737
1738         if (tb[IFLA_NUM_RX_QUEUES])
1739                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1740         else if (ops->get_num_rx_queues)
1741                 num_rx_queues = ops->get_num_rx_queues();
1742
1743         err = -ENOMEM;
1744         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1745                                num_tx_queues, num_rx_queues);
1746         if (!dev)
1747                 goto err;
1748
1749         dev_net_set(dev, net);
1750         dev->rtnl_link_ops = ops;
1751         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1752
1753         if (tb[IFLA_MTU])
1754                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1755         if (tb[IFLA_ADDRESS]) {
1756                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1757                                 nla_len(tb[IFLA_ADDRESS]));
1758                 dev->addr_assign_type = NET_ADDR_SET;
1759         }
1760         if (tb[IFLA_BROADCAST])
1761                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1762                                 nla_len(tb[IFLA_BROADCAST]));
1763         if (tb[IFLA_TXQLEN])
1764                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1765         if (tb[IFLA_OPERSTATE])
1766                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1767         if (tb[IFLA_LINKMODE])
1768                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1769         if (tb[IFLA_GROUP])
1770                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1771
1772         return dev;
1773
1774 err:
1775         return ERR_PTR(err);
1776 }
1777 EXPORT_SYMBOL(rtnl_create_link);
1778
1779 static int rtnl_group_changelink(struct net *net, int group,
1780                 struct ifinfomsg *ifm,
1781                 struct nlattr **tb)
1782 {
1783         struct net_device *dev;
1784         int err;
1785
1786         for_each_netdev(net, dev) {
1787                 if (dev->group == group) {
1788                         err = do_setlink(dev, ifm, tb, NULL, 0);
1789                         if (err < 0)
1790                                 return err;
1791                 }
1792         }
1793
1794         return 0;
1795 }
1796
1797 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1798 {
1799         struct net *net = sock_net(skb->sk);
1800         const struct rtnl_link_ops *ops;
1801         const struct rtnl_link_ops *m_ops = NULL;
1802         struct net_device *dev;
1803         struct net_device *master_dev = NULL;
1804         struct ifinfomsg *ifm;
1805         char kind[MODULE_NAME_LEN];
1806         char ifname[IFNAMSIZ];
1807         struct nlattr *tb[IFLA_MAX+1];
1808         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1809         int err;
1810
1811 #ifdef CONFIG_MODULES
1812 replay:
1813 #endif
1814         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1815         if (err < 0)
1816                 return err;
1817
1818         if (tb[IFLA_IFNAME])
1819                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1820         else
1821                 ifname[0] = '\0';
1822
1823         ifm = nlmsg_data(nlh);
1824         if (ifm->ifi_index > 0)
1825                 dev = __dev_get_by_index(net, ifm->ifi_index);
1826         else {
1827                 if (ifname[0])
1828                         dev = __dev_get_by_name(net, ifname);
1829                 else
1830                         dev = NULL;
1831         }
1832
1833         if (dev) {
1834                 master_dev = netdev_master_upper_dev_get(dev);
1835                 if (master_dev)
1836                         m_ops = master_dev->rtnl_link_ops;
1837         }
1838
1839         err = validate_linkmsg(dev, tb);
1840         if (err < 0)
1841                 return err;
1842
1843         if (tb[IFLA_LINKINFO]) {
1844                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1845                                        tb[IFLA_LINKINFO], ifla_info_policy);
1846                 if (err < 0)
1847                         return err;
1848         } else
1849                 memset(linkinfo, 0, sizeof(linkinfo));
1850
1851         if (linkinfo[IFLA_INFO_KIND]) {
1852                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1853                 ops = rtnl_link_ops_get(kind);
1854         } else {
1855                 kind[0] = '\0';
1856                 ops = NULL;
1857         }
1858
1859         if (1) {
1860                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1861                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1862                 struct nlattr **data = NULL;
1863                 struct nlattr **slave_data = NULL;
1864                 struct net *dest_net;
1865
1866                 if (ops) {
1867                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1868                                 err = nla_parse_nested(attr, ops->maxtype,
1869                                                        linkinfo[IFLA_INFO_DATA],
1870                                                        ops->policy);
1871                                 if (err < 0)
1872                                         return err;
1873                                 data = attr;
1874                         }
1875                         if (ops->validate) {
1876                                 err = ops->validate(tb, data);
1877                                 if (err < 0)
1878                                         return err;
1879                         }
1880                 }
1881
1882                 if (m_ops) {
1883                         if (m_ops->slave_maxtype &&
1884                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
1885                                 err = nla_parse_nested(slave_attr,
1886                                                        m_ops->slave_maxtype,
1887                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
1888                                                        m_ops->slave_policy);
1889                                 if (err < 0)
1890                                         return err;
1891                                 slave_data = slave_attr;
1892                         }
1893                         if (m_ops->slave_validate) {
1894                                 err = m_ops->slave_validate(tb, slave_data);
1895                                 if (err < 0)
1896                                         return err;
1897                         }
1898                 }
1899
1900                 if (dev) {
1901                         int modified = 0;
1902
1903                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1904                                 return -EEXIST;
1905                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1906                                 return -EOPNOTSUPP;
1907
1908                         if (linkinfo[IFLA_INFO_DATA]) {
1909                                 if (!ops || ops != dev->rtnl_link_ops ||
1910                                     !ops->changelink)
1911                                         return -EOPNOTSUPP;
1912
1913                                 err = ops->changelink(dev, tb, data);
1914                                 if (err < 0)
1915                                         return err;
1916                                 modified = 1;
1917                         }
1918
1919                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1920                                 if (!m_ops || !m_ops->slave_changelink)
1921                                         return -EOPNOTSUPP;
1922
1923                                 err = m_ops->slave_changelink(master_dev, dev,
1924                                                               tb, slave_data);
1925                                 if (err < 0)
1926                                         return err;
1927                                 modified = 1;
1928                         }
1929
1930                         return do_setlink(dev, ifm, tb, ifname, modified);
1931                 }
1932
1933                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1934                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1935                                 return rtnl_group_changelink(net,
1936                                                 nla_get_u32(tb[IFLA_GROUP]),
1937                                                 ifm, tb);
1938                         return -ENODEV;
1939                 }
1940
1941                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1942                         return -EOPNOTSUPP;
1943
1944                 if (!ops) {
1945 #ifdef CONFIG_MODULES
1946                         if (kind[0]) {
1947                                 __rtnl_unlock();
1948                                 request_module("rtnl-link-%s", kind);
1949                                 rtnl_lock();
1950                                 ops = rtnl_link_ops_get(kind);
1951                                 if (ops)
1952                                         goto replay;
1953                         }
1954 #endif
1955                         return -EOPNOTSUPP;
1956                 }
1957
1958                 if (!ifname[0])
1959                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1960
1961                 dest_net = rtnl_link_get_net(net, tb);
1962                 if (IS_ERR(dest_net))
1963                         return PTR_ERR(dest_net);
1964
1965                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1966                 if (IS_ERR(dev)) {
1967                         err = PTR_ERR(dev);
1968                         goto out;
1969                 }
1970
1971                 dev->ifindex = ifm->ifi_index;
1972
1973                 if (ops->newlink) {
1974                         err = ops->newlink(net, dev, tb, data);
1975                         /* Drivers should call free_netdev() in ->destructor
1976                          * and unregister it on failure so that device could be
1977                          * finally freed in rtnl_unlock.
1978                          */
1979                         if (err < 0)
1980                                 goto out;
1981                 } else {
1982                         err = register_netdevice(dev);
1983                         if (err < 0) {
1984                                 free_netdev(dev);
1985                                 goto out;
1986                         }
1987                 }
1988                 err = rtnl_configure_link(dev, ifm);
1989                 if (err < 0)
1990                         unregister_netdevice(dev);
1991 out:
1992                 put_net(dest_net);
1993                 return err;
1994         }
1995 }
1996
1997 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1998 {
1999         struct net *net = sock_net(skb->sk);
2000         struct ifinfomsg *ifm;
2001         char ifname[IFNAMSIZ];
2002         struct nlattr *tb[IFLA_MAX+1];
2003         struct net_device *dev = NULL;
2004         struct sk_buff *nskb;
2005         int err;
2006         u32 ext_filter_mask = 0;
2007
2008         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2009         if (err < 0)
2010                 return err;
2011
2012         if (tb[IFLA_IFNAME])
2013                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2014
2015         if (tb[IFLA_EXT_MASK])
2016                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2017
2018         ifm = nlmsg_data(nlh);
2019         if (ifm->ifi_index > 0)
2020                 dev = __dev_get_by_index(net, ifm->ifi_index);
2021         else if (tb[IFLA_IFNAME])
2022                 dev = __dev_get_by_name(net, ifname);
2023         else
2024                 return -EINVAL;
2025
2026         if (dev == NULL)
2027                 return -ENODEV;
2028
2029         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2030         if (nskb == NULL)
2031                 return -ENOBUFS;
2032
2033         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2034                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2035         if (err < 0) {
2036                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2037                 WARN_ON(err == -EMSGSIZE);
2038                 kfree_skb(nskb);
2039         } else
2040                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2041
2042         return err;
2043 }
2044
2045 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2046 {
2047         struct net *net = sock_net(skb->sk);
2048         struct net_device *dev;
2049         struct nlattr *tb[IFLA_MAX+1];
2050         u32 ext_filter_mask = 0;
2051         u16 min_ifinfo_dump_size = 0;
2052
2053         if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2054                         ifla_policy) >= 0) {
2055                 if (tb[IFLA_EXT_MASK])
2056                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2057         }
2058
2059         if (!ext_filter_mask)
2060                 return NLMSG_GOODSIZE;
2061         /*
2062          * traverse the list of net devices and compute the minimum
2063          * buffer size based upon the filter mask.
2064          */
2065         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2066                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2067                                              if_nlmsg_size(dev,
2068                                                            ext_filter_mask));
2069         }
2070
2071         return min_ifinfo_dump_size;
2072 }
2073
2074 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2075 {
2076         int idx;
2077         int s_idx = cb->family;
2078
2079         if (s_idx == 0)
2080                 s_idx = 1;
2081         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2082                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2083                 if (idx < s_idx || idx == PF_PACKET)
2084                         continue;
2085                 if (rtnl_msg_handlers[idx] == NULL ||
2086                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2087                         continue;
2088                 if (idx > s_idx) {
2089                         memset(&cb->args[0], 0, sizeof(cb->args));
2090                         cb->prev_seq = 0;
2091                         cb->seq = 0;
2092                 }
2093                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2094                         break;
2095         }
2096         cb->family = idx;
2097
2098         return skb->len;
2099 }
2100
2101 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2102                   gfp_t flags)
2103 {
2104         struct net *net = dev_net(dev);
2105         struct sk_buff *skb;
2106         int err = -ENOBUFS;
2107         size_t if_info_size;
2108
2109         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2110         if (skb == NULL)
2111                 goto errout;
2112
2113         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2114         if (err < 0) {
2115                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2116                 WARN_ON(err == -EMSGSIZE);
2117                 kfree_skb(skb);
2118                 goto errout;
2119         }
2120         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2121         return;
2122 errout:
2123         if (err < 0)
2124                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2125 }
2126 EXPORT_SYMBOL(rtmsg_ifinfo);
2127
2128 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2129                                    struct net_device *dev,
2130                                    u8 *addr, u32 pid, u32 seq,
2131                                    int type, unsigned int flags,
2132                                    int nlflags)
2133 {
2134         struct nlmsghdr *nlh;
2135         struct ndmsg *ndm;
2136
2137         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2138         if (!nlh)
2139                 return -EMSGSIZE;
2140
2141         ndm = nlmsg_data(nlh);
2142         ndm->ndm_family  = AF_BRIDGE;
2143         ndm->ndm_pad1    = 0;
2144         ndm->ndm_pad2    = 0;
2145         ndm->ndm_flags   = flags;
2146         ndm->ndm_type    = 0;
2147         ndm->ndm_ifindex = dev->ifindex;
2148         ndm->ndm_state   = NUD_PERMANENT;
2149
2150         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2151                 goto nla_put_failure;
2152
2153         return nlmsg_end(skb, nlh);
2154
2155 nla_put_failure:
2156         nlmsg_cancel(skb, nlh);
2157         return -EMSGSIZE;
2158 }
2159
2160 static inline size_t rtnl_fdb_nlmsg_size(void)
2161 {
2162         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2163 }
2164
2165 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2166 {
2167         struct net *net = dev_net(dev);
2168         struct sk_buff *skb;
2169         int err = -ENOBUFS;
2170
2171         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2172         if (!skb)
2173                 goto errout;
2174
2175         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2176         if (err < 0) {
2177                 kfree_skb(skb);
2178                 goto errout;
2179         }
2180
2181         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2182         return;
2183 errout:
2184         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2185 }
2186
2187 /**
2188  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2189  */
2190 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2191                      struct nlattr *tb[],
2192                      struct net_device *dev,
2193                      const unsigned char *addr,
2194                      u16 flags)
2195 {
2196         int err = -EINVAL;
2197
2198         /* If aging addresses are supported device will need to
2199          * implement its own handler for this.
2200          */
2201         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2202                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2203                 return err;
2204         }
2205
2206         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2207                 err = dev_uc_add_excl(dev, addr);
2208         else if (is_multicast_ether_addr(addr))
2209                 err = dev_mc_add_excl(dev, addr);
2210
2211         /* Only return duplicate errors if NLM_F_EXCL is set */
2212         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2213                 err = 0;
2214
2215         return err;
2216 }
2217 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2218
2219 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2220 {
2221         struct net *net = sock_net(skb->sk);
2222         struct ndmsg *ndm;
2223         struct nlattr *tb[NDA_MAX+1];
2224         struct net_device *dev;
2225         u8 *addr;
2226         int err;
2227
2228         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2229         if (err < 0)
2230                 return err;
2231
2232         ndm = nlmsg_data(nlh);
2233         if (ndm->ndm_ifindex == 0) {
2234                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2235                 return -EINVAL;
2236         }
2237
2238         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2239         if (dev == NULL) {
2240                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2241                 return -ENODEV;
2242         }
2243
2244         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2245                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2246                 return -EINVAL;
2247         }
2248
2249         addr = nla_data(tb[NDA_LLADDR]);
2250
2251         err = -EOPNOTSUPP;
2252
2253         /* Support fdb on master device the net/bridge default case */
2254         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2255             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2256                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2257                 const struct net_device_ops *ops = br_dev->netdev_ops;
2258
2259                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2260                 if (err)
2261                         goto out;
2262                 else
2263                         ndm->ndm_flags &= ~NTF_MASTER;
2264         }
2265
2266         /* Embedded bridge, macvlan, and any other device support */
2267         if ((ndm->ndm_flags & NTF_SELF)) {
2268                 if (dev->netdev_ops->ndo_fdb_add)
2269                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2270                                                            nlh->nlmsg_flags);
2271                 else
2272                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2273                                                nlh->nlmsg_flags);
2274
2275                 if (!err) {
2276                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2277                         ndm->ndm_flags &= ~NTF_SELF;
2278                 }
2279         }
2280 out:
2281         return err;
2282 }
2283
2284 /**
2285  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2286  */
2287 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2288                      struct nlattr *tb[],
2289                      struct net_device *dev,
2290                      const unsigned char *addr)
2291 {
2292         int err = -EOPNOTSUPP;
2293
2294         /* If aging addresses are supported device will need to
2295          * implement its own handler for this.
2296          */
2297         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2298                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2299                 return -EINVAL;
2300         }
2301
2302         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2303                 err = dev_uc_del(dev, addr);
2304         else if (is_multicast_ether_addr(addr))
2305                 err = dev_mc_del(dev, addr);
2306         else
2307                 err = -EINVAL;
2308
2309         return err;
2310 }
2311 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2312
2313 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2314 {
2315         struct net *net = sock_net(skb->sk);
2316         struct ndmsg *ndm;
2317         struct nlattr *tb[NDA_MAX+1];
2318         struct net_device *dev;
2319         int err = -EINVAL;
2320         __u8 *addr;
2321
2322         if (!capable(CAP_NET_ADMIN))
2323                 return -EPERM;
2324
2325         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2326         if (err < 0)
2327                 return err;
2328
2329         ndm = nlmsg_data(nlh);
2330         if (ndm->ndm_ifindex == 0) {
2331                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2332                 return -EINVAL;
2333         }
2334
2335         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2336         if (dev == NULL) {
2337                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2338                 return -ENODEV;
2339         }
2340
2341         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2342                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2343                 return -EINVAL;
2344         }
2345
2346         addr = nla_data(tb[NDA_LLADDR]);
2347
2348         err = -EOPNOTSUPP;
2349
2350         /* Support fdb on master device the net/bridge default case */
2351         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2352             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2353                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2354                 const struct net_device_ops *ops = br_dev->netdev_ops;
2355
2356                 if (ops->ndo_fdb_del)
2357                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2358
2359                 if (err)
2360                         goto out;
2361                 else
2362                         ndm->ndm_flags &= ~NTF_MASTER;
2363         }
2364
2365         /* Embedded bridge, macvlan, and any other device support */
2366         if (ndm->ndm_flags & NTF_SELF) {
2367                 if (dev->netdev_ops->ndo_fdb_del)
2368                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2369                 else
2370                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2371
2372                 if (!err) {
2373                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2374                         ndm->ndm_flags &= ~NTF_SELF;
2375                 }
2376         }
2377 out:
2378         return err;
2379 }
2380
2381 static int nlmsg_populate_fdb(struct sk_buff *skb,
2382                               struct netlink_callback *cb,
2383                               struct net_device *dev,
2384                               int *idx,
2385                               struct netdev_hw_addr_list *list)
2386 {
2387         struct netdev_hw_addr *ha;
2388         int err;
2389         u32 portid, seq;
2390
2391         portid = NETLINK_CB(cb->skb).portid;
2392         seq = cb->nlh->nlmsg_seq;
2393
2394         list_for_each_entry(ha, &list->list, list) {
2395                 if (*idx < cb->args[0])
2396                         goto skip;
2397
2398                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2399                                               portid, seq,
2400                                               RTM_NEWNEIGH, NTF_SELF,
2401                                               NLM_F_MULTI);
2402                 if (err < 0)
2403                         return err;
2404 skip:
2405                 *idx += 1;
2406         }
2407         return 0;
2408 }
2409
2410 /**
2411  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2412  * @nlh: netlink message header
2413  * @dev: netdevice
2414  *
2415  * Default netdevice operation to dump the existing unicast address list.
2416  * Returns number of addresses from list put in skb.
2417  */
2418 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2419                       struct netlink_callback *cb,
2420                       struct net_device *dev,
2421                       int idx)
2422 {
2423         int err;
2424
2425         netif_addr_lock_bh(dev);
2426         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2427         if (err)
2428                 goto out;
2429         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2430 out:
2431         netif_addr_unlock_bh(dev);
2432         return idx;
2433 }
2434 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2435
2436 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2437 {
2438         int idx = 0;
2439         struct net *net = sock_net(skb->sk);
2440         struct net_device *dev;
2441
2442         rcu_read_lock();
2443         for_each_netdev_rcu(net, dev) {
2444                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2445                         struct net_device *br_dev;
2446                         const struct net_device_ops *ops;
2447
2448                         br_dev = netdev_master_upper_dev_get(dev);
2449                         ops = br_dev->netdev_ops;
2450                         if (ops->ndo_fdb_dump)
2451                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2452                 }
2453
2454                 if (dev->netdev_ops->ndo_fdb_dump)
2455                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2456                 else
2457                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2458         }
2459         rcu_read_unlock();
2460
2461         cb->args[0] = idx;
2462         return skb->len;
2463 }
2464
2465 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2466                             struct net_device *dev, u16 mode)
2467 {
2468         struct nlmsghdr *nlh;
2469         struct ifinfomsg *ifm;
2470         struct nlattr *br_afspec;
2471         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2472         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2473
2474         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2475         if (nlh == NULL)
2476                 return -EMSGSIZE;
2477
2478         ifm = nlmsg_data(nlh);
2479         ifm->ifi_family = AF_BRIDGE;
2480         ifm->__ifi_pad = 0;
2481         ifm->ifi_type = dev->type;
2482         ifm->ifi_index = dev->ifindex;
2483         ifm->ifi_flags = dev_get_flags(dev);
2484         ifm->ifi_change = 0;
2485
2486
2487         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2488             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2489             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2490             (br_dev &&
2491              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2492             (dev->addr_len &&
2493              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2494             (dev->ifindex != dev->iflink &&
2495              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2496                 goto nla_put_failure;
2497
2498         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2499         if (!br_afspec)
2500                 goto nla_put_failure;
2501
2502         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2503             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2504                 nla_nest_cancel(skb, br_afspec);
2505                 goto nla_put_failure;
2506         }
2507         nla_nest_end(skb, br_afspec);
2508
2509         return nlmsg_end(skb, nlh);
2510 nla_put_failure:
2511         nlmsg_cancel(skb, nlh);
2512         return -EMSGSIZE;
2513 }
2514 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2515
2516 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2517 {
2518         struct net *net = sock_net(skb->sk);
2519         struct net_device *dev;
2520         int idx = 0;
2521         u32 portid = NETLINK_CB(cb->skb).portid;
2522         u32 seq = cb->nlh->nlmsg_seq;
2523         struct nlattr *extfilt;
2524         u32 filter_mask = 0;
2525
2526         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2527                                   IFLA_EXT_MASK);
2528         if (extfilt)
2529                 filter_mask = nla_get_u32(extfilt);
2530
2531         rcu_read_lock();
2532         for_each_netdev_rcu(net, dev) {
2533                 const struct net_device_ops *ops = dev->netdev_ops;
2534                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2535
2536                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2537                         if (idx >= cb->args[0] &&
2538                             br_dev->netdev_ops->ndo_bridge_getlink(
2539                                     skb, portid, seq, dev, filter_mask) < 0)
2540                                 break;
2541                         idx++;
2542                 }
2543
2544                 if (ops->ndo_bridge_getlink) {
2545                         if (idx >= cb->args[0] &&
2546                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2547                                                     filter_mask) < 0)
2548                                 break;
2549                         idx++;
2550                 }
2551         }
2552         rcu_read_unlock();
2553         cb->args[0] = idx;
2554
2555         return skb->len;
2556 }
2557
2558 static inline size_t bridge_nlmsg_size(void)
2559 {
2560         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2561                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2562                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2563                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2564                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2565                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2566                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2567                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2568                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2569                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2570                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2571 }
2572
2573 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2574 {
2575         struct net *net = dev_net(dev);
2576         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2577         struct sk_buff *skb;
2578         int err = -EOPNOTSUPP;
2579
2580         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2581         if (!skb) {
2582                 err = -ENOMEM;
2583                 goto errout;
2584         }
2585
2586         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2587             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2588                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2589                 if (err < 0)
2590                         goto errout;
2591         }
2592
2593         if ((flags & BRIDGE_FLAGS_SELF) &&
2594             dev->netdev_ops->ndo_bridge_getlink) {
2595                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2596                 if (err < 0)
2597                         goto errout;
2598         }
2599
2600         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2601         return 0;
2602 errout:
2603         WARN_ON(err == -EMSGSIZE);
2604         kfree_skb(skb);
2605         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2606         return err;
2607 }
2608
2609 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2610 {
2611         struct net *net = sock_net(skb->sk);
2612         struct ifinfomsg *ifm;
2613         struct net_device *dev;
2614         struct nlattr *br_spec, *attr = NULL;
2615         int rem, err = -EOPNOTSUPP;
2616         u16 oflags, flags = 0;
2617         bool have_flags = false;
2618
2619         if (nlmsg_len(nlh) < sizeof(*ifm))
2620                 return -EINVAL;
2621
2622         ifm = nlmsg_data(nlh);
2623         if (ifm->ifi_family != AF_BRIDGE)
2624                 return -EPFNOSUPPORT;
2625
2626         dev = __dev_get_by_index(net, ifm->ifi_index);
2627         if (!dev) {
2628                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2629                 return -ENODEV;
2630         }
2631
2632         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2633         if (br_spec) {
2634                 nla_for_each_nested(attr, br_spec, rem) {
2635                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2636                                 have_flags = true;
2637                                 flags = nla_get_u16(attr);
2638                                 break;
2639                         }
2640                 }
2641         }
2642
2643         oflags = flags;
2644
2645         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2646                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2647
2648                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2649                         err = -EOPNOTSUPP;
2650                         goto out;
2651                 }
2652
2653                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2654                 if (err)
2655                         goto out;
2656
2657                 flags &= ~BRIDGE_FLAGS_MASTER;
2658         }
2659
2660         if ((flags & BRIDGE_FLAGS_SELF)) {
2661                 if (!dev->netdev_ops->ndo_bridge_setlink)
2662                         err = -EOPNOTSUPP;
2663                 else
2664                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2665
2666                 if (!err)
2667                         flags &= ~BRIDGE_FLAGS_SELF;
2668         }
2669
2670         if (have_flags)
2671                 memcpy(nla_data(attr), &flags, sizeof(flags));
2672         /* Generate event to notify upper layer of bridge change */
2673         if (!err)
2674                 err = rtnl_bridge_notify(dev, oflags);
2675 out:
2676         return err;
2677 }
2678
2679 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2680 {
2681         struct net *net = sock_net(skb->sk);
2682         struct ifinfomsg *ifm;
2683         struct net_device *dev;
2684         struct nlattr *br_spec, *attr = NULL;
2685         int rem, err = -EOPNOTSUPP;
2686         u16 oflags, flags = 0;
2687         bool have_flags = false;
2688
2689         if (nlmsg_len(nlh) < sizeof(*ifm))
2690                 return -EINVAL;
2691
2692         ifm = nlmsg_data(nlh);
2693         if (ifm->ifi_family != AF_BRIDGE)
2694                 return -EPFNOSUPPORT;
2695
2696         dev = __dev_get_by_index(net, ifm->ifi_index);
2697         if (!dev) {
2698                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2699                 return -ENODEV;
2700         }
2701
2702         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2703         if (br_spec) {
2704                 nla_for_each_nested(attr, br_spec, rem) {
2705                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2706                                 have_flags = true;
2707                                 flags = nla_get_u16(attr);
2708                                 break;
2709                         }
2710                 }
2711         }
2712
2713         oflags = flags;
2714
2715         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2716                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2717
2718                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2719                         err = -EOPNOTSUPP;
2720                         goto out;
2721                 }
2722
2723                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2724                 if (err)
2725                         goto out;
2726
2727                 flags &= ~BRIDGE_FLAGS_MASTER;
2728         }
2729
2730         if ((flags & BRIDGE_FLAGS_SELF)) {
2731                 if (!dev->netdev_ops->ndo_bridge_dellink)
2732                         err = -EOPNOTSUPP;
2733                 else
2734                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2735
2736                 if (!err)
2737                         flags &= ~BRIDGE_FLAGS_SELF;
2738         }
2739
2740         if (have_flags)
2741                 memcpy(nla_data(attr), &flags, sizeof(flags));
2742         /* Generate event to notify upper layer of bridge change */
2743         if (!err)
2744                 err = rtnl_bridge_notify(dev, oflags);
2745 out:
2746         return err;
2747 }
2748
2749 /* Process one rtnetlink message. */
2750
2751 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2752 {
2753         struct net *net = sock_net(skb->sk);
2754         rtnl_doit_func doit;
2755         int sz_idx, kind;
2756         int family;
2757         int type;
2758         int err;
2759
2760         type = nlh->nlmsg_type;
2761         if (type > RTM_MAX)
2762                 return -EOPNOTSUPP;
2763
2764         type -= RTM_BASE;
2765
2766         /* All the messages must have at least 1 byte length */
2767         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2768                 return 0;
2769
2770         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2771         sz_idx = type>>2;
2772         kind = type&3;
2773
2774         if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2775                 return -EPERM;
2776
2777         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2778                 struct sock *rtnl;
2779                 rtnl_dumpit_func dumpit;
2780                 rtnl_calcit_func calcit;
2781                 u16 min_dump_alloc = 0;
2782
2783                 dumpit = rtnl_get_dumpit(family, type);
2784                 if (dumpit == NULL)
2785                         return -EOPNOTSUPP;
2786                 calcit = rtnl_get_calcit(family, type);
2787                 if (calcit)
2788                         min_dump_alloc = calcit(skb, nlh);
2789
2790                 __rtnl_unlock();
2791                 rtnl = net->rtnl;
2792                 {
2793                         struct netlink_dump_control c = {
2794                                 .dump           = dumpit,
2795                                 .min_dump_alloc = min_dump_alloc,
2796                         };
2797                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2798                 }
2799                 rtnl_lock();
2800                 return err;
2801         }
2802
2803         doit = rtnl_get_doit(family, type);
2804         if (doit == NULL)
2805                 return -EOPNOTSUPP;
2806
2807         return doit(skb, nlh);
2808 }
2809
2810 static void rtnetlink_rcv(struct sk_buff *skb)
2811 {
2812         rtnl_lock();
2813         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2814         rtnl_unlock();
2815 }
2816
2817 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2818 {
2819         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2820
2821         switch (event) {
2822         case NETDEV_UP:
2823         case NETDEV_DOWN:
2824         case NETDEV_PRE_UP:
2825         case NETDEV_POST_INIT:
2826         case NETDEV_REGISTER:
2827         case NETDEV_CHANGE:
2828         case NETDEV_PRE_TYPE_CHANGE:
2829         case NETDEV_GOING_DOWN:
2830         case NETDEV_UNREGISTER:
2831         case NETDEV_UNREGISTER_FINAL:
2832         case NETDEV_RELEASE:
2833         case NETDEV_JOIN:
2834                 break;
2835         default:
2836                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2837                 break;
2838         }
2839         return NOTIFY_DONE;
2840 }
2841
2842 static struct notifier_block rtnetlink_dev_notifier = {
2843         .notifier_call  = rtnetlink_event,
2844 };
2845
2846
2847 static int __net_init rtnetlink_net_init(struct net *net)
2848 {
2849         struct sock *sk;
2850         struct netlink_kernel_cfg cfg = {
2851                 .groups         = RTNLGRP_MAX,
2852                 .input          = rtnetlink_rcv,
2853                 .cb_mutex       = &rtnl_mutex,
2854                 .flags          = NL_CFG_F_NONROOT_RECV,
2855         };
2856
2857         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2858         if (!sk)
2859                 return -ENOMEM;
2860         net->rtnl = sk;
2861         return 0;
2862 }
2863
2864 static void __net_exit rtnetlink_net_exit(struct net *net)
2865 {
2866         netlink_kernel_release(net->rtnl);
2867         net->rtnl = NULL;
2868 }
2869
2870 static struct pernet_operations rtnetlink_net_ops = {
2871         .init = rtnetlink_net_init,
2872         .exit = rtnetlink_net_exit,
2873 };
2874
2875 void __init rtnetlink_init(void)
2876 {
2877         if (register_pernet_subsys(&rtnetlink_net_ops))
2878                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2879
2880         register_netdevice_notifier(&rtnetlink_dev_notifier);
2881
2882         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2883                       rtnl_dump_ifinfo, rtnl_calcit);
2884         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2885         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2886         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2887
2888         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2889         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2890
2891         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2892         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2893         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2894
2895         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2896         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2897         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2898 }
2899