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                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1757                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1758                 if (err < 0)
1759                         return err;
1760         }
1761
1762         if (tb[IFLA_VF_TX_RATE]) {
1763                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1764                 struct ifla_vf_info ivf;
1765
1766                 err = -EOPNOTSUPP;
1767                 if (ops->ndo_get_vf_config)
1768                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1769                 if (err < 0)
1770                         return err;
1771
1772                 err = -EOPNOTSUPP;
1773                 if (ops->ndo_set_vf_rate)
1774                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1775                                                    ivf.min_tx_rate,
1776                                                    ivt->rate);
1777                 if (err < 0)
1778                         return err;
1779         }
1780
1781         if (tb[IFLA_VF_RATE]) {
1782                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1783
1784                 err = -EOPNOTSUPP;
1785                 if (ops->ndo_set_vf_rate)
1786                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1787                                                    ivt->min_tx_rate,
1788                                                    ivt->max_tx_rate);
1789                 if (err < 0)
1790                         return err;
1791         }
1792
1793         if (tb[IFLA_VF_SPOOFCHK]) {
1794                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1795
1796                 err = -EOPNOTSUPP;
1797                 if (ops->ndo_set_vf_spoofchk)
1798                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1799                                                        ivs->setting);
1800                 if (err < 0)
1801                         return err;
1802         }
1803
1804         if (tb[IFLA_VF_LINK_STATE]) {
1805                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1806
1807                 err = -EOPNOTSUPP;
1808                 if (ops->ndo_set_vf_link_state)
1809                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1810                                                          ivl->link_state);
1811                 if (err < 0)
1812                         return err;
1813         }
1814
1815         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1816                 struct ifla_vf_rss_query_en *ivrssq_en;
1817
1818                 err = -EOPNOTSUPP;
1819                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1820                 if (ops->ndo_set_vf_rss_query_en)
1821                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1822                                                            ivrssq_en->setting);
1823                 if (err < 0)
1824                         return err;
1825         }
1826
1827         if (tb[IFLA_VF_TRUST]) {
1828                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1829
1830                 err = -EOPNOTSUPP;
1831                 if (ops->ndo_set_vf_trust)
1832                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1833                 if (err < 0)
1834                         return err;
1835         }
1836
1837         if (tb[IFLA_VF_IB_NODE_GUID]) {
1838                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1839
1840                 if (!ops->ndo_set_vf_guid)
1841                         return -EOPNOTSUPP;
1842
1843                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1844         }
1845
1846         if (tb[IFLA_VF_IB_PORT_GUID]) {
1847                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1848
1849                 if (!ops->ndo_set_vf_guid)
1850                         return -EOPNOTSUPP;
1851
1852                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1853         }
1854
1855         return err;
1856 }
1857
1858 static int do_set_master(struct net_device *dev, int ifindex)
1859 {
1860         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1861         const struct net_device_ops *ops;
1862         int err;
1863
1864         if (upper_dev) {
1865                 if (upper_dev->ifindex == ifindex)
1866                         return 0;
1867                 ops = upper_dev->netdev_ops;
1868                 if (ops->ndo_del_slave) {
1869                         err = ops->ndo_del_slave(upper_dev, dev);
1870                         if (err)
1871                                 return err;
1872                 } else {
1873                         return -EOPNOTSUPP;
1874                 }
1875         }
1876
1877         if (ifindex) {
1878                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1879                 if (!upper_dev)
1880                         return -EINVAL;
1881                 ops = upper_dev->netdev_ops;
1882                 if (ops->ndo_add_slave) {
1883                         err = ops->ndo_add_slave(upper_dev, dev);
1884                         if (err)
1885                                 return err;
1886                 } else {
1887                         return -EOPNOTSUPP;
1888                 }
1889         }
1890         return 0;
1891 }
1892
1893 #define DO_SETLINK_MODIFIED     0x01
1894 /* notify flag means notify + modified. */
1895 #define DO_SETLINK_NOTIFY       0x03
1896 static int do_setlink(const struct sk_buff *skb,
1897                       struct net_device *dev, struct ifinfomsg *ifm,
1898                       struct nlattr **tb, char *ifname, int status)
1899 {
1900         const struct net_device_ops *ops = dev->netdev_ops;
1901         int err;
1902
1903         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1904                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1905                 if (IS_ERR(net)) {
1906                         err = PTR_ERR(net);
1907                         goto errout;
1908                 }
1909                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1910                         put_net(net);
1911                         err = -EPERM;
1912                         goto errout;
1913                 }
1914                 err = dev_change_net_namespace(dev, net, ifname);
1915                 put_net(net);
1916                 if (err)
1917                         goto errout;
1918                 status |= DO_SETLINK_MODIFIED;
1919         }
1920
1921         if (tb[IFLA_MAP]) {
1922                 struct rtnl_link_ifmap *u_map;
1923                 struct ifmap k_map;
1924
1925                 if (!ops->ndo_set_config) {
1926                         err = -EOPNOTSUPP;
1927                         goto errout;
1928                 }
1929
1930                 if (!netif_device_present(dev)) {
1931                         err = -ENODEV;
1932                         goto errout;
1933                 }
1934
1935                 u_map = nla_data(tb[IFLA_MAP]);
1936                 k_map.mem_start = (unsigned long) u_map->mem_start;
1937                 k_map.mem_end = (unsigned long) u_map->mem_end;
1938                 k_map.base_addr = (unsigned short) u_map->base_addr;
1939                 k_map.irq = (unsigned char) u_map->irq;
1940                 k_map.dma = (unsigned char) u_map->dma;
1941                 k_map.port = (unsigned char) u_map->port;
1942
1943                 err = ops->ndo_set_config(dev, &k_map);
1944                 if (err < 0)
1945                         goto errout;
1946
1947                 status |= DO_SETLINK_NOTIFY;
1948         }
1949
1950         if (tb[IFLA_ADDRESS]) {
1951                 struct sockaddr *sa;
1952                 int len;
1953
1954                 len = sizeof(sa_family_t) + dev->addr_len;
1955                 sa = kmalloc(len, GFP_KERNEL);
1956                 if (!sa) {
1957                         err = -ENOMEM;
1958                         goto errout;
1959                 }
1960                 sa->sa_family = dev->type;
1961                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1962                        dev->addr_len);
1963                 err = dev_set_mac_address(dev, sa);
1964                 kfree(sa);
1965                 if (err)
1966                         goto errout;
1967                 status |= DO_SETLINK_MODIFIED;
1968         }
1969
1970         if (tb[IFLA_MTU]) {
1971                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1972                 if (err < 0)
1973                         goto errout;
1974                 status |= DO_SETLINK_MODIFIED;
1975         }
1976
1977         if (tb[IFLA_GROUP]) {
1978                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1979                 status |= DO_SETLINK_NOTIFY;
1980         }
1981
1982         /*
1983          * Interface selected by interface index but interface
1984          * name provided implies that a name change has been
1985          * requested.
1986          */
1987         if (ifm->ifi_index > 0 && ifname[0]) {
1988                 err = dev_change_name(dev, ifname);
1989                 if (err < 0)
1990                         goto errout;
1991                 status |= DO_SETLINK_MODIFIED;
1992         }
1993
1994         if (tb[IFLA_IFALIAS]) {
1995                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1996                                     nla_len(tb[IFLA_IFALIAS]));
1997                 if (err < 0)
1998                         goto errout;
1999                 status |= DO_SETLINK_NOTIFY;
2000         }
2001
2002         if (tb[IFLA_BROADCAST]) {
2003                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2004                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2005         }
2006
2007         if (ifm->ifi_flags || ifm->ifi_change) {
2008                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2009                 if (err < 0)
2010                         goto errout;
2011         }
2012
2013         if (tb[IFLA_MASTER]) {
2014                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2015                 if (err)
2016                         goto errout;
2017                 status |= DO_SETLINK_MODIFIED;
2018         }
2019
2020         if (tb[IFLA_CARRIER]) {
2021                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2022                 if (err)
2023                         goto errout;
2024                 status |= DO_SETLINK_MODIFIED;
2025         }
2026
2027         if (tb[IFLA_TXQLEN]) {
2028                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2029                 unsigned long orig_len = dev->tx_queue_len;
2030
2031                 if (dev->tx_queue_len ^ value) {
2032                         dev->tx_queue_len = value;
2033                         err = call_netdevice_notifiers(
2034                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2035                         err = notifier_to_errno(err);
2036                         if (err) {
2037                                 dev->tx_queue_len = orig_len;
2038                                 goto errout;
2039                         }
2040                         status |= DO_SETLINK_NOTIFY;
2041                 }
2042         }
2043
2044         if (tb[IFLA_OPERSTATE])
2045                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2046
2047         if (tb[IFLA_LINKMODE]) {
2048                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2049
2050                 write_lock_bh(&dev_base_lock);
2051                 if (dev->link_mode ^ value)
2052                         status |= DO_SETLINK_NOTIFY;
2053                 dev->link_mode = value;
2054                 write_unlock_bh(&dev_base_lock);
2055         }
2056
2057         if (tb[IFLA_VFINFO_LIST]) {
2058                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2059                 struct nlattr *attr;
2060                 int rem;
2061
2062                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2063                         if (nla_type(attr) != IFLA_VF_INFO ||
2064                             nla_len(attr) < NLA_HDRLEN) {
2065                                 err = -EINVAL;
2066                                 goto errout;
2067                         }
2068                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2069                                                ifla_vf_policy);
2070                         if (err < 0)
2071                                 goto errout;
2072                         err = do_setvfinfo(dev, vfinfo);
2073                         if (err < 0)
2074                                 goto errout;
2075                         status |= DO_SETLINK_NOTIFY;
2076                 }
2077         }
2078         err = 0;
2079
2080         if (tb[IFLA_VF_PORTS]) {
2081                 struct nlattr *port[IFLA_PORT_MAX+1];
2082                 struct nlattr *attr;
2083                 int vf;
2084                 int rem;
2085
2086                 err = -EOPNOTSUPP;
2087                 if (!ops->ndo_set_vf_port)
2088                         goto errout;
2089
2090                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2091                         if (nla_type(attr) != IFLA_VF_PORT ||
2092                             nla_len(attr) < NLA_HDRLEN) {
2093                                 err = -EINVAL;
2094                                 goto errout;
2095                         }
2096                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2097                                                ifla_port_policy);
2098                         if (err < 0)
2099                                 goto errout;
2100                         if (!port[IFLA_PORT_VF]) {
2101                                 err = -EOPNOTSUPP;
2102                                 goto errout;
2103                         }
2104                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2105                         err = ops->ndo_set_vf_port(dev, vf, port);
2106                         if (err < 0)
2107                                 goto errout;
2108                         status |= DO_SETLINK_NOTIFY;
2109                 }
2110         }
2111         err = 0;
2112
2113         if (tb[IFLA_PORT_SELF]) {
2114                 struct nlattr *port[IFLA_PORT_MAX+1];
2115
2116                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2117                         tb[IFLA_PORT_SELF], ifla_port_policy);
2118                 if (err < 0)
2119                         goto errout;
2120
2121                 err = -EOPNOTSUPP;
2122                 if (ops->ndo_set_vf_port)
2123                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2124                 if (err < 0)
2125                         goto errout;
2126                 status |= DO_SETLINK_NOTIFY;
2127         }
2128
2129         if (tb[IFLA_AF_SPEC]) {
2130                 struct nlattr *af;
2131                 int rem;
2132
2133                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2134                         const struct rtnl_af_ops *af_ops;
2135
2136                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2137                                 BUG();
2138
2139                         err = af_ops->set_link_af(dev, af);
2140                         if (err < 0)
2141                                 goto errout;
2142
2143                         status |= DO_SETLINK_NOTIFY;
2144                 }
2145         }
2146         err = 0;
2147
2148         if (tb[IFLA_PROTO_DOWN]) {
2149                 err = dev_change_proto_down(dev,
2150                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2151                 if (err)
2152                         goto errout;
2153                 status |= DO_SETLINK_NOTIFY;
2154         }
2155
2156         if (tb[IFLA_XDP]) {
2157                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2158
2159                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2160                                        ifla_xdp_policy);
2161                 if (err < 0)
2162                         goto errout;
2163
2164                 if (xdp[IFLA_XDP_ATTACHED]) {
2165                         err = -EINVAL;
2166                         goto errout;
2167                 }
2168                 if (xdp[IFLA_XDP_FD]) {
2169                         err = dev_change_xdp_fd(dev,
2170                                                 nla_get_s32(xdp[IFLA_XDP_FD]));
2171                         if (err)
2172                                 goto errout;
2173                         status |= DO_SETLINK_NOTIFY;
2174                 }
2175         }
2176
2177 errout:
2178         if (status & DO_SETLINK_MODIFIED) {
2179                 if (status & DO_SETLINK_NOTIFY)
2180                         netdev_state_change(dev);
2181
2182                 if (err < 0)
2183                         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",
2184                                              dev->name);
2185         }
2186
2187         return err;
2188 }
2189
2190 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2191 {
2192         struct net *net = sock_net(skb->sk);
2193         struct ifinfomsg *ifm;
2194         struct net_device *dev;
2195         int err;
2196         struct nlattr *tb[IFLA_MAX+1];
2197         char ifname[IFNAMSIZ];
2198
2199         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2200         if (err < 0)
2201                 goto errout;
2202
2203         if (tb[IFLA_IFNAME])
2204                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2205         else
2206                 ifname[0] = '\0';
2207
2208         err = -EINVAL;
2209         ifm = nlmsg_data(nlh);
2210         if (ifm->ifi_index > 0)
2211                 dev = __dev_get_by_index(net, ifm->ifi_index);
2212         else if (tb[IFLA_IFNAME])
2213                 dev = __dev_get_by_name(net, ifname);
2214         else
2215                 goto errout;
2216
2217         if (dev == NULL) {
2218                 err = -ENODEV;
2219                 goto errout;
2220         }
2221
2222         err = validate_linkmsg(dev, tb);
2223         if (err < 0)
2224                 goto errout;
2225
2226         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2227 errout:
2228         return err;
2229 }
2230
2231 static int rtnl_group_dellink(const struct net *net, int group)
2232 {
2233         struct net_device *dev, *aux;
2234         LIST_HEAD(list_kill);
2235         bool found = false;
2236
2237         if (!group)
2238                 return -EPERM;
2239
2240         for_each_netdev(net, dev) {
2241                 if (dev->group == group) {
2242                         const struct rtnl_link_ops *ops;
2243
2244                         found = true;
2245                         ops = dev->rtnl_link_ops;
2246                         if (!ops || !ops->dellink)
2247                                 return -EOPNOTSUPP;
2248                 }
2249         }
2250
2251         if (!found)
2252                 return -ENODEV;
2253
2254         for_each_netdev_safe(net, dev, aux) {
2255                 if (dev->group == group) {
2256                         const struct rtnl_link_ops *ops;
2257
2258                         ops = dev->rtnl_link_ops;
2259                         ops->dellink(dev, &list_kill);
2260                 }
2261         }
2262         unregister_netdevice_many(&list_kill);
2263
2264         return 0;
2265 }
2266
2267 int rtnl_delete_link(struct net_device *dev)
2268 {
2269         const struct rtnl_link_ops *ops;
2270         LIST_HEAD(list_kill);
2271
2272         ops = dev->rtnl_link_ops;
2273         if (!ops || !ops->dellink)
2274                 return -EOPNOTSUPP;
2275
2276         ops->dellink(dev, &list_kill);
2277         unregister_netdevice_many(&list_kill);
2278
2279         return 0;
2280 }
2281 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2282
2283 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2284 {
2285         struct net *net = sock_net(skb->sk);
2286         struct net_device *dev;
2287         struct ifinfomsg *ifm;
2288         char ifname[IFNAMSIZ];
2289         struct nlattr *tb[IFLA_MAX+1];
2290         int err;
2291
2292         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2293         if (err < 0)
2294                 return err;
2295
2296         if (tb[IFLA_IFNAME])
2297                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2298
2299         ifm = nlmsg_data(nlh);
2300         if (ifm->ifi_index > 0)
2301                 dev = __dev_get_by_index(net, ifm->ifi_index);
2302         else if (tb[IFLA_IFNAME])
2303                 dev = __dev_get_by_name(net, ifname);
2304         else if (tb[IFLA_GROUP])
2305                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2306         else
2307                 return -EINVAL;
2308
2309         if (!dev)
2310                 return -ENODEV;
2311
2312         return rtnl_delete_link(dev);
2313 }
2314
2315 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2316 {
2317         unsigned int old_flags;
2318         int err;
2319
2320         old_flags = dev->flags;
2321         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2322                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2323                 if (err < 0)
2324                         return err;
2325         }
2326
2327         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2328
2329         __dev_notify_flags(dev, old_flags, ~0U);
2330         return 0;
2331 }
2332 EXPORT_SYMBOL(rtnl_configure_link);
2333
2334 struct net_device *rtnl_create_link(struct net *net,
2335         const char *ifname, unsigned char name_assign_type,
2336         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2337 {
2338         int err;
2339         struct net_device *dev;
2340         unsigned int num_tx_queues = 1;
2341         unsigned int num_rx_queues = 1;
2342
2343         if (tb[IFLA_NUM_TX_QUEUES])
2344                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2345         else if (ops->get_num_tx_queues)
2346                 num_tx_queues = ops->get_num_tx_queues();
2347
2348         if (tb[IFLA_NUM_RX_QUEUES])
2349                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2350         else if (ops->get_num_rx_queues)
2351                 num_rx_queues = ops->get_num_rx_queues();
2352
2353         err = -ENOMEM;
2354         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2355                                ops->setup, num_tx_queues, num_rx_queues);
2356         if (!dev)
2357                 goto err;
2358
2359         dev_net_set(dev, net);
2360         dev->rtnl_link_ops = ops;
2361         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2362
2363         if (tb[IFLA_MTU])
2364                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2365         if (tb[IFLA_ADDRESS]) {
2366                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2367                                 nla_len(tb[IFLA_ADDRESS]));
2368                 dev->addr_assign_type = NET_ADDR_SET;
2369         }
2370         if (tb[IFLA_BROADCAST])
2371                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2372                                 nla_len(tb[IFLA_BROADCAST]));
2373         if (tb[IFLA_TXQLEN])
2374                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2375         if (tb[IFLA_OPERSTATE])
2376                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2377         if (tb[IFLA_LINKMODE])
2378                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2379         if (tb[IFLA_GROUP])
2380                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2381
2382         return dev;
2383
2384 err:
2385         return ERR_PTR(err);
2386 }
2387 EXPORT_SYMBOL(rtnl_create_link);
2388
2389 static int rtnl_group_changelink(const struct sk_buff *skb,
2390                 struct net *net, int group,
2391                 struct ifinfomsg *ifm,
2392                 struct nlattr **tb)
2393 {
2394         struct net_device *dev, *aux;
2395         int err;
2396
2397         for_each_netdev_safe(net, dev, aux) {
2398                 if (dev->group == group) {
2399                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2400                         if (err < 0)
2401                                 return err;
2402                 }
2403         }
2404
2405         return 0;
2406 }
2407
2408 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2409 {
2410         struct net *net = sock_net(skb->sk);
2411         const struct rtnl_link_ops *ops;
2412         const struct rtnl_link_ops *m_ops = NULL;
2413         struct net_device *dev;
2414         struct net_device *master_dev = NULL;
2415         struct ifinfomsg *ifm;
2416         char kind[MODULE_NAME_LEN];
2417         char ifname[IFNAMSIZ];
2418         struct nlattr *tb[IFLA_MAX+1];
2419         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2420         unsigned char name_assign_type = NET_NAME_USER;
2421         int err;
2422
2423 #ifdef CONFIG_MODULES
2424 replay:
2425 #endif
2426         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2427         if (err < 0)
2428                 return err;
2429
2430         if (tb[IFLA_IFNAME])
2431                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2432         else
2433                 ifname[0] = '\0';
2434
2435         ifm = nlmsg_data(nlh);
2436         if (ifm->ifi_index > 0)
2437                 dev = __dev_get_by_index(net, ifm->ifi_index);
2438         else {
2439                 if (ifname[0])
2440                         dev = __dev_get_by_name(net, ifname);
2441                 else
2442                         dev = NULL;
2443         }
2444
2445         if (dev) {
2446                 master_dev = netdev_master_upper_dev_get(dev);
2447                 if (master_dev)
2448                         m_ops = master_dev->rtnl_link_ops;
2449         }
2450
2451         err = validate_linkmsg(dev, tb);
2452         if (err < 0)
2453                 return err;
2454
2455         if (tb[IFLA_LINKINFO]) {
2456                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2457                                        tb[IFLA_LINKINFO], ifla_info_policy);
2458                 if (err < 0)
2459                         return err;
2460         } else
2461                 memset(linkinfo, 0, sizeof(linkinfo));
2462
2463         if (linkinfo[IFLA_INFO_KIND]) {
2464                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2465                 ops = rtnl_link_ops_get(kind);
2466         } else {
2467                 kind[0] = '\0';
2468                 ops = NULL;
2469         }
2470
2471         if (1) {
2472                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2473                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2474                 struct nlattr **data = NULL;
2475                 struct nlattr **slave_data = NULL;
2476                 struct net *dest_net, *link_net = NULL;
2477
2478                 if (ops) {
2479                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2480                                 err = nla_parse_nested(attr, ops->maxtype,
2481                                                        linkinfo[IFLA_INFO_DATA],
2482                                                        ops->policy);
2483                                 if (err < 0)
2484                                         return err;
2485                                 data = attr;
2486                         }
2487                         if (ops->validate) {
2488                                 err = ops->validate(tb, data);
2489                                 if (err < 0)
2490                                         return err;
2491                         }
2492                 }
2493
2494                 if (m_ops) {
2495                         if (m_ops->slave_maxtype &&
2496                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2497                                 err = nla_parse_nested(slave_attr,
2498                                                        m_ops->slave_maxtype,
2499                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2500                                                        m_ops->slave_policy);
2501                                 if (err < 0)
2502                                         return err;
2503                                 slave_data = slave_attr;
2504                         }
2505                         if (m_ops->slave_validate) {
2506                                 err = m_ops->slave_validate(tb, slave_data);
2507                                 if (err < 0)
2508                                         return err;
2509                         }
2510                 }
2511
2512                 if (dev) {
2513                         int status = 0;
2514
2515                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2516                                 return -EEXIST;
2517                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2518                                 return -EOPNOTSUPP;
2519
2520                         if (linkinfo[IFLA_INFO_DATA]) {
2521                                 if (!ops || ops != dev->rtnl_link_ops ||
2522                                     !ops->changelink)
2523                                         return -EOPNOTSUPP;
2524
2525                                 err = ops->changelink(dev, tb, data);
2526                                 if (err < 0)
2527                                         return err;
2528                                 status |= DO_SETLINK_NOTIFY;
2529                         }
2530
2531                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2532                                 if (!m_ops || !m_ops->slave_changelink)
2533                                         return -EOPNOTSUPP;
2534
2535                                 err = m_ops->slave_changelink(master_dev, dev,
2536                                                               tb, slave_data);
2537                                 if (err < 0)
2538                                         return err;
2539                                 status |= DO_SETLINK_NOTIFY;
2540                         }
2541
2542                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2543                 }
2544
2545                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2546                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2547                                 return rtnl_group_changelink(skb, net,
2548                                                 nla_get_u32(tb[IFLA_GROUP]),
2549                                                 ifm, tb);
2550                         return -ENODEV;
2551                 }
2552
2553                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2554                         return -EOPNOTSUPP;
2555
2556                 if (!ops) {
2557 #ifdef CONFIG_MODULES
2558                         if (kind[0]) {
2559                                 __rtnl_unlock();
2560                                 request_module("rtnl-link-%s", kind);
2561                                 rtnl_lock();
2562                                 ops = rtnl_link_ops_get(kind);
2563                                 if (ops)
2564                                         goto replay;
2565                         }
2566 #endif
2567                         return -EOPNOTSUPP;
2568                 }
2569
2570                 if (!ops->setup)
2571                         return -EOPNOTSUPP;
2572
2573                 if (!ifname[0]) {
2574                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2575                         name_assign_type = NET_NAME_ENUM;
2576                 }
2577
2578                 dest_net = rtnl_link_get_net(net, tb);
2579                 if (IS_ERR(dest_net))
2580                         return PTR_ERR(dest_net);
2581
2582                 err = -EPERM;
2583                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2584                         goto out;
2585
2586                 if (tb[IFLA_LINK_NETNSID]) {
2587                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2588
2589                         link_net = get_net_ns_by_id(dest_net, id);
2590                         if (!link_net) {
2591                                 err =  -EINVAL;
2592                                 goto out;
2593                         }
2594                         err = -EPERM;
2595                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2596                                 goto out;
2597                 }
2598
2599                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2600                                        name_assign_type, ops, tb);
2601                 if (IS_ERR(dev)) {
2602                         err = PTR_ERR(dev);
2603                         goto out;
2604                 }
2605
2606                 dev->ifindex = ifm->ifi_index;
2607
2608                 if (ops->newlink) {
2609                         err = ops->newlink(link_net ? : net, dev, tb, data);
2610                         /* Drivers should call free_netdev() in ->destructor
2611                          * and unregister it on failure after registration
2612                          * so that device could be finally freed in rtnl_unlock.
2613                          */
2614                         if (err < 0) {
2615                                 /* If device is not registered at all, free it now */
2616                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2617                                         free_netdev(dev);
2618                                 goto out;
2619                         }
2620                 } else {
2621                         err = register_netdevice(dev);
2622                         if (err < 0) {
2623                                 free_netdev(dev);
2624                                 goto out;
2625                         }
2626                 }
2627                 err = rtnl_configure_link(dev, ifm);
2628                 if (err < 0)
2629                         goto out_unregister;
2630                 if (link_net) {
2631                         err = dev_change_net_namespace(dev, dest_net, ifname);
2632                         if (err < 0)
2633                                 goto out_unregister;
2634                 }
2635 out:
2636                 if (link_net)
2637                         put_net(link_net);
2638                 put_net(dest_net);
2639                 return err;
2640 out_unregister:
2641                 if (ops->newlink) {
2642                         LIST_HEAD(list_kill);
2643
2644                         ops->dellink(dev, &list_kill);
2645                         unregister_netdevice_many(&list_kill);
2646                 } else {
2647                         unregister_netdevice(dev);
2648                 }
2649                 goto out;
2650         }
2651 }
2652
2653 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2654 {
2655         struct net *net = sock_net(skb->sk);
2656         struct ifinfomsg *ifm;
2657         char ifname[IFNAMSIZ];
2658         struct nlattr *tb[IFLA_MAX+1];
2659         struct net_device *dev = NULL;
2660         struct sk_buff *nskb;
2661         int err;
2662         u32 ext_filter_mask = 0;
2663
2664         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2665         if (err < 0)
2666                 return err;
2667
2668         if (tb[IFLA_IFNAME])
2669                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2670
2671         if (tb[IFLA_EXT_MASK])
2672                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2673
2674         ifm = nlmsg_data(nlh);
2675         if (ifm->ifi_index > 0)
2676                 dev = __dev_get_by_index(net, ifm->ifi_index);
2677         else if (tb[IFLA_IFNAME])
2678                 dev = __dev_get_by_name(net, ifname);
2679         else
2680                 return -EINVAL;
2681
2682         if (dev == NULL)
2683                 return -ENODEV;
2684
2685         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2686         if (nskb == NULL)
2687                 return -ENOBUFS;
2688
2689         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2690                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2691         if (err < 0) {
2692                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2693                 WARN_ON(err == -EMSGSIZE);
2694                 kfree_skb(nskb);
2695         } else
2696                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2697
2698         return err;
2699 }
2700
2701 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2702 {
2703         struct net *net = sock_net(skb->sk);
2704         struct net_device *dev;
2705         struct nlattr *tb[IFLA_MAX+1];
2706         u32 ext_filter_mask = 0;
2707         u16 min_ifinfo_dump_size = 0;
2708         int hdrlen;
2709
2710         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2711         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2712                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2713
2714         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2715                 if (tb[IFLA_EXT_MASK])
2716                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2717         }
2718
2719         if (!ext_filter_mask)
2720                 return NLMSG_GOODSIZE;
2721         /*
2722          * traverse the list of net devices and compute the minimum
2723          * buffer size based upon the filter mask.
2724          */
2725         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2726                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2727                                              if_nlmsg_size(dev,
2728                                                            ext_filter_mask));
2729         }
2730
2731         return min_ifinfo_dump_size;
2732 }
2733
2734 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2735 {
2736         int idx;
2737         int s_idx = cb->family;
2738
2739         if (s_idx == 0)
2740                 s_idx = 1;
2741         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2742                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2743                 if (idx < s_idx || idx == PF_PACKET)
2744                         continue;
2745                 if (rtnl_msg_handlers[idx] == NULL ||
2746                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2747                         continue;
2748                 if (idx > s_idx) {
2749                         memset(&cb->args[0], 0, sizeof(cb->args));
2750                         cb->prev_seq = 0;
2751                         cb->seq = 0;
2752                 }
2753                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2754                         break;
2755         }
2756         cb->family = idx;
2757
2758         return skb->len;
2759 }
2760
2761 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2762                                        unsigned int change, gfp_t flags)
2763 {
2764         struct net *net = dev_net(dev);
2765         struct sk_buff *skb;
2766         int err = -ENOBUFS;
2767         size_t if_info_size;
2768
2769         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2770         if (skb == NULL)
2771                 goto errout;
2772
2773         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2774         if (err < 0) {
2775                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2776                 WARN_ON(err == -EMSGSIZE);
2777                 kfree_skb(skb);
2778                 goto errout;
2779         }
2780         return skb;
2781 errout:
2782         if (err < 0)
2783                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2784         return NULL;
2785 }
2786
2787 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2788 {
2789         struct net *net = dev_net(dev);
2790
2791         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2792 }
2793
2794 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2795                   gfp_t flags)
2796 {
2797         struct sk_buff *skb;
2798
2799         if (dev->reg_state != NETREG_REGISTERED)
2800                 return;
2801
2802         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2803         if (skb)
2804                 rtmsg_ifinfo_send(skb, dev, flags);
2805 }
2806 EXPORT_SYMBOL(rtmsg_ifinfo);
2807
2808 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2809                                    struct net_device *dev,
2810                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2811                                    int type, unsigned int flags,
2812                                    int nlflags, u16 ndm_state)
2813 {
2814         struct nlmsghdr *nlh;
2815         struct ndmsg *ndm;
2816
2817         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2818         if (!nlh)
2819                 return -EMSGSIZE;
2820
2821         ndm = nlmsg_data(nlh);
2822         ndm->ndm_family  = AF_BRIDGE;
2823         ndm->ndm_pad1    = 0;
2824         ndm->ndm_pad2    = 0;
2825         ndm->ndm_flags   = flags;
2826         ndm->ndm_type    = 0;
2827         ndm->ndm_ifindex = dev->ifindex;
2828         ndm->ndm_state   = ndm_state;
2829
2830         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2831                 goto nla_put_failure;
2832         if (vid)
2833                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2834                         goto nla_put_failure;
2835
2836         nlmsg_end(skb, nlh);
2837         return 0;
2838
2839 nla_put_failure:
2840         nlmsg_cancel(skb, nlh);
2841         return -EMSGSIZE;
2842 }
2843
2844 static inline size_t rtnl_fdb_nlmsg_size(void)
2845 {
2846         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2847 }
2848
2849 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2850                             u16 ndm_state)
2851 {
2852         struct net *net = dev_net(dev);
2853         struct sk_buff *skb;
2854         int err = -ENOBUFS;
2855
2856         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2857         if (!skb)
2858                 goto errout;
2859
2860         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2861                                       0, 0, type, NTF_SELF, 0, ndm_state);
2862         if (err < 0) {
2863                 kfree_skb(skb);
2864                 goto errout;
2865         }
2866
2867         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2868         return;
2869 errout:
2870         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2871 }
2872
2873 /**
2874  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2875  */
2876 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2877                      struct nlattr *tb[],
2878                      struct net_device *dev,
2879                      const unsigned char *addr, u16 vid,
2880                      u16 flags)
2881 {
2882         int err = -EINVAL;
2883
2884         /* If aging addresses are supported device will need to
2885          * implement its own handler for this.
2886          */
2887         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2888                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2889                 return err;
2890         }
2891
2892         if (vid) {
2893                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2894                 return err;
2895         }
2896
2897         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2898                 err = dev_uc_add_excl(dev, addr);
2899         else if (is_multicast_ether_addr(addr))
2900                 err = dev_mc_add_excl(dev, addr);
2901
2902         /* Only return duplicate errors if NLM_F_EXCL is set */
2903         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2904                 err = 0;
2905
2906         return err;
2907 }
2908 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2909
2910 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2911 {
2912         u16 vid = 0;
2913
2914         if (vlan_attr) {
2915                 if (nla_len(vlan_attr) != sizeof(u16)) {
2916                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2917                         return -EINVAL;
2918                 }
2919
2920                 vid = nla_get_u16(vlan_attr);
2921
2922                 if (!vid || vid >= VLAN_VID_MASK) {
2923                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2924                                 vid);
2925                         return -EINVAL;
2926                 }
2927         }
2928         *p_vid = vid;
2929         return 0;
2930 }
2931
2932 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2933 {
2934         struct net *net = sock_net(skb->sk);
2935         struct ndmsg *ndm;
2936         struct nlattr *tb[NDA_MAX+1];
2937         struct net_device *dev;
2938         u8 *addr;
2939         u16 vid;
2940         int err;
2941
2942         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2943         if (err < 0)
2944                 return err;
2945
2946         ndm = nlmsg_data(nlh);
2947         if (ndm->ndm_ifindex == 0) {
2948                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2949                 return -EINVAL;
2950         }
2951
2952         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2953         if (dev == NULL) {
2954                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2955                 return -ENODEV;
2956         }
2957
2958         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2959                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2960                 return -EINVAL;
2961         }
2962
2963         addr = nla_data(tb[NDA_LLADDR]);
2964
2965         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2966         if (err)
2967                 return err;
2968
2969         err = -EOPNOTSUPP;
2970
2971         /* Support fdb on master device the net/bridge default case */
2972         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2973             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2974                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2975                 const struct net_device_ops *ops = br_dev->netdev_ops;
2976
2977                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2978                                        nlh->nlmsg_flags);
2979                 if (err)
2980                         goto out;
2981                 else
2982                         ndm->ndm_flags &= ~NTF_MASTER;
2983         }
2984
2985         /* Embedded bridge, macvlan, and any other device support */
2986         if ((ndm->ndm_flags & NTF_SELF)) {
2987                 if (dev->netdev_ops->ndo_fdb_add)
2988                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2989                                                            vid,
2990                                                            nlh->nlmsg_flags);
2991                 else
2992                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2993                                                nlh->nlmsg_flags);
2994
2995                 if (!err) {
2996                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2997                                         ndm->ndm_state);
2998                         ndm->ndm_flags &= ~NTF_SELF;
2999                 }
3000         }
3001 out:
3002         return err;
3003 }
3004
3005 /**
3006  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3007  */
3008 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3009                      struct nlattr *tb[],
3010                      struct net_device *dev,
3011                      const unsigned char *addr, u16 vid)
3012 {
3013         int err = -EINVAL;
3014
3015         /* If aging addresses are supported device will need to
3016          * implement its own handler for this.
3017          */
3018         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3019                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3020                 return err;
3021         }
3022
3023         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3024                 err = dev_uc_del(dev, addr);
3025         else if (is_multicast_ether_addr(addr))
3026                 err = dev_mc_del(dev, addr);
3027
3028         return err;
3029 }
3030 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3031
3032 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3033 {
3034         struct net *net = sock_net(skb->sk);
3035         struct ndmsg *ndm;
3036         struct nlattr *tb[NDA_MAX+1];
3037         struct net_device *dev;
3038         int err = -EINVAL;
3039         __u8 *addr;
3040         u16 vid;
3041
3042         if (!netlink_capable(skb, CAP_NET_ADMIN))
3043                 return -EPERM;
3044
3045         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3046         if (err < 0)
3047                 return err;
3048
3049         ndm = nlmsg_data(nlh);
3050         if (ndm->ndm_ifindex == 0) {
3051                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3052                 return -EINVAL;
3053         }
3054
3055         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3056         if (dev == NULL) {
3057                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3058                 return -ENODEV;
3059         }
3060
3061         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3062                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3063                 return -EINVAL;
3064         }
3065
3066         addr = nla_data(tb[NDA_LLADDR]);
3067
3068         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3069         if (err)
3070                 return err;
3071
3072         err = -EOPNOTSUPP;
3073
3074         /* Support fdb on master device the net/bridge default case */
3075         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3076             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3077                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3078                 const struct net_device_ops *ops = br_dev->netdev_ops;
3079
3080                 if (ops->ndo_fdb_del)
3081                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3082
3083                 if (err)
3084                         goto out;
3085                 else
3086                         ndm->ndm_flags &= ~NTF_MASTER;
3087         }
3088
3089         /* Embedded bridge, macvlan, and any other device support */
3090         if (ndm->ndm_flags & NTF_SELF) {
3091                 if (dev->netdev_ops->ndo_fdb_del)
3092                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3093                                                            vid);
3094                 else
3095                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3096
3097                 if (!err) {
3098                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3099                                         ndm->ndm_state);
3100                         ndm->ndm_flags &= ~NTF_SELF;
3101                 }
3102         }
3103 out:
3104         return err;
3105 }
3106
3107 static int nlmsg_populate_fdb(struct sk_buff *skb,
3108                               struct netlink_callback *cb,
3109                               struct net_device *dev,
3110                               int *idx,
3111                               struct netdev_hw_addr_list *list)
3112 {
3113         struct netdev_hw_addr *ha;
3114         int err;
3115         u32 portid, seq;
3116
3117         portid = NETLINK_CB(cb->skb).portid;
3118         seq = cb->nlh->nlmsg_seq;
3119
3120         list_for_each_entry(ha, &list->list, list) {
3121                 if (*idx < cb->args[2])
3122                         goto skip;
3123
3124                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3125                                               portid, seq,
3126                                               RTM_NEWNEIGH, NTF_SELF,
3127                                               NLM_F_MULTI, NUD_PERMANENT);
3128                 if (err < 0)
3129                         return err;
3130 skip:
3131                 *idx += 1;
3132         }
3133         return 0;
3134 }
3135
3136 /**
3137  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3138  * @nlh: netlink message header
3139  * @dev: netdevice
3140  *
3141  * Default netdevice operation to dump the existing unicast address list.
3142  * Returns number of addresses from list put in skb.
3143  */
3144 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3145                       struct netlink_callback *cb,
3146                       struct net_device *dev,
3147                       struct net_device *filter_dev,
3148                       int *idx)
3149 {
3150         int err;
3151
3152         netif_addr_lock_bh(dev);
3153         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3154         if (err)
3155                 goto out;
3156         nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3157 out:
3158         netif_addr_unlock_bh(dev);
3159         return err;
3160 }
3161 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3162
3163 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3164 {
3165         struct net_device *dev;
3166         struct nlattr *tb[IFLA_MAX+1];
3167         struct net_device *br_dev = NULL;
3168         const struct net_device_ops *ops = NULL;
3169         const struct net_device_ops *cops = NULL;
3170         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3171         struct net *net = sock_net(skb->sk);
3172         struct hlist_head *head;
3173         int brport_idx = 0;
3174         int br_idx = 0;
3175         int h, s_h;
3176         int idx = 0, s_idx;
3177         int err = 0;
3178         int fidx = 0;
3179
3180         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3181                         ifla_policy) == 0) {
3182                 if (tb[IFLA_MASTER])
3183                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3184         }
3185
3186         brport_idx = ifm->ifi_index;
3187
3188         if (br_idx) {
3189                 br_dev = __dev_get_by_index(net, br_idx);
3190                 if (!br_dev)
3191                         return -ENODEV;
3192
3193                 ops = br_dev->netdev_ops;
3194         }
3195
3196         s_h = cb->args[0];
3197         s_idx = cb->args[1];
3198
3199         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3200                 idx = 0;
3201                 head = &net->dev_index_head[h];
3202                 hlist_for_each_entry(dev, head, index_hlist) {
3203
3204                         if (brport_idx && (dev->ifindex != brport_idx))
3205                                 continue;
3206
3207                         if (!br_idx) { /* user did not specify a specific bridge */
3208                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3209                                         br_dev = netdev_master_upper_dev_get(dev);
3210                                         cops = br_dev->netdev_ops;
3211                                 }
3212                         } else {
3213                                 if (dev != br_dev &&
3214                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3215                                         continue;
3216
3217                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3218                                     !(dev->priv_flags & IFF_EBRIDGE))
3219                                         continue;
3220                                 cops = ops;
3221                         }
3222
3223                         if (idx < s_idx)
3224                                 goto cont;
3225
3226                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3227                                 if (cops && cops->ndo_fdb_dump) {
3228                                         err = cops->ndo_fdb_dump(skb, cb,
3229                                                                 br_dev, dev,
3230                                                                 &fidx);
3231                                         if (err == -EMSGSIZE)
3232                                                 goto out;
3233                                 }
3234                         }
3235
3236                         if (dev->netdev_ops->ndo_fdb_dump)
3237                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3238                                                                     dev, NULL,
3239                                                                     &fidx);
3240                         else
3241                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3242                                                         &fidx);
3243                         if (err == -EMSGSIZE)
3244                                 goto out;
3245
3246                         cops = NULL;
3247
3248                         /* reset fdb offset to 0 for rest of the interfaces */
3249                         cb->args[2] = 0;
3250                         fidx = 0;
3251 cont:
3252                         idx++;
3253                 }
3254         }
3255
3256 out:
3257         cb->args[0] = h;
3258         cb->args[1] = idx;
3259         cb->args[2] = fidx;
3260
3261         return skb->len;
3262 }
3263
3264 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3265                                unsigned int attrnum, unsigned int flag)
3266 {
3267         if (mask & flag)
3268                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3269         return 0;
3270 }
3271
3272 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3273                             struct net_device *dev, u16 mode,
3274                             u32 flags, u32 mask, int nlflags,
3275                             u32 filter_mask,
3276                             int (*vlan_fill)(struct sk_buff *skb,
3277                                              struct net_device *dev,
3278                                              u32 filter_mask))
3279 {
3280         struct nlmsghdr *nlh;
3281         struct ifinfomsg *ifm;
3282         struct nlattr *br_afspec;
3283         struct nlattr *protinfo;
3284         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3285         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3286         int err = 0;
3287
3288         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3289         if (nlh == NULL)
3290                 return -EMSGSIZE;
3291
3292         ifm = nlmsg_data(nlh);
3293         ifm->ifi_family = AF_BRIDGE;
3294         ifm->__ifi_pad = 0;
3295         ifm->ifi_type = dev->type;
3296         ifm->ifi_index = dev->ifindex;
3297         ifm->ifi_flags = dev_get_flags(dev);
3298         ifm->ifi_change = 0;
3299
3300
3301         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3302             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3303             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3304             (br_dev &&
3305              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3306             (dev->addr_len &&
3307              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3308             (dev->ifindex != dev_get_iflink(dev) &&
3309              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3310                 goto nla_put_failure;
3311
3312         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3313         if (!br_afspec)
3314                 goto nla_put_failure;
3315
3316         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3317                 nla_nest_cancel(skb, br_afspec);
3318                 goto nla_put_failure;
3319         }
3320
3321         if (mode != BRIDGE_MODE_UNDEF) {
3322                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3323                         nla_nest_cancel(skb, br_afspec);
3324                         goto nla_put_failure;
3325                 }
3326         }
3327         if (vlan_fill) {
3328                 err = vlan_fill(skb, dev, filter_mask);
3329                 if (err) {
3330                         nla_nest_cancel(skb, br_afspec);
3331                         goto nla_put_failure;
3332                 }
3333         }
3334         nla_nest_end(skb, br_afspec);
3335
3336         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3337         if (!protinfo)
3338                 goto nla_put_failure;
3339
3340         if (brport_nla_put_flag(skb, flags, mask,
3341                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3342             brport_nla_put_flag(skb, flags, mask,
3343                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3344             brport_nla_put_flag(skb, flags, mask,
3345                                 IFLA_BRPORT_FAST_LEAVE,
3346                                 BR_MULTICAST_FAST_LEAVE) ||
3347             brport_nla_put_flag(skb, flags, mask,
3348                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3349             brport_nla_put_flag(skb, flags, mask,
3350                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3351             brport_nla_put_flag(skb, flags, mask,
3352                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3353             brport_nla_put_flag(skb, flags, mask,
3354                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3355             brport_nla_put_flag(skb, flags, mask,
3356                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3357                 nla_nest_cancel(skb, protinfo);
3358                 goto nla_put_failure;
3359         }
3360
3361         nla_nest_end(skb, protinfo);
3362
3363         nlmsg_end(skb, nlh);
3364         return 0;
3365 nla_put_failure:
3366         nlmsg_cancel(skb, nlh);
3367         return err ? err : -EMSGSIZE;
3368 }
3369 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3370
3371 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3372 {
3373         struct net *net = sock_net(skb->sk);
3374         struct net_device *dev;
3375         int idx = 0;
3376         u32 portid = NETLINK_CB(cb->skb).portid;
3377         u32 seq = cb->nlh->nlmsg_seq;
3378         u32 filter_mask = 0;
3379         int err;
3380
3381         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3382                 struct nlattr *extfilt;
3383
3384                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3385                                           IFLA_EXT_MASK);
3386                 if (extfilt) {
3387                         if (nla_len(extfilt) < sizeof(filter_mask))
3388                                 return -EINVAL;
3389
3390                         filter_mask = nla_get_u32(extfilt);
3391                 }
3392         }
3393
3394         rcu_read_lock();
3395         for_each_netdev_rcu(net, dev) {
3396                 const struct net_device_ops *ops = dev->netdev_ops;
3397                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3398
3399                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3400                         if (idx >= cb->args[0]) {
3401                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3402                                                 skb, portid, seq, dev,
3403                                                 filter_mask, NLM_F_MULTI);
3404                                 if (err < 0 && err != -EOPNOTSUPP)
3405                                         break;
3406                         }
3407                         idx++;
3408                 }
3409
3410                 if (ops->ndo_bridge_getlink) {
3411                         if (idx >= cb->args[0]) {
3412                                 err = ops->ndo_bridge_getlink(skb, portid,
3413                                                               seq, dev,
3414                                                               filter_mask,
3415                                                               NLM_F_MULTI);
3416                                 if (err < 0 && err != -EOPNOTSUPP)
3417                                         break;
3418                         }
3419                         idx++;
3420                 }
3421         }
3422         rcu_read_unlock();
3423         cb->args[0] = idx;
3424
3425         return skb->len;
3426 }
3427
3428 static inline size_t bridge_nlmsg_size(void)
3429 {
3430         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3431                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3432                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3433                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3434                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3435                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3436                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3437                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3438                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3439                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3440                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3441 }
3442
3443 static int rtnl_bridge_notify(struct net_device *dev)
3444 {
3445         struct net *net = dev_net(dev);
3446         struct sk_buff *skb;
3447         int err = -EOPNOTSUPP;
3448
3449         if (!dev->netdev_ops->ndo_bridge_getlink)
3450                 return 0;
3451
3452         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3453         if (!skb) {
3454                 err = -ENOMEM;
3455                 goto errout;
3456         }
3457
3458         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3459         if (err < 0)
3460                 goto errout;
3461
3462         if (!skb->len)
3463                 goto errout;
3464
3465         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3466         return 0;
3467 errout:
3468         WARN_ON(err == -EMSGSIZE);
3469         kfree_skb(skb);
3470         if (err)
3471                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3472         return err;
3473 }
3474
3475 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3476 {
3477         struct net *net = sock_net(skb->sk);
3478         struct ifinfomsg *ifm;
3479         struct net_device *dev;
3480         struct nlattr *br_spec, *attr = NULL;
3481         int rem, err = -EOPNOTSUPP;
3482         u16 flags = 0;
3483         bool have_flags = false;
3484
3485         if (nlmsg_len(nlh) < sizeof(*ifm))
3486                 return -EINVAL;
3487
3488         ifm = nlmsg_data(nlh);
3489         if (ifm->ifi_family != AF_BRIDGE)
3490                 return -EPFNOSUPPORT;
3491
3492         dev = __dev_get_by_index(net, ifm->ifi_index);
3493         if (!dev) {
3494                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3495                 return -ENODEV;
3496         }
3497
3498         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3499         if (br_spec) {
3500                 nla_for_each_nested(attr, br_spec, rem) {
3501                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3502                                 if (nla_len(attr) < sizeof(flags))
3503                                         return -EINVAL;
3504
3505                                 have_flags = true;
3506                                 flags = nla_get_u16(attr);
3507                                 break;
3508                         }
3509                 }
3510         }
3511
3512         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3513                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3514
3515                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3516                         err = -EOPNOTSUPP;
3517                         goto out;
3518                 }
3519
3520                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3521                 if (err)
3522                         goto out;
3523
3524                 flags &= ~BRIDGE_FLAGS_MASTER;
3525         }
3526
3527         if ((flags & BRIDGE_FLAGS_SELF)) {
3528                 if (!dev->netdev_ops->ndo_bridge_setlink)
3529                         err = -EOPNOTSUPP;
3530                 else
3531                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3532                                                                   flags);
3533                 if (!err) {
3534                         flags &= ~BRIDGE_FLAGS_SELF;
3535
3536                         /* Generate event to notify upper layer of bridge
3537                          * change
3538                          */
3539                         err = rtnl_bridge_notify(dev);
3540                 }
3541         }
3542
3543         if (have_flags)
3544                 memcpy(nla_data(attr), &flags, sizeof(flags));
3545 out:
3546         return err;
3547 }
3548
3549 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3550 {
3551         struct net *net = sock_net(skb->sk);
3552         struct ifinfomsg *ifm;
3553         struct net_device *dev;
3554         struct nlattr *br_spec, *attr = NULL;
3555         int rem, err = -EOPNOTSUPP;
3556         u16 flags = 0;
3557         bool have_flags = false;
3558
3559         if (nlmsg_len(nlh) < sizeof(*ifm))
3560                 return -EINVAL;
3561
3562         ifm = nlmsg_data(nlh);
3563         if (ifm->ifi_family != AF_BRIDGE)
3564                 return -EPFNOSUPPORT;
3565
3566         dev = __dev_get_by_index(net, ifm->ifi_index);
3567         if (!dev) {
3568                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3569                 return -ENODEV;
3570         }
3571
3572         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3573         if (br_spec) {
3574                 nla_for_each_nested(attr, br_spec, rem) {
3575                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3576                                 if (nla_len(attr) < sizeof(flags))
3577                                         return -EINVAL;
3578
3579                                 have_flags = true;
3580                                 flags = nla_get_u16(attr);
3581                                 break;
3582                         }
3583                 }
3584         }
3585
3586         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3587                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3588
3589                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3590                         err = -EOPNOTSUPP;
3591                         goto out;
3592                 }
3593
3594                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3595                 if (err)
3596                         goto out;
3597
3598                 flags &= ~BRIDGE_FLAGS_MASTER;
3599         }
3600
3601         if ((flags & BRIDGE_FLAGS_SELF)) {
3602                 if (!dev->netdev_ops->ndo_bridge_dellink)
3603                         err = -EOPNOTSUPP;
3604                 else
3605                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3606                                                                   flags);
3607
3608                 if (!err) {
3609                         flags &= ~BRIDGE_FLAGS_SELF;
3610
3611                         /* Generate event to notify upper layer of bridge
3612                          * change
3613                          */
3614                         err = rtnl_bridge_notify(dev);
3615                 }
3616         }
3617
3618         if (have_flags)
3619                 memcpy(nla_data(attr), &flags, sizeof(flags));
3620 out:
3621         return err;
3622 }
3623
3624 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3625 {
3626         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3627                (!idxattr || idxattr == attrid);
3628 }
3629
3630 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3631 static int rtnl_get_offload_stats_attr_size(int attr_id)
3632 {
3633         switch (attr_id) {
3634         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3635                 return sizeof(struct rtnl_link_stats64);
3636         }
3637
3638         return 0;
3639 }
3640
3641 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3642                                   int *prividx)
3643 {
3644         struct nlattr *attr = NULL;
3645         int attr_id, size;
3646         void *attr_data;
3647         int err;
3648
3649         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3650               dev->netdev_ops->ndo_get_offload_stats))
3651                 return -ENODATA;
3652
3653         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3654              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3655                 if (attr_id < *prividx)
3656                         continue;
3657
3658                 size = rtnl_get_offload_stats_attr_size(attr_id);
3659                 if (!size)
3660                         continue;
3661
3662                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3663                         continue;
3664
3665                 attr = nla_reserve_64bit(skb, attr_id, size,
3666                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3667                 if (!attr)
3668                         goto nla_put_failure;
3669
3670                 attr_data = nla_data(attr);
3671                 memset(attr_data, 0, size);
3672                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3673                                                              attr_data);
3674                 if (err)
3675                         goto get_offload_stats_failure;
3676         }
3677
3678         if (!attr)
3679                 return -ENODATA;
3680
3681         *prividx = 0;
3682         return 0;
3683
3684 nla_put_failure:
3685         err = -EMSGSIZE;
3686 get_offload_stats_failure:
3687         *prividx = attr_id;
3688         return err;
3689 }
3690
3691 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3692 {
3693         int nla_size = 0;
3694         int attr_id;
3695         int size;
3696
3697         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3698               dev->netdev_ops->ndo_get_offload_stats))
3699                 return 0;
3700
3701         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3702              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3703                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3704                         continue;
3705                 size = rtnl_get_offload_stats_attr_size(attr_id);
3706                 nla_size += nla_total_size_64bit(size);
3707         }
3708
3709         if (nla_size != 0)
3710                 nla_size += nla_total_size(0);
3711
3712         return nla_size;
3713 }
3714
3715 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3716                                int type, u32 pid, u32 seq, u32 change,
3717                                unsigned int flags, unsigned int filter_mask,
3718                                int *idxattr, int *prividx)
3719 {
3720         struct if_stats_msg *ifsm;
3721         struct nlmsghdr *nlh;
3722         struct nlattr *attr;
3723         int s_prividx = *prividx;
3724         int err;
3725
3726         ASSERT_RTNL();
3727
3728         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3729         if (!nlh)
3730                 return -EMSGSIZE;
3731
3732         ifsm = nlmsg_data(nlh);
3733         ifsm->ifindex = dev->ifindex;
3734         ifsm->filter_mask = filter_mask;
3735
3736         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3737                 struct rtnl_link_stats64 *sp;
3738
3739                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3740                                          sizeof(struct rtnl_link_stats64),
3741                                          IFLA_STATS_UNSPEC);
3742                 if (!attr)
3743                         goto nla_put_failure;
3744
3745                 sp = nla_data(attr);
3746                 dev_get_stats(dev, sp);
3747         }
3748
3749         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3750                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3751
3752                 if (ops && ops->fill_linkxstats) {
3753                         *idxattr = IFLA_STATS_LINK_XSTATS;
3754                         attr = nla_nest_start(skb,
3755                                               IFLA_STATS_LINK_XSTATS);
3756                         if (!attr)
3757                                 goto nla_put_failure;
3758
3759                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3760                         nla_nest_end(skb, attr);
3761                         if (err)
3762                                 goto nla_put_failure;
3763                         *idxattr = 0;
3764                 }
3765         }
3766
3767         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3768                              *idxattr)) {
3769                 const struct rtnl_link_ops *ops = NULL;
3770                 const struct net_device *master;
3771
3772                 master = netdev_master_upper_dev_get(dev);
3773                 if (master)
3774                         ops = master->rtnl_link_ops;
3775                 if (ops && ops->fill_linkxstats) {
3776                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3777                         attr = nla_nest_start(skb,
3778                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3779                         if (!attr)
3780                                 goto nla_put_failure;
3781
3782                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3783                         nla_nest_end(skb, attr);
3784                         if (err)
3785                                 goto nla_put_failure;
3786                         *idxattr = 0;
3787                 }
3788         }
3789
3790         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3791                              *idxattr)) {
3792                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3793                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3794                 if (!attr)
3795                         goto nla_put_failure;
3796
3797                 err = rtnl_get_offload_stats(skb, dev, prividx);
3798                 if (err == -ENODATA)
3799                         nla_nest_cancel(skb, attr);
3800                 else
3801                         nla_nest_end(skb, attr);
3802
3803                 if (err && err != -ENODATA)
3804                         goto nla_put_failure;
3805                 *idxattr = 0;
3806         }
3807
3808         nlmsg_end(skb, nlh);
3809
3810         return 0;
3811
3812 nla_put_failure:
3813         /* not a multi message or no progress mean a real error */
3814         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3815                 nlmsg_cancel(skb, nlh);
3816         else
3817                 nlmsg_end(skb, nlh);
3818
3819         return -EMSGSIZE;
3820 }
3821
3822 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3823                                   u32 filter_mask)
3824 {
3825         size_t size = 0;
3826
3827         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3828                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3829
3830         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3831                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3832                 int attr = IFLA_STATS_LINK_XSTATS;
3833
3834                 if (ops && ops->get_linkxstats_size) {
3835                         size += nla_total_size(ops->get_linkxstats_size(dev,
3836                                                                         attr));
3837                         /* for IFLA_STATS_LINK_XSTATS */
3838                         size += nla_total_size(0);
3839                 }
3840         }
3841
3842         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3843                 struct net_device *_dev = (struct net_device *)dev;
3844                 const struct rtnl_link_ops *ops = NULL;
3845                 const struct net_device *master;
3846
3847                 /* netdev_master_upper_dev_get can't take const */
3848                 master = netdev_master_upper_dev_get(_dev);
3849                 if (master)
3850                         ops = master->rtnl_link_ops;
3851                 if (ops && ops->get_linkxstats_size) {
3852                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3853
3854                         size += nla_total_size(ops->get_linkxstats_size(dev,
3855                                                                         attr));
3856                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3857                         size += nla_total_size(0);
3858                 }
3859         }
3860
3861         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3862                 size += rtnl_get_offload_stats_size(dev);
3863
3864         return size;
3865 }
3866
3867 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3868 {
3869         struct net *net = sock_net(skb->sk);
3870         struct net_device *dev = NULL;
3871         int idxattr = 0, prividx = 0;
3872         struct if_stats_msg *ifsm;
3873         struct sk_buff *nskb;
3874         u32 filter_mask;
3875         int err;
3876
3877         ifsm = nlmsg_data(nlh);
3878         if (ifsm->ifindex > 0)
3879                 dev = __dev_get_by_index(net, ifsm->ifindex);
3880         else
3881                 return -EINVAL;
3882
3883         if (!dev)
3884                 return -ENODEV;
3885
3886         filter_mask = ifsm->filter_mask;
3887         if (!filter_mask)
3888                 return -EINVAL;
3889
3890         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3891         if (!nskb)
3892                 return -ENOBUFS;
3893
3894         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3895                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3896                                   0, filter_mask, &idxattr, &prividx);
3897         if (err < 0) {
3898                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3899                 WARN_ON(err == -EMSGSIZE);
3900                 kfree_skb(nskb);
3901         } else {
3902                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3903         }
3904
3905         return err;
3906 }
3907
3908 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3909 {
3910         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3911         struct net *net = sock_net(skb->sk);
3912         unsigned int flags = NLM_F_MULTI;
3913         struct if_stats_msg *ifsm;
3914         struct hlist_head *head;
3915         struct net_device *dev;
3916         u32 filter_mask = 0;
3917         int idx = 0;
3918
3919         s_h = cb->args[0];
3920         s_idx = cb->args[1];
3921         s_idxattr = cb->args[2];
3922         s_prividx = cb->args[3];
3923
3924         cb->seq = net->dev_base_seq;
3925
3926         ifsm = nlmsg_data(cb->nlh);
3927         filter_mask = ifsm->filter_mask;
3928         if (!filter_mask)
3929                 return -EINVAL;
3930
3931         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3932                 idx = 0;
3933                 head = &net->dev_index_head[h];
3934                 hlist_for_each_entry(dev, head, index_hlist) {
3935                         if (idx < s_idx)
3936                                 goto cont;
3937                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3938                                                   NETLINK_CB(cb->skb).portid,
3939                                                   cb->nlh->nlmsg_seq, 0,
3940                                                   flags, filter_mask,
3941                                                   &s_idxattr, &s_prividx);
3942                         /* If we ran out of room on the first message,
3943                          * we're in trouble
3944                          */
3945                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3946
3947                         if (err < 0)
3948                                 goto out;
3949                         s_prividx = 0;
3950                         s_idxattr = 0;
3951                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3952 cont:
3953                         idx++;
3954                 }
3955         }
3956 out:
3957         cb->args[3] = s_prividx;
3958         cb->args[2] = s_idxattr;
3959         cb->args[1] = idx;
3960         cb->args[0] = h;
3961
3962         return skb->len;
3963 }
3964
3965 /* Process one rtnetlink message. */
3966
3967 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3968 {
3969         struct net *net = sock_net(skb->sk);
3970         rtnl_doit_func doit;
3971         int kind;
3972         int family;
3973         int type;
3974         int err;
3975
3976         type = nlh->nlmsg_type;
3977         if (type > RTM_MAX)
3978                 return -EOPNOTSUPP;
3979
3980         type -= RTM_BASE;
3981
3982         /* All the messages must have at least 1 byte length */
3983         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3984                 return 0;
3985
3986         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3987         kind = type&3;
3988
3989         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3990                 return -EPERM;
3991
3992         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3993                 struct sock *rtnl;
3994                 rtnl_dumpit_func dumpit;
3995                 rtnl_calcit_func calcit;
3996                 u16 min_dump_alloc = 0;
3997
3998                 dumpit = rtnl_get_dumpit(family, type);
3999                 if (dumpit == NULL)
4000                         return -EOPNOTSUPP;
4001                 calcit = rtnl_get_calcit(family, type);
4002                 if (calcit)
4003                         min_dump_alloc = calcit(skb, nlh);
4004
4005                 __rtnl_unlock();
4006                 rtnl = net->rtnl;
4007                 {
4008                         struct netlink_dump_control c = {
4009                                 .dump           = dumpit,
4010                                 .min_dump_alloc = min_dump_alloc,
4011                         };
4012                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4013                 }
4014                 rtnl_lock();
4015                 return err;
4016         }
4017
4018         doit = rtnl_get_doit(family, type);
4019         if (doit == NULL)
4020                 return -EOPNOTSUPP;
4021
4022         return doit(skb, nlh);
4023 }
4024
4025 static void rtnetlink_rcv(struct sk_buff *skb)
4026 {
4027         rtnl_lock();
4028         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4029         rtnl_unlock();
4030 }
4031
4032 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4033 {
4034         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4035
4036         switch (event) {
4037         case NETDEV_UP:
4038         case NETDEV_DOWN:
4039         case NETDEV_PRE_UP:
4040         case NETDEV_POST_INIT:
4041         case NETDEV_REGISTER:
4042         case NETDEV_CHANGE:
4043         case NETDEV_PRE_TYPE_CHANGE:
4044         case NETDEV_GOING_DOWN:
4045         case NETDEV_UNREGISTER:
4046         case NETDEV_UNREGISTER_FINAL:
4047         case NETDEV_RELEASE:
4048         case NETDEV_JOIN:
4049         case NETDEV_BONDING_INFO:
4050                 break;
4051         default:
4052                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4053                 break;
4054         }
4055         return NOTIFY_DONE;
4056 }
4057
4058 static struct notifier_block rtnetlink_dev_notifier = {
4059         .notifier_call  = rtnetlink_event,
4060 };
4061
4062
4063 static int __net_init rtnetlink_net_init(struct net *net)
4064 {
4065         struct sock *sk;
4066         struct netlink_kernel_cfg cfg = {
4067                 .groups         = RTNLGRP_MAX,
4068                 .input          = rtnetlink_rcv,
4069                 .cb_mutex       = &rtnl_mutex,
4070                 .flags          = NL_CFG_F_NONROOT_RECV,
4071         };
4072
4073         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4074         if (!sk)
4075                 return -ENOMEM;
4076         net->rtnl = sk;
4077         return 0;
4078 }
4079
4080 static void __net_exit rtnetlink_net_exit(struct net *net)
4081 {
4082         netlink_kernel_release(net->rtnl);
4083         net->rtnl = NULL;
4084 }
4085
4086 static struct pernet_operations rtnetlink_net_ops = {
4087         .init = rtnetlink_net_init,
4088         .exit = rtnetlink_net_exit,
4089 };
4090
4091 void __init rtnetlink_init(void)
4092 {
4093         if (register_pernet_subsys(&rtnetlink_net_ops))
4094                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4095
4096         register_netdevice_notifier(&rtnetlink_dev_notifier);
4097
4098         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4099                       rtnl_dump_ifinfo, rtnl_calcit);
4100         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4101         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4102         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4103
4104         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4105         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4106
4107         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4108         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4109         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4110
4111         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4112         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4113         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4114
4115         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4116                       NULL);
4117 }