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/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 static struct sk_buff *defer_kfree_skb_list;
  75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  76 {
  77         if (head && tail) {
  78                 tail->next = defer_kfree_skb_list;
  79                 defer_kfree_skb_list = head;
  80         }
  81 }
  82 EXPORT_SYMBOL(rtnl_kfree_skbs);
  83
  84 void __rtnl_unlock(void)
  85 {
  86         struct sk_buff *head = defer_kfree_skb_list;
  87
  88         defer_kfree_skb_list = NULL;
  89
  90         mutex_unlock(&rtnl_mutex);
  91
  92         while (head) {
  93                 struct sk_buff *next = head->next;
  94
  95                 kfree_skb(head);
  96                 cond_resched();
  97                 head = next;
  98         }
  99 }
 100
 101 void rtnl_unlock(void)
 102 {
 103         /* This fellow will unlock it for us. */
 104         netdev_run_todo();
 105 }
 106 EXPORT_SYMBOL(rtnl_unlock);
 107
 108 int rtnl_trylock(void)
 109 {
 110         return mutex_trylock(&rtnl_mutex);
 111 }
 112 EXPORT_SYMBOL(rtnl_trylock);
 113
 114 int rtnl_is_locked(void)
 115 {
 116         return mutex_is_locked(&rtnl_mutex);
 117 }
 118 EXPORT_SYMBOL(rtnl_is_locked);
 119
 120 #ifdef CONFIG_PROVE_LOCKING
 121 bool lockdep_rtnl_is_held(void)
 122 {
 123         return lockdep_is_held(&rtnl_mutex);
 124 }
 125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 127
 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 129
 130 static inline int rtm_msgindex(int msgtype)
 131 {
 132         int msgindex = msgtype - RTM_BASE;
 133
 134         /*
 135          * msgindex < 0 implies someone tried to register a netlink
 136          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 137          * the message type has not been added to linux/rtnetlink.h
 138          */
 139         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 140
 141         return msgindex;
 142 }
 143
 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 145 {
 146         struct rtnl_link *tab;
 147
 148         if (protocol <= RTNL_FAMILY_MAX)
 149                 tab = rtnl_msg_handlers[protocol];
 150         else
 151                 tab = NULL;
 152
 153         if (tab == NULL || tab[msgindex].doit == NULL)
 154                 tab = rtnl_msg_handlers[PF_UNSPEC];
 155
 156         return tab[msgindex].doit;
 157 }
 158
 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 160 {
 161         struct rtnl_link *tab;
 162
 163         if (protocol <= RTNL_FAMILY_MAX)
 164                 tab = rtnl_msg_handlers[protocol];
 165         else
 166                 tab = NULL;
 167
 168         if (tab == NULL || tab[msgindex].dumpit == NULL)
 169                 tab = rtnl_msg_handlers[PF_UNSPEC];
 170
 171         return tab[msgindex].dumpit;
 172 }
 173
 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 175 {
 176         struct rtnl_link *tab;
 177
 178         if (protocol <= RTNL_FAMILY_MAX)
 179                 tab = rtnl_msg_handlers[protocol];
 180         else
 181                 tab = NULL;
 182
 183         if (tab == NULL || tab[msgindex].calcit == NULL)
 184                 tab = rtnl_msg_handlers[PF_UNSPEC];
 185
 186         return tab[msgindex].calcit;
 187 }
 188
 189 /**
 190  * __rtnl_register - Register a rtnetlink message type
 191  * @protocol: Protocol family or PF_UNSPEC
 192  * @msgtype: rtnetlink message type
 193  * @doit: Function pointer called for each request message
 194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 195  * @calcit: Function pointer to calc size of dump message
 196  *
 197  * Registers the specified function pointers (at least one of them has
 198  * to be non-NULL) to be called whenever a request message for the
 199  * specified protocol family and message type is received.
 200  *
 201  * The special protocol family PF_UNSPEC may be used to define fallback
 202  * function pointers for the case when no entry for the specific protocol
 203  * family exists.
 204  *
 205  * Returns 0 on success or a negative error code.
 206  */
 207 int __rtnl_register(int protocol, int msgtype,
 208                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 209                     rtnl_calcit_func calcit)
 210 {
 211         struct rtnl_link *tab;
 212         int msgindex;
 213
 214         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 215         msgindex = rtm_msgindex(msgtype);
 216
 217         tab = rtnl_msg_handlers[protocol];
 218         if (tab == NULL) {
 219                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 220                 if (tab == NULL)
 221                         return -ENOBUFS;
 222
 223                 rtnl_msg_handlers[protocol] = tab;
 224         }
 225
 226         if (doit)
 227                 tab[msgindex].doit = doit;
 228
 229         if (dumpit)
 230                 tab[msgindex].dumpit = dumpit;
 231
 232         if (calcit)
 233                 tab[msgindex].calcit = calcit;
 234
 235         return 0;
 236 }
 237 EXPORT_SYMBOL_GPL(__rtnl_register);
 238
 239 /**
 240  * rtnl_register - Register a rtnetlink message type
 241  *
 242  * Identical to __rtnl_register() but panics on failure. This is useful
 243  * as failure of this function is very unlikely, it can only happen due
 244  * to lack of memory when allocating the chain to store all message
 245  * handlers for a protocol. Meant for use in init functions where lack
 246  * of memory implies no sense in continuing.
 247  */
 248 void rtnl_register(int protocol, int msgtype,
 249                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 250                    rtnl_calcit_func calcit)
 251 {
 252         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 253                 panic("Unable to register rtnetlink message handler, "
 254                       "protocol = %d, message type = %d\n",
 255                       protocol, msgtype);
 256 }
 257 EXPORT_SYMBOL_GPL(rtnl_register);
 258
 259 /**
 260  * rtnl_unregister - Unregister a rtnetlink message type
 261  * @protocol: Protocol family or PF_UNSPEC
 262  * @msgtype: rtnetlink message type
 263  *
 264  * Returns 0 on success or a negative error code.
 265  */
 266 int rtnl_unregister(int protocol, int msgtype)
 267 {
 268         int msgindex;
 269
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271         msgindex = rtm_msgindex(msgtype);
 272
 273         if (rtnl_msg_handlers[protocol] == NULL)
 274                 return -ENOENT;
 275
 276         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 277         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 278
 279         return 0;
 280 }
 281 EXPORT_SYMBOL_GPL(rtnl_unregister);
 282
 283 /**
 284  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 285  * @protocol : Protocol family or PF_UNSPEC
 286  *
 287  * Identical to calling rtnl_unregster() for all registered message types
 288  * of a certain protocol family.
 289  */
 290 void rtnl_unregister_all(int protocol)
 291 {
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293
 294         kfree(rtnl_msg_handlers[protocol]);
 295         rtnl_msg_handlers[protocol] = NULL;
 296 }
 297 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 298
 299 static LIST_HEAD(link_ops);
 300
 301 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 302 {
 303         const struct rtnl_link_ops *ops;
 304
 305         list_for_each_entry(ops, &link_ops, list) {
 306                 if (!strcmp(ops->kind, kind))
 307                         return ops;
 308         }
 309         return NULL;
 310 }
 311
 312 /**
 313  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 314  * @ops: struct rtnl_link_ops * to register
 315  *
 316  * The caller must hold the rtnl_mutex. This function should be used
 317  * by drivers that create devices during module initialization. It
 318  * must be called before registering the devices.
 319  *
 320  * Returns 0 on success or a negative error code.
 321  */
 322 int __rtnl_link_register(struct rtnl_link_ops *ops)
 323 {
 324         if (rtnl_link_ops_get(ops->kind))
 325                 return -EEXIST;
 326
 327         /* The check for setup is here because if ops
 328          * does not have that filled up, it is not possible
 329          * to use the ops for creating device. So do not
 330          * fill up dellink as well. That disables rtnl_dellink.
 331          */
 332         if (ops->setup && !ops->dellink)
 333                 ops->dellink = unregister_netdevice_queue;
 334
 335         list_add_tail(&ops->list, &link_ops);
 336         return 0;
 337 }
 338 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 339
 340 /**
 341  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 342  * @ops: struct rtnl_link_ops * to register
 343  *
 344  * Returns 0 on success or a negative error code.
 345  */
 346 int rtnl_link_register(struct rtnl_link_ops *ops)
 347 {
 348         int err;
 349
 350         rtnl_lock();
 351         err = __rtnl_link_register(ops);
 352         rtnl_unlock();
 353         return err;
 354 }
 355 EXPORT_SYMBOL_GPL(rtnl_link_register);
 356
 357 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 358 {
 359         struct net_device *dev;
 360         LIST_HEAD(list_kill);
 361
 362         for_each_netdev(net, dev) {
 363                 if (dev->rtnl_link_ops == ops)
 364                         ops->dellink(dev, &list_kill);
 365         }
 366         unregister_netdevice_many(&list_kill);
 367 }
 368
 369 /**
 370  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 371  * @ops: struct rtnl_link_ops * to unregister
 372  *
 373  * The caller must hold the rtnl_mutex.
 374  */
 375 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 376 {
 377         struct net *net;
 378
 379         for_each_net(net) {
 380                 __rtnl_kill_links(net, ops);
 381         }
 382         list_del(&ops->list);
 383 }
 384 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 385
 386 /* Return with the rtnl_lock held when there are no network
 387  * devices unregistering in any network namespace.
 388  */
 389 static void rtnl_lock_unregistering_all(void)
 390 {
 391         struct net *net;
 392         bool unregistering;
 393         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 394
 395         add_wait_queue(&netdev_unregistering_wq, &wait);
 396         for (;;) {
 397                 unregistering = false;
 398                 rtnl_lock();
 399                 for_each_net(net) {
 400                         if (net->dev_unreg_count > 0) {
 401                                 unregistering = true;
 402                                 break;
 403                         }
 404                 }
 405                 if (!unregistering)
 406                         break;
 407                 __rtnl_unlock();
 408
 409                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 410         }
 411         remove_wait_queue(&netdev_unregistering_wq, &wait);
 412 }
 413
 414 /**
 415  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 416  * @ops: struct rtnl_link_ops * to unregister
 417  */
 418 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 419 {
 420         /* Close the race with cleanup_net() */
 421         mutex_lock(&net_mutex);
 422         rtnl_lock_unregistering_all();
 423         __rtnl_link_unregister(ops);
 424         rtnl_unlock();
 425         mutex_unlock(&net_mutex);
 426 }
 427 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 428
 429 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 430 {
 431         struct net_device *master_dev;
 432         const struct rtnl_link_ops *ops;
 433
 434         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 435         if (!master_dev)
 436                 return 0;
 437         ops = master_dev->rtnl_link_ops;
 438         if (!ops || !ops->get_slave_size)
 439                 return 0;
 440         /* IFLA_INFO_SLAVE_DATA + nested data */
 441         return nla_total_size(sizeof(struct nlattr)) +
 442                ops->get_slave_size(master_dev, dev);
 443 }
 444
 445 static size_t rtnl_link_get_size(const struct net_device *dev)
 446 {
 447         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 448         size_t size;
 449
 450         if (!ops)
 451                 return 0;
 452
 453         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 454                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 455
 456         if (ops->get_size)
 457                 /* IFLA_INFO_DATA + nested data */
 458                 size += nla_total_size(sizeof(struct nlattr)) +
 459                         ops->get_size(dev);
 460
 461         if (ops->get_xstats_size)
 462                 /* IFLA_INFO_XSTATS */
 463                 size += nla_total_size(ops->get_xstats_size(dev));
 464
 465         size += rtnl_link_get_slave_info_data_size(dev);
 466
 467         return size;
 468 }
 469
 470 static LIST_HEAD(rtnl_af_ops);
 471
 472 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 473 {
 474         const struct rtnl_af_ops *ops;
 475
 476         list_for_each_entry(ops, &rtnl_af_ops, list) {
 477                 if (ops->family == family)
 478                         return ops;
 479         }
 480
 481         return NULL;
 482 }
 483
 484 /**
 485  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 486  * @ops: struct rtnl_af_ops * to register
 487  *
 488  * Returns 0 on success or a negative error code.
 489  */
 490 void rtnl_af_register(struct rtnl_af_ops *ops)
 491 {
 492         rtnl_lock();
 493         list_add_tail(&ops->list, &rtnl_af_ops);
 494         rtnl_unlock();
 495 }
 496 EXPORT_SYMBOL_GPL(rtnl_af_register);
 497
 498 /**
 499  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 500  * @ops: struct rtnl_af_ops * to unregister
 501  *
 502  * The caller must hold the rtnl_mutex.
 503  */
 504 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 505 {
 506         list_del(&ops->list);
 507 }
 508 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 509
 510 /**
 511  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 512  * @ops: struct rtnl_af_ops * to unregister
 513  */
 514 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 515 {
 516         rtnl_lock();
 517         __rtnl_af_unregister(ops);
 518         rtnl_unlock();
 519 }
 520 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 521
 522 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 523                                     u32 ext_filter_mask)
 524 {
 525         struct rtnl_af_ops *af_ops;
 526         size_t size;
 527
 528         /* IFLA_AF_SPEC */
 529         size = nla_total_size(sizeof(struct nlattr));
 530
 531         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 532                 if (af_ops->get_link_af_size) {
 533                         /* AF_* + nested data */
 534                         size += nla_total_size(sizeof(struct nlattr)) +
 535                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 536                 }
 537         }
 538
 539         return size;
 540 }
 541
 542 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 543 {
 544         struct net_device *master_dev;
 545
 546         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 547         if (master_dev && master_dev->rtnl_link_ops)
 548                 return true;
 549         return false;
 550 }
 551
 552 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 553                                      const struct net_device *dev)
 554 {
 555         struct net_device *master_dev;
 556         const struct rtnl_link_ops *ops;
 557         struct nlattr *slave_data;
 558         int err;
 559
 560         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 561         if (!master_dev)
 562                 return 0;
 563         ops = master_dev->rtnl_link_ops;
 564         if (!ops)
 565                 return 0;
 566         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 567                 return -EMSGSIZE;
 568         if (ops->fill_slave_info) {
 569                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 570                 if (!slave_data)
 571                         return -EMSGSIZE;
 572                 err = ops->fill_slave_info(skb, master_dev, dev);
 573                 if (err < 0)
 574                         goto err_cancel_slave_data;
 575                 nla_nest_end(skb, slave_data);
 576         }
 577         return 0;
 578
 579 err_cancel_slave_data:
 580         nla_nest_cancel(skb, slave_data);
 581         return err;
 582 }
 583
 584 static int rtnl_link_info_fill(struct sk_buff *skb,
 585                                const struct net_device *dev)
 586 {
 587         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 588         struct nlattr *data;
 589         int err;
 590
 591         if (!ops)
 592                 return 0;
 593         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 594                 return -EMSGSIZE;
 595         if (ops->fill_xstats) {
 596                 err = ops->fill_xstats(skb, dev);
 597                 if (err < 0)
 598                         return err;
 599         }
 600         if (ops->fill_info) {
 601                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 602                 if (data == NULL)
 603                         return -EMSGSIZE;
 604                 err = ops->fill_info(skb, dev);
 605                 if (err < 0)
 606                         goto err_cancel_data;
 607                 nla_nest_end(skb, data);
 608         }
 609         return 0;
 610
 611 err_cancel_data:
 612         nla_nest_cancel(skb, data);
 613         return err;
 614 }
 615
 616 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 617 {
 618         struct nlattr *linkinfo;
 619         int err = -EMSGSIZE;
 620
 621         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 622         if (linkinfo == NULL)
 623                 goto out;
 624
 625         err = rtnl_link_info_fill(skb, dev);
 626         if (err < 0)
 627                 goto err_cancel_link;
 628
 629         err = rtnl_link_slave_info_fill(skb, dev);
 630         if (err < 0)
 631                 goto err_cancel_link;
 632
 633         nla_nest_end(skb, linkinfo);
 634         return 0;
 635
 636 err_cancel_link:
 637         nla_nest_cancel(skb, linkinfo);
 638 out:
 639         return err;
 640 }
 641
 642 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 643 {
 644         struct sock *rtnl = net->rtnl;
 645         int err = 0;
 646
 647         NETLINK_CB(skb).dst_group = group;
 648         if (echo)
 649                 atomic_inc(&skb->users);
 650         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 651         if (echo)
 652                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 653         return err;
 654 }
 655
 656 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 657 {
 658         struct sock *rtnl = net->rtnl;
 659
 660         return nlmsg_unicast(rtnl, skb, pid);
 661 }
 662 EXPORT_SYMBOL(rtnl_unicast);
 663
 664 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 665                  struct nlmsghdr *nlh, gfp_t flags)
 666 {
 667         struct sock *rtnl = net->rtnl;
 668         int report = 0;
 669
 670         if (nlh)
 671                 report = nlmsg_report(nlh);
 672
 673         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 674 }
 675 EXPORT_SYMBOL(rtnl_notify);
 676
 677 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 678 {
 679         struct sock *rtnl = net->rtnl;
 680
 681         netlink_set_err(rtnl, 0, group, error);
 682 }
 683 EXPORT_SYMBOL(rtnl_set_sk_err);
 684
 685 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 686 {
 687         struct nlattr *mx;
 688         int i, valid = 0;
 689
 690         mx = nla_nest_start(skb, RTA_METRICS);
 691         if (mx == NULL)
 692                 return -ENOBUFS;
 693
 694         for (i = 0; i < RTAX_MAX; i++) {
 695                 if (metrics[i]) {
 696                         if (i == RTAX_CC_ALGO - 1) {
 697                                 char tmp[TCP_CA_NAME_MAX], *name;
 698
 699                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 700                                 if (!name)
 701                                         continue;
 702                                 if (nla_put_string(skb, i + 1, name))
 703                                         goto nla_put_failure;
 704                         } else if (i == RTAX_FEATURES - 1) {
 705                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 706
 707                                 if (!user_features)
 708                                         continue;
 709                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 710                                 if (nla_put_u32(skb, i + 1, user_features))
 711                                         goto nla_put_failure;
 712                         } else {
 713                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 714                                         goto nla_put_failure;
 715                         }
 716                         valid++;
 717                 }
 718         }
 719
 720         if (!valid) {
 721                 nla_nest_cancel(skb, mx);
 722                 return 0;
 723         }
 724
 725         return nla_nest_end(skb, mx);
 726
 727 nla_put_failure:
 728         nla_nest_cancel(skb, mx);
 729         return -EMSGSIZE;
 730 }
 731 EXPORT_SYMBOL(rtnetlink_put_metrics);
 732
 733 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 734                        long expires, u32 error)
 735 {
 736         struct rta_cacheinfo ci = {
 737                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 738                 .rta_used = dst->__use,
 739                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 740                 .rta_error = error,
 741                 .rta_id =  id,
 742         };
 743
 744         if (expires) {
 745                 unsigned long clock;
 746
 747                 clock = jiffies_to_clock_t(abs(expires));
 748                 clock = min_t(unsigned long, clock, INT_MAX);
 749                 ci.rta_expires = (expires > 0) ? clock : -clock;
 750         }
 751         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 752 }
 753 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 754
 755 static void set_operstate(struct net_device *dev, unsigned char transition)
 756 {
 757         unsigned char operstate = dev->operstate;
 758
 759         switch (transition) {
 760         case IF_OPER_UP:
 761                 if ((operstate == IF_OPER_DORMANT ||
 762                      operstate == IF_OPER_UNKNOWN) &&
 763                     !netif_dormant(dev))
 764                         operstate = IF_OPER_UP;
 765                 break;
 766
 767         case IF_OPER_DORMANT:
 768                 if (operstate == IF_OPER_UP ||
 769                     operstate == IF_OPER_UNKNOWN)
 770                         operstate = IF_OPER_DORMANT;
 771                 break;
 772         }
 773
 774         if (dev->operstate != operstate) {
 775                 write_lock_bh(&dev_base_lock);
 776                 dev->operstate = operstate;
 777                 write_unlock_bh(&dev_base_lock);
 778                 netdev_state_change(dev);
 779         }
 780 }
 781
 782 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 783 {
 784         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 785                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 786 }
 787
 788 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 789                                            const struct ifinfomsg *ifm)
 790 {
 791         unsigned int flags = ifm->ifi_flags;
 792
 793         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 794         if (ifm->ifi_change)
 795                 flags = (flags & ifm->ifi_change) |
 796                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 797
 798         return flags;
 799 }
 800
 801 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 802                                  const struct rtnl_link_stats64 *b)
 803 {
 804         a->rx_packets = b->rx_packets;
 805         a->tx_packets = b->tx_packets;
 806         a->rx_bytes = b->rx_bytes;
 807         a->tx_bytes = b->tx_bytes;
 808         a->rx_errors = b->rx_errors;
 809         a->tx_errors = b->tx_errors;
 810         a->rx_dropped = b->rx_dropped;
 811         a->tx_dropped = b->tx_dropped;
 812
 813         a->multicast = b->multicast;
 814         a->collisions = b->collisions;
 815
 816         a->rx_length_errors = b->rx_length_errors;
 817         a->rx_over_errors = b->rx_over_errors;
 818         a->rx_crc_errors = b->rx_crc_errors;
 819         a->rx_frame_errors = b->rx_frame_errors;
 820         a->rx_fifo_errors = b->rx_fifo_errors;
 821         a->rx_missed_errors = b->rx_missed_errors;
 822
 823         a->tx_aborted_errors = b->tx_aborted_errors;
 824         a->tx_carrier_errors = b->tx_carrier_errors;
 825         a->tx_fifo_errors = b->tx_fifo_errors;
 826         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 827         a->tx_window_errors = b->tx_window_errors;
 828
 829         a->rx_compressed = b->rx_compressed;
 830         a->tx_compressed = b->tx_compressed;
 831
 832         a->rx_nohandler = b->rx_nohandler;
 833 }
 834
 835 /* All VF info */
 836 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 837                                    u32 ext_filter_mask)
 838 {
 839         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 840             (ext_filter_mask & RTEXT_FILTER_VF)) {
 841                 int num_vfs = dev_num_vf(dev->dev.parent);
 842                 size_t size = nla_total_size(sizeof(struct nlattr));
 843                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 844                 size += num_vfs *
 845                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 846                          nla_total_size(MAX_VLAN_LIST_LEN *
 847                                         sizeof(struct nlattr)) +
 848                          nla_total_size(MAX_VLAN_LIST_LEN *
 849                                         sizeof(struct ifla_vf_vlan_info)) +
 850                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 851                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 852                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 853                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 854                          /* IFLA_VF_STATS_RX_PACKETS */
 855                          nla_total_size_64bit(sizeof(__u64)) +
 856                          /* IFLA_VF_STATS_TX_PACKETS */
 857                          nla_total_size_64bit(sizeof(__u64)) +
 858                          /* IFLA_VF_STATS_RX_BYTES */
 859                          nla_total_size_64bit(sizeof(__u64)) +
 860                          /* IFLA_VF_STATS_TX_BYTES */
 861                          nla_total_size_64bit(sizeof(__u64)) +
 862                          /* IFLA_VF_STATS_BROADCAST */
 863                          nla_total_size_64bit(sizeof(__u64)) +
 864                          /* IFLA_VF_STATS_MULTICAST */
 865                          nla_total_size_64bit(sizeof(__u64)) +
 866                          nla_total_size(sizeof(struct ifla_vf_trust)));
 867                 return size;
 868         } else
 869                 return 0;
 870 }
 871
 872 static size_t rtnl_port_size(const struct net_device *dev,
 873                              u32 ext_filter_mask)
 874 {
 875         size_t port_size = nla_total_size(4)            /* PORT_VF */
 876                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 877                 + nla_total_size(sizeof(struct ifla_port_vsi))
 878                                                         /* PORT_VSI_TYPE */
 879                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 880                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 881                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 882                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 883         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 884         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 885                 + port_size;
 886         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 887                 + port_size;
 888
 889         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 890             !(ext_filter_mask & RTEXT_FILTER_VF))
 891                 return 0;
 892         if (dev_num_vf(dev->dev.parent))
 893                 return port_self_size + vf_ports_size +
 894                         vf_port_size * dev_num_vf(dev->dev.parent);
 895         else
 896                 return port_self_size;
 897 }
 898
 899 static size_t rtnl_xdp_size(const struct net_device *dev)
 900 {
 901         size_t xdp_size = nla_total_size(1);    /* XDP_ATTACHED */
 902
 903         if (!dev->netdev_ops->ndo_xdp)
 904                 return 0;
 905         else
 906                 return xdp_size;
 907 }
 908
 909 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 910                                      u32 ext_filter_mask)
 911 {
 912         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 913                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 914                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 915                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 916                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 917                + nla_total_size(sizeof(struct rtnl_link_stats))
 918                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 919                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 920                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 921                + nla_total_size(4) /* IFLA_TXQLEN */
 922                + nla_total_size(4) /* IFLA_WEIGHT */
 923                + nla_total_size(4) /* IFLA_MTU */
 924                + nla_total_size(4) /* IFLA_LINK */
 925                + nla_total_size(4) /* IFLA_MASTER */
 926                + nla_total_size(1) /* IFLA_CARRIER */
 927                + nla_total_size(4) /* IFLA_PROMISCUITY */
 928                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 929                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 930                + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
 931                + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
 932                + nla_total_size(1) /* IFLA_OPERSTATE */
 933                + nla_total_size(1) /* IFLA_LINKMODE */
 934                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 935                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 936                + nla_total_size(ext_filter_mask
 937                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 938                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 939                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 940                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 941                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 942                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 943                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 944                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 945                + rtnl_xdp_size(dev) /* IFLA_XDP */
 946                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 947
 948 }
 949
 950 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 951 {
 952         struct nlattr *vf_ports;
 953         struct nlattr *vf_port;
 954         int vf;
 955         int err;
 956
 957         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 958         if (!vf_ports)
 959                 return -EMSGSIZE;
 960
 961         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 962                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 963                 if (!vf_port)
 964                         goto nla_put_failure;
 965                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 966                         goto nla_put_failure;
 967                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 968                 if (err == -EMSGSIZE)
 969                         goto nla_put_failure;
 970                 if (err) {
 971                         nla_nest_cancel(skb, vf_port);
 972                         continue;
 973                 }
 974                 nla_nest_end(skb, vf_port);
 975         }
 976
 977         nla_nest_end(skb, vf_ports);
 978
 979         return 0;
 980
 981 nla_put_failure:
 982         nla_nest_cancel(skb, vf_ports);
 983         return -EMSGSIZE;
 984 }
 985
 986 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 987 {
 988         struct nlattr *port_self;
 989         int err;
 990
 991         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 992         if (!port_self)
 993                 return -EMSGSIZE;
 994
 995         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 996         if (err) {
 997                 nla_nest_cancel(skb, port_self);
 998                 return (err == -EMSGSIZE) ? err : 0;
 999         }
1000
1001         nla_nest_end(skb, port_self);
1002
1003         return 0;
1004 }
1005
1006 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1007                           u32 ext_filter_mask)
1008 {
1009         int err;
1010
1011         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1012             !(ext_filter_mask & RTEXT_FILTER_VF))
1013                 return 0;
1014
1015         err = rtnl_port_self_fill(skb, dev);
1016         if (err)
1017                 return err;
1018
1019         if (dev_num_vf(dev->dev.parent)) {
1020                 err = rtnl_vf_ports_fill(skb, dev);
1021                 if (err)
1022                         return err;
1023         }
1024
1025         return 0;
1026 }
1027
1028 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1029 {
1030         int err;
1031         struct netdev_phys_item_id ppid;
1032
1033         err = dev_get_phys_port_id(dev, &ppid);
1034         if (err) {
1035                 if (err == -EOPNOTSUPP)
1036                         return 0;
1037                 return err;
1038         }
1039
1040         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1041                 return -EMSGSIZE;
1042
1043         return 0;
1044 }
1045
1046 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1047 {
1048         char name[IFNAMSIZ];
1049         int err;
1050
1051         err = dev_get_phys_port_name(dev, name, sizeof(name));
1052         if (err) {
1053                 if (err == -EOPNOTSUPP)
1054                         return 0;
1055                 return err;
1056         }
1057
1058         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1059                 return -EMSGSIZE;
1060
1061         return 0;
1062 }
1063
1064 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1065 {
1066         int err;
1067         struct switchdev_attr attr = {
1068                 .orig_dev = dev,
1069                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1070                 .flags = SWITCHDEV_F_NO_RECURSE,
1071         };
1072
1073         err = switchdev_port_attr_get(dev, &attr);
1074         if (err) {
1075                 if (err == -EOPNOTSUPP)
1076                         return 0;
1077                 return err;
1078         }
1079
1080         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1081                     attr.u.ppid.id))
1082                 return -EMSGSIZE;
1083
1084         return 0;
1085 }
1086
1087 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1088                                               struct net_device *dev)
1089 {
1090         struct rtnl_link_stats64 *sp;
1091         struct nlattr *attr;
1092
1093         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1094                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1095         if (!attr)
1096                 return -EMSGSIZE;
1097
1098         sp = nla_data(attr);
1099         dev_get_stats(dev, sp);
1100
1101         attr = nla_reserve(skb, IFLA_STATS,
1102                            sizeof(struct rtnl_link_stats));
1103         if (!attr)
1104                 return -EMSGSIZE;
1105
1106         copy_rtnl_link_stats(nla_data(attr), sp);
1107
1108         return 0;
1109 }
1110
1111 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1112                                                struct net_device *dev,
1113                                                int vfs_num,
1114                                                struct nlattr *vfinfo)
1115 {
1116         struct ifla_vf_rss_query_en vf_rss_query_en;
1117         struct nlattr *vf, *vfstats, *vfvlanlist;
1118         struct ifla_vf_link_state vf_linkstate;
1119         struct ifla_vf_vlan_info vf_vlan_info;
1120         struct ifla_vf_spoofchk vf_spoofchk;
1121         struct ifla_vf_tx_rate vf_tx_rate;
1122         struct ifla_vf_stats vf_stats;
1123         struct ifla_vf_trust vf_trust;
1124         struct ifla_vf_vlan vf_vlan;
1125         struct ifla_vf_rate vf_rate;
1126         struct ifla_vf_mac vf_mac;
1127         struct ifla_vf_info ivi;
1128
1129         /* Not all SR-IOV capable drivers support the
1130          * spoofcheck and "RSS query enable" query.  Preset to
1131          * -1 so the user space tool can detect that the driver
1132          * didn't report anything.
1133          */
1134         ivi.spoofchk = -1;
1135         ivi.rss_query_en = -1;
1136         ivi.trusted = -1;
1137         memset(ivi.mac, 0, sizeof(ivi.mac));
1138         /* The default value for VF link state is "auto"
1139          * IFLA_VF_LINK_STATE_AUTO which equals zero
1140          */
1141         ivi.linkstate = 0;
1142         /* VLAN Protocol by default is 802.1Q */
1143         ivi.vlan_proto = htons(ETH_P_8021Q);
1144         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1145                 return 0;
1146
1147         vf_mac.vf =
1148                 vf_vlan.vf =
1149                 vf_vlan_info.vf =
1150                 vf_rate.vf =
1151                 vf_tx_rate.vf =
1152                 vf_spoofchk.vf =
1153                 vf_linkstate.vf =
1154                 vf_rss_query_en.vf =
1155                 vf_trust.vf = ivi.vf;
1156
1157         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1158         vf_vlan.vlan = ivi.vlan;
1159         vf_vlan.qos = ivi.qos;
1160         vf_vlan_info.vlan = ivi.vlan;
1161         vf_vlan_info.qos = ivi.qos;
1162         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1163         vf_tx_rate.rate = ivi.max_tx_rate;
1164         vf_rate.min_tx_rate = ivi.min_tx_rate;
1165         vf_rate.max_tx_rate = ivi.max_tx_rate;
1166         vf_spoofchk.setting = ivi.spoofchk;
1167         vf_linkstate.link_state = ivi.linkstate;
1168         vf_rss_query_en.setting = ivi.rss_query_en;
1169         vf_trust.setting = ivi.trusted;
1170         vf = nla_nest_start(skb, IFLA_VF_INFO);
1171         if (!vf)
1172                 goto nla_put_vfinfo_failure;
1173         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1174             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1175             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1176                     &vf_rate) ||
1177             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1178                     &vf_tx_rate) ||
1179             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1180                     &vf_spoofchk) ||
1181             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1182                     &vf_linkstate) ||
1183             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1184                     sizeof(vf_rss_query_en),
1185                     &vf_rss_query_en) ||
1186             nla_put(skb, IFLA_VF_TRUST,
1187                     sizeof(vf_trust), &vf_trust))
1188                 goto nla_put_vf_failure;
1189         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1190         if (!vfvlanlist)
1191                 goto nla_put_vf_failure;
1192         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1193                     &vf_vlan_info)) {
1194                 nla_nest_cancel(skb, vfvlanlist);
1195                 goto nla_put_vf_failure;
1196         }
1197         nla_nest_end(skb, vfvlanlist);
1198         memset(&vf_stats, 0, sizeof(vf_stats));
1199         if (dev->netdev_ops->ndo_get_vf_stats)
1200                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1201                                                 &vf_stats);
1202         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1203         if (!vfstats)
1204                 goto nla_put_vf_failure;
1205         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1206                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1207             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1208                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1209             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1210                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1211             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1212                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1214                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1216                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1217                 nla_nest_cancel(skb, vfstats);
1218                 goto nla_put_vf_failure;
1219         }
1220         nla_nest_end(skb, vfstats);
1221         nla_nest_end(skb, vf);
1222         return 0;
1223
1224 nla_put_vf_failure:
1225         nla_nest_cancel(skb, vf);
1226 nla_put_vfinfo_failure:
1227         nla_nest_cancel(skb, vfinfo);
1228         return -EMSGSIZE;
1229 }
1230
1231 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1232 {
1233         struct rtnl_link_ifmap map;
1234
1235         memset(&map, 0, sizeof(map));
1236         map.mem_start   = dev->mem_start;
1237         map.mem_end     = dev->mem_end;
1238         map.base_addr   = dev->base_addr;
1239         map.irq         = dev->irq;
1240         map.dma         = dev->dma;
1241         map.port        = dev->if_port;
1242
1243         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1244                 return -EMSGSIZE;
1245
1246         return 0;
1247 }
1248
1249 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1250 {
1251         struct netdev_xdp xdp_op = {};
1252         struct nlattr *xdp;
1253         int err;
1254
1255         if (!dev->netdev_ops->ndo_xdp)
1256                 return 0;
1257         xdp = nla_nest_start(skb, IFLA_XDP);
1258         if (!xdp)
1259                 return -EMSGSIZE;
1260         xdp_op.command = XDP_QUERY_PROG;
1261         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1262         if (err)
1263                 goto err_cancel;
1264         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1265         if (err)
1266                 goto err_cancel;
1267
1268         nla_nest_end(skb, xdp);
1269         return 0;
1270
1271 err_cancel:
1272         nla_nest_cancel(skb, xdp);
1273         return err;
1274 }
1275
1276 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1277                             int type, u32 pid, u32 seq, u32 change,
1278                             unsigned int flags, u32 ext_filter_mask)
1279 {
1280         struct ifinfomsg *ifm;
1281         struct nlmsghdr *nlh;
1282         struct nlattr *af_spec;
1283         struct rtnl_af_ops *af_ops;
1284         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1285
1286         ASSERT_RTNL();
1287         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1288         if (nlh == NULL)
1289                 return -EMSGSIZE;
1290
1291         ifm = nlmsg_data(nlh);
1292         ifm->ifi_family = AF_UNSPEC;
1293         ifm->__ifi_pad = 0;
1294         ifm->ifi_type = dev->type;
1295         ifm->ifi_index = dev->ifindex;
1296         ifm->ifi_flags = dev_get_flags(dev);
1297         ifm->ifi_change = change;
1298
1299         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1300             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1301             nla_put_u8(skb, IFLA_OPERSTATE,
1302                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1303             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1304             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1305             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1306             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1307             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1308             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1309             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1310 #ifdef CONFIG_RPS
1311             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1312 #endif
1313             (dev->ifindex != dev_get_iflink(dev) &&
1314              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1315             (upper_dev &&
1316              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1317             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1318             (dev->qdisc &&
1319              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1320             (dev->ifalias &&
1321              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1322             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1323                         atomic_read(&dev->carrier_changes)) ||
1324             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1325                 goto nla_put_failure;
1326
1327         if (rtnl_fill_link_ifmap(skb, dev))
1328                 goto nla_put_failure;
1329
1330         if (dev->addr_len) {
1331                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1332                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1333                         goto nla_put_failure;
1334         }
1335
1336         if (rtnl_phys_port_id_fill(skb, dev))
1337                 goto nla_put_failure;
1338
1339         if (rtnl_phys_port_name_fill(skb, dev))
1340                 goto nla_put_failure;
1341
1342         if (rtnl_phys_switch_id_fill(skb, dev))
1343                 goto nla_put_failure;
1344
1345         if (rtnl_fill_stats(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1349             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1350                 goto nla_put_failure;
1351
1352         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1353             ext_filter_mask & RTEXT_FILTER_VF) {
1354                 int i;
1355                 struct nlattr *vfinfo;
1356                 int num_vfs = dev_num_vf(dev->dev.parent);
1357
1358                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1359                 if (!vfinfo)
1360                         goto nla_put_failure;
1361                 for (i = 0; i < num_vfs; i++) {
1362                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1363                                 goto nla_put_failure;
1364                 }
1365
1366                 nla_nest_end(skb, vfinfo);
1367         }
1368
1369         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1370                 goto nla_put_failure;
1371
1372         if (rtnl_xdp_fill(skb, dev))
1373                 goto nla_put_failure;
1374
1375         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1376                 if (rtnl_link_fill(skb, dev) < 0)
1377                         goto nla_put_failure;
1378         }
1379
1380         if (dev->rtnl_link_ops &&
1381             dev->rtnl_link_ops->get_link_net) {
1382                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1383
1384                 if (!net_eq(dev_net(dev), link_net)) {
1385                         int id = peernet2id_alloc(dev_net(dev), link_net);
1386
1387                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1388                                 goto nla_put_failure;
1389                 }
1390         }
1391
1392         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1393                 goto nla_put_failure;
1394
1395         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1396                 if (af_ops->fill_link_af) {
1397                         struct nlattr *af;
1398                         int err;
1399
1400                         if (!(af = nla_nest_start(skb, af_ops->family)))
1401                                 goto nla_put_failure;
1402
1403                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1404
1405                         /*
1406                          * Caller may return ENODATA to indicate that there
1407                          * was no data to be dumped. This is not an error, it
1408                          * means we should trim the attribute header and
1409                          * continue.
1410                          */
1411                         if (err == -ENODATA)
1412                                 nla_nest_cancel(skb, af);
1413                         else if (err < 0)
1414                                 goto nla_put_failure;
1415
1416                         nla_nest_end(skb, af);
1417                 }
1418         }
1419
1420         nla_nest_end(skb, af_spec);
1421
1422         nlmsg_end(skb, nlh);
1423         return 0;
1424
1425 nla_put_failure:
1426         nlmsg_cancel(skb, nlh);
1427         return -EMSGSIZE;
1428 }
1429
1430 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1431         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1432         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1433         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1434         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1435         [IFLA_MTU]              = { .type = NLA_U32 },
1436         [IFLA_LINK]             = { .type = NLA_U32 },
1437         [IFLA_MASTER]           = { .type = NLA_U32 },
1438         [IFLA_CARRIER]          = { .type = NLA_U8 },
1439         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1440         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1441         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1442         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1443         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1444         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1445         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1446         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1447         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1448         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1449         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1450         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1451         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1452         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1453         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1454         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1455         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1456         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1457         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1458         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1459         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1460         [IFLA_XDP]              = { .type = NLA_NESTED },
1461 };
1462
1463 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1464         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1465         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1466         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1467         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1468 };
1469
1470 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1471         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1472         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1473         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1474         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1475         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1476         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1477         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1478         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1479         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1480         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1481         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1482         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1483 };
1484
1485 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1486         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1487         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1488                                     .len = PORT_PROFILE_MAX },
1489         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1490                                     .len = sizeof(struct ifla_port_vsi)},
1491         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1492                                       .len = PORT_UUID_MAX },
1493         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1494                                     .len = PORT_UUID_MAX },
1495         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1496         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1497 };
1498
1499 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1500         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1501         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1502 };
1503
1504 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1505 {
1506         const struct rtnl_link_ops *ops = NULL;
1507         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1508
1509         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1510                 return NULL;
1511
1512         if (linfo[IFLA_INFO_KIND]) {
1513                 char kind[MODULE_NAME_LEN];
1514
1515                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1516                 ops = rtnl_link_ops_get(kind);
1517         }
1518
1519         return ops;
1520 }
1521
1522 static bool link_master_filtered(struct net_device *dev, int master_idx)
1523 {
1524         struct net_device *master;
1525
1526         if (!master_idx)
1527                 return false;
1528
1529         master = netdev_master_upper_dev_get(dev);
1530         if (!master || master->ifindex != master_idx)
1531                 return true;
1532
1533         return false;
1534 }
1535
1536 static bool link_kind_filtered(const struct net_device *dev,
1537                                const struct rtnl_link_ops *kind_ops)
1538 {
1539         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1540                 return true;
1541
1542         return false;
1543 }
1544
1545 static bool link_dump_filtered(struct net_device *dev,
1546                                int master_idx,
1547                                const struct rtnl_link_ops *kind_ops)
1548 {
1549         if (link_master_filtered(dev, master_idx) ||
1550             link_kind_filtered(dev, kind_ops))
1551                 return true;
1552
1553         return false;
1554 }
1555
1556 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1557 {
1558         struct net *net = sock_net(skb->sk);
1559         int h, s_h;
1560         int idx = 0, s_idx;
1561         struct net_device *dev;
1562         struct hlist_head *head;
1563         struct nlattr *tb[IFLA_MAX+1];
1564         u32 ext_filter_mask = 0;
1565         const struct rtnl_link_ops *kind_ops = NULL;
1566         unsigned int flags = NLM_F_MULTI;
1567         int master_idx = 0;
1568         int err;
1569         int hdrlen;
1570
1571         s_h = cb->args[0];
1572         s_idx = cb->args[1];
1573
1574         cb->seq = net->dev_base_seq;
1575
1576         /* A hack to preserve kernel<->userspace interface.
1577          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1578          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1579          * what iproute2 < v3.9.0 used.
1580          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1581          * attribute, its netlink message is shorter than struct ifinfomsg.
1582          */
1583         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1584                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1585
1586         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1587
1588                 if (tb[IFLA_EXT_MASK])
1589                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1590
1591                 if (tb[IFLA_MASTER])
1592                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1593
1594                 if (tb[IFLA_LINKINFO])
1595                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1596
1597                 if (master_idx || kind_ops)
1598                         flags |= NLM_F_DUMP_FILTERED;
1599         }
1600
1601         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1602                 idx = 0;
1603                 head = &net->dev_index_head[h];
1604                 hlist_for_each_entry(dev, head, index_hlist) {
1605                         if (link_dump_filtered(dev, master_idx, kind_ops))
1606                                 continue;
1607                         if (idx < s_idx)
1608                                 goto cont;
1609                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1610                                                NETLINK_CB(cb->skb).portid,
1611                                                cb->nlh->nlmsg_seq, 0,
1612                                                flags,
1613                                                ext_filter_mask);
1614                         /* If we ran out of room on the first message,
1615                          * we're in trouble
1616                          */
1617                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1618
1619                         if (err < 0)
1620                                 goto out;
1621
1622                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1623 cont:
1624                         idx++;
1625                 }
1626         }
1627 out:
1628         cb->args[1] = idx;
1629         cb->args[0] = h;
1630
1631         return skb->len;
1632 }
1633
1634 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1635 {
1636         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1637 }
1638 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1639
1640 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1641 {
1642         struct net *net;
1643         /* Examine the link attributes and figure out which
1644          * network namespace we are talking about.
1645          */
1646         if (tb[IFLA_NET_NS_PID])
1647                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1648         else if (tb[IFLA_NET_NS_FD])
1649                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1650         else
1651                 net = get_net(src_net);
1652         return net;
1653 }
1654 EXPORT_SYMBOL(rtnl_link_get_net);
1655
1656 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1657 {
1658         if (dev) {
1659                 if (tb[IFLA_ADDRESS] &&
1660                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1661                         return -EINVAL;
1662
1663                 if (tb[IFLA_BROADCAST] &&
1664                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1665                         return -EINVAL;
1666         }
1667
1668         if (tb[IFLA_AF_SPEC]) {
1669                 struct nlattr *af;
1670                 int rem, err;
1671
1672                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1673                         const struct rtnl_af_ops *af_ops;
1674
1675                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1676                                 return -EAFNOSUPPORT;
1677
1678                         if (!af_ops->set_link_af)
1679                                 return -EOPNOTSUPP;
1680
1681                         if (af_ops->validate_link_af) {
1682                                 err = af_ops->validate_link_af(dev, af);
1683                                 if (err < 0)
1684                                         return err;
1685                         }
1686                 }
1687         }
1688
1689         return 0;
1690 }
1691
1692 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1693                                   int guid_type)
1694 {
1695         const struct net_device_ops *ops = dev->netdev_ops;
1696
1697         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1698 }
1699
1700 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1701 {
1702         if (dev->type != ARPHRD_INFINIBAND)
1703                 return -EOPNOTSUPP;
1704
1705         return handle_infiniband_guid(dev, ivt, guid_type);
1706 }
1707
1708 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1709 {
1710         const struct net_device_ops *ops = dev->netdev_ops;
1711         int err = -EINVAL;
1712
1713         if (tb[IFLA_VF_MAC]) {
1714                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1715
1716                 err = -EOPNOTSUPP;
1717                 if (ops->ndo_set_vf_mac)
1718                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1719                                                   ivm->mac);
1720                 if (err < 0)
1721                         return err;
1722         }
1723
1724         if (tb[IFLA_VF_VLAN]) {
1725                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1726
1727                 err = -EOPNOTSUPP;
1728                 if (ops->ndo_set_vf_vlan)
1729                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1730                                                    ivv->qos,
1731                                                    htons(ETH_P_8021Q));
1732                 if (err < 0)
1733                         return err;
1734         }
1735
1736         if (tb[IFLA_VF_VLAN_LIST]) {
1737                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1738                 struct nlattr *attr;
1739                 int rem, len = 0;
1740
1741                 err = -EOPNOTSUPP;
1742                 if (!ops->ndo_set_vf_vlan)
1743                         return err;
1744
1745                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1746                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1747                             nla_len(attr) < NLA_HDRLEN) {
1748                                 return -EINVAL;
1749                         }
1750                         if (len >= MAX_VLAN_LIST_LEN)
1751                                 return -EOPNOTSUPP;
1752                         ivvl[len] = nla_data(attr);
1753
1754                         len++;
1755                 }
1756                 if (len == 0)
1757                         return -EINVAL;
1758
1759                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1760                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1761                 if (err < 0)
1762                         return err;
1763         }
1764
1765         if (tb[IFLA_VF_TX_RATE]) {
1766                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1767                 struct ifla_vf_info ivf;
1768
1769                 err = -EOPNOTSUPP;
1770                 if (ops->ndo_get_vf_config)
1771                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1772                 if (err < 0)
1773                         return err;
1774
1775                 err = -EOPNOTSUPP;
1776                 if (ops->ndo_set_vf_rate)
1777                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1778                                                    ivf.min_tx_rate,
1779                                                    ivt->rate);
1780                 if (err < 0)
1781                         return err;
1782         }
1783
1784         if (tb[IFLA_VF_RATE]) {
1785                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1786
1787                 err = -EOPNOTSUPP;
1788                 if (ops->ndo_set_vf_rate)
1789                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1790                                                    ivt->min_tx_rate,
1791                                                    ivt->max_tx_rate);
1792                 if (err < 0)
1793                         return err;
1794         }
1795
1796         if (tb[IFLA_VF_SPOOFCHK]) {
1797                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1798
1799                 err = -EOPNOTSUPP;
1800                 if (ops->ndo_set_vf_spoofchk)
1801                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1802                                                        ivs->setting);
1803                 if (err < 0)
1804                         return err;
1805         }
1806
1807         if (tb[IFLA_VF_LINK_STATE]) {
1808                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1809
1810                 err = -EOPNOTSUPP;
1811                 if (ops->ndo_set_vf_link_state)
1812                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1813                                                          ivl->link_state);
1814                 if (err < 0)
1815                         return err;
1816         }
1817
1818         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1819                 struct ifla_vf_rss_query_en *ivrssq_en;
1820
1821                 err = -EOPNOTSUPP;
1822                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1823                 if (ops->ndo_set_vf_rss_query_en)
1824                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1825                                                            ivrssq_en->setting);
1826                 if (err < 0)
1827                         return err;
1828         }
1829
1830         if (tb[IFLA_VF_TRUST]) {
1831                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1832
1833                 err = -EOPNOTSUPP;
1834                 if (ops->ndo_set_vf_trust)
1835                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1836                 if (err < 0)
1837                         return err;
1838         }
1839
1840         if (tb[IFLA_VF_IB_NODE_GUID]) {
1841                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1842
1843                 if (!ops->ndo_set_vf_guid)
1844                         return -EOPNOTSUPP;
1845
1846                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1847         }
1848
1849         if (tb[IFLA_VF_IB_PORT_GUID]) {
1850                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1851
1852                 if (!ops->ndo_set_vf_guid)
1853                         return -EOPNOTSUPP;
1854
1855                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1856         }
1857
1858         return err;
1859 }
1860
1861 static int do_set_master(struct net_device *dev, int ifindex)
1862 {
1863         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1864         const struct net_device_ops *ops;
1865         int err;
1866
1867         if (upper_dev) {
1868                 if (upper_dev->ifindex == ifindex)
1869                         return 0;
1870                 ops = upper_dev->netdev_ops;
1871                 if (ops->ndo_del_slave) {
1872                         err = ops->ndo_del_slave(upper_dev, dev);
1873                         if (err)
1874                                 return err;
1875                 } else {
1876                         return -EOPNOTSUPP;
1877                 }
1878         }
1879
1880         if (ifindex) {
1881                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1882                 if (!upper_dev)
1883                         return -EINVAL;
1884                 ops = upper_dev->netdev_ops;
1885                 if (ops->ndo_add_slave) {
1886                         err = ops->ndo_add_slave(upper_dev, dev);
1887                         if (err)
1888                                 return err;
1889                 } else {
1890                         return -EOPNOTSUPP;
1891                 }
1892         }
1893         return 0;
1894 }
1895
1896 #define DO_SETLINK_MODIFIED     0x01
1897 /* notify flag means notify + modified. */
1898 #define DO_SETLINK_NOTIFY       0x03
1899 static int do_setlink(const struct sk_buff *skb,
1900                       struct net_device *dev, struct ifinfomsg *ifm,
1901                       struct nlattr **tb, char *ifname, int status)
1902 {
1903         const struct net_device_ops *ops = dev->netdev_ops;
1904         int err;
1905
1906         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1907                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1908                 if (IS_ERR(net)) {
1909                         err = PTR_ERR(net);
1910                         goto errout;
1911                 }
1912                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1913                         put_net(net);
1914                         err = -EPERM;
1915                         goto errout;
1916                 }
1917                 err = dev_change_net_namespace(dev, net, ifname);
1918                 put_net(net);
1919                 if (err)
1920                         goto errout;
1921                 status |= DO_SETLINK_MODIFIED;
1922         }
1923
1924         if (tb[IFLA_MAP]) {
1925                 struct rtnl_link_ifmap *u_map;
1926                 struct ifmap k_map;
1927
1928                 if (!ops->ndo_set_config) {
1929                         err = -EOPNOTSUPP;
1930                         goto errout;
1931                 }
1932
1933                 if (!netif_device_present(dev)) {
1934                         err = -ENODEV;
1935                         goto errout;
1936                 }
1937
1938                 u_map = nla_data(tb[IFLA_MAP]);
1939                 k_map.mem_start = (unsigned long) u_map->mem_start;
1940                 k_map.mem_end = (unsigned long) u_map->mem_end;
1941                 k_map.base_addr = (unsigned short) u_map->base_addr;
1942                 k_map.irq = (unsigned char) u_map->irq;
1943                 k_map.dma = (unsigned char) u_map->dma;
1944                 k_map.port = (unsigned char) u_map->port;
1945
1946                 err = ops->ndo_set_config(dev, &k_map);
1947                 if (err < 0)
1948                         goto errout;
1949
1950                 status |= DO_SETLINK_NOTIFY;
1951         }
1952
1953         if (tb[IFLA_ADDRESS]) {
1954                 struct sockaddr *sa;
1955                 int len;
1956
1957                 len = sizeof(sa_family_t) + dev->addr_len;
1958                 sa = kmalloc(len, GFP_KERNEL);
1959                 if (!sa) {
1960                         err = -ENOMEM;
1961                         goto errout;
1962                 }
1963                 sa->sa_family = dev->type;
1964                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1965                        dev->addr_len);
1966                 err = dev_set_mac_address(dev, sa);
1967                 kfree(sa);
1968                 if (err)
1969                         goto errout;
1970                 status |= DO_SETLINK_MODIFIED;
1971         }
1972
1973         if (tb[IFLA_MTU]) {
1974                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1975                 if (err < 0)
1976                         goto errout;
1977                 status |= DO_SETLINK_MODIFIED;
1978         }
1979
1980         if (tb[IFLA_GROUP]) {
1981                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1982                 status |= DO_SETLINK_NOTIFY;
1983         }
1984
1985         /*
1986          * Interface selected by interface index but interface
1987          * name provided implies that a name change has been
1988          * requested.
1989          */
1990         if (ifm->ifi_index > 0 && ifname[0]) {
1991                 err = dev_change_name(dev, ifname);
1992                 if (err < 0)
1993                         goto errout;
1994                 status |= DO_SETLINK_MODIFIED;
1995         }
1996
1997         if (tb[IFLA_IFALIAS]) {
1998                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1999                                     nla_len(tb[IFLA_IFALIAS]));
2000                 if (err < 0)
2001                         goto errout;
2002                 status |= DO_SETLINK_NOTIFY;
2003         }
2004
2005         if (tb[IFLA_BROADCAST]) {
2006                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2007                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2008         }
2009
2010         if (ifm->ifi_flags || ifm->ifi_change) {
2011                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2012                 if (err < 0)
2013                         goto errout;
2014         }
2015
2016         if (tb[IFLA_MASTER]) {
2017                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2018                 if (err)
2019                         goto errout;
2020                 status |= DO_SETLINK_MODIFIED;
2021         }
2022
2023         if (tb[IFLA_CARRIER]) {
2024                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2025                 if (err)
2026                         goto errout;
2027                 status |= DO_SETLINK_MODIFIED;
2028         }
2029
2030         if (tb[IFLA_TXQLEN]) {
2031                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2032                 unsigned long orig_len = dev->tx_queue_len;
2033
2034                 if (dev->tx_queue_len ^ value) {
2035                         dev->tx_queue_len = value;
2036                         err = call_netdevice_notifiers(
2037                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2038                         err = notifier_to_errno(err);
2039                         if (err) {
2040                                 dev->tx_queue_len = orig_len;
2041                                 goto errout;
2042                         }
2043                         status |= DO_SETLINK_NOTIFY;
2044                 }
2045         }
2046
2047         if (tb[IFLA_OPERSTATE])
2048                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2049
2050         if (tb[IFLA_LINKMODE]) {
2051                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2052
2053                 write_lock_bh(&dev_base_lock);
2054                 if (dev->link_mode ^ value)
2055                         status |= DO_SETLINK_NOTIFY;
2056                 dev->link_mode = value;
2057                 write_unlock_bh(&dev_base_lock);
2058         }
2059
2060         if (tb[IFLA_VFINFO_LIST]) {
2061                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2062                 struct nlattr *attr;
2063                 int rem;
2064
2065                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2066                         if (nla_type(attr) != IFLA_VF_INFO ||
2067                             nla_len(attr) < NLA_HDRLEN) {
2068                                 err = -EINVAL;
2069                                 goto errout;
2070                         }
2071                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2072                                                ifla_vf_policy);
2073                         if (err < 0)
2074                                 goto errout;
2075                         err = do_setvfinfo(dev, vfinfo);
2076                         if (err < 0)
2077                                 goto errout;
2078                         status |= DO_SETLINK_NOTIFY;
2079                 }
2080         }
2081         err = 0;
2082
2083         if (tb[IFLA_VF_PORTS]) {
2084                 struct nlattr *port[IFLA_PORT_MAX+1];
2085                 struct nlattr *attr;
2086                 int vf;
2087                 int rem;
2088
2089                 err = -EOPNOTSUPP;
2090                 if (!ops->ndo_set_vf_port)
2091                         goto errout;
2092
2093                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2094                         if (nla_type(attr) != IFLA_VF_PORT ||
2095                             nla_len(attr) < NLA_HDRLEN) {
2096                                 err = -EINVAL;
2097                                 goto errout;
2098                         }
2099                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2100                                                ifla_port_policy);
2101                         if (err < 0)
2102                                 goto errout;
2103                         if (!port[IFLA_PORT_VF]) {
2104                                 err = -EOPNOTSUPP;
2105                                 goto errout;
2106                         }
2107                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2108                         err = ops->ndo_set_vf_port(dev, vf, port);
2109                         if (err < 0)
2110                                 goto errout;
2111                         status |= DO_SETLINK_NOTIFY;
2112                 }
2113         }
2114         err = 0;
2115
2116         if (tb[IFLA_PORT_SELF]) {
2117                 struct nlattr *port[IFLA_PORT_MAX+1];
2118
2119                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2120                         tb[IFLA_PORT_SELF], ifla_port_policy);
2121                 if (err < 0)
2122                         goto errout;
2123
2124                 err = -EOPNOTSUPP;
2125                 if (ops->ndo_set_vf_port)
2126                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2127                 if (err < 0)
2128                         goto errout;
2129                 status |= DO_SETLINK_NOTIFY;
2130         }
2131
2132         if (tb[IFLA_AF_SPEC]) {
2133                 struct nlattr *af;
2134                 int rem;
2135
2136                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2137                         const struct rtnl_af_ops *af_ops;
2138
2139                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2140                                 BUG();
2141
2142                         err = af_ops->set_link_af(dev, af);
2143                         if (err < 0)
2144                                 goto errout;
2145
2146                         status |= DO_SETLINK_NOTIFY;
2147                 }
2148         }
2149         err = 0;
2150
2151         if (tb[IFLA_PROTO_DOWN]) {
2152                 err = dev_change_proto_down(dev,
2153                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2154                 if (err)
2155                         goto errout;
2156                 status |= DO_SETLINK_NOTIFY;
2157         }
2158
2159         if (tb[IFLA_XDP]) {
2160                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2161
2162                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2163                                        ifla_xdp_policy);
2164                 if (err < 0)
2165                         goto errout;
2166
2167                 if (xdp[IFLA_XDP_ATTACHED]) {
2168                         err = -EINVAL;
2169                         goto errout;
2170                 }
2171                 if (xdp[IFLA_XDP_FD]) {
2172                         err = dev_change_xdp_fd(dev,
2173                                                 nla_get_s32(xdp[IFLA_XDP_FD]));
2174                         if (err)
2175                                 goto errout;
2176                         status |= DO_SETLINK_NOTIFY;
2177                 }
2178         }
2179
2180 errout:
2181         if (status & DO_SETLINK_MODIFIED) {
2182                 if (status & DO_SETLINK_NOTIFY)
2183                         netdev_state_change(dev);
2184
2185                 if (err < 0)
2186                         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",
2187                                              dev->name);
2188         }
2189
2190         return err;
2191 }
2192
2193 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2194 {
2195         struct net *net = sock_net(skb->sk);
2196         struct ifinfomsg *ifm;
2197         struct net_device *dev;
2198         int err;
2199         struct nlattr *tb[IFLA_MAX+1];
2200         char ifname[IFNAMSIZ];
2201
2202         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2203         if (err < 0)
2204                 goto errout;
2205
2206         if (tb[IFLA_IFNAME])
2207                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2208         else
2209                 ifname[0] = '\0';
2210
2211         err = -EINVAL;
2212         ifm = nlmsg_data(nlh);
2213         if (ifm->ifi_index > 0)
2214                 dev = __dev_get_by_index(net, ifm->ifi_index);
2215         else if (tb[IFLA_IFNAME])
2216                 dev = __dev_get_by_name(net, ifname);
2217         else
2218                 goto errout;
2219
2220         if (dev == NULL) {
2221                 err = -ENODEV;
2222                 goto errout;
2223         }
2224
2225         err = validate_linkmsg(dev, tb);
2226         if (err < 0)
2227                 goto errout;
2228
2229         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2230 errout:
2231         return err;
2232 }
2233
2234 static int rtnl_group_dellink(const struct net *net, int group)
2235 {
2236         struct net_device *dev, *aux;
2237         LIST_HEAD(list_kill);
2238         bool found = false;
2239
2240         if (!group)
2241                 return -EPERM;
2242
2243         for_each_netdev(net, dev) {
2244                 if (dev->group == group) {
2245                         const struct rtnl_link_ops *ops;
2246
2247                         found = true;
2248                         ops = dev->rtnl_link_ops;
2249                         if (!ops || !ops->dellink)
2250                                 return -EOPNOTSUPP;
2251                 }
2252         }
2253
2254         if (!found)
2255                 return -ENODEV;
2256
2257         for_each_netdev_safe(net, dev, aux) {
2258                 if (dev->group == group) {
2259                         const struct rtnl_link_ops *ops;
2260
2261                         ops = dev->rtnl_link_ops;
2262                         ops->dellink(dev, &list_kill);
2263                 }
2264         }
2265         unregister_netdevice_many(&list_kill);
2266
2267         return 0;
2268 }
2269
2270 int rtnl_delete_link(struct net_device *dev)
2271 {
2272         const struct rtnl_link_ops *ops;
2273         LIST_HEAD(list_kill);
2274
2275         ops = dev->rtnl_link_ops;
2276         if (!ops || !ops->dellink)
2277                 return -EOPNOTSUPP;
2278
2279         ops->dellink(dev, &list_kill);
2280         unregister_netdevice_many(&list_kill);
2281
2282         return 0;
2283 }
2284 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2285
2286 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2287 {
2288         struct net *net = sock_net(skb->sk);
2289         struct net_device *dev;
2290         struct ifinfomsg *ifm;
2291         char ifname[IFNAMSIZ];
2292         struct nlattr *tb[IFLA_MAX+1];
2293         int err;
2294
2295         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2296         if (err < 0)
2297                 return err;
2298
2299         if (tb[IFLA_IFNAME])
2300                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2301
2302         ifm = nlmsg_data(nlh);
2303         if (ifm->ifi_index > 0)
2304                 dev = __dev_get_by_index(net, ifm->ifi_index);
2305         else if (tb[IFLA_IFNAME])
2306                 dev = __dev_get_by_name(net, ifname);
2307         else if (tb[IFLA_GROUP])
2308                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2309         else
2310                 return -EINVAL;
2311
2312         if (!dev)
2313                 return -ENODEV;
2314
2315         return rtnl_delete_link(dev);
2316 }
2317
2318 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2319 {
2320         unsigned int old_flags;
2321         int err;
2322
2323         old_flags = dev->flags;
2324         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2325                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2326                 if (err < 0)
2327                         return err;
2328         }
2329
2330         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2331
2332         __dev_notify_flags(dev, old_flags, ~0U);
2333         return 0;
2334 }
2335 EXPORT_SYMBOL(rtnl_configure_link);
2336
2337 struct net_device *rtnl_create_link(struct net *net,
2338         const char *ifname, unsigned char name_assign_type,
2339         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2340 {
2341         int err;
2342         struct net_device *dev;
2343         unsigned int num_tx_queues = 1;
2344         unsigned int num_rx_queues = 1;
2345
2346         if (tb[IFLA_NUM_TX_QUEUES])
2347                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2348         else if (ops->get_num_tx_queues)
2349                 num_tx_queues = ops->get_num_tx_queues();
2350
2351         if (tb[IFLA_NUM_RX_QUEUES])
2352                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2353         else if (ops->get_num_rx_queues)
2354                 num_rx_queues = ops->get_num_rx_queues();
2355
2356         err = -ENOMEM;
2357         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2358                                ops->setup, num_tx_queues, num_rx_queues);
2359         if (!dev)
2360                 goto err;
2361
2362         dev_net_set(dev, net);
2363         dev->rtnl_link_ops = ops;
2364         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2365
2366         if (tb[IFLA_MTU])
2367                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2368         if (tb[IFLA_ADDRESS]) {
2369                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2370                                 nla_len(tb[IFLA_ADDRESS]));
2371                 dev->addr_assign_type = NET_ADDR_SET;
2372         }
2373         if (tb[IFLA_BROADCAST])
2374                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2375                                 nla_len(tb[IFLA_BROADCAST]));
2376         if (tb[IFLA_TXQLEN])
2377                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2378         if (tb[IFLA_OPERSTATE])
2379                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2380         if (tb[IFLA_LINKMODE])
2381                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2382         if (tb[IFLA_GROUP])
2383                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2384
2385         return dev;
2386
2387 err:
2388         return ERR_PTR(err);
2389 }
2390 EXPORT_SYMBOL(rtnl_create_link);
2391
2392 static int rtnl_group_changelink(const struct sk_buff *skb,
2393                 struct net *net, int group,
2394                 struct ifinfomsg *ifm,
2395                 struct nlattr **tb)
2396 {
2397         struct net_device *dev, *aux;
2398         int err;
2399
2400         for_each_netdev_safe(net, dev, aux) {
2401                 if (dev->group == group) {
2402                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2403                         if (err < 0)
2404                                 return err;
2405                 }
2406         }
2407
2408         return 0;
2409 }
2410
2411 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2412 {
2413         struct net *net = sock_net(skb->sk);
2414         const struct rtnl_link_ops *ops;
2415         const struct rtnl_link_ops *m_ops = NULL;
2416         struct net_device *dev;
2417         struct net_device *master_dev = NULL;
2418         struct ifinfomsg *ifm;
2419         char kind[MODULE_NAME_LEN];
2420         char ifname[IFNAMSIZ];
2421         struct nlattr *tb[IFLA_MAX+1];
2422         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2423         unsigned char name_assign_type = NET_NAME_USER;
2424         int err;
2425
2426 #ifdef CONFIG_MODULES
2427 replay:
2428 #endif
2429         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2430         if (err < 0)
2431                 return err;
2432
2433         if (tb[IFLA_IFNAME])
2434                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2435         else
2436                 ifname[0] = '\0';
2437
2438         ifm = nlmsg_data(nlh);
2439         if (ifm->ifi_index > 0)
2440                 dev = __dev_get_by_index(net, ifm->ifi_index);
2441         else {
2442                 if (ifname[0])
2443                         dev = __dev_get_by_name(net, ifname);
2444                 else
2445                         dev = NULL;
2446         }
2447
2448         if (dev) {
2449                 master_dev = netdev_master_upper_dev_get(dev);
2450                 if (master_dev)
2451                         m_ops = master_dev->rtnl_link_ops;
2452         }
2453
2454         err = validate_linkmsg(dev, tb);
2455         if (err < 0)
2456                 return err;
2457
2458         if (tb[IFLA_LINKINFO]) {
2459                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2460                                        tb[IFLA_LINKINFO], ifla_info_policy);
2461                 if (err < 0)
2462                         return err;
2463         } else
2464                 memset(linkinfo, 0, sizeof(linkinfo));
2465
2466         if (linkinfo[IFLA_INFO_KIND]) {
2467                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2468                 ops = rtnl_link_ops_get(kind);
2469         } else {
2470                 kind[0] = '\0';
2471                 ops = NULL;
2472         }
2473
2474         if (1) {
2475                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2476                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2477                 struct nlattr **data = NULL;
2478                 struct nlattr **slave_data = NULL;
2479                 struct net *dest_net, *link_net = NULL;
2480
2481                 if (ops) {
2482                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2483                                 err = nla_parse_nested(attr, ops->maxtype,
2484                                                        linkinfo[IFLA_INFO_DATA],
2485                                                        ops->policy);
2486                                 if (err < 0)
2487                                         return err;
2488                                 data = attr;
2489                         }
2490                         if (ops->validate) {
2491                                 err = ops->validate(tb, data);
2492                                 if (err < 0)
2493                                         return err;
2494                         }
2495                 }
2496
2497                 if (m_ops) {
2498                         if (m_ops->slave_maxtype &&
2499                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2500                                 err = nla_parse_nested(slave_attr,
2501                                                        m_ops->slave_maxtype,
2502                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2503                                                        m_ops->slave_policy);
2504                                 if (err < 0)
2505                                         return err;
2506                                 slave_data = slave_attr;
2507                         }
2508                         if (m_ops->slave_validate) {
2509                                 err = m_ops->slave_validate(tb, slave_data);
2510                                 if (err < 0)
2511                                         return err;
2512                         }
2513                 }
2514
2515                 if (dev) {
2516                         int status = 0;
2517
2518                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2519                                 return -EEXIST;
2520                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2521                                 return -EOPNOTSUPP;
2522
2523                         if (linkinfo[IFLA_INFO_DATA]) {
2524                                 if (!ops || ops != dev->rtnl_link_ops ||
2525                                     !ops->changelink)
2526                                         return -EOPNOTSUPP;
2527
2528                                 err = ops->changelink(dev, tb, data);
2529                                 if (err < 0)
2530                                         return err;
2531                                 status |= DO_SETLINK_NOTIFY;
2532                         }
2533
2534                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2535                                 if (!m_ops || !m_ops->slave_changelink)
2536                                         return -EOPNOTSUPP;
2537
2538                                 err = m_ops->slave_changelink(master_dev, dev,
2539                                                               tb, slave_data);
2540                                 if (err < 0)
2541                                         return err;
2542                                 status |= DO_SETLINK_NOTIFY;
2543                         }
2544
2545                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2546                 }
2547
2548                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2549                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2550                                 return rtnl_group_changelink(skb, net,
2551                                                 nla_get_u32(tb[IFLA_GROUP]),
2552                                                 ifm, tb);
2553                         return -ENODEV;
2554                 }
2555
2556                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2557                         return -EOPNOTSUPP;
2558
2559                 if (!ops) {
2560 #ifdef CONFIG_MODULES
2561                         if (kind[0]) {
2562                                 __rtnl_unlock();
2563                                 request_module("rtnl-link-%s", kind);
2564                                 rtnl_lock();
2565                                 ops = rtnl_link_ops_get(kind);
2566                                 if (ops)
2567                                         goto replay;
2568                         }
2569 #endif
2570                         return -EOPNOTSUPP;
2571                 }
2572
2573                 if (!ops->setup)
2574                         return -EOPNOTSUPP;
2575
2576                 if (!ifname[0]) {
2577                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2578                         name_assign_type = NET_NAME_ENUM;
2579                 }
2580
2581                 dest_net = rtnl_link_get_net(net, tb);
2582                 if (IS_ERR(dest_net))
2583                         return PTR_ERR(dest_net);
2584
2585                 err = -EPERM;
2586                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2587                         goto out;
2588
2589                 if (tb[IFLA_LINK_NETNSID]) {
2590                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2591
2592                         link_net = get_net_ns_by_id(dest_net, id);
2593                         if (!link_net) {
2594                                 err =  -EINVAL;
2595                                 goto out;
2596                         }
2597                         err = -EPERM;
2598                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2599                                 goto out;
2600                 }
2601
2602                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2603                                        name_assign_type, ops, tb);
2604                 if (IS_ERR(dev)) {
2605                         err = PTR_ERR(dev);
2606                         goto out;
2607                 }
2608
2609                 dev->ifindex = ifm->ifi_index;
2610
2611                 if (ops->newlink) {
2612                         err = ops->newlink(link_net ? : net, dev, tb, data);
2613                         /* Drivers should call free_netdev() in ->destructor
2614                          * and unregister it on failure after registration
2615                          * so that device could be finally freed in rtnl_unlock.
2616                          */
2617                         if (err < 0) {
2618                                 /* If device is not registered at all, free it now */
2619                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2620                                         free_netdev(dev);
2621                                 goto out;
2622                         }
2623                 } else {
2624                         err = register_netdevice(dev);
2625                         if (err < 0) {
2626                                 free_netdev(dev);
2627                                 goto out;
2628                         }
2629                 }
2630                 err = rtnl_configure_link(dev, ifm);
2631                 if (err < 0)
2632                         goto out_unregister;
2633                 if (link_net) {
2634                         err = dev_change_net_namespace(dev, dest_net, ifname);
2635                         if (err < 0)
2636                                 goto out_unregister;
2637                 }
2638 out:
2639                 if (link_net)
2640                         put_net(link_net);
2641                 put_net(dest_net);
2642                 return err;
2643 out_unregister:
2644                 if (ops->newlink) {
2645                         LIST_HEAD(list_kill);
2646
2647                         ops->dellink(dev, &list_kill);
2648                         unregister_netdevice_many(&list_kill);
2649                 } else {
2650                         unregister_netdevice(dev);
2651                 }
2652                 goto out;
2653         }
2654 }
2655
2656 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2657 {
2658         struct net *net = sock_net(skb->sk);
2659         struct ifinfomsg *ifm;
2660         char ifname[IFNAMSIZ];
2661         struct nlattr *tb[IFLA_MAX+1];
2662         struct net_device *dev = NULL;
2663         struct sk_buff *nskb;
2664         int err;
2665         u32 ext_filter_mask = 0;
2666
2667         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2668         if (err < 0)
2669                 return err;
2670
2671         if (tb[IFLA_IFNAME])
2672                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2673
2674         if (tb[IFLA_EXT_MASK])
2675                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2676
2677         ifm = nlmsg_data(nlh);
2678         if (ifm->ifi_index > 0)
2679                 dev = __dev_get_by_index(net, ifm->ifi_index);
2680         else if (tb[IFLA_IFNAME])
2681                 dev = __dev_get_by_name(net, ifname);
2682         else
2683                 return -EINVAL;
2684
2685         if (dev == NULL)
2686                 return -ENODEV;
2687
2688         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2689         if (nskb == NULL)
2690                 return -ENOBUFS;
2691
2692         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2693                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2694         if (err < 0) {
2695                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2696                 WARN_ON(err == -EMSGSIZE);
2697                 kfree_skb(nskb);
2698         } else
2699                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2700
2701         return err;
2702 }
2703
2704 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2705 {
2706         struct net *net = sock_net(skb->sk);
2707         struct net_device *dev;
2708         struct nlattr *tb[IFLA_MAX+1];
2709         u32 ext_filter_mask = 0;
2710         u16 min_ifinfo_dump_size = 0;
2711         int hdrlen;
2712
2713         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2714         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2715                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2716
2717         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2718                 if (tb[IFLA_EXT_MASK])
2719                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2720         }
2721
2722         if (!ext_filter_mask)
2723                 return NLMSG_GOODSIZE;
2724         /*
2725          * traverse the list of net devices and compute the minimum
2726          * buffer size based upon the filter mask.
2727          */
2728         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2729                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2730                                              if_nlmsg_size(dev,
2731                                                            ext_filter_mask));
2732         }
2733
2734         return min_ifinfo_dump_size;
2735 }
2736
2737 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2738 {
2739         int idx;
2740         int s_idx = cb->family;
2741
2742         if (s_idx == 0)
2743                 s_idx = 1;
2744         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2745                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2746                 if (idx < s_idx || idx == PF_PACKET)
2747                         continue;
2748                 if (rtnl_msg_handlers[idx] == NULL ||
2749                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2750                         continue;
2751                 if (idx > s_idx) {
2752                         memset(&cb->args[0], 0, sizeof(cb->args));
2753                         cb->prev_seq = 0;
2754                         cb->seq = 0;
2755                 }
2756                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2757                         break;
2758         }
2759         cb->family = idx;
2760
2761         return skb->len;
2762 }
2763
2764 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2765                                        unsigned int change, gfp_t flags)
2766 {
2767         struct net *net = dev_net(dev);
2768         struct sk_buff *skb;
2769         int err = -ENOBUFS;
2770         size_t if_info_size;
2771
2772         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2773         if (skb == NULL)
2774                 goto errout;
2775
2776         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2777         if (err < 0) {
2778                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2779                 WARN_ON(err == -EMSGSIZE);
2780                 kfree_skb(skb);
2781                 goto errout;
2782         }
2783         return skb;
2784 errout:
2785         if (err < 0)
2786                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2787         return NULL;
2788 }
2789
2790 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2791 {
2792         struct net *net = dev_net(dev);
2793
2794         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2795 }
2796
2797 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2798                   gfp_t flags)
2799 {
2800         struct sk_buff *skb;
2801
2802         if (dev->reg_state != NETREG_REGISTERED)
2803                 return;
2804
2805         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2806         if (skb)
2807                 rtmsg_ifinfo_send(skb, dev, flags);
2808 }
2809 EXPORT_SYMBOL(rtmsg_ifinfo);
2810
2811 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2812                                    struct net_device *dev,
2813                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2814                                    int type, unsigned int flags,
2815                                    int nlflags, u16 ndm_state)
2816 {
2817         struct nlmsghdr *nlh;
2818         struct ndmsg *ndm;
2819
2820         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2821         if (!nlh)
2822                 return -EMSGSIZE;
2823
2824         ndm = nlmsg_data(nlh);
2825         ndm->ndm_family  = AF_BRIDGE;
2826         ndm->ndm_pad1    = 0;
2827         ndm->ndm_pad2    = 0;
2828         ndm->ndm_flags   = flags;
2829         ndm->ndm_type    = 0;
2830         ndm->ndm_ifindex = dev->ifindex;
2831         ndm->ndm_state   = ndm_state;
2832
2833         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2834                 goto nla_put_failure;
2835         if (vid)
2836                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2837                         goto nla_put_failure;
2838
2839         nlmsg_end(skb, nlh);
2840         return 0;
2841
2842 nla_put_failure:
2843         nlmsg_cancel(skb, nlh);
2844         return -EMSGSIZE;
2845 }
2846
2847 static inline size_t rtnl_fdb_nlmsg_size(void)
2848 {
2849         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2850 }
2851
2852 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2853                             u16 ndm_state)
2854 {
2855         struct net *net = dev_net(dev);
2856         struct sk_buff *skb;
2857         int err = -ENOBUFS;
2858
2859         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2860         if (!skb)
2861                 goto errout;
2862
2863         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2864                                       0, 0, type, NTF_SELF, 0, ndm_state);
2865         if (err < 0) {
2866                 kfree_skb(skb);
2867                 goto errout;
2868         }
2869
2870         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2871         return;
2872 errout:
2873         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2874 }
2875
2876 /**
2877  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2878  */
2879 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2880                      struct nlattr *tb[],
2881                      struct net_device *dev,
2882                      const unsigned char *addr, u16 vid,
2883                      u16 flags)
2884 {
2885         int err = -EINVAL;
2886
2887         /* If aging addresses are supported device will need to
2888          * implement its own handler for this.
2889          */
2890         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2891                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2892                 return err;
2893         }
2894
2895         if (vid) {
2896                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2897                 return err;
2898         }
2899
2900         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2901                 err = dev_uc_add_excl(dev, addr);
2902         else if (is_multicast_ether_addr(addr))
2903                 err = dev_mc_add_excl(dev, addr);
2904
2905         /* Only return duplicate errors if NLM_F_EXCL is set */
2906         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2907                 err = 0;
2908
2909         return err;
2910 }
2911 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2912
2913 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2914 {
2915         u16 vid = 0;
2916
2917         if (vlan_attr) {
2918                 if (nla_len(vlan_attr) != sizeof(u16)) {
2919                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2920                         return -EINVAL;
2921                 }
2922
2923                 vid = nla_get_u16(vlan_attr);
2924
2925                 if (!vid || vid >= VLAN_VID_MASK) {
2926                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2927                                 vid);
2928                         return -EINVAL;
2929                 }
2930         }
2931         *p_vid = vid;
2932         return 0;
2933 }
2934
2935 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2936 {
2937         struct net *net = sock_net(skb->sk);
2938         struct ndmsg *ndm;
2939         struct nlattr *tb[NDA_MAX+1];
2940         struct net_device *dev;
2941         u8 *addr;
2942         u16 vid;
2943         int err;
2944
2945         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2946         if (err < 0)
2947                 return err;
2948
2949         ndm = nlmsg_data(nlh);
2950         if (ndm->ndm_ifindex == 0) {
2951                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2952                 return -EINVAL;
2953         }
2954
2955         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2956         if (dev == NULL) {
2957                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2958                 return -ENODEV;
2959         }
2960
2961         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2962                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2963                 return -EINVAL;
2964         }
2965
2966         addr = nla_data(tb[NDA_LLADDR]);
2967
2968         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2969         if (err)
2970                 return err;
2971
2972         err = -EOPNOTSUPP;
2973
2974         /* Support fdb on master device the net/bridge default case */
2975         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2976             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2977                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2978                 const struct net_device_ops *ops = br_dev->netdev_ops;
2979
2980                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2981                                        nlh->nlmsg_flags);
2982                 if (err)
2983                         goto out;
2984                 else
2985                         ndm->ndm_flags &= ~NTF_MASTER;
2986         }
2987
2988         /* Embedded bridge, macvlan, and any other device support */
2989         if ((ndm->ndm_flags & NTF_SELF)) {
2990                 if (dev->netdev_ops->ndo_fdb_add)
2991                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2992                                                            vid,
2993                                                            nlh->nlmsg_flags);
2994                 else
2995                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2996                                                nlh->nlmsg_flags);
2997
2998                 if (!err) {
2999                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3000                                         ndm->ndm_state);
3001                         ndm->ndm_flags &= ~NTF_SELF;
3002                 }
3003         }
3004 out:
3005         return err;
3006 }
3007
3008 /**
3009  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3010  */
3011 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3012                      struct nlattr *tb[],
3013                      struct net_device *dev,
3014                      const unsigned char *addr, u16 vid)
3015 {
3016         int err = -EINVAL;
3017
3018         /* If aging addresses are supported device will need to
3019          * implement its own handler for this.
3020          */
3021         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3022                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3023                 return err;
3024         }
3025
3026         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3027                 err = dev_uc_del(dev, addr);
3028         else if (is_multicast_ether_addr(addr))
3029                 err = dev_mc_del(dev, addr);
3030
3031         return err;
3032 }
3033 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3034
3035 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3036 {
3037         struct net *net = sock_net(skb->sk);
3038         struct ndmsg *ndm;
3039         struct nlattr *tb[NDA_MAX+1];
3040         struct net_device *dev;
3041         int err = -EINVAL;
3042         __u8 *addr;
3043         u16 vid;
3044
3045         if (!netlink_capable(skb, CAP_NET_ADMIN))
3046                 return -EPERM;
3047
3048         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3049         if (err < 0)
3050                 return err;
3051
3052         ndm = nlmsg_data(nlh);
3053         if (ndm->ndm_ifindex == 0) {
3054                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3055                 return -EINVAL;
3056         }
3057
3058         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3059         if (dev == NULL) {
3060                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3061                 return -ENODEV;
3062         }
3063
3064         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3065                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3066                 return -EINVAL;
3067         }
3068
3069         addr = nla_data(tb[NDA_LLADDR]);
3070
3071         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3072         if (err)
3073                 return err;
3074
3075         err = -EOPNOTSUPP;
3076
3077         /* Support fdb on master device the net/bridge default case */
3078         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3079             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3080                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3081                 const struct net_device_ops *ops = br_dev->netdev_ops;
3082
3083                 if (ops->ndo_fdb_del)
3084                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3085
3086                 if (err)
3087                         goto out;
3088                 else
3089                         ndm->ndm_flags &= ~NTF_MASTER;
3090         }
3091
3092         /* Embedded bridge, macvlan, and any other device support */
3093         if (ndm->ndm_flags & NTF_SELF) {
3094                 if (dev->netdev_ops->ndo_fdb_del)
3095                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3096                                                            vid);
3097                 else
3098                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3099
3100                 if (!err) {
3101                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3102                                         ndm->ndm_state);
3103                         ndm->ndm_flags &= ~NTF_SELF;
3104                 }
3105         }
3106 out:
3107         return err;
3108 }
3109
3110 static int nlmsg_populate_fdb(struct sk_buff *skb,
3111                               struct netlink_callback *cb,
3112                               struct net_device *dev,
3113                               int *idx,
3114                               struct netdev_hw_addr_list *list)
3115 {
3116         struct netdev_hw_addr *ha;
3117         int err;
3118         u32 portid, seq;
3119
3120         portid = NETLINK_CB(cb->skb).portid;
3121         seq = cb->nlh->nlmsg_seq;
3122
3123         list_for_each_entry(ha, &list->list, list) {
3124                 if (*idx < cb->args[2])
3125                         goto skip;
3126
3127                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3128                                               portid, seq,
3129                                               RTM_NEWNEIGH, NTF_SELF,
3130                                               NLM_F_MULTI, NUD_PERMANENT);
3131                 if (err < 0)
3132                         return err;
3133 skip:
3134                 *idx += 1;
3135         }
3136         return 0;
3137 }
3138
3139 /**
3140  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3141  * @nlh: netlink message header
3142  * @dev: netdevice
3143  *
3144  * Default netdevice operation to dump the existing unicast address list.
3145  * Returns number of addresses from list put in skb.
3146  */
3147 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3148                       struct netlink_callback *cb,
3149                       struct net_device *dev,
3150                       struct net_device *filter_dev,
3151                       int *idx)
3152 {
3153         int err;
3154
3155         netif_addr_lock_bh(dev);
3156         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3157         if (err)
3158                 goto out;
3159         nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3160 out:
3161         netif_addr_unlock_bh(dev);
3162         return err;
3163 }
3164 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3165
3166 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3167 {
3168         struct net_device *dev;
3169         struct nlattr *tb[IFLA_MAX+1];
3170         struct net_device *br_dev = NULL;
3171         const struct net_device_ops *ops = NULL;
3172         const struct net_device_ops *cops = NULL;
3173         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3174         struct net *net = sock_net(skb->sk);
3175         struct hlist_head *head;
3176         int brport_idx = 0;
3177         int br_idx = 0;
3178         int h, s_h;
3179         int idx = 0, s_idx;
3180         int err = 0;
3181         int fidx = 0;
3182
3183         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3184                         ifla_policy) == 0) {
3185                 if (tb[IFLA_MASTER])
3186                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3187         }
3188
3189         brport_idx = ifm->ifi_index;
3190
3191         if (br_idx) {
3192                 br_dev = __dev_get_by_index(net, br_idx);
3193                 if (!br_dev)
3194                         return -ENODEV;
3195
3196                 ops = br_dev->netdev_ops;
3197         }
3198
3199         s_h = cb->args[0];
3200         s_idx = cb->args[1];
3201
3202         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3203                 idx = 0;
3204                 head = &net->dev_index_head[h];
3205                 hlist_for_each_entry(dev, head, index_hlist) {
3206
3207                         if (brport_idx && (dev->ifindex != brport_idx))
3208                                 continue;
3209
3210                         if (!br_idx) { /* user did not specify a specific bridge */
3211                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3212                                         br_dev = netdev_master_upper_dev_get(dev);
3213                                         cops = br_dev->netdev_ops;
3214                                 }
3215                         } else {
3216                                 if (dev != br_dev &&
3217                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3218                                         continue;
3219
3220                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3221                                     !(dev->priv_flags & IFF_EBRIDGE))
3222                                         continue;
3223                                 cops = ops;
3224                         }
3225
3226                         if (idx < s_idx)
3227                                 goto cont;
3228
3229                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3230                                 if (cops && cops->ndo_fdb_dump) {
3231                                         err = cops->ndo_fdb_dump(skb, cb,
3232                                                                 br_dev, dev,
3233                                                                 &fidx);
3234                                         if (err == -EMSGSIZE)
3235                                                 goto out;
3236                                 }
3237                         }
3238
3239                         if (dev->netdev_ops->ndo_fdb_dump)
3240                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3241                                                                     dev, NULL,
3242                                                                     &fidx);
3243                         else
3244                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3245                                                         &fidx);
3246                         if (err == -EMSGSIZE)
3247                                 goto out;
3248
3249                         cops = NULL;
3250
3251                         /* reset fdb offset to 0 for rest of the interfaces */
3252                         cb->args[2] = 0;
3253                         fidx = 0;
3254 cont:
3255                         idx++;
3256                 }
3257         }
3258
3259 out:
3260         cb->args[0] = h;
3261         cb->args[1] = idx;
3262         cb->args[2] = fidx;
3263
3264         return skb->len;
3265 }
3266
3267 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3268                                unsigned int attrnum, unsigned int flag)
3269 {
3270         if (mask & flag)
3271                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3272         return 0;
3273 }
3274
3275 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3276                             struct net_device *dev, u16 mode,
3277                             u32 flags, u32 mask, int nlflags,
3278                             u32 filter_mask,
3279                             int (*vlan_fill)(struct sk_buff *skb,
3280                                              struct net_device *dev,
3281                                              u32 filter_mask))
3282 {
3283         struct nlmsghdr *nlh;
3284         struct ifinfomsg *ifm;
3285         struct nlattr *br_afspec;
3286         struct nlattr *protinfo;
3287         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3288         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3289         int err = 0;
3290
3291         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3292         if (nlh == NULL)
3293                 return -EMSGSIZE;
3294
3295         ifm = nlmsg_data(nlh);
3296         ifm->ifi_family = AF_BRIDGE;
3297         ifm->__ifi_pad = 0;
3298         ifm->ifi_type = dev->type;
3299         ifm->ifi_index = dev->ifindex;
3300         ifm->ifi_flags = dev_get_flags(dev);
3301         ifm->ifi_change = 0;
3302
3303
3304         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3305             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3306             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3307             (br_dev &&
3308              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3309             (dev->addr_len &&
3310              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3311             (dev->ifindex != dev_get_iflink(dev) &&
3312              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3313                 goto nla_put_failure;
3314
3315         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3316         if (!br_afspec)
3317                 goto nla_put_failure;
3318
3319         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3320                 nla_nest_cancel(skb, br_afspec);
3321                 goto nla_put_failure;
3322         }
3323
3324         if (mode != BRIDGE_MODE_UNDEF) {
3325                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3326                         nla_nest_cancel(skb, br_afspec);
3327                         goto nla_put_failure;
3328                 }
3329         }
3330         if (vlan_fill) {
3331                 err = vlan_fill(skb, dev, filter_mask);
3332                 if (err) {
3333                         nla_nest_cancel(skb, br_afspec);
3334                         goto nla_put_failure;
3335                 }
3336         }
3337         nla_nest_end(skb, br_afspec);
3338
3339         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3340         if (!protinfo)
3341                 goto nla_put_failure;
3342
3343         if (brport_nla_put_flag(skb, flags, mask,
3344                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3345             brport_nla_put_flag(skb, flags, mask,
3346                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3347             brport_nla_put_flag(skb, flags, mask,
3348                                 IFLA_BRPORT_FAST_LEAVE,
3349                                 BR_MULTICAST_FAST_LEAVE) ||
3350             brport_nla_put_flag(skb, flags, mask,
3351                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3352             brport_nla_put_flag(skb, flags, mask,
3353                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3354             brport_nla_put_flag(skb, flags, mask,
3355                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3356             brport_nla_put_flag(skb, flags, mask,
3357                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3358             brport_nla_put_flag(skb, flags, mask,
3359                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3360                 nla_nest_cancel(skb, protinfo);
3361                 goto nla_put_failure;
3362         }
3363
3364         nla_nest_end(skb, protinfo);
3365
3366         nlmsg_end(skb, nlh);
3367         return 0;
3368 nla_put_failure:
3369         nlmsg_cancel(skb, nlh);
3370         return err ? err : -EMSGSIZE;
3371 }
3372 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3373
3374 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3375 {
3376         struct net *net = sock_net(skb->sk);
3377         struct net_device *dev;
3378         int idx = 0;
3379         u32 portid = NETLINK_CB(cb->skb).portid;
3380         u32 seq = cb->nlh->nlmsg_seq;
3381         u32 filter_mask = 0;
3382         int err;
3383
3384         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3385                 struct nlattr *extfilt;
3386
3387                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3388                                           IFLA_EXT_MASK);
3389                 if (extfilt) {
3390                         if (nla_len(extfilt) < sizeof(filter_mask))
3391                                 return -EINVAL;
3392
3393                         filter_mask = nla_get_u32(extfilt);
3394                 }
3395         }
3396
3397         rcu_read_lock();
3398         for_each_netdev_rcu(net, dev) {
3399                 const struct net_device_ops *ops = dev->netdev_ops;
3400                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3401
3402                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3403                         if (idx >= cb->args[0]) {
3404                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3405                                                 skb, portid, seq, dev,
3406                                                 filter_mask, NLM_F_MULTI);
3407                                 if (err < 0 && err != -EOPNOTSUPP)
3408                                         break;
3409                         }
3410                         idx++;
3411                 }
3412
3413                 if (ops->ndo_bridge_getlink) {
3414                         if (idx >= cb->args[0]) {
3415                                 err = ops->ndo_bridge_getlink(skb, portid,
3416                                                               seq, dev,
3417                                                               filter_mask,
3418                                                               NLM_F_MULTI);
3419                                 if (err < 0 && err != -EOPNOTSUPP)
3420                                         break;
3421                         }
3422                         idx++;
3423                 }
3424         }
3425         rcu_read_unlock();
3426         cb->args[0] = idx;
3427
3428         return skb->len;
3429 }
3430
3431 static inline size_t bridge_nlmsg_size(void)
3432 {
3433         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3434                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3435                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3436                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3437                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3438                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3439                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3440                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3441                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3442                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3443                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3444 }
3445
3446 static int rtnl_bridge_notify(struct net_device *dev)
3447 {
3448         struct net *net = dev_net(dev);
3449         struct sk_buff *skb;
3450         int err = -EOPNOTSUPP;
3451
3452         if (!dev->netdev_ops->ndo_bridge_getlink)
3453                 return 0;
3454
3455         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3456         if (!skb) {
3457                 err = -ENOMEM;
3458                 goto errout;
3459         }
3460
3461         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3462         if (err < 0)
3463                 goto errout;
3464
3465         if (!skb->len)
3466                 goto errout;
3467
3468         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3469         return 0;
3470 errout:
3471         WARN_ON(err == -EMSGSIZE);
3472         kfree_skb(skb);
3473         if (err)
3474                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3475         return err;
3476 }
3477
3478 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3479 {
3480         struct net *net = sock_net(skb->sk);
3481         struct ifinfomsg *ifm;
3482         struct net_device *dev;
3483         struct nlattr *br_spec, *attr = NULL;
3484         int rem, err = -EOPNOTSUPP;
3485         u16 flags = 0;
3486         bool have_flags = false;
3487
3488         if (nlmsg_len(nlh) < sizeof(*ifm))
3489                 return -EINVAL;
3490
3491         ifm = nlmsg_data(nlh);
3492         if (ifm->ifi_family != AF_BRIDGE)
3493                 return -EPFNOSUPPORT;
3494
3495         dev = __dev_get_by_index(net, ifm->ifi_index);
3496         if (!dev) {
3497                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3498                 return -ENODEV;
3499         }
3500
3501         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3502         if (br_spec) {
3503                 nla_for_each_nested(attr, br_spec, rem) {
3504                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3505                                 if (nla_len(attr) < sizeof(flags))
3506                                         return -EINVAL;
3507
3508                                 have_flags = true;
3509                                 flags = nla_get_u16(attr);
3510                                 break;
3511                         }
3512                 }
3513         }
3514
3515         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3516                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3517
3518                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3519                         err = -EOPNOTSUPP;
3520                         goto out;
3521                 }
3522
3523                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3524                 if (err)
3525                         goto out;
3526
3527                 flags &= ~BRIDGE_FLAGS_MASTER;
3528         }
3529
3530         if ((flags & BRIDGE_FLAGS_SELF)) {
3531                 if (!dev->netdev_ops->ndo_bridge_setlink)
3532                         err = -EOPNOTSUPP;
3533                 else
3534                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3535                                                                   flags);
3536                 if (!err) {
3537                         flags &= ~BRIDGE_FLAGS_SELF;
3538
3539                         /* Generate event to notify upper layer of bridge
3540                          * change
3541                          */
3542                         err = rtnl_bridge_notify(dev);
3543                 }
3544         }
3545
3546         if (have_flags)
3547                 memcpy(nla_data(attr), &flags, sizeof(flags));
3548 out:
3549         return err;
3550 }
3551
3552 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3553 {
3554         struct net *net = sock_net(skb->sk);
3555         struct ifinfomsg *ifm;
3556         struct net_device *dev;
3557         struct nlattr *br_spec, *attr = NULL;
3558         int rem, err = -EOPNOTSUPP;
3559         u16 flags = 0;
3560         bool have_flags = false;
3561
3562         if (nlmsg_len(nlh) < sizeof(*ifm))
3563                 return -EINVAL;
3564
3565         ifm = nlmsg_data(nlh);
3566         if (ifm->ifi_family != AF_BRIDGE)
3567                 return -EPFNOSUPPORT;
3568
3569         dev = __dev_get_by_index(net, ifm->ifi_index);
3570         if (!dev) {
3571                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3572                 return -ENODEV;
3573         }
3574
3575         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3576         if (br_spec) {
3577                 nla_for_each_nested(attr, br_spec, rem) {
3578                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3579                                 if (nla_len(attr) < sizeof(flags))
3580                                         return -EINVAL;
3581
3582                                 have_flags = true;
3583                                 flags = nla_get_u16(attr);
3584                                 break;
3585                         }
3586                 }
3587         }
3588
3589         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3590                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3591
3592                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3593                         err = -EOPNOTSUPP;
3594                         goto out;
3595                 }
3596
3597                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3598                 if (err)
3599                         goto out;
3600
3601                 flags &= ~BRIDGE_FLAGS_MASTER;
3602         }
3603
3604         if ((flags & BRIDGE_FLAGS_SELF)) {
3605                 if (!dev->netdev_ops->ndo_bridge_dellink)
3606                         err = -EOPNOTSUPP;
3607                 else
3608                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3609                                                                   flags);
3610
3611                 if (!err) {
3612                         flags &= ~BRIDGE_FLAGS_SELF;
3613
3614                         /* Generate event to notify upper layer of bridge
3615                          * change
3616                          */
3617                         err = rtnl_bridge_notify(dev);
3618                 }
3619         }
3620
3621         if (have_flags)
3622                 memcpy(nla_data(attr), &flags, sizeof(flags));
3623 out:
3624         return err;
3625 }
3626
3627 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3628 {
3629         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3630                (!idxattr || idxattr == attrid);
3631 }
3632
3633 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3634 static int rtnl_get_offload_stats_attr_size(int attr_id)
3635 {
3636         switch (attr_id) {
3637         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3638                 return sizeof(struct rtnl_link_stats64);
3639         }
3640
3641         return 0;
3642 }
3643
3644 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3645                                   int *prividx)
3646 {
3647         struct nlattr *attr = NULL;
3648         int attr_id, size;
3649         void *attr_data;
3650         int err;
3651
3652         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3653               dev->netdev_ops->ndo_get_offload_stats))
3654                 return -ENODATA;
3655
3656         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3657              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3658                 if (attr_id < *prividx)
3659                         continue;
3660
3661                 size = rtnl_get_offload_stats_attr_size(attr_id);
3662                 if (!size)
3663                         continue;
3664
3665                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3666                         continue;
3667
3668                 attr = nla_reserve_64bit(skb, attr_id, size,
3669                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3670                 if (!attr)
3671                         goto nla_put_failure;
3672
3673                 attr_data = nla_data(attr);
3674                 memset(attr_data, 0, size);
3675                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3676                                                              attr_data);
3677                 if (err)
3678                         goto get_offload_stats_failure;
3679         }
3680
3681         if (!attr)
3682                 return -ENODATA;
3683
3684         *prividx = 0;
3685         return 0;
3686
3687 nla_put_failure:
3688         err = -EMSGSIZE;
3689 get_offload_stats_failure:
3690         *prividx = attr_id;
3691         return err;
3692 }
3693
3694 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3695 {
3696         int nla_size = 0;
3697         int attr_id;
3698         int size;
3699
3700         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3701               dev->netdev_ops->ndo_get_offload_stats))
3702                 return 0;
3703
3704         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3705              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3706                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3707                         continue;
3708                 size = rtnl_get_offload_stats_attr_size(attr_id);
3709                 nla_size += nla_total_size_64bit(size);
3710         }
3711
3712         if (nla_size != 0)
3713                 nla_size += nla_total_size(0);
3714
3715         return nla_size;
3716 }
3717
3718 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3719                                int type, u32 pid, u32 seq, u32 change,
3720                                unsigned int flags, unsigned int filter_mask,
3721                                int *idxattr, int *prividx)
3722 {
3723         struct if_stats_msg *ifsm;
3724         struct nlmsghdr *nlh;
3725         struct nlattr *attr;
3726         int s_prividx = *prividx;
3727         int err;
3728
3729         ASSERT_RTNL();
3730
3731         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3732         if (!nlh)
3733                 return -EMSGSIZE;
3734
3735         ifsm = nlmsg_data(nlh);
3736         ifsm->ifindex = dev->ifindex;
3737         ifsm->filter_mask = filter_mask;
3738
3739         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3740                 struct rtnl_link_stats64 *sp;
3741
3742                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3743                                          sizeof(struct rtnl_link_stats64),
3744                                          IFLA_STATS_UNSPEC);
3745                 if (!attr)
3746                         goto nla_put_failure;
3747
3748                 sp = nla_data(attr);
3749                 dev_get_stats(dev, sp);
3750         }
3751
3752         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3753                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3754
3755                 if (ops && ops->fill_linkxstats) {
3756                         *idxattr = IFLA_STATS_LINK_XSTATS;
3757                         attr = nla_nest_start(skb,
3758                                               IFLA_STATS_LINK_XSTATS);
3759                         if (!attr)
3760                                 goto nla_put_failure;
3761
3762                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3763                         nla_nest_end(skb, attr);
3764                         if (err)
3765                                 goto nla_put_failure;
3766                         *idxattr = 0;
3767                 }
3768         }
3769
3770         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3771                              *idxattr)) {
3772                 const struct rtnl_link_ops *ops = NULL;
3773                 const struct net_device *master;
3774
3775                 master = netdev_master_upper_dev_get(dev);
3776                 if (master)
3777                         ops = master->rtnl_link_ops;
3778                 if (ops && ops->fill_linkxstats) {
3779                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3780                         attr = nla_nest_start(skb,
3781                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3782                         if (!attr)
3783                                 goto nla_put_failure;
3784
3785                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3786                         nla_nest_end(skb, attr);
3787                         if (err)
3788                                 goto nla_put_failure;
3789                         *idxattr = 0;
3790                 }
3791         }
3792
3793         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3794                              *idxattr)) {
3795                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3796                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3797                 if (!attr)
3798                         goto nla_put_failure;
3799
3800                 err = rtnl_get_offload_stats(skb, dev, prividx);
3801                 if (err == -ENODATA)
3802                         nla_nest_cancel(skb, attr);
3803                 else
3804                         nla_nest_end(skb, attr);
3805
3806                 if (err && err != -ENODATA)
3807                         goto nla_put_failure;
3808                 *idxattr = 0;
3809         }
3810
3811         nlmsg_end(skb, nlh);
3812
3813         return 0;
3814
3815 nla_put_failure:
3816         /* not a multi message or no progress mean a real error */
3817         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3818                 nlmsg_cancel(skb, nlh);
3819         else
3820                 nlmsg_end(skb, nlh);
3821
3822         return -EMSGSIZE;
3823 }
3824
3825 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3826                                   u32 filter_mask)
3827 {
3828         size_t size = 0;
3829
3830         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3831                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3832
3833         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3834                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3835                 int attr = IFLA_STATS_LINK_XSTATS;
3836
3837                 if (ops && ops->get_linkxstats_size) {
3838                         size += nla_total_size(ops->get_linkxstats_size(dev,
3839                                                                         attr));
3840                         /* for IFLA_STATS_LINK_XSTATS */
3841                         size += nla_total_size(0);
3842                 }
3843         }
3844
3845         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3846                 struct net_device *_dev = (struct net_device *)dev;
3847                 const struct rtnl_link_ops *ops = NULL;
3848                 const struct net_device *master;
3849
3850                 /* netdev_master_upper_dev_get can't take const */
3851                 master = netdev_master_upper_dev_get(_dev);
3852                 if (master)
3853                         ops = master->rtnl_link_ops;
3854                 if (ops && ops->get_linkxstats_size) {
3855                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3856
3857                         size += nla_total_size(ops->get_linkxstats_size(dev,
3858                                                                         attr));
3859                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3860                         size += nla_total_size(0);
3861                 }
3862         }
3863
3864         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3865                 size += rtnl_get_offload_stats_size(dev);
3866
3867         return size;
3868 }
3869
3870 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3871 {
3872         struct net *net = sock_net(skb->sk);
3873         struct net_device *dev = NULL;
3874         int idxattr = 0, prividx = 0;
3875         struct if_stats_msg *ifsm;
3876         struct sk_buff *nskb;
3877         u32 filter_mask;
3878         int err;
3879
3880         ifsm = nlmsg_data(nlh);
3881         if (ifsm->ifindex > 0)
3882                 dev = __dev_get_by_index(net, ifsm->ifindex);
3883         else
3884                 return -EINVAL;
3885
3886         if (!dev)
3887                 return -ENODEV;
3888
3889         filter_mask = ifsm->filter_mask;
3890         if (!filter_mask)
3891                 return -EINVAL;
3892
3893         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3894         if (!nskb)
3895                 return -ENOBUFS;
3896
3897         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3898                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3899                                   0, filter_mask, &idxattr, &prividx);
3900         if (err < 0) {
3901                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3902                 WARN_ON(err == -EMSGSIZE);
3903                 kfree_skb(nskb);
3904         } else {
3905                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3906         }
3907
3908         return err;
3909 }
3910
3911 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3912 {
3913         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3914         struct net *net = sock_net(skb->sk);
3915         unsigned int flags = NLM_F_MULTI;
3916         struct if_stats_msg *ifsm;
3917         struct hlist_head *head;
3918         struct net_device *dev;
3919         u32 filter_mask = 0;
3920         int idx = 0;
3921
3922         s_h = cb->args[0];
3923         s_idx = cb->args[1];
3924         s_idxattr = cb->args[2];
3925         s_prividx = cb->args[3];
3926
3927         cb->seq = net->dev_base_seq;
3928
3929         ifsm = nlmsg_data(cb->nlh);
3930         filter_mask = ifsm->filter_mask;
3931         if (!filter_mask)
3932                 return -EINVAL;
3933
3934         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3935                 idx = 0;
3936                 head = &net->dev_index_head[h];
3937                 hlist_for_each_entry(dev, head, index_hlist) {
3938                         if (idx < s_idx)
3939                                 goto cont;
3940                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3941                                                   NETLINK_CB(cb->skb).portid,
3942                                                   cb->nlh->nlmsg_seq, 0,
3943                                                   flags, filter_mask,
3944                                                   &s_idxattr, &s_prividx);
3945                         /* If we ran out of room on the first message,
3946                          * we're in trouble
3947                          */
3948                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3949
3950                         if (err < 0)
3951                                 goto out;
3952                         s_prividx = 0;
3953                         s_idxattr = 0;
3954                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3955 cont:
3956                         idx++;
3957                 }
3958         }
3959 out:
3960         cb->args[3] = s_prividx;
3961         cb->args[2] = s_idxattr;
3962         cb->args[1] = idx;
3963         cb->args[0] = h;
3964
3965         return skb->len;
3966 }
3967
3968 /* Process one rtnetlink message. */
3969
3970 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3971 {
3972         struct net *net = sock_net(skb->sk);
3973         rtnl_doit_func doit;
3974         int kind;
3975         int family;
3976         int type;
3977         int err;
3978
3979         type = nlh->nlmsg_type;
3980         if (type > RTM_MAX)
3981                 return -EOPNOTSUPP;
3982
3983         type -= RTM_BASE;
3984
3985         /* All the messages must have at least 1 byte length */
3986         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3987                 return 0;
3988
3989         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3990         kind = type&3;
3991
3992         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3993                 return -EPERM;
3994
3995         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3996                 struct sock *rtnl;
3997                 rtnl_dumpit_func dumpit;
3998                 rtnl_calcit_func calcit;
3999                 u16 min_dump_alloc = 0;
4000
4001                 dumpit = rtnl_get_dumpit(family, type);
4002                 if (dumpit == NULL)
4003                         return -EOPNOTSUPP;
4004                 calcit = rtnl_get_calcit(family, type);
4005                 if (calcit)
4006                         min_dump_alloc = calcit(skb, nlh);
4007
4008                 __rtnl_unlock();
4009                 rtnl = net->rtnl;
4010                 {
4011                         struct netlink_dump_control c = {
4012                                 .dump           = dumpit,
4013                                 .min_dump_alloc = min_dump_alloc,
4014                         };
4015                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4016                 }
4017                 rtnl_lock();
4018                 return err;
4019         }
4020
4021         doit = rtnl_get_doit(family, type);
4022         if (doit == NULL)
4023                 return -EOPNOTSUPP;
4024
4025         return doit(skb, nlh);
4026 }
4027
4028 static void rtnetlink_rcv(struct sk_buff *skb)
4029 {
4030         rtnl_lock();
4031         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4032         rtnl_unlock();
4033 }
4034
4035 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4036 {
4037         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4038
4039         switch (event) {
4040         case NETDEV_UP:
4041         case NETDEV_DOWN:
4042         case NETDEV_PRE_UP:
4043         case NETDEV_POST_INIT:
4044         case NETDEV_REGISTER:
4045         case NETDEV_CHANGE:
4046         case NETDEV_PRE_TYPE_CHANGE:
4047         case NETDEV_GOING_DOWN:
4048         case NETDEV_UNREGISTER:
4049         case NETDEV_UNREGISTER_FINAL:
4050         case NETDEV_RELEASE:
4051         case NETDEV_JOIN:
4052         case NETDEV_BONDING_INFO:
4053                 break;
4054         default:
4055                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4056                 break;
4057         }
4058         return NOTIFY_DONE;
4059 }
4060
4061 static struct notifier_block rtnetlink_dev_notifier = {
4062         .notifier_call  = rtnetlink_event,
4063 };
4064
4065
4066 static int __net_init rtnetlink_net_init(struct net *net)
4067 {
4068         struct sock *sk;
4069         struct netlink_kernel_cfg cfg = {
4070                 .groups         = RTNLGRP_MAX,
4071                 .input          = rtnetlink_rcv,
4072                 .cb_mutex       = &rtnl_mutex,
4073                 .flags          = NL_CFG_F_NONROOT_RECV,
4074         };
4075
4076         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4077         if (!sk)
4078                 return -ENOMEM;
4079         net->rtnl = sk;
4080         return 0;
4081 }
4082
4083 static void __net_exit rtnetlink_net_exit(struct net *net)
4084 {
4085         netlink_kernel_release(net->rtnl);
4086         net->rtnl = NULL;
4087 }
4088
4089 static struct pernet_operations rtnetlink_net_ops = {
4090         .init = rtnetlink_net_init,
4091         .exit = rtnetlink_net_exit,
4092 };
4093
4094 void __init rtnetlink_init(void)
4095 {
4096         if (register_pernet_subsys(&rtnetlink_net_ops))
4097                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4098
4099         register_netdevice_notifier(&rtnetlink_dev_notifier);
4100
4101         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4102                       rtnl_dump_ifinfo, rtnl_calcit);
4103         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4104         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4105         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4106
4107         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4108         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4109
4110         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4111         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4112         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4113
4114         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4115         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4116         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4117
4118         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4119                       NULL);
4120 }