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         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1148
1149         vf_mac.vf =
1150                 vf_vlan.vf =
1151                 vf_vlan_info.vf =
1152                 vf_rate.vf =
1153                 vf_tx_rate.vf =
1154                 vf_spoofchk.vf =
1155                 vf_linkstate.vf =
1156                 vf_rss_query_en.vf =
1157                 vf_trust.vf = ivi.vf;
1158
1159         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1160         vf_vlan.vlan = ivi.vlan;
1161         vf_vlan.qos = ivi.qos;
1162         vf_vlan_info.vlan = ivi.vlan;
1163         vf_vlan_info.qos = ivi.qos;
1164         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1165         vf_tx_rate.rate = ivi.max_tx_rate;
1166         vf_rate.min_tx_rate = ivi.min_tx_rate;
1167         vf_rate.max_tx_rate = ivi.max_tx_rate;
1168         vf_spoofchk.setting = ivi.spoofchk;
1169         vf_linkstate.link_state = ivi.linkstate;
1170         vf_rss_query_en.setting = ivi.rss_query_en;
1171         vf_trust.setting = ivi.trusted;
1172         vf = nla_nest_start(skb, IFLA_VF_INFO);
1173         if (!vf)
1174                 goto nla_put_vfinfo_failure;
1175         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1176             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1177             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1178                     &vf_rate) ||
1179             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1180                     &vf_tx_rate) ||
1181             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1182                     &vf_spoofchk) ||
1183             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1184                     &vf_linkstate) ||
1185             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1186                     sizeof(vf_rss_query_en),
1187                     &vf_rss_query_en) ||
1188             nla_put(skb, IFLA_VF_TRUST,
1189                     sizeof(vf_trust), &vf_trust))
1190                 goto nla_put_vf_failure;
1191         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1192         if (!vfvlanlist)
1193                 goto nla_put_vf_failure;
1194         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1195                     &vf_vlan_info)) {
1196                 nla_nest_cancel(skb, vfvlanlist);
1197                 goto nla_put_vf_failure;
1198         }
1199         nla_nest_end(skb, vfvlanlist);
1200         memset(&vf_stats, 0, sizeof(vf_stats));
1201         if (dev->netdev_ops->ndo_get_vf_stats)
1202                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1203                                                 &vf_stats);
1204         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1205         if (!vfstats)
1206                 goto nla_put_vf_failure;
1207         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1208                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1209             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1210                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1211             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1212                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1214                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1216                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1217             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1218                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1219                 nla_nest_cancel(skb, vfstats);
1220                 goto nla_put_vf_failure;
1221         }
1222         nla_nest_end(skb, vfstats);
1223         nla_nest_end(skb, vf);
1224         return 0;
1225
1226 nla_put_vf_failure:
1227         nla_nest_cancel(skb, vf);
1228 nla_put_vfinfo_failure:
1229         nla_nest_cancel(skb, vfinfo);
1230         return -EMSGSIZE;
1231 }
1232
1233 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1234 {
1235         struct rtnl_link_ifmap map;
1236
1237         memset(&map, 0, sizeof(map));
1238         map.mem_start   = dev->mem_start;
1239         map.mem_end     = dev->mem_end;
1240         map.base_addr   = dev->base_addr;
1241         map.irq         = dev->irq;
1242         map.dma         = dev->dma;
1243         map.port        = dev->if_port;
1244
1245         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1246                 return -EMSGSIZE;
1247
1248         return 0;
1249 }
1250
1251 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1252 {
1253         struct netdev_xdp xdp_op = {};
1254         struct nlattr *xdp;
1255         int err;
1256
1257         if (!dev->netdev_ops->ndo_xdp)
1258                 return 0;
1259         xdp = nla_nest_start(skb, IFLA_XDP);
1260         if (!xdp)
1261                 return -EMSGSIZE;
1262         xdp_op.command = XDP_QUERY_PROG;
1263         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1264         if (err)
1265                 goto err_cancel;
1266         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1267         if (err)
1268                 goto err_cancel;
1269
1270         nla_nest_end(skb, xdp);
1271         return 0;
1272
1273 err_cancel:
1274         nla_nest_cancel(skb, xdp);
1275         return err;
1276 }
1277
1278 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1279                             int type, u32 pid, u32 seq, u32 change,
1280                             unsigned int flags, u32 ext_filter_mask)
1281 {
1282         struct ifinfomsg *ifm;
1283         struct nlmsghdr *nlh;
1284         struct nlattr *af_spec;
1285         struct rtnl_af_ops *af_ops;
1286         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1287
1288         ASSERT_RTNL();
1289         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1290         if (nlh == NULL)
1291                 return -EMSGSIZE;
1292
1293         ifm = nlmsg_data(nlh);
1294         ifm->ifi_family = AF_UNSPEC;
1295         ifm->__ifi_pad = 0;
1296         ifm->ifi_type = dev->type;
1297         ifm->ifi_index = dev->ifindex;
1298         ifm->ifi_flags = dev_get_flags(dev);
1299         ifm->ifi_change = change;
1300
1301         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1302             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1303             nla_put_u8(skb, IFLA_OPERSTATE,
1304                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1305             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1306             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1307             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1308             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1309             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1310             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1311             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1312 #ifdef CONFIG_RPS
1313             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1314 #endif
1315             (dev->ifindex != dev_get_iflink(dev) &&
1316              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1317             (upper_dev &&
1318              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1319             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1320             (dev->qdisc &&
1321              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1322             (dev->ifalias &&
1323              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1324             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1325                         atomic_read(&dev->carrier_changes)) ||
1326             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1327                 goto nla_put_failure;
1328
1329         if (rtnl_fill_link_ifmap(skb, dev))
1330                 goto nla_put_failure;
1331
1332         if (dev->addr_len) {
1333                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1334                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1335                         goto nla_put_failure;
1336         }
1337
1338         if (rtnl_phys_port_id_fill(skb, dev))
1339                 goto nla_put_failure;
1340
1341         if (rtnl_phys_port_name_fill(skb, dev))
1342                 goto nla_put_failure;
1343
1344         if (rtnl_phys_switch_id_fill(skb, dev))
1345                 goto nla_put_failure;
1346
1347         if (rtnl_fill_stats(skb, dev))
1348                 goto nla_put_failure;
1349
1350         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1351             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1352                 goto nla_put_failure;
1353
1354         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1355             ext_filter_mask & RTEXT_FILTER_VF) {
1356                 int i;
1357                 struct nlattr *vfinfo;
1358                 int num_vfs = dev_num_vf(dev->dev.parent);
1359
1360                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1361                 if (!vfinfo)
1362                         goto nla_put_failure;
1363                 for (i = 0; i < num_vfs; i++) {
1364                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1365                                 goto nla_put_failure;
1366                 }
1367
1368                 nla_nest_end(skb, vfinfo);
1369         }
1370
1371         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1372                 goto nla_put_failure;
1373
1374         if (rtnl_xdp_fill(skb, dev))
1375                 goto nla_put_failure;
1376
1377         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1378                 if (rtnl_link_fill(skb, dev) < 0)
1379                         goto nla_put_failure;
1380         }
1381
1382         if (dev->rtnl_link_ops &&
1383             dev->rtnl_link_ops->get_link_net) {
1384                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1385
1386                 if (!net_eq(dev_net(dev), link_net)) {
1387                         int id = peernet2id_alloc(dev_net(dev), link_net);
1388
1389                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1390                                 goto nla_put_failure;
1391                 }
1392         }
1393
1394         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1395                 goto nla_put_failure;
1396
1397         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1398                 if (af_ops->fill_link_af) {
1399                         struct nlattr *af;
1400                         int err;
1401
1402                         if (!(af = nla_nest_start(skb, af_ops->family)))
1403                                 goto nla_put_failure;
1404
1405                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1406
1407                         /*
1408                          * Caller may return ENODATA to indicate that there
1409                          * was no data to be dumped. This is not an error, it
1410                          * means we should trim the attribute header and
1411                          * continue.
1412                          */
1413                         if (err == -ENODATA)
1414                                 nla_nest_cancel(skb, af);
1415                         else if (err < 0)
1416                                 goto nla_put_failure;
1417
1418                         nla_nest_end(skb, af);
1419                 }
1420         }
1421
1422         nla_nest_end(skb, af_spec);
1423
1424         nlmsg_end(skb, nlh);
1425         return 0;
1426
1427 nla_put_failure:
1428         nlmsg_cancel(skb, nlh);
1429         return -EMSGSIZE;
1430 }
1431
1432 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1433         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1434         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1435         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1436         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1437         [IFLA_MTU]              = { .type = NLA_U32 },
1438         [IFLA_LINK]             = { .type = NLA_U32 },
1439         [IFLA_MASTER]           = { .type = NLA_U32 },
1440         [IFLA_CARRIER]          = { .type = NLA_U8 },
1441         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1442         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1443         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1444         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1445         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1446         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1447         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1448         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1449         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1450         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1451         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1452         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1453         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1454         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1455         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1456         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1457         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1458         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1459         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1460         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1461         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1462         [IFLA_XDP]              = { .type = NLA_NESTED },
1463 };
1464
1465 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1466         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1467         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1468         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1469         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1470 };
1471
1472 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1473         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1474         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1475         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1476         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1477         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1478         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1479         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1480         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1481         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1482         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1483         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1484         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1485 };
1486
1487 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1488         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1489         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1490                                     .len = PORT_PROFILE_MAX },
1491         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1492                                     .len = sizeof(struct ifla_port_vsi)},
1493         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1494                                       .len = PORT_UUID_MAX },
1495         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1496                                     .len = PORT_UUID_MAX },
1497         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1498         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1499 };
1500
1501 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1502         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1503         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1504 };
1505
1506 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1507 {
1508         const struct rtnl_link_ops *ops = NULL;
1509         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1510
1511         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1512                 return NULL;
1513
1514         if (linfo[IFLA_INFO_KIND]) {
1515                 char kind[MODULE_NAME_LEN];
1516
1517                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1518                 ops = rtnl_link_ops_get(kind);
1519         }
1520
1521         return ops;
1522 }
1523
1524 static bool link_master_filtered(struct net_device *dev, int master_idx)
1525 {
1526         struct net_device *master;
1527
1528         if (!master_idx)
1529                 return false;
1530
1531         master = netdev_master_upper_dev_get(dev);
1532         if (!master || master->ifindex != master_idx)
1533                 return true;
1534
1535         return false;
1536 }
1537
1538 static bool link_kind_filtered(const struct net_device *dev,
1539                                const struct rtnl_link_ops *kind_ops)
1540 {
1541         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1542                 return true;
1543
1544         return false;
1545 }
1546
1547 static bool link_dump_filtered(struct net_device *dev,
1548                                int master_idx,
1549                                const struct rtnl_link_ops *kind_ops)
1550 {
1551         if (link_master_filtered(dev, master_idx) ||
1552             link_kind_filtered(dev, kind_ops))
1553                 return true;
1554
1555         return false;
1556 }
1557
1558 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1559 {
1560         struct net *net = sock_net(skb->sk);
1561         int h, s_h;
1562         int idx = 0, s_idx;
1563         struct net_device *dev;
1564         struct hlist_head *head;
1565         struct nlattr *tb[IFLA_MAX+1];
1566         u32 ext_filter_mask = 0;
1567         const struct rtnl_link_ops *kind_ops = NULL;
1568         unsigned int flags = NLM_F_MULTI;
1569         int master_idx = 0;
1570         int err;
1571         int hdrlen;
1572
1573         s_h = cb->args[0];
1574         s_idx = cb->args[1];
1575
1576         cb->seq = net->dev_base_seq;
1577
1578         /* A hack to preserve kernel<->userspace interface.
1579          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1580          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1581          * what iproute2 < v3.9.0 used.
1582          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1583          * attribute, its netlink message is shorter than struct ifinfomsg.
1584          */
1585         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1586                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1587
1588         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1589
1590                 if (tb[IFLA_EXT_MASK])
1591                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1592
1593                 if (tb[IFLA_MASTER])
1594                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1595
1596                 if (tb[IFLA_LINKINFO])
1597                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1598
1599                 if (master_idx || kind_ops)
1600                         flags |= NLM_F_DUMP_FILTERED;
1601         }
1602
1603         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1604                 idx = 0;
1605                 head = &net->dev_index_head[h];
1606                 hlist_for_each_entry(dev, head, index_hlist) {
1607                         if (link_dump_filtered(dev, master_idx, kind_ops))
1608                                 continue;
1609                         if (idx < s_idx)
1610                                 goto cont;
1611                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1612                                                NETLINK_CB(cb->skb).portid,
1613                                                cb->nlh->nlmsg_seq, 0,
1614                                                flags,
1615                                                ext_filter_mask);
1616                         /* If we ran out of room on the first message,
1617                          * we're in trouble
1618                          */
1619                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1620
1621                         if (err < 0)
1622                                 goto out;
1623
1624                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1625 cont:
1626                         idx++;
1627                 }
1628         }
1629 out:
1630         cb->args[1] = idx;
1631         cb->args[0] = h;
1632
1633         return skb->len;
1634 }
1635
1636 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1637 {
1638         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1639 }
1640 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1641
1642 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1643 {
1644         struct net *net;
1645         /* Examine the link attributes and figure out which
1646          * network namespace we are talking about.
1647          */
1648         if (tb[IFLA_NET_NS_PID])
1649                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1650         else if (tb[IFLA_NET_NS_FD])
1651                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1652         else
1653                 net = get_net(src_net);
1654         return net;
1655 }
1656 EXPORT_SYMBOL(rtnl_link_get_net);
1657
1658 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1659 {
1660         if (dev) {
1661                 if (tb[IFLA_ADDRESS] &&
1662                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1663                         return -EINVAL;
1664
1665                 if (tb[IFLA_BROADCAST] &&
1666                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1667                         return -EINVAL;
1668         }
1669
1670         if (tb[IFLA_AF_SPEC]) {
1671                 struct nlattr *af;
1672                 int rem, err;
1673
1674                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1675                         const struct rtnl_af_ops *af_ops;
1676
1677                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1678                                 return -EAFNOSUPPORT;
1679
1680                         if (!af_ops->set_link_af)
1681                                 return -EOPNOTSUPP;
1682
1683                         if (af_ops->validate_link_af) {
1684                                 err = af_ops->validate_link_af(dev, af);
1685                                 if (err < 0)
1686                                         return err;
1687                         }
1688                 }
1689         }
1690
1691         return 0;
1692 }
1693
1694 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1695                                   int guid_type)
1696 {
1697         const struct net_device_ops *ops = dev->netdev_ops;
1698
1699         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1700 }
1701
1702 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1703 {
1704         if (dev->type != ARPHRD_INFINIBAND)
1705                 return -EOPNOTSUPP;
1706
1707         return handle_infiniband_guid(dev, ivt, guid_type);
1708 }
1709
1710 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1711 {
1712         const struct net_device_ops *ops = dev->netdev_ops;
1713         int err = -EINVAL;
1714
1715         if (tb[IFLA_VF_MAC]) {
1716                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1717
1718                 err = -EOPNOTSUPP;
1719                 if (ops->ndo_set_vf_mac)
1720                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1721                                                   ivm->mac);
1722                 if (err < 0)
1723                         return err;
1724         }
1725
1726         if (tb[IFLA_VF_VLAN]) {
1727                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1728
1729                 err = -EOPNOTSUPP;
1730                 if (ops->ndo_set_vf_vlan)
1731                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1732                                                    ivv->qos,
1733                                                    htons(ETH_P_8021Q));
1734                 if (err < 0)
1735                         return err;
1736         }
1737
1738         if (tb[IFLA_VF_VLAN_LIST]) {
1739                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1740                 struct nlattr *attr;
1741                 int rem, len = 0;
1742
1743                 err = -EOPNOTSUPP;
1744                 if (!ops->ndo_set_vf_vlan)
1745                         return err;
1746
1747                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1748                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1749                             nla_len(attr) < NLA_HDRLEN) {
1750                                 return -EINVAL;
1751                         }
1752                         if (len >= MAX_VLAN_LIST_LEN)
1753                                 return -EOPNOTSUPP;
1754                         ivvl[len] = nla_data(attr);
1755
1756                         len++;
1757                 }
1758                 if (len == 0)
1759                         return -EINVAL;
1760
1761                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1762                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1763                 if (err < 0)
1764                         return err;
1765         }
1766
1767         if (tb[IFLA_VF_TX_RATE]) {
1768                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1769                 struct ifla_vf_info ivf;
1770
1771                 err = -EOPNOTSUPP;
1772                 if (ops->ndo_get_vf_config)
1773                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1774                 if (err < 0)
1775                         return err;
1776
1777                 err = -EOPNOTSUPP;
1778                 if (ops->ndo_set_vf_rate)
1779                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1780                                                    ivf.min_tx_rate,
1781                                                    ivt->rate);
1782                 if (err < 0)
1783                         return err;
1784         }
1785
1786         if (tb[IFLA_VF_RATE]) {
1787                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1788
1789                 err = -EOPNOTSUPP;
1790                 if (ops->ndo_set_vf_rate)
1791                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1792                                                    ivt->min_tx_rate,
1793                                                    ivt->max_tx_rate);
1794                 if (err < 0)
1795                         return err;
1796         }
1797
1798         if (tb[IFLA_VF_SPOOFCHK]) {
1799                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1800
1801                 err = -EOPNOTSUPP;
1802                 if (ops->ndo_set_vf_spoofchk)
1803                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1804                                                        ivs->setting);
1805                 if (err < 0)
1806                         return err;
1807         }
1808
1809         if (tb[IFLA_VF_LINK_STATE]) {
1810                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1811
1812                 err = -EOPNOTSUPP;
1813                 if (ops->ndo_set_vf_link_state)
1814                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1815                                                          ivl->link_state);
1816                 if (err < 0)
1817                         return err;
1818         }
1819
1820         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1821                 struct ifla_vf_rss_query_en *ivrssq_en;
1822
1823                 err = -EOPNOTSUPP;
1824                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1825                 if (ops->ndo_set_vf_rss_query_en)
1826                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1827                                                            ivrssq_en->setting);
1828                 if (err < 0)
1829                         return err;
1830         }
1831
1832         if (tb[IFLA_VF_TRUST]) {
1833                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1834
1835                 err = -EOPNOTSUPP;
1836                 if (ops->ndo_set_vf_trust)
1837                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1838                 if (err < 0)
1839                         return err;
1840         }
1841
1842         if (tb[IFLA_VF_IB_NODE_GUID]) {
1843                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1844
1845                 if (!ops->ndo_set_vf_guid)
1846                         return -EOPNOTSUPP;
1847
1848                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1849         }
1850
1851         if (tb[IFLA_VF_IB_PORT_GUID]) {
1852                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1853
1854                 if (!ops->ndo_set_vf_guid)
1855                         return -EOPNOTSUPP;
1856
1857                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1858         }
1859
1860         return err;
1861 }
1862
1863 static int do_set_master(struct net_device *dev, int ifindex)
1864 {
1865         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1866         const struct net_device_ops *ops;
1867         int err;
1868
1869         if (upper_dev) {
1870                 if (upper_dev->ifindex == ifindex)
1871                         return 0;
1872                 ops = upper_dev->netdev_ops;
1873                 if (ops->ndo_del_slave) {
1874                         err = ops->ndo_del_slave(upper_dev, dev);
1875                         if (err)
1876                                 return err;
1877                 } else {
1878                         return -EOPNOTSUPP;
1879                 }
1880         }
1881
1882         if (ifindex) {
1883                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1884                 if (!upper_dev)
1885                         return -EINVAL;
1886                 ops = upper_dev->netdev_ops;
1887                 if (ops->ndo_add_slave) {
1888                         err = ops->ndo_add_slave(upper_dev, dev);
1889                         if (err)
1890                                 return err;
1891                 } else {
1892                         return -EOPNOTSUPP;
1893                 }
1894         }
1895         return 0;
1896 }
1897
1898 #define DO_SETLINK_MODIFIED     0x01
1899 /* notify flag means notify + modified. */
1900 #define DO_SETLINK_NOTIFY       0x03
1901 static int do_setlink(const struct sk_buff *skb,
1902                       struct net_device *dev, struct ifinfomsg *ifm,
1903                       struct nlattr **tb, char *ifname, int status)
1904 {
1905         const struct net_device_ops *ops = dev->netdev_ops;
1906         int err;
1907
1908         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1909                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1910                 if (IS_ERR(net)) {
1911                         err = PTR_ERR(net);
1912                         goto errout;
1913                 }
1914                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1915                         put_net(net);
1916                         err = -EPERM;
1917                         goto errout;
1918                 }
1919                 err = dev_change_net_namespace(dev, net, ifname);
1920                 put_net(net);
1921                 if (err)
1922                         goto errout;
1923                 status |= DO_SETLINK_MODIFIED;
1924         }
1925
1926         if (tb[IFLA_MAP]) {
1927                 struct rtnl_link_ifmap *u_map;
1928                 struct ifmap k_map;
1929
1930                 if (!ops->ndo_set_config) {
1931                         err = -EOPNOTSUPP;
1932                         goto errout;
1933                 }
1934
1935                 if (!netif_device_present(dev)) {
1936                         err = -ENODEV;
1937                         goto errout;
1938                 }
1939
1940                 u_map = nla_data(tb[IFLA_MAP]);
1941                 k_map.mem_start = (unsigned long) u_map->mem_start;
1942                 k_map.mem_end = (unsigned long) u_map->mem_end;
1943                 k_map.base_addr = (unsigned short) u_map->base_addr;
1944                 k_map.irq = (unsigned char) u_map->irq;
1945                 k_map.dma = (unsigned char) u_map->dma;
1946                 k_map.port = (unsigned char) u_map->port;
1947
1948                 err = ops->ndo_set_config(dev, &k_map);
1949                 if (err < 0)
1950                         goto errout;
1951
1952                 status |= DO_SETLINK_NOTIFY;
1953         }
1954
1955         if (tb[IFLA_ADDRESS]) {
1956                 struct sockaddr *sa;
1957                 int len;
1958
1959                 len = sizeof(sa_family_t) + dev->addr_len;
1960                 sa = kmalloc(len, GFP_KERNEL);
1961                 if (!sa) {
1962                         err = -ENOMEM;
1963                         goto errout;
1964                 }
1965                 sa->sa_family = dev->type;
1966                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1967                        dev->addr_len);
1968                 err = dev_set_mac_address(dev, sa);
1969                 kfree(sa);
1970                 if (err)
1971                         goto errout;
1972                 status |= DO_SETLINK_MODIFIED;
1973         }
1974
1975         if (tb[IFLA_MTU]) {
1976                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1977                 if (err < 0)
1978                         goto errout;
1979                 status |= DO_SETLINK_MODIFIED;
1980         }
1981
1982         if (tb[IFLA_GROUP]) {
1983                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1984                 status |= DO_SETLINK_NOTIFY;
1985         }
1986
1987         /*
1988          * Interface selected by interface index but interface
1989          * name provided implies that a name change has been
1990          * requested.
1991          */
1992         if (ifm->ifi_index > 0 && ifname[0]) {
1993                 err = dev_change_name(dev, ifname);
1994                 if (err < 0)
1995                         goto errout;
1996                 status |= DO_SETLINK_MODIFIED;
1997         }
1998
1999         if (tb[IFLA_IFALIAS]) {
2000                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2001                                     nla_len(tb[IFLA_IFALIAS]));
2002                 if (err < 0)
2003                         goto errout;
2004                 status |= DO_SETLINK_NOTIFY;
2005         }
2006
2007         if (tb[IFLA_BROADCAST]) {
2008                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2009                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2010         }
2011
2012         if (ifm->ifi_flags || ifm->ifi_change) {
2013                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2014                 if (err < 0)
2015                         goto errout;
2016         }
2017
2018         if (tb[IFLA_MASTER]) {
2019                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2020                 if (err)
2021                         goto errout;
2022                 status |= DO_SETLINK_MODIFIED;
2023         }
2024
2025         if (tb[IFLA_CARRIER]) {
2026                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2027                 if (err)
2028                         goto errout;
2029                 status |= DO_SETLINK_MODIFIED;
2030         }
2031
2032         if (tb[IFLA_TXQLEN]) {
2033                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2034                 unsigned long orig_len = dev->tx_queue_len;
2035
2036                 if (dev->tx_queue_len ^ value) {
2037                         dev->tx_queue_len = value;
2038                         err = call_netdevice_notifiers(
2039                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2040                         err = notifier_to_errno(err);
2041                         if (err) {
2042                                 dev->tx_queue_len = orig_len;
2043                                 goto errout;
2044                         }
2045                         status |= DO_SETLINK_NOTIFY;
2046                 }
2047         }
2048
2049         if (tb[IFLA_OPERSTATE])
2050                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2051
2052         if (tb[IFLA_LINKMODE]) {
2053                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2054
2055                 write_lock_bh(&dev_base_lock);
2056                 if (dev->link_mode ^ value)
2057                         status |= DO_SETLINK_NOTIFY;
2058                 dev->link_mode = value;
2059                 write_unlock_bh(&dev_base_lock);
2060         }
2061
2062         if (tb[IFLA_VFINFO_LIST]) {
2063                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2064                 struct nlattr *attr;
2065                 int rem;
2066
2067                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2068                         if (nla_type(attr) != IFLA_VF_INFO ||
2069                             nla_len(attr) < NLA_HDRLEN) {
2070                                 err = -EINVAL;
2071                                 goto errout;
2072                         }
2073                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2074                                                ifla_vf_policy);
2075                         if (err < 0)
2076                                 goto errout;
2077                         err = do_setvfinfo(dev, vfinfo);
2078                         if (err < 0)
2079                                 goto errout;
2080                         status |= DO_SETLINK_NOTIFY;
2081                 }
2082         }
2083         err = 0;
2084
2085         if (tb[IFLA_VF_PORTS]) {
2086                 struct nlattr *port[IFLA_PORT_MAX+1];
2087                 struct nlattr *attr;
2088                 int vf;
2089                 int rem;
2090
2091                 err = -EOPNOTSUPP;
2092                 if (!ops->ndo_set_vf_port)
2093                         goto errout;
2094
2095                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2096                         if (nla_type(attr) != IFLA_VF_PORT ||
2097                             nla_len(attr) < NLA_HDRLEN) {
2098                                 err = -EINVAL;
2099                                 goto errout;
2100                         }
2101                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2102                                                ifla_port_policy);
2103                         if (err < 0)
2104                                 goto errout;
2105                         if (!port[IFLA_PORT_VF]) {
2106                                 err = -EOPNOTSUPP;
2107                                 goto errout;
2108                         }
2109                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2110                         err = ops->ndo_set_vf_port(dev, vf, port);
2111                         if (err < 0)
2112                                 goto errout;
2113                         status |= DO_SETLINK_NOTIFY;
2114                 }
2115         }
2116         err = 0;
2117
2118         if (tb[IFLA_PORT_SELF]) {
2119                 struct nlattr *port[IFLA_PORT_MAX+1];
2120
2121                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2122                         tb[IFLA_PORT_SELF], ifla_port_policy);
2123                 if (err < 0)
2124                         goto errout;
2125
2126                 err = -EOPNOTSUPP;
2127                 if (ops->ndo_set_vf_port)
2128                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2129                 if (err < 0)
2130                         goto errout;
2131                 status |= DO_SETLINK_NOTIFY;
2132         }
2133
2134         if (tb[IFLA_AF_SPEC]) {
2135                 struct nlattr *af;
2136                 int rem;
2137
2138                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2139                         const struct rtnl_af_ops *af_ops;
2140
2141                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2142                                 BUG();
2143
2144                         err = af_ops->set_link_af(dev, af);
2145                         if (err < 0)
2146                                 goto errout;
2147
2148                         status |= DO_SETLINK_NOTIFY;
2149                 }
2150         }
2151         err = 0;
2152
2153         if (tb[IFLA_PROTO_DOWN]) {
2154                 err = dev_change_proto_down(dev,
2155                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2156                 if (err)
2157                         goto errout;
2158                 status |= DO_SETLINK_NOTIFY;
2159         }
2160
2161         if (tb[IFLA_XDP]) {
2162                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2163
2164                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2165                                        ifla_xdp_policy);
2166                 if (err < 0)
2167                         goto errout;
2168
2169                 if (xdp[IFLA_XDP_ATTACHED]) {
2170                         err = -EINVAL;
2171                         goto errout;
2172                 }
2173                 if (xdp[IFLA_XDP_FD]) {
2174                         err = dev_change_xdp_fd(dev,
2175                                                 nla_get_s32(xdp[IFLA_XDP_FD]));
2176                         if (err)
2177                                 goto errout;
2178                         status |= DO_SETLINK_NOTIFY;
2179                 }
2180         }
2181
2182 errout:
2183         if (status & DO_SETLINK_MODIFIED) {
2184                 if (status & DO_SETLINK_NOTIFY)
2185                         netdev_state_change(dev);
2186
2187                 if (err < 0)
2188                         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",
2189                                              dev->name);
2190         }
2191
2192         return err;
2193 }
2194
2195 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2196 {
2197         struct net *net = sock_net(skb->sk);
2198         struct ifinfomsg *ifm;
2199         struct net_device *dev;
2200         int err;
2201         struct nlattr *tb[IFLA_MAX+1];
2202         char ifname[IFNAMSIZ];
2203
2204         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2205         if (err < 0)
2206                 goto errout;
2207
2208         if (tb[IFLA_IFNAME])
2209                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2210         else
2211                 ifname[0] = '\0';
2212
2213         err = -EINVAL;
2214         ifm = nlmsg_data(nlh);
2215         if (ifm->ifi_index > 0)
2216                 dev = __dev_get_by_index(net, ifm->ifi_index);
2217         else if (tb[IFLA_IFNAME])
2218                 dev = __dev_get_by_name(net, ifname);
2219         else
2220                 goto errout;
2221
2222         if (dev == NULL) {
2223                 err = -ENODEV;
2224                 goto errout;
2225         }
2226
2227         err = validate_linkmsg(dev, tb);
2228         if (err < 0)
2229                 goto errout;
2230
2231         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2232 errout:
2233         return err;
2234 }
2235
2236 static int rtnl_group_dellink(const struct net *net, int group)
2237 {
2238         struct net_device *dev, *aux;
2239         LIST_HEAD(list_kill);
2240         bool found = false;
2241
2242         if (!group)
2243                 return -EPERM;
2244
2245         for_each_netdev(net, dev) {
2246                 if (dev->group == group) {
2247                         const struct rtnl_link_ops *ops;
2248
2249                         found = true;
2250                         ops = dev->rtnl_link_ops;
2251                         if (!ops || !ops->dellink)
2252                                 return -EOPNOTSUPP;
2253                 }
2254         }
2255
2256         if (!found)
2257                 return -ENODEV;
2258
2259         for_each_netdev_safe(net, dev, aux) {
2260                 if (dev->group == group) {
2261                         const struct rtnl_link_ops *ops;
2262
2263                         ops = dev->rtnl_link_ops;
2264                         ops->dellink(dev, &list_kill);
2265                 }
2266         }
2267         unregister_netdevice_many(&list_kill);
2268
2269         return 0;
2270 }
2271
2272 int rtnl_delete_link(struct net_device *dev)
2273 {
2274         const struct rtnl_link_ops *ops;
2275         LIST_HEAD(list_kill);
2276
2277         ops = dev->rtnl_link_ops;
2278         if (!ops || !ops->dellink)
2279                 return -EOPNOTSUPP;
2280
2281         ops->dellink(dev, &list_kill);
2282         unregister_netdevice_many(&list_kill);
2283
2284         return 0;
2285 }
2286 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2287
2288 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2289 {
2290         struct net *net = sock_net(skb->sk);
2291         struct net_device *dev;
2292         struct ifinfomsg *ifm;
2293         char ifname[IFNAMSIZ];
2294         struct nlattr *tb[IFLA_MAX+1];
2295         int err;
2296
2297         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2298         if (err < 0)
2299                 return err;
2300
2301         if (tb[IFLA_IFNAME])
2302                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2303
2304         ifm = nlmsg_data(nlh);
2305         if (ifm->ifi_index > 0)
2306                 dev = __dev_get_by_index(net, ifm->ifi_index);
2307         else if (tb[IFLA_IFNAME])
2308                 dev = __dev_get_by_name(net, ifname);
2309         else if (tb[IFLA_GROUP])
2310                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2311         else
2312                 return -EINVAL;
2313
2314         if (!dev)
2315                 return -ENODEV;
2316
2317         return rtnl_delete_link(dev);
2318 }
2319
2320 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2321 {
2322         unsigned int old_flags;
2323         int err;
2324
2325         old_flags = dev->flags;
2326         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2327                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2328                 if (err < 0)
2329                         return err;
2330         }
2331
2332         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2333
2334         __dev_notify_flags(dev, old_flags, ~0U);
2335         return 0;
2336 }
2337 EXPORT_SYMBOL(rtnl_configure_link);
2338
2339 struct net_device *rtnl_create_link(struct net *net,
2340         const char *ifname, unsigned char name_assign_type,
2341         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2342 {
2343         int err;
2344         struct net_device *dev;
2345         unsigned int num_tx_queues = 1;
2346         unsigned int num_rx_queues = 1;
2347
2348         if (tb[IFLA_NUM_TX_QUEUES])
2349                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2350         else if (ops->get_num_tx_queues)
2351                 num_tx_queues = ops->get_num_tx_queues();
2352
2353         if (tb[IFLA_NUM_RX_QUEUES])
2354                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2355         else if (ops->get_num_rx_queues)
2356                 num_rx_queues = ops->get_num_rx_queues();
2357
2358         err = -ENOMEM;
2359         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2360                                ops->setup, num_tx_queues, num_rx_queues);
2361         if (!dev)
2362                 goto err;
2363
2364         dev_net_set(dev, net);
2365         dev->rtnl_link_ops = ops;
2366         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2367
2368         if (tb[IFLA_MTU])
2369                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2370         if (tb[IFLA_ADDRESS]) {
2371                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2372                                 nla_len(tb[IFLA_ADDRESS]));
2373                 dev->addr_assign_type = NET_ADDR_SET;
2374         }
2375         if (tb[IFLA_BROADCAST])
2376                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2377                                 nla_len(tb[IFLA_BROADCAST]));
2378         if (tb[IFLA_TXQLEN])
2379                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2380         if (tb[IFLA_OPERSTATE])
2381                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2382         if (tb[IFLA_LINKMODE])
2383                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2384         if (tb[IFLA_GROUP])
2385                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2386
2387         return dev;
2388
2389 err:
2390         return ERR_PTR(err);
2391 }
2392 EXPORT_SYMBOL(rtnl_create_link);
2393
2394 static int rtnl_group_changelink(const struct sk_buff *skb,
2395                 struct net *net, int group,
2396                 struct ifinfomsg *ifm,
2397                 struct nlattr **tb)
2398 {
2399         struct net_device *dev, *aux;
2400         int err;
2401
2402         for_each_netdev_safe(net, dev, aux) {
2403                 if (dev->group == group) {
2404                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2405                         if (err < 0)
2406                                 return err;
2407                 }
2408         }
2409
2410         return 0;
2411 }
2412
2413 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2414 {
2415         struct net *net = sock_net(skb->sk);
2416         const struct rtnl_link_ops *ops;
2417         const struct rtnl_link_ops *m_ops = NULL;
2418         struct net_device *dev;
2419         struct net_device *master_dev = NULL;
2420         struct ifinfomsg *ifm;
2421         char kind[MODULE_NAME_LEN];
2422         char ifname[IFNAMSIZ];
2423         struct nlattr *tb[IFLA_MAX+1];
2424         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2425         unsigned char name_assign_type = NET_NAME_USER;
2426         int err;
2427
2428 #ifdef CONFIG_MODULES
2429 replay:
2430 #endif
2431         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2432         if (err < 0)
2433                 return err;
2434
2435         if (tb[IFLA_IFNAME])
2436                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2437         else
2438                 ifname[0] = '\0';
2439
2440         ifm = nlmsg_data(nlh);
2441         if (ifm->ifi_index > 0)
2442                 dev = __dev_get_by_index(net, ifm->ifi_index);
2443         else {
2444                 if (ifname[0])
2445                         dev = __dev_get_by_name(net, ifname);
2446                 else
2447                         dev = NULL;
2448         }
2449
2450         if (dev) {
2451                 master_dev = netdev_master_upper_dev_get(dev);
2452                 if (master_dev)
2453                         m_ops = master_dev->rtnl_link_ops;
2454         }
2455
2456         err = validate_linkmsg(dev, tb);
2457         if (err < 0)
2458                 return err;
2459
2460         if (tb[IFLA_LINKINFO]) {
2461                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2462                                        tb[IFLA_LINKINFO], ifla_info_policy);
2463                 if (err < 0)
2464                         return err;
2465         } else
2466                 memset(linkinfo, 0, sizeof(linkinfo));
2467
2468         if (linkinfo[IFLA_INFO_KIND]) {
2469                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2470                 ops = rtnl_link_ops_get(kind);
2471         } else {
2472                 kind[0] = '\0';
2473                 ops = NULL;
2474         }
2475
2476         if (1) {
2477                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2478                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2479                 struct nlattr **data = NULL;
2480                 struct nlattr **slave_data = NULL;
2481                 struct net *dest_net, *link_net = NULL;
2482
2483                 if (ops) {
2484                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2485                                 err = nla_parse_nested(attr, ops->maxtype,
2486                                                        linkinfo[IFLA_INFO_DATA],
2487                                                        ops->policy);
2488                                 if (err < 0)
2489                                         return err;
2490                                 data = attr;
2491                         }
2492                         if (ops->validate) {
2493                                 err = ops->validate(tb, data);
2494                                 if (err < 0)
2495                                         return err;
2496                         }
2497                 }
2498
2499                 if (m_ops) {
2500                         if (m_ops->slave_maxtype &&
2501                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2502                                 err = nla_parse_nested(slave_attr,
2503                                                        m_ops->slave_maxtype,
2504                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2505                                                        m_ops->slave_policy);
2506                                 if (err < 0)
2507                                         return err;
2508                                 slave_data = slave_attr;
2509                         }
2510                         if (m_ops->slave_validate) {
2511                                 err = m_ops->slave_validate(tb, slave_data);
2512                                 if (err < 0)
2513                                         return err;
2514                         }
2515                 }
2516
2517                 if (dev) {
2518                         int status = 0;
2519
2520                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2521                                 return -EEXIST;
2522                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2523                                 return -EOPNOTSUPP;
2524
2525                         if (linkinfo[IFLA_INFO_DATA]) {
2526                                 if (!ops || ops != dev->rtnl_link_ops ||
2527                                     !ops->changelink)
2528                                         return -EOPNOTSUPP;
2529
2530                                 err = ops->changelink(dev, tb, data);
2531                                 if (err < 0)
2532                                         return err;
2533                                 status |= DO_SETLINK_NOTIFY;
2534                         }
2535
2536                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2537                                 if (!m_ops || !m_ops->slave_changelink)
2538                                         return -EOPNOTSUPP;
2539
2540                                 err = m_ops->slave_changelink(master_dev, dev,
2541                                                               tb, slave_data);
2542                                 if (err < 0)
2543                                         return err;
2544                                 status |= DO_SETLINK_NOTIFY;
2545                         }
2546
2547                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2548                 }
2549
2550                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2551                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2552                                 return rtnl_group_changelink(skb, net,
2553                                                 nla_get_u32(tb[IFLA_GROUP]),
2554                                                 ifm, tb);
2555                         return -ENODEV;
2556                 }
2557
2558                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2559                         return -EOPNOTSUPP;
2560
2561                 if (!ops) {
2562 #ifdef CONFIG_MODULES
2563                         if (kind[0]) {
2564                                 __rtnl_unlock();
2565                                 request_module("rtnl-link-%s", kind);
2566                                 rtnl_lock();
2567                                 ops = rtnl_link_ops_get(kind);
2568                                 if (ops)
2569                                         goto replay;
2570                         }
2571 #endif
2572                         return -EOPNOTSUPP;
2573                 }
2574
2575                 if (!ops->setup)
2576                         return -EOPNOTSUPP;
2577
2578                 if (!ifname[0]) {
2579                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2580                         name_assign_type = NET_NAME_ENUM;
2581                 }
2582
2583                 dest_net = rtnl_link_get_net(net, tb);
2584                 if (IS_ERR(dest_net))
2585                         return PTR_ERR(dest_net);
2586
2587                 err = -EPERM;
2588                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2589                         goto out;
2590
2591                 if (tb[IFLA_LINK_NETNSID]) {
2592                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2593
2594                         link_net = get_net_ns_by_id(dest_net, id);
2595                         if (!link_net) {
2596                                 err =  -EINVAL;
2597                                 goto out;
2598                         }
2599                         err = -EPERM;
2600                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2601                                 goto out;
2602                 }
2603
2604                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2605                                        name_assign_type, ops, tb);
2606                 if (IS_ERR(dev)) {
2607                         err = PTR_ERR(dev);
2608                         goto out;
2609                 }
2610
2611                 dev->ifindex = ifm->ifi_index;
2612
2613                 if (ops->newlink) {
2614                         err = ops->newlink(link_net ? : net, dev, tb, data);
2615                         /* Drivers should call free_netdev() in ->destructor
2616                          * and unregister it on failure after registration
2617                          * so that device could be finally freed in rtnl_unlock.
2618                          */
2619                         if (err < 0) {
2620                                 /* If device is not registered at all, free it now */
2621                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2622                                         free_netdev(dev);
2623                                 goto out;
2624                         }
2625                 } else {
2626                         err = register_netdevice(dev);
2627                         if (err < 0) {
2628                                 free_netdev(dev);
2629                                 goto out;
2630                         }
2631                 }
2632                 err = rtnl_configure_link(dev, ifm);
2633                 if (err < 0)
2634                         goto out_unregister;
2635                 if (link_net) {
2636                         err = dev_change_net_namespace(dev, dest_net, ifname);
2637                         if (err < 0)
2638                                 goto out_unregister;
2639                 }
2640 out:
2641                 if (link_net)
2642                         put_net(link_net);
2643                 put_net(dest_net);
2644                 return err;
2645 out_unregister:
2646                 if (ops->newlink) {
2647                         LIST_HEAD(list_kill);
2648
2649                         ops->dellink(dev, &list_kill);
2650                         unregister_netdevice_many(&list_kill);
2651                 } else {
2652                         unregister_netdevice(dev);
2653                 }
2654                 goto out;
2655         }
2656 }
2657
2658 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2659 {
2660         struct net *net = sock_net(skb->sk);
2661         struct ifinfomsg *ifm;
2662         char ifname[IFNAMSIZ];
2663         struct nlattr *tb[IFLA_MAX+1];
2664         struct net_device *dev = NULL;
2665         struct sk_buff *nskb;
2666         int err;
2667         u32 ext_filter_mask = 0;
2668
2669         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2670         if (err < 0)
2671                 return err;
2672
2673         if (tb[IFLA_IFNAME])
2674                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2675
2676         if (tb[IFLA_EXT_MASK])
2677                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2678
2679         ifm = nlmsg_data(nlh);
2680         if (ifm->ifi_index > 0)
2681                 dev = __dev_get_by_index(net, ifm->ifi_index);
2682         else if (tb[IFLA_IFNAME])
2683                 dev = __dev_get_by_name(net, ifname);
2684         else
2685                 return -EINVAL;
2686
2687         if (dev == NULL)
2688                 return -ENODEV;
2689
2690         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2691         if (nskb == NULL)
2692                 return -ENOBUFS;
2693
2694         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2695                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2696         if (err < 0) {
2697                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2698                 WARN_ON(err == -EMSGSIZE);
2699                 kfree_skb(nskb);
2700         } else
2701                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2702
2703         return err;
2704 }
2705
2706 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2707 {
2708         struct net *net = sock_net(skb->sk);
2709         struct net_device *dev;
2710         struct nlattr *tb[IFLA_MAX+1];
2711         u32 ext_filter_mask = 0;
2712         u16 min_ifinfo_dump_size = 0;
2713         int hdrlen;
2714
2715         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2716         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2717                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2718
2719         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2720                 if (tb[IFLA_EXT_MASK])
2721                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2722         }
2723
2724         if (!ext_filter_mask)
2725                 return NLMSG_GOODSIZE;
2726         /*
2727          * traverse the list of net devices and compute the minimum
2728          * buffer size based upon the filter mask.
2729          */
2730         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2731                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2732                                              if_nlmsg_size(dev,
2733                                                            ext_filter_mask));
2734         }
2735
2736         return min_ifinfo_dump_size;
2737 }
2738
2739 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2740 {
2741         int idx;
2742         int s_idx = cb->family;
2743
2744         if (s_idx == 0)
2745                 s_idx = 1;
2746         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2747                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2748                 if (idx < s_idx || idx == PF_PACKET)
2749                         continue;
2750                 if (rtnl_msg_handlers[idx] == NULL ||
2751                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2752                         continue;
2753                 if (idx > s_idx) {
2754                         memset(&cb->args[0], 0, sizeof(cb->args));
2755                         cb->prev_seq = 0;
2756                         cb->seq = 0;
2757                 }
2758                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2759                         break;
2760         }
2761         cb->family = idx;
2762
2763         return skb->len;
2764 }
2765
2766 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2767                                        unsigned int change, gfp_t flags)
2768 {
2769         struct net *net = dev_net(dev);
2770         struct sk_buff *skb;
2771         int err = -ENOBUFS;
2772         size_t if_info_size;
2773
2774         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2775         if (skb == NULL)
2776                 goto errout;
2777
2778         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2779         if (err < 0) {
2780                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2781                 WARN_ON(err == -EMSGSIZE);
2782                 kfree_skb(skb);
2783                 goto errout;
2784         }
2785         return skb;
2786 errout:
2787         if (err < 0)
2788                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2789         return NULL;
2790 }
2791
2792 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2793 {
2794         struct net *net = dev_net(dev);
2795
2796         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2797 }
2798
2799 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2800                   gfp_t flags)
2801 {
2802         struct sk_buff *skb;
2803
2804         if (dev->reg_state != NETREG_REGISTERED)
2805                 return;
2806
2807         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2808         if (skb)
2809                 rtmsg_ifinfo_send(skb, dev, flags);
2810 }
2811 EXPORT_SYMBOL(rtmsg_ifinfo);
2812
2813 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2814                                    struct net_device *dev,
2815                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2816                                    int type, unsigned int flags,
2817                                    int nlflags, u16 ndm_state)
2818 {
2819         struct nlmsghdr *nlh;
2820         struct ndmsg *ndm;
2821
2822         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2823         if (!nlh)
2824                 return -EMSGSIZE;
2825
2826         ndm = nlmsg_data(nlh);
2827         ndm->ndm_family  = AF_BRIDGE;
2828         ndm->ndm_pad1    = 0;
2829         ndm->ndm_pad2    = 0;
2830         ndm->ndm_flags   = flags;
2831         ndm->ndm_type    = 0;
2832         ndm->ndm_ifindex = dev->ifindex;
2833         ndm->ndm_state   = ndm_state;
2834
2835         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2836                 goto nla_put_failure;
2837         if (vid)
2838                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2839                         goto nla_put_failure;
2840
2841         nlmsg_end(skb, nlh);
2842         return 0;
2843
2844 nla_put_failure:
2845         nlmsg_cancel(skb, nlh);
2846         return -EMSGSIZE;
2847 }
2848
2849 static inline size_t rtnl_fdb_nlmsg_size(void)
2850 {
2851         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2852 }
2853
2854 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2855                             u16 ndm_state)
2856 {
2857         struct net *net = dev_net(dev);
2858         struct sk_buff *skb;
2859         int err = -ENOBUFS;
2860
2861         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2862         if (!skb)
2863                 goto errout;
2864
2865         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2866                                       0, 0, type, NTF_SELF, 0, ndm_state);
2867         if (err < 0) {
2868                 kfree_skb(skb);
2869                 goto errout;
2870         }
2871
2872         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2873         return;
2874 errout:
2875         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2876 }
2877
2878 /**
2879  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2880  */
2881 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2882                      struct nlattr *tb[],
2883                      struct net_device *dev,
2884                      const unsigned char *addr, u16 vid,
2885                      u16 flags)
2886 {
2887         int err = -EINVAL;
2888
2889         /* If aging addresses are supported device will need to
2890          * implement its own handler for this.
2891          */
2892         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2893                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2894                 return err;
2895         }
2896
2897         if (vid) {
2898                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2899                 return err;
2900         }
2901
2902         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2903                 err = dev_uc_add_excl(dev, addr);
2904         else if (is_multicast_ether_addr(addr))
2905                 err = dev_mc_add_excl(dev, addr);
2906
2907         /* Only return duplicate errors if NLM_F_EXCL is set */
2908         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2909                 err = 0;
2910
2911         return err;
2912 }
2913 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2914
2915 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2916 {
2917         u16 vid = 0;
2918
2919         if (vlan_attr) {
2920                 if (nla_len(vlan_attr) != sizeof(u16)) {
2921                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2922                         return -EINVAL;
2923                 }
2924
2925                 vid = nla_get_u16(vlan_attr);
2926
2927                 if (!vid || vid >= VLAN_VID_MASK) {
2928                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2929                                 vid);
2930                         return -EINVAL;
2931                 }
2932         }
2933         *p_vid = vid;
2934         return 0;
2935 }
2936
2937 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2938 {
2939         struct net *net = sock_net(skb->sk);
2940         struct ndmsg *ndm;
2941         struct nlattr *tb[NDA_MAX+1];
2942         struct net_device *dev;
2943         u8 *addr;
2944         u16 vid;
2945         int err;
2946
2947         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2948         if (err < 0)
2949                 return err;
2950
2951         ndm = nlmsg_data(nlh);
2952         if (ndm->ndm_ifindex == 0) {
2953                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2954                 return -EINVAL;
2955         }
2956
2957         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2958         if (dev == NULL) {
2959                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2960                 return -ENODEV;
2961         }
2962
2963         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2964                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2965                 return -EINVAL;
2966         }
2967
2968         addr = nla_data(tb[NDA_LLADDR]);
2969
2970         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2971         if (err)
2972                 return err;
2973
2974         err = -EOPNOTSUPP;
2975
2976         /* Support fdb on master device the net/bridge default case */
2977         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2978             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2979                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2980                 const struct net_device_ops *ops = br_dev->netdev_ops;
2981
2982                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2983                                        nlh->nlmsg_flags);
2984                 if (err)
2985                         goto out;
2986                 else
2987                         ndm->ndm_flags &= ~NTF_MASTER;
2988         }
2989
2990         /* Embedded bridge, macvlan, and any other device support */
2991         if ((ndm->ndm_flags & NTF_SELF)) {
2992                 if (dev->netdev_ops->ndo_fdb_add)
2993                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2994                                                            vid,
2995                                                            nlh->nlmsg_flags);
2996                 else
2997                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2998                                                nlh->nlmsg_flags);
2999
3000                 if (!err) {
3001                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3002                                         ndm->ndm_state);
3003                         ndm->ndm_flags &= ~NTF_SELF;
3004                 }
3005         }
3006 out:
3007         return err;
3008 }
3009
3010 /**
3011  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3012  */
3013 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3014                      struct nlattr *tb[],
3015                      struct net_device *dev,
3016                      const unsigned char *addr, u16 vid)
3017 {
3018         int err = -EINVAL;
3019
3020         /* If aging addresses are supported device will need to
3021          * implement its own handler for this.
3022          */
3023         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3024                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3025                 return err;
3026         }
3027
3028         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3029                 err = dev_uc_del(dev, addr);
3030         else if (is_multicast_ether_addr(addr))
3031                 err = dev_mc_del(dev, addr);
3032
3033         return err;
3034 }
3035 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3036
3037 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3038 {
3039         struct net *net = sock_net(skb->sk);
3040         struct ndmsg *ndm;
3041         struct nlattr *tb[NDA_MAX+1];
3042         struct net_device *dev;
3043         int err = -EINVAL;
3044         __u8 *addr;
3045         u16 vid;
3046
3047         if (!netlink_capable(skb, CAP_NET_ADMIN))
3048                 return -EPERM;
3049
3050         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3051         if (err < 0)
3052                 return err;
3053
3054         ndm = nlmsg_data(nlh);
3055         if (ndm->ndm_ifindex == 0) {
3056                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3057                 return -EINVAL;
3058         }
3059
3060         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3061         if (dev == NULL) {
3062                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3063                 return -ENODEV;
3064         }
3065
3066         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3067                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3068                 return -EINVAL;
3069         }
3070
3071         addr = nla_data(tb[NDA_LLADDR]);
3072
3073         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3074         if (err)
3075                 return err;
3076
3077         err = -EOPNOTSUPP;
3078
3079         /* Support fdb on master device the net/bridge default case */
3080         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3081             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3082                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3083                 const struct net_device_ops *ops = br_dev->netdev_ops;
3084
3085                 if (ops->ndo_fdb_del)
3086                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3087
3088                 if (err)
3089                         goto out;
3090                 else
3091                         ndm->ndm_flags &= ~NTF_MASTER;
3092         }
3093
3094         /* Embedded bridge, macvlan, and any other device support */
3095         if (ndm->ndm_flags & NTF_SELF) {
3096                 if (dev->netdev_ops->ndo_fdb_del)
3097                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3098                                                            vid);
3099                 else
3100                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3101
3102                 if (!err) {
3103                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3104                                         ndm->ndm_state);
3105                         ndm->ndm_flags &= ~NTF_SELF;
3106                 }
3107         }
3108 out:
3109         return err;
3110 }
3111
3112 static int nlmsg_populate_fdb(struct sk_buff *skb,
3113                               struct netlink_callback *cb,
3114                               struct net_device *dev,
3115                               int *idx,
3116                               struct netdev_hw_addr_list *list)
3117 {
3118         struct netdev_hw_addr *ha;
3119         int err;
3120         u32 portid, seq;
3121
3122         portid = NETLINK_CB(cb->skb).portid;
3123         seq = cb->nlh->nlmsg_seq;
3124
3125         list_for_each_entry(ha, &list->list, list) {
3126                 if (*idx < cb->args[2])
3127                         goto skip;
3128
3129                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3130                                               portid, seq,
3131                                               RTM_NEWNEIGH, NTF_SELF,
3132                                               NLM_F_MULTI, NUD_PERMANENT);
3133                 if (err < 0)
3134                         return err;
3135 skip:
3136                 *idx += 1;
3137         }
3138         return 0;
3139 }
3140
3141 /**
3142  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3143  * @nlh: netlink message header
3144  * @dev: netdevice
3145  *
3146  * Default netdevice operation to dump the existing unicast address list.
3147  * Returns number of addresses from list put in skb.
3148  */
3149 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3150                       struct netlink_callback *cb,
3151                       struct net_device *dev,
3152                       struct net_device *filter_dev,
3153                       int *idx)
3154 {
3155         int err;
3156
3157         netif_addr_lock_bh(dev);
3158         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3159         if (err)
3160                 goto out;
3161         nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3162 out:
3163         netif_addr_unlock_bh(dev);
3164         return err;
3165 }
3166 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3167
3168 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3169 {
3170         struct net_device *dev;
3171         struct nlattr *tb[IFLA_MAX+1];
3172         struct net_device *br_dev = NULL;
3173         const struct net_device_ops *ops = NULL;
3174         const struct net_device_ops *cops = NULL;
3175         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3176         struct net *net = sock_net(skb->sk);
3177         struct hlist_head *head;
3178         int brport_idx = 0;
3179         int br_idx = 0;
3180         int h, s_h;
3181         int idx = 0, s_idx;
3182         int err = 0;
3183         int fidx = 0;
3184
3185         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3186                         ifla_policy) == 0) {
3187                 if (tb[IFLA_MASTER])
3188                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3189         }
3190
3191         brport_idx = ifm->ifi_index;
3192
3193         if (br_idx) {
3194                 br_dev = __dev_get_by_index(net, br_idx);
3195                 if (!br_dev)
3196                         return -ENODEV;
3197
3198                 ops = br_dev->netdev_ops;
3199         }
3200
3201         s_h = cb->args[0];
3202         s_idx = cb->args[1];
3203
3204         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3205                 idx = 0;
3206                 head = &net->dev_index_head[h];
3207                 hlist_for_each_entry(dev, head, index_hlist) {
3208
3209                         if (brport_idx && (dev->ifindex != brport_idx))
3210                                 continue;
3211
3212                         if (!br_idx) { /* user did not specify a specific bridge */
3213                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3214                                         br_dev = netdev_master_upper_dev_get(dev);
3215                                         cops = br_dev->netdev_ops;
3216                                 }
3217                         } else {
3218                                 if (dev != br_dev &&
3219                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3220                                         continue;
3221
3222                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3223                                     !(dev->priv_flags & IFF_EBRIDGE))
3224                                         continue;
3225                                 cops = ops;
3226                         }
3227
3228                         if (idx < s_idx)
3229                                 goto cont;
3230
3231                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3232                                 if (cops && cops->ndo_fdb_dump) {
3233                                         err = cops->ndo_fdb_dump(skb, cb,
3234                                                                 br_dev, dev,
3235                                                                 &fidx);
3236                                         if (err == -EMSGSIZE)
3237                                                 goto out;
3238                                 }
3239                         }
3240
3241                         if (dev->netdev_ops->ndo_fdb_dump)
3242                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3243                                                                     dev, NULL,
3244                                                                     &fidx);
3245                         else
3246                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3247                                                         &fidx);
3248                         if (err == -EMSGSIZE)
3249                                 goto out;
3250
3251                         cops = NULL;
3252
3253                         /* reset fdb offset to 0 for rest of the interfaces */
3254                         cb->args[2] = 0;
3255                         fidx = 0;
3256 cont:
3257                         idx++;
3258                 }
3259         }
3260
3261 out:
3262         cb->args[0] = h;
3263         cb->args[1] = idx;
3264         cb->args[2] = fidx;
3265
3266         return skb->len;
3267 }
3268
3269 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3270                                unsigned int attrnum, unsigned int flag)
3271 {
3272         if (mask & flag)
3273                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3274         return 0;
3275 }
3276
3277 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3278                             struct net_device *dev, u16 mode,
3279                             u32 flags, u32 mask, int nlflags,
3280                             u32 filter_mask,
3281                             int (*vlan_fill)(struct sk_buff *skb,
3282                                              struct net_device *dev,
3283                                              u32 filter_mask))
3284 {
3285         struct nlmsghdr *nlh;
3286         struct ifinfomsg *ifm;
3287         struct nlattr *br_afspec;
3288         struct nlattr *protinfo;
3289         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3290         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3291         int err = 0;
3292
3293         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3294         if (nlh == NULL)
3295                 return -EMSGSIZE;
3296
3297         ifm = nlmsg_data(nlh);
3298         ifm->ifi_family = AF_BRIDGE;
3299         ifm->__ifi_pad = 0;
3300         ifm->ifi_type = dev->type;
3301         ifm->ifi_index = dev->ifindex;
3302         ifm->ifi_flags = dev_get_flags(dev);
3303         ifm->ifi_change = 0;
3304
3305
3306         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3307             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3308             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3309             (br_dev &&
3310              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3311             (dev->addr_len &&
3312              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3313             (dev->ifindex != dev_get_iflink(dev) &&
3314              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3315                 goto nla_put_failure;
3316
3317         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3318         if (!br_afspec)
3319                 goto nla_put_failure;
3320
3321         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3322                 nla_nest_cancel(skb, br_afspec);
3323                 goto nla_put_failure;
3324         }
3325
3326         if (mode != BRIDGE_MODE_UNDEF) {
3327                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3328                         nla_nest_cancel(skb, br_afspec);
3329                         goto nla_put_failure;
3330                 }
3331         }
3332         if (vlan_fill) {
3333                 err = vlan_fill(skb, dev, filter_mask);
3334                 if (err) {
3335                         nla_nest_cancel(skb, br_afspec);
3336                         goto nla_put_failure;
3337                 }
3338         }
3339         nla_nest_end(skb, br_afspec);
3340
3341         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3342         if (!protinfo)
3343                 goto nla_put_failure;
3344
3345         if (brport_nla_put_flag(skb, flags, mask,
3346                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3347             brport_nla_put_flag(skb, flags, mask,
3348                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3349             brport_nla_put_flag(skb, flags, mask,
3350                                 IFLA_BRPORT_FAST_LEAVE,
3351                                 BR_MULTICAST_FAST_LEAVE) ||
3352             brport_nla_put_flag(skb, flags, mask,
3353                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3354             brport_nla_put_flag(skb, flags, mask,
3355                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3356             brport_nla_put_flag(skb, flags, mask,
3357                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3358             brport_nla_put_flag(skb, flags, mask,
3359                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3360             brport_nla_put_flag(skb, flags, mask,
3361                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3362                 nla_nest_cancel(skb, protinfo);
3363                 goto nla_put_failure;
3364         }
3365
3366         nla_nest_end(skb, protinfo);
3367
3368         nlmsg_end(skb, nlh);
3369         return 0;
3370 nla_put_failure:
3371         nlmsg_cancel(skb, nlh);
3372         return err ? err : -EMSGSIZE;
3373 }
3374 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3375
3376 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3377 {
3378         struct net *net = sock_net(skb->sk);
3379         struct net_device *dev;
3380         int idx = 0;
3381         u32 portid = NETLINK_CB(cb->skb).portid;
3382         u32 seq = cb->nlh->nlmsg_seq;
3383         u32 filter_mask = 0;
3384         int err;
3385
3386         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3387                 struct nlattr *extfilt;
3388
3389                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3390                                           IFLA_EXT_MASK);
3391                 if (extfilt) {
3392                         if (nla_len(extfilt) < sizeof(filter_mask))
3393                                 return -EINVAL;
3394
3395                         filter_mask = nla_get_u32(extfilt);
3396                 }
3397         }
3398
3399         rcu_read_lock();
3400         for_each_netdev_rcu(net, dev) {
3401                 const struct net_device_ops *ops = dev->netdev_ops;
3402                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3403
3404                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3405                         if (idx >= cb->args[0]) {
3406                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3407                                                 skb, portid, seq, dev,
3408                                                 filter_mask, NLM_F_MULTI);
3409                                 if (err < 0 && err != -EOPNOTSUPP)
3410                                         break;
3411                         }
3412                         idx++;
3413                 }
3414
3415                 if (ops->ndo_bridge_getlink) {
3416                         if (idx >= cb->args[0]) {
3417                                 err = ops->ndo_bridge_getlink(skb, portid,
3418                                                               seq, dev,
3419                                                               filter_mask,
3420                                                               NLM_F_MULTI);
3421                                 if (err < 0 && err != -EOPNOTSUPP)
3422                                         break;
3423                         }
3424                         idx++;
3425                 }
3426         }
3427         rcu_read_unlock();
3428         cb->args[0] = idx;
3429
3430         return skb->len;
3431 }
3432
3433 static inline size_t bridge_nlmsg_size(void)
3434 {
3435         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3436                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3437                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3438                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3439                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3440                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3441                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3442                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3443                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3444                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3445                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3446 }
3447
3448 static int rtnl_bridge_notify(struct net_device *dev)
3449 {
3450         struct net *net = dev_net(dev);
3451         struct sk_buff *skb;
3452         int err = -EOPNOTSUPP;
3453
3454         if (!dev->netdev_ops->ndo_bridge_getlink)
3455                 return 0;
3456
3457         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3458         if (!skb) {
3459                 err = -ENOMEM;
3460                 goto errout;
3461         }
3462
3463         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3464         if (err < 0)
3465                 goto errout;
3466
3467         if (!skb->len)
3468                 goto errout;
3469
3470         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3471         return 0;
3472 errout:
3473         WARN_ON(err == -EMSGSIZE);
3474         kfree_skb(skb);
3475         if (err)
3476                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3477         return err;
3478 }
3479
3480 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3481 {
3482         struct net *net = sock_net(skb->sk);
3483         struct ifinfomsg *ifm;
3484         struct net_device *dev;
3485         struct nlattr *br_spec, *attr = NULL;
3486         int rem, err = -EOPNOTSUPP;
3487         u16 flags = 0;
3488         bool have_flags = false;
3489
3490         if (nlmsg_len(nlh) < sizeof(*ifm))
3491                 return -EINVAL;
3492
3493         ifm = nlmsg_data(nlh);
3494         if (ifm->ifi_family != AF_BRIDGE)
3495                 return -EPFNOSUPPORT;
3496
3497         dev = __dev_get_by_index(net, ifm->ifi_index);
3498         if (!dev) {
3499                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3500                 return -ENODEV;
3501         }
3502
3503         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3504         if (br_spec) {
3505                 nla_for_each_nested(attr, br_spec, rem) {
3506                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3507                                 if (nla_len(attr) < sizeof(flags))
3508                                         return -EINVAL;
3509
3510                                 have_flags = true;
3511                                 flags = nla_get_u16(attr);
3512                                 break;
3513                         }
3514                 }
3515         }
3516
3517         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3518                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3519
3520                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3521                         err = -EOPNOTSUPP;
3522                         goto out;
3523                 }
3524
3525                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3526                 if (err)
3527                         goto out;
3528
3529                 flags &= ~BRIDGE_FLAGS_MASTER;
3530         }
3531
3532         if ((flags & BRIDGE_FLAGS_SELF)) {
3533                 if (!dev->netdev_ops->ndo_bridge_setlink)
3534                         err = -EOPNOTSUPP;
3535                 else
3536                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3537                                                                   flags);
3538                 if (!err) {
3539                         flags &= ~BRIDGE_FLAGS_SELF;
3540
3541                         /* Generate event to notify upper layer of bridge
3542                          * change
3543                          */
3544                         err = rtnl_bridge_notify(dev);
3545                 }
3546         }
3547
3548         if (have_flags)
3549                 memcpy(nla_data(attr), &flags, sizeof(flags));
3550 out:
3551         return err;
3552 }
3553
3554 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3555 {
3556         struct net *net = sock_net(skb->sk);
3557         struct ifinfomsg *ifm;
3558         struct net_device *dev;
3559         struct nlattr *br_spec, *attr = NULL;
3560         int rem, err = -EOPNOTSUPP;
3561         u16 flags = 0;
3562         bool have_flags = false;
3563
3564         if (nlmsg_len(nlh) < sizeof(*ifm))
3565                 return -EINVAL;
3566
3567         ifm = nlmsg_data(nlh);
3568         if (ifm->ifi_family != AF_BRIDGE)
3569                 return -EPFNOSUPPORT;
3570
3571         dev = __dev_get_by_index(net, ifm->ifi_index);
3572         if (!dev) {
3573                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3574                 return -ENODEV;
3575         }
3576
3577         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3578         if (br_spec) {
3579                 nla_for_each_nested(attr, br_spec, rem) {
3580                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3581                                 if (nla_len(attr) < sizeof(flags))
3582                                         return -EINVAL;
3583
3584                                 have_flags = true;
3585                                 flags = nla_get_u16(attr);
3586                                 break;
3587                         }
3588                 }
3589         }
3590
3591         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3592                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3593
3594                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3595                         err = -EOPNOTSUPP;
3596                         goto out;
3597                 }
3598
3599                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3600                 if (err)
3601                         goto out;
3602
3603                 flags &= ~BRIDGE_FLAGS_MASTER;
3604         }
3605
3606         if ((flags & BRIDGE_FLAGS_SELF)) {
3607                 if (!dev->netdev_ops->ndo_bridge_dellink)
3608                         err = -EOPNOTSUPP;
3609                 else
3610                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3611                                                                   flags);
3612
3613                 if (!err) {
3614                         flags &= ~BRIDGE_FLAGS_SELF;
3615
3616                         /* Generate event to notify upper layer of bridge
3617                          * change
3618                          */
3619                         err = rtnl_bridge_notify(dev);
3620                 }
3621         }
3622
3623         if (have_flags)
3624                 memcpy(nla_data(attr), &flags, sizeof(flags));
3625 out:
3626         return err;
3627 }
3628
3629 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3630 {
3631         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3632                (!idxattr || idxattr == attrid);
3633 }
3634
3635 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3636 static int rtnl_get_offload_stats_attr_size(int attr_id)
3637 {
3638         switch (attr_id) {
3639         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3640                 return sizeof(struct rtnl_link_stats64);
3641         }
3642
3643         return 0;
3644 }
3645
3646 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3647                                   int *prividx)
3648 {
3649         struct nlattr *attr = NULL;
3650         int attr_id, size;
3651         void *attr_data;
3652         int err;
3653
3654         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3655               dev->netdev_ops->ndo_get_offload_stats))
3656                 return -ENODATA;
3657
3658         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3659              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3660                 if (attr_id < *prividx)
3661                         continue;
3662
3663                 size = rtnl_get_offload_stats_attr_size(attr_id);
3664                 if (!size)
3665                         continue;
3666
3667                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3668                         continue;
3669
3670                 attr = nla_reserve_64bit(skb, attr_id, size,
3671                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3672                 if (!attr)
3673                         goto nla_put_failure;
3674
3675                 attr_data = nla_data(attr);
3676                 memset(attr_data, 0, size);
3677                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3678                                                              attr_data);
3679                 if (err)
3680                         goto get_offload_stats_failure;
3681         }
3682
3683         if (!attr)
3684                 return -ENODATA;
3685
3686         *prividx = 0;
3687         return 0;
3688
3689 nla_put_failure:
3690         err = -EMSGSIZE;
3691 get_offload_stats_failure:
3692         *prividx = attr_id;
3693         return err;
3694 }
3695
3696 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3697 {
3698         int nla_size = 0;
3699         int attr_id;
3700         int size;
3701
3702         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3703               dev->netdev_ops->ndo_get_offload_stats))
3704                 return 0;
3705
3706         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3707              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3708                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3709                         continue;
3710                 size = rtnl_get_offload_stats_attr_size(attr_id);
3711                 nla_size += nla_total_size_64bit(size);
3712         }
3713
3714         if (nla_size != 0)
3715                 nla_size += nla_total_size(0);
3716
3717         return nla_size;
3718 }
3719
3720 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3721                                int type, u32 pid, u32 seq, u32 change,
3722                                unsigned int flags, unsigned int filter_mask,
3723                                int *idxattr, int *prividx)
3724 {
3725         struct if_stats_msg *ifsm;
3726         struct nlmsghdr *nlh;
3727         struct nlattr *attr;
3728         int s_prividx = *prividx;
3729         int err;
3730
3731         ASSERT_RTNL();
3732
3733         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3734         if (!nlh)
3735                 return -EMSGSIZE;
3736
3737         ifsm = nlmsg_data(nlh);
3738         ifsm->ifindex = dev->ifindex;
3739         ifsm->filter_mask = filter_mask;
3740
3741         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3742                 struct rtnl_link_stats64 *sp;
3743
3744                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3745                                          sizeof(struct rtnl_link_stats64),
3746                                          IFLA_STATS_UNSPEC);
3747                 if (!attr)
3748                         goto nla_put_failure;
3749
3750                 sp = nla_data(attr);
3751                 dev_get_stats(dev, sp);
3752         }
3753
3754         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3755                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3756
3757                 if (ops && ops->fill_linkxstats) {
3758                         *idxattr = IFLA_STATS_LINK_XSTATS;
3759                         attr = nla_nest_start(skb,
3760                                               IFLA_STATS_LINK_XSTATS);
3761                         if (!attr)
3762                                 goto nla_put_failure;
3763
3764                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3765                         nla_nest_end(skb, attr);
3766                         if (err)
3767                                 goto nla_put_failure;
3768                         *idxattr = 0;
3769                 }
3770         }
3771
3772         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3773                              *idxattr)) {
3774                 const struct rtnl_link_ops *ops = NULL;
3775                 const struct net_device *master;
3776
3777                 master = netdev_master_upper_dev_get(dev);
3778                 if (master)
3779                         ops = master->rtnl_link_ops;
3780                 if (ops && ops->fill_linkxstats) {
3781                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3782                         attr = nla_nest_start(skb,
3783                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3784                         if (!attr)
3785                                 goto nla_put_failure;
3786
3787                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3788                         nla_nest_end(skb, attr);
3789                         if (err)
3790                                 goto nla_put_failure;
3791                         *idxattr = 0;
3792                 }
3793         }
3794
3795         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3796                              *idxattr)) {
3797                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3798                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3799                 if (!attr)
3800                         goto nla_put_failure;
3801
3802                 err = rtnl_get_offload_stats(skb, dev, prividx);
3803                 if (err == -ENODATA)
3804                         nla_nest_cancel(skb, attr);
3805                 else
3806                         nla_nest_end(skb, attr);
3807
3808                 if (err && err != -ENODATA)
3809                         goto nla_put_failure;
3810                 *idxattr = 0;
3811         }
3812
3813         nlmsg_end(skb, nlh);
3814
3815         return 0;
3816
3817 nla_put_failure:
3818         /* not a multi message or no progress mean a real error */
3819         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3820                 nlmsg_cancel(skb, nlh);
3821         else
3822                 nlmsg_end(skb, nlh);
3823
3824         return -EMSGSIZE;
3825 }
3826
3827 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3828                                   u32 filter_mask)
3829 {
3830         size_t size = 0;
3831
3832         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3833                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3834
3835         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3836                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3837                 int attr = IFLA_STATS_LINK_XSTATS;
3838
3839                 if (ops && ops->get_linkxstats_size) {
3840                         size += nla_total_size(ops->get_linkxstats_size(dev,
3841                                                                         attr));
3842                         /* for IFLA_STATS_LINK_XSTATS */
3843                         size += nla_total_size(0);
3844                 }
3845         }
3846
3847         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3848                 struct net_device *_dev = (struct net_device *)dev;
3849                 const struct rtnl_link_ops *ops = NULL;
3850                 const struct net_device *master;
3851
3852                 /* netdev_master_upper_dev_get can't take const */
3853                 master = netdev_master_upper_dev_get(_dev);
3854                 if (master)
3855                         ops = master->rtnl_link_ops;
3856                 if (ops && ops->get_linkxstats_size) {
3857                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3858
3859                         size += nla_total_size(ops->get_linkxstats_size(dev,
3860                                                                         attr));
3861                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3862                         size += nla_total_size(0);
3863                 }
3864         }
3865
3866         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3867                 size += rtnl_get_offload_stats_size(dev);
3868
3869         return size;
3870 }
3871
3872 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3873 {
3874         struct net *net = sock_net(skb->sk);
3875         struct net_device *dev = NULL;
3876         int idxattr = 0, prividx = 0;
3877         struct if_stats_msg *ifsm;
3878         struct sk_buff *nskb;
3879         u32 filter_mask;
3880         int err;
3881
3882         ifsm = nlmsg_data(nlh);
3883         if (ifsm->ifindex > 0)
3884                 dev = __dev_get_by_index(net, ifsm->ifindex);
3885         else
3886                 return -EINVAL;
3887
3888         if (!dev)
3889                 return -ENODEV;
3890
3891         filter_mask = ifsm->filter_mask;
3892         if (!filter_mask)
3893                 return -EINVAL;
3894
3895         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3896         if (!nskb)
3897                 return -ENOBUFS;
3898
3899         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3900                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3901                                   0, filter_mask, &idxattr, &prividx);
3902         if (err < 0) {
3903                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3904                 WARN_ON(err == -EMSGSIZE);
3905                 kfree_skb(nskb);
3906         } else {
3907                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3908         }
3909
3910         return err;
3911 }
3912
3913 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3914 {
3915         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3916         struct net *net = sock_net(skb->sk);
3917         unsigned int flags = NLM_F_MULTI;
3918         struct if_stats_msg *ifsm;
3919         struct hlist_head *head;
3920         struct net_device *dev;
3921         u32 filter_mask = 0;
3922         int idx = 0;
3923
3924         s_h = cb->args[0];
3925         s_idx = cb->args[1];
3926         s_idxattr = cb->args[2];
3927         s_prividx = cb->args[3];
3928
3929         cb->seq = net->dev_base_seq;
3930
3931         ifsm = nlmsg_data(cb->nlh);
3932         filter_mask = ifsm->filter_mask;
3933         if (!filter_mask)
3934                 return -EINVAL;
3935
3936         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3937                 idx = 0;
3938                 head = &net->dev_index_head[h];
3939                 hlist_for_each_entry(dev, head, index_hlist) {
3940                         if (idx < s_idx)
3941                                 goto cont;
3942                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3943                                                   NETLINK_CB(cb->skb).portid,
3944                                                   cb->nlh->nlmsg_seq, 0,
3945                                                   flags, filter_mask,
3946                                                   &s_idxattr, &s_prividx);
3947                         /* If we ran out of room on the first message,
3948                          * we're in trouble
3949                          */
3950                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3951
3952                         if (err < 0)
3953                                 goto out;
3954                         s_prividx = 0;
3955                         s_idxattr = 0;
3956                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3957 cont:
3958                         idx++;
3959                 }
3960         }
3961 out:
3962         cb->args[3] = s_prividx;
3963         cb->args[2] = s_idxattr;
3964         cb->args[1] = idx;
3965         cb->args[0] = h;
3966
3967         return skb->len;
3968 }
3969
3970 /* Process one rtnetlink message. */
3971
3972 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3973 {
3974         struct net *net = sock_net(skb->sk);
3975         rtnl_doit_func doit;
3976         int kind;
3977         int family;
3978         int type;
3979         int err;
3980
3981         type = nlh->nlmsg_type;
3982         if (type > RTM_MAX)
3983                 return -EOPNOTSUPP;
3984
3985         type -= RTM_BASE;
3986
3987         /* All the messages must have at least 1 byte length */
3988         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3989                 return 0;
3990
3991         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3992         kind = type&3;
3993
3994         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3995                 return -EPERM;
3996
3997         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3998                 struct sock *rtnl;
3999                 rtnl_dumpit_func dumpit;
4000                 rtnl_calcit_func calcit;
4001                 u16 min_dump_alloc = 0;
4002
4003                 dumpit = rtnl_get_dumpit(family, type);
4004                 if (dumpit == NULL)
4005                         return -EOPNOTSUPP;
4006                 calcit = rtnl_get_calcit(family, type);
4007                 if (calcit)
4008                         min_dump_alloc = calcit(skb, nlh);
4009
4010                 __rtnl_unlock();
4011                 rtnl = net->rtnl;
4012                 {
4013                         struct netlink_dump_control c = {
4014                                 .dump           = dumpit,
4015                                 .min_dump_alloc = min_dump_alloc,
4016                         };
4017                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4018                 }
4019                 rtnl_lock();
4020                 return err;
4021         }
4022
4023         doit = rtnl_get_doit(family, type);
4024         if (doit == NULL)
4025                 return -EOPNOTSUPP;
4026
4027         return doit(skb, nlh);
4028 }
4029
4030 static void rtnetlink_rcv(struct sk_buff *skb)
4031 {
4032         rtnl_lock();
4033         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4034         rtnl_unlock();
4035 }
4036
4037 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4038 {
4039         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4040
4041         switch (event) {
4042         case NETDEV_UP:
4043         case NETDEV_DOWN:
4044         case NETDEV_PRE_UP:
4045         case NETDEV_POST_INIT:
4046         case NETDEV_REGISTER:
4047         case NETDEV_CHANGE:
4048         case NETDEV_PRE_TYPE_CHANGE:
4049         case NETDEV_GOING_DOWN:
4050         case NETDEV_UNREGISTER:
4051         case NETDEV_UNREGISTER_FINAL:
4052         case NETDEV_RELEASE:
4053         case NETDEV_JOIN:
4054         case NETDEV_BONDING_INFO:
4055                 break;
4056         default:
4057                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4058                 break;
4059         }
4060         return NOTIFY_DONE;
4061 }
4062
4063 static struct notifier_block rtnetlink_dev_notifier = {
4064         .notifier_call  = rtnetlink_event,
4065 };
4066
4067
4068 static int __net_init rtnetlink_net_init(struct net *net)
4069 {
4070         struct sock *sk;
4071         struct netlink_kernel_cfg cfg = {
4072                 .groups         = RTNLGRP_MAX,
4073                 .input          = rtnetlink_rcv,
4074                 .cb_mutex       = &rtnl_mutex,
4075                 .flags          = NL_CFG_F_NONROOT_RECV,
4076         };
4077
4078         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4079         if (!sk)
4080                 return -ENOMEM;
4081         net->rtnl = sk;
4082         return 0;
4083 }
4084
4085 static void __net_exit rtnetlink_net_exit(struct net *net)
4086 {
4087         netlink_kernel_release(net->rtnl);
4088         net->rtnl = NULL;
4089 }
4090
4091 static struct pernet_operations rtnetlink_net_ops = {
4092         .init = rtnetlink_net_init,
4093         .exit = rtnetlink_net_exit,
4094 };
4095
4096 void __init rtnetlink_init(void)
4097 {
4098         if (register_pernet_subsys(&rtnetlink_net_ops))
4099                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4100
4101         register_netdevice_notifier(&rtnetlink_dev_notifier);
4102
4103         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4104                       rtnl_dump_ifinfo, rtnl_calcit);
4105         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4106         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4107         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4108
4109         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4110         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4111
4112         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4113         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4114         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4115
4116         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4117         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4118         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4119
4120         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4121                       NULL);
4122 }