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 <linux/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         rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
 279
 280         return 0;
 281 }
 282 EXPORT_SYMBOL_GPL(rtnl_unregister);
 283
 284 /**
 285  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 286  * @protocol : Protocol family or PF_UNSPEC
 287  *
 288  * Identical to calling rtnl_unregster() for all registered message types
 289  * of a certain protocol family.
 290  */
 291 void rtnl_unregister_all(int protocol)
 292 {
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294
 295         kfree(rtnl_msg_handlers[protocol]);
 296         rtnl_msg_handlers[protocol] = NULL;
 297 }
 298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 299
 300 static LIST_HEAD(link_ops);
 301
 302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 303 {
 304         const struct rtnl_link_ops *ops;
 305
 306         list_for_each_entry(ops, &link_ops, list) {
 307                 if (!strcmp(ops->kind, kind))
 308                         return ops;
 309         }
 310         return NULL;
 311 }
 312
 313 /**
 314  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 315  * @ops: struct rtnl_link_ops * to register
 316  *
 317  * The caller must hold the rtnl_mutex. This function should be used
 318  * by drivers that create devices during module initialization. It
 319  * must be called before registering the devices.
 320  *
 321  * Returns 0 on success or a negative error code.
 322  */
 323 int __rtnl_link_register(struct rtnl_link_ops *ops)
 324 {
 325         if (rtnl_link_ops_get(ops->kind))
 326                 return -EEXIST;
 327
 328         /* The check for setup is here because if ops
 329          * does not have that filled up, it is not possible
 330          * to use the ops for creating device. So do not
 331          * fill up dellink as well. That disables rtnl_dellink.
 332          */
 333         if (ops->setup && !ops->dellink)
 334                 ops->dellink = unregister_netdevice_queue;
 335
 336         list_add_tail(&ops->list, &link_ops);
 337         return 0;
 338 }
 339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 340
 341 /**
 342  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 343  * @ops: struct rtnl_link_ops * to register
 344  *
 345  * Returns 0 on success or a negative error code.
 346  */
 347 int rtnl_link_register(struct rtnl_link_ops *ops)
 348 {
 349         int err;
 350
 351         rtnl_lock();
 352         err = __rtnl_link_register(ops);
 353         rtnl_unlock();
 354         return err;
 355 }
 356 EXPORT_SYMBOL_GPL(rtnl_link_register);
 357
 358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 359 {
 360         struct net_device *dev;
 361         LIST_HEAD(list_kill);
 362
 363         for_each_netdev(net, dev) {
 364                 if (dev->rtnl_link_ops == ops)
 365                         ops->dellink(dev, &list_kill);
 366         }
 367         unregister_netdevice_many(&list_kill);
 368 }
 369
 370 /**
 371  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 372  * @ops: struct rtnl_link_ops * to unregister
 373  *
 374  * The caller must hold the rtnl_mutex.
 375  */
 376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 377 {
 378         struct net *net;
 379
 380         for_each_net(net) {
 381                 __rtnl_kill_links(net, ops);
 382         }
 383         list_del(&ops->list);
 384 }
 385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 386
 387 /* Return with the rtnl_lock held when there are no network
 388  * devices unregistering in any network namespace.
 389  */
 390 static void rtnl_lock_unregistering_all(void)
 391 {
 392         struct net *net;
 393         bool unregistering;
 394         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 395
 396         add_wait_queue(&netdev_unregistering_wq, &wait);
 397         for (;;) {
 398                 unregistering = false;
 399                 rtnl_lock();
 400                 for_each_net(net) {
 401                         if (net->dev_unreg_count > 0) {
 402                                 unregistering = true;
 403                                 break;
 404                         }
 405                 }
 406                 if (!unregistering)
 407                         break;
 408                 __rtnl_unlock();
 409
 410                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 411         }
 412         remove_wait_queue(&netdev_unregistering_wq, &wait);
 413 }
 414
 415 /**
 416  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 417  * @ops: struct rtnl_link_ops * to unregister
 418  */
 419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 420 {
 421         /* Close the race with cleanup_net() */
 422         mutex_lock(&net_mutex);
 423         rtnl_lock_unregistering_all();
 424         __rtnl_link_unregister(ops);
 425         rtnl_unlock();
 426         mutex_unlock(&net_mutex);
 427 }
 428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 429
 430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 431 {
 432         struct net_device *master_dev;
 433         const struct rtnl_link_ops *ops;
 434
 435         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 436         if (!master_dev)
 437                 return 0;
 438         ops = master_dev->rtnl_link_ops;
 439         if (!ops || !ops->get_slave_size)
 440                 return 0;
 441         /* IFLA_INFO_SLAVE_DATA + nested data */
 442         return nla_total_size(sizeof(struct nlattr)) +
 443                ops->get_slave_size(master_dev, dev);
 444 }
 445
 446 static size_t rtnl_link_get_size(const struct net_device *dev)
 447 {
 448         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 449         size_t size;
 450
 451         if (!ops)
 452                 return 0;
 453
 454         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 455                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 456
 457         if (ops->get_size)
 458                 /* IFLA_INFO_DATA + nested data */
 459                 size += nla_total_size(sizeof(struct nlattr)) +
 460                         ops->get_size(dev);
 461
 462         if (ops->get_xstats_size)
 463                 /* IFLA_INFO_XSTATS */
 464                 size += nla_total_size(ops->get_xstats_size(dev));
 465
 466         size += rtnl_link_get_slave_info_data_size(dev);
 467
 468         return size;
 469 }
 470
 471 static LIST_HEAD(rtnl_af_ops);
 472
 473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 474 {
 475         const struct rtnl_af_ops *ops;
 476
 477         list_for_each_entry(ops, &rtnl_af_ops, list) {
 478                 if (ops->family == family)
 479                         return ops;
 480         }
 481
 482         return NULL;
 483 }
 484
 485 /**
 486  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 487  * @ops: struct rtnl_af_ops * to register
 488  *
 489  * Returns 0 on success or a negative error code.
 490  */
 491 void rtnl_af_register(struct rtnl_af_ops *ops)
 492 {
 493         rtnl_lock();
 494         list_add_tail(&ops->list, &rtnl_af_ops);
 495         rtnl_unlock();
 496 }
 497 EXPORT_SYMBOL_GPL(rtnl_af_register);
 498
 499 /**
 500  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 501  * @ops: struct rtnl_af_ops * to unregister
 502  *
 503  * The caller must hold the rtnl_mutex.
 504  */
 505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 506 {
 507         list_del(&ops->list);
 508 }
 509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 510
 511 /**
 512  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 513  * @ops: struct rtnl_af_ops * to unregister
 514  */
 515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 516 {
 517         rtnl_lock();
 518         __rtnl_af_unregister(ops);
 519         rtnl_unlock();
 520 }
 521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 522
 523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 524                                     u32 ext_filter_mask)
 525 {
 526         struct rtnl_af_ops *af_ops;
 527         size_t size;
 528
 529         /* IFLA_AF_SPEC */
 530         size = nla_total_size(sizeof(struct nlattr));
 531
 532         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 533                 if (af_ops->get_link_af_size) {
 534                         /* AF_* + nested data */
 535                         size += nla_total_size(sizeof(struct nlattr)) +
 536                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 537                 }
 538         }
 539
 540         return size;
 541 }
 542
 543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 544 {
 545         struct net_device *master_dev;
 546
 547         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 548         if (master_dev && master_dev->rtnl_link_ops)
 549                 return true;
 550         return false;
 551 }
 552
 553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 554                                      const struct net_device *dev)
 555 {
 556         struct net_device *master_dev;
 557         const struct rtnl_link_ops *ops;
 558         struct nlattr *slave_data;
 559         int err;
 560
 561         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 562         if (!master_dev)
 563                 return 0;
 564         ops = master_dev->rtnl_link_ops;
 565         if (!ops)
 566                 return 0;
 567         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 568                 return -EMSGSIZE;
 569         if (ops->fill_slave_info) {
 570                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 571                 if (!slave_data)
 572                         return -EMSGSIZE;
 573                 err = ops->fill_slave_info(skb, master_dev, dev);
 574                 if (err < 0)
 575                         goto err_cancel_slave_data;
 576                 nla_nest_end(skb, slave_data);
 577         }
 578         return 0;
 579
 580 err_cancel_slave_data:
 581         nla_nest_cancel(skb, slave_data);
 582         return err;
 583 }
 584
 585 static int rtnl_link_info_fill(struct sk_buff *skb,
 586                                const struct net_device *dev)
 587 {
 588         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 589         struct nlattr *data;
 590         int err;
 591
 592         if (!ops)
 593                 return 0;
 594         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 595                 return -EMSGSIZE;
 596         if (ops->fill_xstats) {
 597                 err = ops->fill_xstats(skb, dev);
 598                 if (err < 0)
 599                         return err;
 600         }
 601         if (ops->fill_info) {
 602                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 603                 if (data == NULL)
 604                         return -EMSGSIZE;
 605                 err = ops->fill_info(skb, dev);
 606                 if (err < 0)
 607                         goto err_cancel_data;
 608                 nla_nest_end(skb, data);
 609         }
 610         return 0;
 611
 612 err_cancel_data:
 613         nla_nest_cancel(skb, data);
 614         return err;
 615 }
 616
 617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 618 {
 619         struct nlattr *linkinfo;
 620         int err = -EMSGSIZE;
 621
 622         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 623         if (linkinfo == NULL)
 624                 goto out;
 625
 626         err = rtnl_link_info_fill(skb, dev);
 627         if (err < 0)
 628                 goto err_cancel_link;
 629
 630         err = rtnl_link_slave_info_fill(skb, dev);
 631         if (err < 0)
 632                 goto err_cancel_link;
 633
 634         nla_nest_end(skb, linkinfo);
 635         return 0;
 636
 637 err_cancel_link:
 638         nla_nest_cancel(skb, linkinfo);
 639 out:
 640         return err;
 641 }
 642
 643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 644 {
 645         struct sock *rtnl = net->rtnl;
 646         int err = 0;
 647
 648         NETLINK_CB(skb).dst_group = group;
 649         if (echo)
 650                 atomic_inc(&skb->users);
 651         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 652         if (echo)
 653                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 654         return err;
 655 }
 656
 657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 658 {
 659         struct sock *rtnl = net->rtnl;
 660
 661         return nlmsg_unicast(rtnl, skb, pid);
 662 }
 663 EXPORT_SYMBOL(rtnl_unicast);
 664
 665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 666                  struct nlmsghdr *nlh, gfp_t flags)
 667 {
 668         struct sock *rtnl = net->rtnl;
 669         int report = 0;
 670
 671         if (nlh)
 672                 report = nlmsg_report(nlh);
 673
 674         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 675 }
 676 EXPORT_SYMBOL(rtnl_notify);
 677
 678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 679 {
 680         struct sock *rtnl = net->rtnl;
 681
 682         netlink_set_err(rtnl, 0, group, error);
 683 }
 684 EXPORT_SYMBOL(rtnl_set_sk_err);
 685
 686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 687 {
 688         struct nlattr *mx;
 689         int i, valid = 0;
 690
 691         mx = nla_nest_start(skb, RTA_METRICS);
 692         if (mx == NULL)
 693                 return -ENOBUFS;
 694
 695         for (i = 0; i < RTAX_MAX; i++) {
 696                 if (metrics[i]) {
 697                         if (i == RTAX_CC_ALGO - 1) {
 698                                 char tmp[TCP_CA_NAME_MAX], *name;
 699
 700                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 701                                 if (!name)
 702                                         continue;
 703                                 if (nla_put_string(skb, i + 1, name))
 704                                         goto nla_put_failure;
 705                         } else if (i == RTAX_FEATURES - 1) {
 706                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 707
 708                                 if (!user_features)
 709                                         continue;
 710                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 711                                 if (nla_put_u32(skb, i + 1, user_features))
 712                                         goto nla_put_failure;
 713                         } else {
 714                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 715                                         goto nla_put_failure;
 716                         }
 717                         valid++;
 718                 }
 719         }
 720
 721         if (!valid) {
 722                 nla_nest_cancel(skb, mx);
 723                 return 0;
 724         }
 725
 726         return nla_nest_end(skb, mx);
 727
 728 nla_put_failure:
 729         nla_nest_cancel(skb, mx);
 730         return -EMSGSIZE;
 731 }
 732 EXPORT_SYMBOL(rtnetlink_put_metrics);
 733
 734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 735                        long expires, u32 error)
 736 {
 737         struct rta_cacheinfo ci = {
 738                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 739                 .rta_used = dst->__use,
 740                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 741                 .rta_error = error,
 742                 .rta_id =  id,
 743         };
 744
 745         if (expires) {
 746                 unsigned long clock;
 747
 748                 clock = jiffies_to_clock_t(abs(expires));
 749                 clock = min_t(unsigned long, clock, INT_MAX);
 750                 ci.rta_expires = (expires > 0) ? clock : -clock;
 751         }
 752         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 753 }
 754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 755
 756 static void set_operstate(struct net_device *dev, unsigned char transition)
 757 {
 758         unsigned char operstate = dev->operstate;
 759
 760         switch (transition) {
 761         case IF_OPER_UP:
 762                 if ((operstate == IF_OPER_DORMANT ||
 763                      operstate == IF_OPER_UNKNOWN) &&
 764                     !netif_dormant(dev))
 765                         operstate = IF_OPER_UP;
 766                 break;
 767
 768         case IF_OPER_DORMANT:
 769                 if (operstate == IF_OPER_UP ||
 770                     operstate == IF_OPER_UNKNOWN)
 771                         operstate = IF_OPER_DORMANT;
 772                 break;
 773         }
 774
 775         if (dev->operstate != operstate) {
 776                 write_lock_bh(&dev_base_lock);
 777                 dev->operstate = operstate;
 778                 write_unlock_bh(&dev_base_lock);
 779                 netdev_state_change(dev);
 780         }
 781 }
 782
 783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 784 {
 785         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 786                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 787 }
 788
 789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 790                                            const struct ifinfomsg *ifm)
 791 {
 792         unsigned int flags = ifm->ifi_flags;
 793
 794         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 795         if (ifm->ifi_change)
 796                 flags = (flags & ifm->ifi_change) |
 797                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 798
 799         return flags;
 800 }
 801
 802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 803                                  const struct rtnl_link_stats64 *b)
 804 {
 805         a->rx_packets = b->rx_packets;
 806         a->tx_packets = b->tx_packets;
 807         a->rx_bytes = b->rx_bytes;
 808         a->tx_bytes = b->tx_bytes;
 809         a->rx_errors = b->rx_errors;
 810         a->tx_errors = b->tx_errors;
 811         a->rx_dropped = b->rx_dropped;
 812         a->tx_dropped = b->tx_dropped;
 813
 814         a->multicast = b->multicast;
 815         a->collisions = b->collisions;
 816
 817         a->rx_length_errors = b->rx_length_errors;
 818         a->rx_over_errors = b->rx_over_errors;
 819         a->rx_crc_errors = b->rx_crc_errors;
 820         a->rx_frame_errors = b->rx_frame_errors;
 821         a->rx_fifo_errors = b->rx_fifo_errors;
 822         a->rx_missed_errors = b->rx_missed_errors;
 823
 824         a->tx_aborted_errors = b->tx_aborted_errors;
 825         a->tx_carrier_errors = b->tx_carrier_errors;
 826         a->tx_fifo_errors = b->tx_fifo_errors;
 827         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 828         a->tx_window_errors = b->tx_window_errors;
 829
 830         a->rx_compressed = b->rx_compressed;
 831         a->tx_compressed = b->tx_compressed;
 832
 833         a->rx_nohandler = b->rx_nohandler;
 834 }
 835
 836 /* All VF info */
 837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 838                                    u32 ext_filter_mask)
 839 {
 840         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 841                 int num_vfs = dev_num_vf(dev->dev.parent);
 842                 size_t size = nla_total_size(0);
 843                 size += num_vfs *
 844                         (nla_total_size(0) +
 845                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 846                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 847                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 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_tx_rate)) +
 852                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 853                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 854                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 855                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 856                          /* IFLA_VF_STATS_RX_PACKETS */
 857                          nla_total_size_64bit(sizeof(__u64)) +
 858                          /* IFLA_VF_STATS_TX_PACKETS */
 859                          nla_total_size_64bit(sizeof(__u64)) +
 860                          /* IFLA_VF_STATS_RX_BYTES */
 861                          nla_total_size_64bit(sizeof(__u64)) +
 862                          /* IFLA_VF_STATS_TX_BYTES */
 863                          nla_total_size_64bit(sizeof(__u64)) +
 864                          /* IFLA_VF_STATS_BROADCAST */
 865                          nla_total_size_64bit(sizeof(__u64)) +
 866                          /* IFLA_VF_STATS_MULTICAST */
 867                          nla_total_size_64bit(sizeof(__u64)) +
 868                          nla_total_size(sizeof(struct ifla_vf_trust)));
 869                 return size;
 870         } else
 871                 return 0;
 872 }
 873
 874 static size_t rtnl_port_size(const struct net_device *dev,
 875                              u32 ext_filter_mask)
 876 {
 877         size_t port_size = nla_total_size(4)            /* PORT_VF */
 878                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 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(0) +   /* nest IFLA_XDP */
 902                           nla_total_size(1);    /* XDP_ATTACHED */
 903
 904         if (!dev->netdev_ops->ndo_xdp)
 905                 return 0;
 906         else
 907                 return xdp_size;
 908 }
 909
 910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 911                                      u32 ext_filter_mask)
 912 {
 913         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 914                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 915                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 916                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 917                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 918                + nla_total_size(sizeof(struct rtnl_link_stats))
 919                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 920                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 921                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 922                + nla_total_size(4) /* IFLA_TXQLEN */
 923                + nla_total_size(4) /* IFLA_WEIGHT */
 924                + nla_total_size(4) /* IFLA_MTU */
 925                + nla_total_size(4) /* IFLA_LINK */
 926                + nla_total_size(4) /* IFLA_MASTER */
 927                + nla_total_size(1) /* IFLA_CARRIER */
 928                + nla_total_size(4) /* IFLA_PROMISCUITY */
 929                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 930                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 931                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 932                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 933                + nla_total_size(1) /* IFLA_OPERSTATE */
 934                + nla_total_size(1) /* IFLA_LINKMODE */
 935                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 936                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 937                + nla_total_size(ext_filter_mask
 938                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 939                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 940                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 941                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 942                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 943                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 944                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 945                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 946                + rtnl_xdp_size(dev) /* IFLA_XDP */
 947                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 948
 949 }
 950
 951 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 952 {
 953         struct nlattr *vf_ports;
 954         struct nlattr *vf_port;
 955         int vf;
 956         int err;
 957
 958         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 959         if (!vf_ports)
 960                 return -EMSGSIZE;
 961
 962         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 963                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 964                 if (!vf_port)
 965                         goto nla_put_failure;
 966                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 967                         goto nla_put_failure;
 968                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 969                 if (err == -EMSGSIZE)
 970                         goto nla_put_failure;
 971                 if (err) {
 972                         nla_nest_cancel(skb, vf_port);
 973                         continue;
 974                 }
 975                 nla_nest_end(skb, vf_port);
 976         }
 977
 978         nla_nest_end(skb, vf_ports);
 979
 980         return 0;
 981
 982 nla_put_failure:
 983         nla_nest_cancel(skb, vf_ports);
 984         return -EMSGSIZE;
 985 }
 986
 987 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 988 {
 989         struct nlattr *port_self;
 990         int err;
 991
 992         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 993         if (!port_self)
 994                 return -EMSGSIZE;
 995
 996         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 997         if (err) {
 998                 nla_nest_cancel(skb, port_self);
 999                 return (err == -EMSGSIZE) ? err : 0;
1000         }
1001
1002         nla_nest_end(skb, port_self);
1003
1004         return 0;
1005 }
1006
1007 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1008                           u32 ext_filter_mask)
1009 {
1010         int err;
1011
1012         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1013             !(ext_filter_mask & RTEXT_FILTER_VF))
1014                 return 0;
1015
1016         err = rtnl_port_self_fill(skb, dev);
1017         if (err)
1018                 return err;
1019
1020         if (dev_num_vf(dev->dev.parent)) {
1021                 err = rtnl_vf_ports_fill(skb, dev);
1022                 if (err)
1023                         return err;
1024         }
1025
1026         return 0;
1027 }
1028
1029 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1030 {
1031         int err;
1032         struct netdev_phys_item_id ppid;
1033
1034         err = dev_get_phys_port_id(dev, &ppid);
1035         if (err) {
1036                 if (err == -EOPNOTSUPP)
1037                         return 0;
1038                 return err;
1039         }
1040
1041         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1042                 return -EMSGSIZE;
1043
1044         return 0;
1045 }
1046
1047 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1048 {
1049         char name[IFNAMSIZ];
1050         int err;
1051
1052         err = dev_get_phys_port_name(dev, name, sizeof(name));
1053         if (err) {
1054                 if (err == -EOPNOTSUPP)
1055                         return 0;
1056                 return err;
1057         }
1058
1059         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1060                 return -EMSGSIZE;
1061
1062         return 0;
1063 }
1064
1065 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1066 {
1067         int err;
1068         struct switchdev_attr attr = {
1069                 .orig_dev = dev,
1070                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1071                 .flags = SWITCHDEV_F_NO_RECURSE,
1072         };
1073
1074         err = switchdev_port_attr_get(dev, &attr);
1075         if (err) {
1076                 if (err == -EOPNOTSUPP)
1077                         return 0;
1078                 return err;
1079         }
1080
1081         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1082                     attr.u.ppid.id))
1083                 return -EMSGSIZE;
1084
1085         return 0;
1086 }
1087
1088 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1089                                               struct net_device *dev)
1090 {
1091         struct rtnl_link_stats64 *sp;
1092         struct nlattr *attr;
1093
1094         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1095                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1096         if (!attr)
1097                 return -EMSGSIZE;
1098
1099         sp = nla_data(attr);
1100         dev_get_stats(dev, sp);
1101
1102         attr = nla_reserve(skb, IFLA_STATS,
1103                            sizeof(struct rtnl_link_stats));
1104         if (!attr)
1105                 return -EMSGSIZE;
1106
1107         copy_rtnl_link_stats(nla_data(attr), sp);
1108
1109         return 0;
1110 }
1111
1112 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1113                                                struct net_device *dev,
1114                                                int vfs_num,
1115                                                struct nlattr *vfinfo)
1116 {
1117         struct ifla_vf_rss_query_en vf_rss_query_en;
1118         struct nlattr *vf, *vfstats, *vfvlanlist;
1119         struct ifla_vf_link_state vf_linkstate;
1120         struct ifla_vf_vlan_info vf_vlan_info;
1121         struct ifla_vf_spoofchk vf_spoofchk;
1122         struct ifla_vf_tx_rate vf_tx_rate;
1123         struct ifla_vf_stats vf_stats;
1124         struct ifla_vf_trust vf_trust;
1125         struct ifla_vf_vlan vf_vlan;
1126         struct ifla_vf_rate vf_rate;
1127         struct ifla_vf_mac vf_mac;
1128         struct ifla_vf_info ivi;
1129
1130         /* Not all SR-IOV capable drivers support the
1131          * spoofcheck and "RSS query enable" query.  Preset to
1132          * -1 so the user space tool can detect that the driver
1133          * didn't report anything.
1134          */
1135         ivi.spoofchk = -1;
1136         ivi.rss_query_en = -1;
1137         ivi.trusted = -1;
1138         memset(ivi.mac, 0, sizeof(ivi.mac));
1139         /* The default value for VF link state is "auto"
1140          * IFLA_VF_LINK_STATE_AUTO which equals zero
1141          */
1142         ivi.linkstate = 0;
1143         /* VLAN Protocol by default is 802.1Q */
1144         ivi.vlan_proto = htons(ETH_P_8021Q);
1145         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1146                 return 0;
1147
1148         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1149
1150         vf_mac.vf =
1151                 vf_vlan.vf =
1152                 vf_vlan_info.vf =
1153                 vf_rate.vf =
1154                 vf_tx_rate.vf =
1155                 vf_spoofchk.vf =
1156                 vf_linkstate.vf =
1157                 vf_rss_query_en.vf =
1158                 vf_trust.vf = ivi.vf;
1159
1160         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1161         vf_vlan.vlan = ivi.vlan;
1162         vf_vlan.qos = ivi.qos;
1163         vf_vlan_info.vlan = ivi.vlan;
1164         vf_vlan_info.qos = ivi.qos;
1165         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1166         vf_tx_rate.rate = ivi.max_tx_rate;
1167         vf_rate.min_tx_rate = ivi.min_tx_rate;
1168         vf_rate.max_tx_rate = ivi.max_tx_rate;
1169         vf_spoofchk.setting = ivi.spoofchk;
1170         vf_linkstate.link_state = ivi.linkstate;
1171         vf_rss_query_en.setting = ivi.rss_query_en;
1172         vf_trust.setting = ivi.trusted;
1173         vf = nla_nest_start(skb, IFLA_VF_INFO);
1174         if (!vf)
1175                 goto nla_put_vfinfo_failure;
1176         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1177             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1178             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1179                     &vf_rate) ||
1180             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1181                     &vf_tx_rate) ||
1182             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1183                     &vf_spoofchk) ||
1184             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1185                     &vf_linkstate) ||
1186             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1187                     sizeof(vf_rss_query_en),
1188                     &vf_rss_query_en) ||
1189             nla_put(skb, IFLA_VF_TRUST,
1190                     sizeof(vf_trust), &vf_trust))
1191                 goto nla_put_vf_failure;
1192         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1193         if (!vfvlanlist)
1194                 goto nla_put_vf_failure;
1195         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1196                     &vf_vlan_info)) {
1197                 nla_nest_cancel(skb, vfvlanlist);
1198                 goto nla_put_vf_failure;
1199         }
1200         nla_nest_end(skb, vfvlanlist);
1201         memset(&vf_stats, 0, sizeof(vf_stats));
1202         if (dev->netdev_ops->ndo_get_vf_stats)
1203                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1204                                                 &vf_stats);
1205         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1206         if (!vfstats)
1207                 goto nla_put_vf_failure;
1208         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1209                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1210             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1211                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1212             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1213                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1214             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1215                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1216             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1217                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1218             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1219                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1220                 nla_nest_cancel(skb, vfstats);
1221                 goto nla_put_vf_failure;
1222         }
1223         nla_nest_end(skb, vfstats);
1224         nla_nest_end(skb, vf);
1225         return 0;
1226
1227 nla_put_vf_failure:
1228         nla_nest_cancel(skb, vf);
1229 nla_put_vfinfo_failure:
1230         nla_nest_cancel(skb, vfinfo);
1231         return -EMSGSIZE;
1232 }
1233
1234 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1235 {
1236         struct rtnl_link_ifmap map;
1237
1238         memset(&map, 0, sizeof(map));
1239         map.mem_start   = dev->mem_start;
1240         map.mem_end     = dev->mem_end;
1241         map.base_addr   = dev->base_addr;
1242         map.irq         = dev->irq;
1243         map.dma         = dev->dma;
1244         map.port        = dev->if_port;
1245
1246         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1247                 return -EMSGSIZE;
1248
1249         return 0;
1250 }
1251
1252 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1253 {
1254         struct netdev_xdp xdp_op = {};
1255         struct nlattr *xdp;
1256         int err;
1257
1258         if (!dev->netdev_ops->ndo_xdp)
1259                 return 0;
1260         xdp = nla_nest_start(skb, IFLA_XDP);
1261         if (!xdp)
1262                 return -EMSGSIZE;
1263         xdp_op.command = XDP_QUERY_PROG;
1264         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1265         if (err)
1266                 goto err_cancel;
1267         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1268         if (err)
1269                 goto err_cancel;
1270
1271         nla_nest_end(skb, xdp);
1272         return 0;
1273
1274 err_cancel:
1275         nla_nest_cancel(skb, xdp);
1276         return err;
1277 }
1278
1279 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1280                             int type, u32 pid, u32 seq, u32 change,
1281                             unsigned int flags, u32 ext_filter_mask)
1282 {
1283         struct ifinfomsg *ifm;
1284         struct nlmsghdr *nlh;
1285         struct nlattr *af_spec;
1286         struct rtnl_af_ops *af_ops;
1287         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1288
1289         ASSERT_RTNL();
1290         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1291         if (nlh == NULL)
1292                 return -EMSGSIZE;
1293
1294         ifm = nlmsg_data(nlh);
1295         ifm->ifi_family = AF_UNSPEC;
1296         ifm->__ifi_pad = 0;
1297         ifm->ifi_type = dev->type;
1298         ifm->ifi_index = dev->ifindex;
1299         ifm->ifi_flags = dev_get_flags(dev);
1300         ifm->ifi_change = change;
1301
1302         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1303             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1304             nla_put_u8(skb, IFLA_OPERSTATE,
1305                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1306             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1307             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1308             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1309             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1310             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1311             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1312             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1313 #ifdef CONFIG_RPS
1314             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1315 #endif
1316             (dev->ifindex != dev_get_iflink(dev) &&
1317              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1318             (upper_dev &&
1319              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1320             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1321             (dev->qdisc &&
1322              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1323             (dev->ifalias &&
1324              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1325             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1326                         atomic_read(&dev->carrier_changes)) ||
1327             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1328                 goto nla_put_failure;
1329
1330         if (rtnl_fill_link_ifmap(skb, dev))
1331                 goto nla_put_failure;
1332
1333         if (dev->addr_len) {
1334                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1335                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1336                         goto nla_put_failure;
1337         }
1338
1339         if (rtnl_phys_port_id_fill(skb, dev))
1340                 goto nla_put_failure;
1341
1342         if (rtnl_phys_port_name_fill(skb, dev))
1343                 goto nla_put_failure;
1344
1345         if (rtnl_phys_switch_id_fill(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (rtnl_fill_stats(skb, dev))
1349                 goto nla_put_failure;
1350
1351         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1352             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1353                 goto nla_put_failure;
1354
1355         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1356             ext_filter_mask & RTEXT_FILTER_VF) {
1357                 int i;
1358                 struct nlattr *vfinfo;
1359                 int num_vfs = dev_num_vf(dev->dev.parent);
1360
1361                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1362                 if (!vfinfo)
1363                         goto nla_put_failure;
1364                 for (i = 0; i < num_vfs; i++) {
1365                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1366                                 goto nla_put_failure;
1367                 }
1368
1369                 nla_nest_end(skb, vfinfo);
1370         }
1371
1372         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1373                 goto nla_put_failure;
1374
1375         if (rtnl_xdp_fill(skb, dev))
1376                 goto nla_put_failure;
1377
1378         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1379                 if (rtnl_link_fill(skb, dev) < 0)
1380                         goto nla_put_failure;
1381         }
1382
1383         if (dev->rtnl_link_ops &&
1384             dev->rtnl_link_ops->get_link_net) {
1385                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1386
1387                 if (!net_eq(dev_net(dev), link_net)) {
1388                         int id = peernet2id_alloc(dev_net(dev), link_net);
1389
1390                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1391                                 goto nla_put_failure;
1392                 }
1393         }
1394
1395         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1396                 goto nla_put_failure;
1397
1398         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1399                 if (af_ops->fill_link_af) {
1400                         struct nlattr *af;
1401                         int err;
1402
1403                         if (!(af = nla_nest_start(skb, af_ops->family)))
1404                                 goto nla_put_failure;
1405
1406                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1407
1408                         /*
1409                          * Caller may return ENODATA to indicate that there
1410                          * was no data to be dumped. This is not an error, it
1411                          * means we should trim the attribute header and
1412                          * continue.
1413                          */
1414                         if (err == -ENODATA)
1415                                 nla_nest_cancel(skb, af);
1416                         else if (err < 0)
1417                                 goto nla_put_failure;
1418
1419                         nla_nest_end(skb, af);
1420                 }
1421         }
1422
1423         nla_nest_end(skb, af_spec);
1424
1425         nlmsg_end(skb, nlh);
1426         return 0;
1427
1428 nla_put_failure:
1429         nlmsg_cancel(skb, nlh);
1430         return -EMSGSIZE;
1431 }
1432
1433 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1434         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1435         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1436         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1437         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1438         [IFLA_MTU]              = { .type = NLA_U32 },
1439         [IFLA_LINK]             = { .type = NLA_U32 },
1440         [IFLA_MASTER]           = { .type = NLA_U32 },
1441         [IFLA_CARRIER]          = { .type = NLA_U8 },
1442         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1443         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1444         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1445         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1446         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1447         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1448         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1449         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1450         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1451         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1452         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1453         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1454         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1455         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1456         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1457         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1458         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1459         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1460         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1461         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1462         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1463         [IFLA_XDP]              = { .type = NLA_NESTED },
1464 };
1465
1466 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1467         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1468         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1469         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1470         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1471 };
1472
1473 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1474         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1475         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1476         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1477         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1478         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1479         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1480         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1481         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1482         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1483         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1484         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1485         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1486 };
1487
1488 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1489         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1490         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1491                                     .len = PORT_PROFILE_MAX },
1492         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1493                                       .len = PORT_UUID_MAX },
1494         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1495                                     .len = PORT_UUID_MAX },
1496         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1497         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1498
1499         /* Unused, but we need to keep it here since user space could
1500          * fill it. It's also broken with regard to NLA_BINARY use in
1501          * combination with structs.
1502          */
1503         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1504                                     .len = sizeof(struct ifla_port_vsi) },
1505 };
1506
1507 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1508         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1509         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1510         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1511 };
1512
1513 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1514 {
1515         const struct rtnl_link_ops *ops = NULL;
1516         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1517
1518         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1519                 return NULL;
1520
1521         if (linfo[IFLA_INFO_KIND]) {
1522                 char kind[MODULE_NAME_LEN];
1523
1524                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1525                 ops = rtnl_link_ops_get(kind);
1526         }
1527
1528         return ops;
1529 }
1530
1531 static bool link_master_filtered(struct net_device *dev, int master_idx)
1532 {
1533         struct net_device *master;
1534
1535         if (!master_idx)
1536                 return false;
1537
1538         master = netdev_master_upper_dev_get(dev);
1539         if (!master || master->ifindex != master_idx)
1540                 return true;
1541
1542         return false;
1543 }
1544
1545 static bool link_kind_filtered(const struct net_device *dev,
1546                                const struct rtnl_link_ops *kind_ops)
1547 {
1548         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1549                 return true;
1550
1551         return false;
1552 }
1553
1554 static bool link_dump_filtered(struct net_device *dev,
1555                                int master_idx,
1556                                const struct rtnl_link_ops *kind_ops)
1557 {
1558         if (link_master_filtered(dev, master_idx) ||
1559             link_kind_filtered(dev, kind_ops))
1560                 return true;
1561
1562         return false;
1563 }
1564
1565 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1566 {
1567         struct net *net = sock_net(skb->sk);
1568         int h, s_h;
1569         int idx = 0, s_idx;
1570         struct net_device *dev;
1571         struct hlist_head *head;
1572         struct nlattr *tb[IFLA_MAX+1];
1573         u32 ext_filter_mask = 0;
1574         const struct rtnl_link_ops *kind_ops = NULL;
1575         unsigned int flags = NLM_F_MULTI;
1576         int master_idx = 0;
1577         int err;
1578         int hdrlen;
1579
1580         s_h = cb->args[0];
1581         s_idx = cb->args[1];
1582
1583         cb->seq = net->dev_base_seq;
1584
1585         /* A hack to preserve kernel<->userspace interface.
1586          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1587          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1588          * what iproute2 < v3.9.0 used.
1589          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1590          * attribute, its netlink message is shorter than struct ifinfomsg.
1591          */
1592         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1593                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1594
1595         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1596
1597                 if (tb[IFLA_EXT_MASK])
1598                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1599
1600                 if (tb[IFLA_MASTER])
1601                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1602
1603                 if (tb[IFLA_LINKINFO])
1604                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1605
1606                 if (master_idx || kind_ops)
1607                         flags |= NLM_F_DUMP_FILTERED;
1608         }
1609
1610         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1611                 idx = 0;
1612                 head = &net->dev_index_head[h];
1613                 hlist_for_each_entry(dev, head, index_hlist) {
1614                         if (link_dump_filtered(dev, master_idx, kind_ops))
1615                                 goto cont;
1616                         if (idx < s_idx)
1617                                 goto cont;
1618                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1619                                                NETLINK_CB(cb->skb).portid,
1620                                                cb->nlh->nlmsg_seq, 0,
1621                                                flags,
1622                                                ext_filter_mask);
1623                         /* If we ran out of room on the first message,
1624                          * we're in trouble
1625                          */
1626                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1627
1628                         if (err < 0)
1629                                 goto out;
1630
1631                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1632 cont:
1633                         idx++;
1634                 }
1635         }
1636 out:
1637         cb->args[1] = idx;
1638         cb->args[0] = h;
1639
1640         return skb->len;
1641 }
1642
1643 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1644 {
1645         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1646 }
1647 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1648
1649 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1650 {
1651         struct net *net;
1652         /* Examine the link attributes and figure out which
1653          * network namespace we are talking about.
1654          */
1655         if (tb[IFLA_NET_NS_PID])
1656                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1657         else if (tb[IFLA_NET_NS_FD])
1658                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1659         else
1660                 net = get_net(src_net);
1661         return net;
1662 }
1663 EXPORT_SYMBOL(rtnl_link_get_net);
1664
1665 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1666 {
1667         if (dev) {
1668                 if (tb[IFLA_ADDRESS] &&
1669                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1670                         return -EINVAL;
1671
1672                 if (tb[IFLA_BROADCAST] &&
1673                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1674                         return -EINVAL;
1675         }
1676
1677         if (tb[IFLA_AF_SPEC]) {
1678                 struct nlattr *af;
1679                 int rem, err;
1680
1681                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1682                         const struct rtnl_af_ops *af_ops;
1683
1684                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1685                                 return -EAFNOSUPPORT;
1686
1687                         if (!af_ops->set_link_af)
1688                                 return -EOPNOTSUPP;
1689
1690                         if (af_ops->validate_link_af) {
1691                                 err = af_ops->validate_link_af(dev, af);
1692                                 if (err < 0)
1693                                         return err;
1694                         }
1695                 }
1696         }
1697
1698         return 0;
1699 }
1700
1701 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1702                                   int guid_type)
1703 {
1704         const struct net_device_ops *ops = dev->netdev_ops;
1705
1706         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1707 }
1708
1709 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1710 {
1711         if (dev->type != ARPHRD_INFINIBAND)
1712                 return -EOPNOTSUPP;
1713
1714         return handle_infiniband_guid(dev, ivt, guid_type);
1715 }
1716
1717 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1718 {
1719         const struct net_device_ops *ops = dev->netdev_ops;
1720         int err = -EINVAL;
1721
1722         if (tb[IFLA_VF_MAC]) {
1723                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1724
1725                 err = -EOPNOTSUPP;
1726                 if (ops->ndo_set_vf_mac)
1727                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1728                                                   ivm->mac);
1729                 if (err < 0)
1730                         return err;
1731         }
1732
1733         if (tb[IFLA_VF_VLAN]) {
1734                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1735
1736                 err = -EOPNOTSUPP;
1737                 if (ops->ndo_set_vf_vlan)
1738                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1739                                                    ivv->qos,
1740                                                    htons(ETH_P_8021Q));
1741                 if (err < 0)
1742                         return err;
1743         }
1744
1745         if (tb[IFLA_VF_VLAN_LIST]) {
1746                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1747                 struct nlattr *attr;
1748                 int rem, len = 0;
1749
1750                 err = -EOPNOTSUPP;
1751                 if (!ops->ndo_set_vf_vlan)
1752                         return err;
1753
1754                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1755                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1756                             nla_len(attr) < NLA_HDRLEN) {
1757                                 return -EINVAL;
1758                         }
1759                         if (len >= MAX_VLAN_LIST_LEN)
1760                                 return -EOPNOTSUPP;
1761                         ivvl[len] = nla_data(attr);
1762
1763                         len++;
1764                 }
1765                 if (len == 0)
1766                         return -EINVAL;
1767
1768                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1769                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1770                 if (err < 0)
1771                         return err;
1772         }
1773
1774         if (tb[IFLA_VF_TX_RATE]) {
1775                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1776                 struct ifla_vf_info ivf;
1777
1778                 err = -EOPNOTSUPP;
1779                 if (ops->ndo_get_vf_config)
1780                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1781                 if (err < 0)
1782                         return err;
1783
1784                 err = -EOPNOTSUPP;
1785                 if (ops->ndo_set_vf_rate)
1786                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1787                                                    ivf.min_tx_rate,
1788                                                    ivt->rate);
1789                 if (err < 0)
1790                         return err;
1791         }
1792
1793         if (tb[IFLA_VF_RATE]) {
1794                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1795
1796                 err = -EOPNOTSUPP;
1797                 if (ops->ndo_set_vf_rate)
1798                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1799                                                    ivt->min_tx_rate,
1800                                                    ivt->max_tx_rate);
1801                 if (err < 0)
1802                         return err;
1803         }
1804
1805         if (tb[IFLA_VF_SPOOFCHK]) {
1806                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1807
1808                 err = -EOPNOTSUPP;
1809                 if (ops->ndo_set_vf_spoofchk)
1810                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1811                                                        ivs->setting);
1812                 if (err < 0)
1813                         return err;
1814         }
1815
1816         if (tb[IFLA_VF_LINK_STATE]) {
1817                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1818
1819                 err = -EOPNOTSUPP;
1820                 if (ops->ndo_set_vf_link_state)
1821                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1822                                                          ivl->link_state);
1823                 if (err < 0)
1824                         return err;
1825         }
1826
1827         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1828                 struct ifla_vf_rss_query_en *ivrssq_en;
1829
1830                 err = -EOPNOTSUPP;
1831                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1832                 if (ops->ndo_set_vf_rss_query_en)
1833                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1834                                                            ivrssq_en->setting);
1835                 if (err < 0)
1836                         return err;
1837         }
1838
1839         if (tb[IFLA_VF_TRUST]) {
1840                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1841
1842                 err = -EOPNOTSUPP;
1843                 if (ops->ndo_set_vf_trust)
1844                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1845                 if (err < 0)
1846                         return err;
1847         }
1848
1849         if (tb[IFLA_VF_IB_NODE_GUID]) {
1850                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1851
1852                 if (!ops->ndo_set_vf_guid)
1853                         return -EOPNOTSUPP;
1854
1855                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1856         }
1857
1858         if (tb[IFLA_VF_IB_PORT_GUID]) {
1859                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1860
1861                 if (!ops->ndo_set_vf_guid)
1862                         return -EOPNOTSUPP;
1863
1864                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1865         }
1866
1867         return err;
1868 }
1869
1870 static int do_set_master(struct net_device *dev, int ifindex)
1871 {
1872         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1873         const struct net_device_ops *ops;
1874         int err;
1875
1876         if (upper_dev) {
1877                 if (upper_dev->ifindex == ifindex)
1878                         return 0;
1879                 ops = upper_dev->netdev_ops;
1880                 if (ops->ndo_del_slave) {
1881                         err = ops->ndo_del_slave(upper_dev, dev);
1882                         if (err)
1883                                 return err;
1884                 } else {
1885                         return -EOPNOTSUPP;
1886                 }
1887         }
1888
1889         if (ifindex) {
1890                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1891                 if (!upper_dev)
1892                         return -EINVAL;
1893                 ops = upper_dev->netdev_ops;
1894                 if (ops->ndo_add_slave) {
1895                         err = ops->ndo_add_slave(upper_dev, dev);
1896                         if (err)
1897                                 return err;
1898                 } else {
1899                         return -EOPNOTSUPP;
1900                 }
1901         }
1902         return 0;
1903 }
1904
1905 #define DO_SETLINK_MODIFIED     0x01
1906 /* notify flag means notify + modified. */
1907 #define DO_SETLINK_NOTIFY       0x03
1908 static int do_setlink(const struct sk_buff *skb,
1909                       struct net_device *dev, struct ifinfomsg *ifm,
1910                       struct nlattr **tb, char *ifname, int status)
1911 {
1912         const struct net_device_ops *ops = dev->netdev_ops;
1913         int err;
1914
1915         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1916                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1917                 if (IS_ERR(net)) {
1918                         err = PTR_ERR(net);
1919                         goto errout;
1920                 }
1921                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1922                         put_net(net);
1923                         err = -EPERM;
1924                         goto errout;
1925                 }
1926                 err = dev_change_net_namespace(dev, net, ifname);
1927                 put_net(net);
1928                 if (err)
1929                         goto errout;
1930                 status |= DO_SETLINK_MODIFIED;
1931         }
1932
1933         if (tb[IFLA_MAP]) {
1934                 struct rtnl_link_ifmap *u_map;
1935                 struct ifmap k_map;
1936
1937                 if (!ops->ndo_set_config) {
1938                         err = -EOPNOTSUPP;
1939                         goto errout;
1940                 }
1941
1942                 if (!netif_device_present(dev)) {
1943                         err = -ENODEV;
1944                         goto errout;
1945                 }
1946
1947                 u_map = nla_data(tb[IFLA_MAP]);
1948                 k_map.mem_start = (unsigned long) u_map->mem_start;
1949                 k_map.mem_end = (unsigned long) u_map->mem_end;
1950                 k_map.base_addr = (unsigned short) u_map->base_addr;
1951                 k_map.irq = (unsigned char) u_map->irq;
1952                 k_map.dma = (unsigned char) u_map->dma;
1953                 k_map.port = (unsigned char) u_map->port;
1954
1955                 err = ops->ndo_set_config(dev, &k_map);
1956                 if (err < 0)
1957                         goto errout;
1958
1959                 status |= DO_SETLINK_NOTIFY;
1960         }
1961
1962         if (tb[IFLA_ADDRESS]) {
1963                 struct sockaddr *sa;
1964                 int len;
1965
1966                 len = sizeof(sa_family_t) + dev->addr_len;
1967                 sa = kmalloc(len, GFP_KERNEL);
1968                 if (!sa) {
1969                         err = -ENOMEM;
1970                         goto errout;
1971                 }
1972                 sa->sa_family = dev->type;
1973                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1974                        dev->addr_len);
1975                 err = dev_set_mac_address(dev, sa);
1976                 kfree(sa);
1977                 if (err)
1978                         goto errout;
1979                 status |= DO_SETLINK_MODIFIED;
1980         }
1981
1982         if (tb[IFLA_MTU]) {
1983                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1984                 if (err < 0)
1985                         goto errout;
1986                 status |= DO_SETLINK_MODIFIED;
1987         }
1988
1989         if (tb[IFLA_GROUP]) {
1990                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1991                 status |= DO_SETLINK_NOTIFY;
1992         }
1993
1994         /*
1995          * Interface selected by interface index but interface
1996          * name provided implies that a name change has been
1997          * requested.
1998          */
1999         if (ifm->ifi_index > 0 && ifname[0]) {
2000                 err = dev_change_name(dev, ifname);
2001                 if (err < 0)
2002                         goto errout;
2003                 status |= DO_SETLINK_MODIFIED;
2004         }
2005
2006         if (tb[IFLA_IFALIAS]) {
2007                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2008                                     nla_len(tb[IFLA_IFALIAS]));
2009                 if (err < 0)
2010                         goto errout;
2011                 status |= DO_SETLINK_NOTIFY;
2012         }
2013
2014         if (tb[IFLA_BROADCAST]) {
2015                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2016                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2017         }
2018
2019         if (ifm->ifi_flags || ifm->ifi_change) {
2020                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2021                 if (err < 0)
2022                         goto errout;
2023         }
2024
2025         if (tb[IFLA_MASTER]) {
2026                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2027                 if (err)
2028                         goto errout;
2029                 status |= DO_SETLINK_MODIFIED;
2030         }
2031
2032         if (tb[IFLA_CARRIER]) {
2033                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2034                 if (err)
2035                         goto errout;
2036                 status |= DO_SETLINK_MODIFIED;
2037         }
2038
2039         if (tb[IFLA_TXQLEN]) {
2040                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2041                 unsigned long orig_len = dev->tx_queue_len;
2042
2043                 if (dev->tx_queue_len ^ value) {
2044                         dev->tx_queue_len = value;
2045                         err = call_netdevice_notifiers(
2046                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2047                         err = notifier_to_errno(err);
2048                         if (err) {
2049                                 dev->tx_queue_len = orig_len;
2050                                 goto errout;
2051                         }
2052                         status |= DO_SETLINK_NOTIFY;
2053                 }
2054         }
2055
2056         if (tb[IFLA_OPERSTATE])
2057                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2058
2059         if (tb[IFLA_LINKMODE]) {
2060                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2061
2062                 write_lock_bh(&dev_base_lock);
2063                 if (dev->link_mode ^ value)
2064                         status |= DO_SETLINK_NOTIFY;
2065                 dev->link_mode = value;
2066                 write_unlock_bh(&dev_base_lock);
2067         }
2068
2069         if (tb[IFLA_VFINFO_LIST]) {
2070                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2071                 struct nlattr *attr;
2072                 int rem;
2073
2074                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2075                         if (nla_type(attr) != IFLA_VF_INFO ||
2076                             nla_len(attr) < NLA_HDRLEN) {
2077                                 err = -EINVAL;
2078                                 goto errout;
2079                         }
2080                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2081                                                ifla_vf_policy);
2082                         if (err < 0)
2083                                 goto errout;
2084                         err = do_setvfinfo(dev, vfinfo);
2085                         if (err < 0)
2086                                 goto errout;
2087                         status |= DO_SETLINK_NOTIFY;
2088                 }
2089         }
2090         err = 0;
2091
2092         if (tb[IFLA_VF_PORTS]) {
2093                 struct nlattr *port[IFLA_PORT_MAX+1];
2094                 struct nlattr *attr;
2095                 int vf;
2096                 int rem;
2097
2098                 err = -EOPNOTSUPP;
2099                 if (!ops->ndo_set_vf_port)
2100                         goto errout;
2101
2102                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2103                         if (nla_type(attr) != IFLA_VF_PORT ||
2104                             nla_len(attr) < NLA_HDRLEN) {
2105                                 err = -EINVAL;
2106                                 goto errout;
2107                         }
2108                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2109                                                ifla_port_policy);
2110                         if (err < 0)
2111                                 goto errout;
2112                         if (!port[IFLA_PORT_VF]) {
2113                                 err = -EOPNOTSUPP;
2114                                 goto errout;
2115                         }
2116                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2117                         err = ops->ndo_set_vf_port(dev, vf, port);
2118                         if (err < 0)
2119                                 goto errout;
2120                         status |= DO_SETLINK_NOTIFY;
2121                 }
2122         }
2123         err = 0;
2124
2125         if (tb[IFLA_PORT_SELF]) {
2126                 struct nlattr *port[IFLA_PORT_MAX+1];
2127
2128                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2129                         tb[IFLA_PORT_SELF], ifla_port_policy);
2130                 if (err < 0)
2131                         goto errout;
2132
2133                 err = -EOPNOTSUPP;
2134                 if (ops->ndo_set_vf_port)
2135                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2136                 if (err < 0)
2137                         goto errout;
2138                 status |= DO_SETLINK_NOTIFY;
2139         }
2140
2141         if (tb[IFLA_AF_SPEC]) {
2142                 struct nlattr *af;
2143                 int rem;
2144
2145                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2146                         const struct rtnl_af_ops *af_ops;
2147
2148                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2149                                 BUG();
2150
2151                         err = af_ops->set_link_af(dev, af);
2152                         if (err < 0)
2153                                 goto errout;
2154
2155                         status |= DO_SETLINK_NOTIFY;
2156                 }
2157         }
2158         err = 0;
2159
2160         if (tb[IFLA_PROTO_DOWN]) {
2161                 err = dev_change_proto_down(dev,
2162                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2163                 if (err)
2164                         goto errout;
2165                 status |= DO_SETLINK_NOTIFY;
2166         }
2167
2168         if (tb[IFLA_XDP]) {
2169                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2170                 u32 xdp_flags = 0;
2171
2172                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2173                                        ifla_xdp_policy);
2174                 if (err < 0)
2175                         goto errout;
2176
2177                 if (xdp[IFLA_XDP_ATTACHED]) {
2178                         err = -EINVAL;
2179                         goto errout;
2180                 }
2181
2182                 if (xdp[IFLA_XDP_FLAGS]) {
2183                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2184                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2185                                 err = -EINVAL;
2186                                 goto errout;
2187                         }
2188                 }
2189
2190                 if (xdp[IFLA_XDP_FD]) {
2191                         err = dev_change_xdp_fd(dev,
2192                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2193                                                 xdp_flags);
2194                         if (err)
2195                                 goto errout;
2196                         status |= DO_SETLINK_NOTIFY;
2197                 }
2198         }
2199
2200 errout:
2201         if (status & DO_SETLINK_MODIFIED) {
2202                 if (status & DO_SETLINK_NOTIFY)
2203                         netdev_state_change(dev);
2204
2205                 if (err < 0)
2206                         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",
2207                                              dev->name);
2208         }
2209
2210         return err;
2211 }
2212
2213 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2214 {
2215         struct net *net = sock_net(skb->sk);
2216         struct ifinfomsg *ifm;
2217         struct net_device *dev;
2218         int err;
2219         struct nlattr *tb[IFLA_MAX+1];
2220         char ifname[IFNAMSIZ];
2221
2222         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2223         if (err < 0)
2224                 goto errout;
2225
2226         if (tb[IFLA_IFNAME])
2227                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2228         else
2229                 ifname[0] = '\0';
2230
2231         err = -EINVAL;
2232         ifm = nlmsg_data(nlh);
2233         if (ifm->ifi_index > 0)
2234                 dev = __dev_get_by_index(net, ifm->ifi_index);
2235         else if (tb[IFLA_IFNAME])
2236                 dev = __dev_get_by_name(net, ifname);
2237         else
2238                 goto errout;
2239
2240         if (dev == NULL) {
2241                 err = -ENODEV;
2242                 goto errout;
2243         }
2244
2245         err = validate_linkmsg(dev, tb);
2246         if (err < 0)
2247                 goto errout;
2248
2249         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2250 errout:
2251         return err;
2252 }
2253
2254 static int rtnl_group_dellink(const struct net *net, int group)
2255 {
2256         struct net_device *dev, *aux;
2257         LIST_HEAD(list_kill);
2258         bool found = false;
2259
2260         if (!group)
2261                 return -EPERM;
2262
2263         for_each_netdev(net, dev) {
2264                 if (dev->group == group) {
2265                         const struct rtnl_link_ops *ops;
2266
2267                         found = true;
2268                         ops = dev->rtnl_link_ops;
2269                         if (!ops || !ops->dellink)
2270                                 return -EOPNOTSUPP;
2271                 }
2272         }
2273
2274         if (!found)
2275                 return -ENODEV;
2276
2277         for_each_netdev_safe(net, dev, aux) {
2278                 if (dev->group == group) {
2279                         const struct rtnl_link_ops *ops;
2280
2281                         ops = dev->rtnl_link_ops;
2282                         ops->dellink(dev, &list_kill);
2283                 }
2284         }
2285         unregister_netdevice_many(&list_kill);
2286
2287         return 0;
2288 }
2289
2290 int rtnl_delete_link(struct net_device *dev)
2291 {
2292         const struct rtnl_link_ops *ops;
2293         LIST_HEAD(list_kill);
2294
2295         ops = dev->rtnl_link_ops;
2296         if (!ops || !ops->dellink)
2297                 return -EOPNOTSUPP;
2298
2299         ops->dellink(dev, &list_kill);
2300         unregister_netdevice_many(&list_kill);
2301
2302         return 0;
2303 }
2304 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2305
2306 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2307 {
2308         struct net *net = sock_net(skb->sk);
2309         struct net_device *dev;
2310         struct ifinfomsg *ifm;
2311         char ifname[IFNAMSIZ];
2312         struct nlattr *tb[IFLA_MAX+1];
2313         int err;
2314
2315         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2316         if (err < 0)
2317                 return err;
2318
2319         if (tb[IFLA_IFNAME])
2320                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2321
2322         ifm = nlmsg_data(nlh);
2323         if (ifm->ifi_index > 0)
2324                 dev = __dev_get_by_index(net, ifm->ifi_index);
2325         else if (tb[IFLA_IFNAME])
2326                 dev = __dev_get_by_name(net, ifname);
2327         else if (tb[IFLA_GROUP])
2328                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2329         else
2330                 return -EINVAL;
2331
2332         if (!dev)
2333                 return -ENODEV;
2334
2335         return rtnl_delete_link(dev);
2336 }
2337
2338 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2339 {
2340         unsigned int old_flags;
2341         int err;
2342
2343         old_flags = dev->flags;
2344         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2345                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2346                 if (err < 0)
2347                         return err;
2348         }
2349
2350         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2351
2352         __dev_notify_flags(dev, old_flags, ~0U);
2353         return 0;
2354 }
2355 EXPORT_SYMBOL(rtnl_configure_link);
2356
2357 struct net_device *rtnl_create_link(struct net *net,
2358         const char *ifname, unsigned char name_assign_type,
2359         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2360 {
2361         struct net_device *dev;
2362         unsigned int num_tx_queues = 1;
2363         unsigned int num_rx_queues = 1;
2364
2365         if (tb[IFLA_NUM_TX_QUEUES])
2366                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2367         else if (ops->get_num_tx_queues)
2368                 num_tx_queues = ops->get_num_tx_queues();
2369
2370         if (tb[IFLA_NUM_RX_QUEUES])
2371                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2372         else if (ops->get_num_rx_queues)
2373                 num_rx_queues = ops->get_num_rx_queues();
2374
2375         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2376                                ops->setup, num_tx_queues, num_rx_queues);
2377         if (!dev)
2378                 return ERR_PTR(-ENOMEM);
2379
2380         dev_net_set(dev, net);
2381         dev->rtnl_link_ops = ops;
2382         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2383
2384         if (tb[IFLA_MTU])
2385                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2386         if (tb[IFLA_ADDRESS]) {
2387                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2388                                 nla_len(tb[IFLA_ADDRESS]));
2389                 dev->addr_assign_type = NET_ADDR_SET;
2390         }
2391         if (tb[IFLA_BROADCAST])
2392                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2393                                 nla_len(tb[IFLA_BROADCAST]));
2394         if (tb[IFLA_TXQLEN])
2395                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2396         if (tb[IFLA_OPERSTATE])
2397                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2398         if (tb[IFLA_LINKMODE])
2399                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2400         if (tb[IFLA_GROUP])
2401                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2402
2403         return dev;
2404 }
2405 EXPORT_SYMBOL(rtnl_create_link);
2406
2407 static int rtnl_group_changelink(const struct sk_buff *skb,
2408                 struct net *net, int group,
2409                 struct ifinfomsg *ifm,
2410                 struct nlattr **tb)
2411 {
2412         struct net_device *dev, *aux;
2413         int err;
2414
2415         for_each_netdev_safe(net, dev, aux) {
2416                 if (dev->group == group) {
2417                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2418                         if (err < 0)
2419                                 return err;
2420                 }
2421         }
2422
2423         return 0;
2424 }
2425
2426 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2427 {
2428         struct net *net = sock_net(skb->sk);
2429         const struct rtnl_link_ops *ops;
2430         const struct rtnl_link_ops *m_ops = NULL;
2431         struct net_device *dev;
2432         struct net_device *master_dev = NULL;
2433         struct ifinfomsg *ifm;
2434         char kind[MODULE_NAME_LEN];
2435         char ifname[IFNAMSIZ];
2436         struct nlattr *tb[IFLA_MAX+1];
2437         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2438         unsigned char name_assign_type = NET_NAME_USER;
2439         int err;
2440
2441 #ifdef CONFIG_MODULES
2442 replay:
2443 #endif
2444         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2445         if (err < 0)
2446                 return err;
2447
2448         if (tb[IFLA_IFNAME])
2449                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2450         else
2451                 ifname[0] = '\0';
2452
2453         ifm = nlmsg_data(nlh);
2454         if (ifm->ifi_index > 0)
2455                 dev = __dev_get_by_index(net, ifm->ifi_index);
2456         else {
2457                 if (ifname[0])
2458                         dev = __dev_get_by_name(net, ifname);
2459                 else
2460                         dev = NULL;
2461         }
2462
2463         if (dev) {
2464                 master_dev = netdev_master_upper_dev_get(dev);
2465                 if (master_dev)
2466                         m_ops = master_dev->rtnl_link_ops;
2467         }
2468
2469         err = validate_linkmsg(dev, tb);
2470         if (err < 0)
2471                 return err;
2472
2473         if (tb[IFLA_LINKINFO]) {
2474                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2475                                        tb[IFLA_LINKINFO], ifla_info_policy);
2476                 if (err < 0)
2477                         return err;
2478         } else
2479                 memset(linkinfo, 0, sizeof(linkinfo));
2480
2481         if (linkinfo[IFLA_INFO_KIND]) {
2482                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2483                 ops = rtnl_link_ops_get(kind);
2484         } else {
2485                 kind[0] = '\0';
2486                 ops = NULL;
2487         }
2488
2489         if (1) {
2490                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2491                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2492                 struct nlattr **data = NULL;
2493                 struct nlattr **slave_data = NULL;
2494                 struct net *dest_net, *link_net = NULL;
2495
2496                 if (ops) {
2497                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2498                                 err = nla_parse_nested(attr, ops->maxtype,
2499                                                        linkinfo[IFLA_INFO_DATA],
2500                                                        ops->policy);
2501                                 if (err < 0)
2502                                         return err;
2503                                 data = attr;
2504                         }
2505                         if (ops->validate) {
2506                                 err = ops->validate(tb, data);
2507                                 if (err < 0)
2508                                         return err;
2509                         }
2510                 }
2511
2512                 if (m_ops) {
2513                         if (m_ops->slave_maxtype &&
2514                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2515                                 err = nla_parse_nested(slave_attr,
2516                                                        m_ops->slave_maxtype,
2517                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2518                                                        m_ops->slave_policy);
2519                                 if (err < 0)
2520                                         return err;
2521                                 slave_data = slave_attr;
2522                         }
2523                         if (m_ops->slave_validate) {
2524                                 err = m_ops->slave_validate(tb, slave_data);
2525                                 if (err < 0)
2526                                         return err;
2527                         }
2528                 }
2529
2530                 if (dev) {
2531                         int status = 0;
2532
2533                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2534                                 return -EEXIST;
2535                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2536                                 return -EOPNOTSUPP;
2537
2538                         if (linkinfo[IFLA_INFO_DATA]) {
2539                                 if (!ops || ops != dev->rtnl_link_ops ||
2540                                     !ops->changelink)
2541                                         return -EOPNOTSUPP;
2542
2543                                 err = ops->changelink(dev, tb, data);
2544                                 if (err < 0)
2545                                         return err;
2546                                 status |= DO_SETLINK_NOTIFY;
2547                         }
2548
2549                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2550                                 if (!m_ops || !m_ops->slave_changelink)
2551                                         return -EOPNOTSUPP;
2552
2553                                 err = m_ops->slave_changelink(master_dev, dev,
2554                                                               tb, slave_data);
2555                                 if (err < 0)
2556                                         return err;
2557                                 status |= DO_SETLINK_NOTIFY;
2558                         }
2559
2560                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2561                 }
2562
2563                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2564                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2565                                 return rtnl_group_changelink(skb, net,
2566                                                 nla_get_u32(tb[IFLA_GROUP]),
2567                                                 ifm, tb);
2568                         return -ENODEV;
2569                 }
2570
2571                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2572                         return -EOPNOTSUPP;
2573
2574                 if (!ops) {
2575 #ifdef CONFIG_MODULES
2576                         if (kind[0]) {
2577                                 __rtnl_unlock();
2578                                 request_module("rtnl-link-%s", kind);
2579                                 rtnl_lock();
2580                                 ops = rtnl_link_ops_get(kind);
2581                                 if (ops)
2582                                         goto replay;
2583                         }
2584 #endif
2585                         return -EOPNOTSUPP;
2586                 }
2587
2588                 if (!ops->setup)
2589                         return -EOPNOTSUPP;
2590
2591                 if (!ifname[0]) {
2592                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2593                         name_assign_type = NET_NAME_ENUM;
2594                 }
2595
2596                 dest_net = rtnl_link_get_net(net, tb);
2597                 if (IS_ERR(dest_net))
2598                         return PTR_ERR(dest_net);
2599
2600                 err = -EPERM;
2601                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2602                         goto out;
2603
2604                 if (tb[IFLA_LINK_NETNSID]) {
2605                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2606
2607                         link_net = get_net_ns_by_id(dest_net, id);
2608                         if (!link_net) {
2609                                 err =  -EINVAL;
2610                                 goto out;
2611                         }
2612                         err = -EPERM;
2613                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2614                                 goto out;
2615                 }
2616
2617                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2618                                        name_assign_type, ops, tb);
2619                 if (IS_ERR(dev)) {
2620                         err = PTR_ERR(dev);
2621                         goto out;
2622                 }
2623
2624                 dev->ifindex = ifm->ifi_index;
2625
2626                 if (ops->newlink) {
2627                         err = ops->newlink(link_net ? : net, dev, tb, data);
2628                         /* Drivers should call free_netdev() in ->destructor
2629                          * and unregister it on failure after registration
2630                          * so that device could be finally freed in rtnl_unlock.
2631                          */
2632                         if (err < 0) {
2633                                 /* If device is not registered at all, free it now */
2634                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2635                                         free_netdev(dev);
2636                                 goto out;
2637                         }
2638                 } else {
2639                         err = register_netdevice(dev);
2640                         if (err < 0) {
2641                                 free_netdev(dev);
2642                                 goto out;
2643                         }
2644                 }
2645                 err = rtnl_configure_link(dev, ifm);
2646                 if (err < 0)
2647                         goto out_unregister;
2648                 if (link_net) {
2649                         err = dev_change_net_namespace(dev, dest_net, ifname);
2650                         if (err < 0)
2651                                 goto out_unregister;
2652                 }
2653                 if (tb[IFLA_MASTER]) {
2654                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2655                         if (err)
2656                                 goto out_unregister;
2657                 }
2658 out:
2659                 if (link_net)
2660                         put_net(link_net);
2661                 put_net(dest_net);
2662                 return err;
2663 out_unregister:
2664                 if (ops->newlink) {
2665                         LIST_HEAD(list_kill);
2666
2667                         ops->dellink(dev, &list_kill);
2668                         unregister_netdevice_many(&list_kill);
2669                 } else {
2670                         unregister_netdevice(dev);
2671                 }
2672                 goto out;
2673         }
2674 }
2675
2676 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2677 {
2678         struct net *net = sock_net(skb->sk);
2679         struct ifinfomsg *ifm;
2680         char ifname[IFNAMSIZ];
2681         struct nlattr *tb[IFLA_MAX+1];
2682         struct net_device *dev = NULL;
2683         struct sk_buff *nskb;
2684         int err;
2685         u32 ext_filter_mask = 0;
2686
2687         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2688         if (err < 0)
2689                 return err;
2690
2691         if (tb[IFLA_IFNAME])
2692                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2693
2694         if (tb[IFLA_EXT_MASK])
2695                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2696
2697         ifm = nlmsg_data(nlh);
2698         if (ifm->ifi_index > 0)
2699                 dev = __dev_get_by_index(net, ifm->ifi_index);
2700         else if (tb[IFLA_IFNAME])
2701                 dev = __dev_get_by_name(net, ifname);
2702         else
2703                 return -EINVAL;
2704
2705         if (dev == NULL)
2706                 return -ENODEV;
2707
2708         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2709         if (nskb == NULL)
2710                 return -ENOBUFS;
2711
2712         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2713                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2714         if (err < 0) {
2715                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2716                 WARN_ON(err == -EMSGSIZE);
2717                 kfree_skb(nskb);
2718         } else
2719                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2720
2721         return err;
2722 }
2723
2724 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2725 {
2726         struct net *net = sock_net(skb->sk);
2727         struct net_device *dev;
2728         struct nlattr *tb[IFLA_MAX+1];
2729         u32 ext_filter_mask = 0;
2730         u16 min_ifinfo_dump_size = 0;
2731         int hdrlen;
2732
2733         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2734         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2735                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2736
2737         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2738                 if (tb[IFLA_EXT_MASK])
2739                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2740         }
2741
2742         if (!ext_filter_mask)
2743                 return NLMSG_GOODSIZE;
2744         /*
2745          * traverse the list of net devices and compute the minimum
2746          * buffer size based upon the filter mask.
2747          */
2748         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2749                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2750                                              if_nlmsg_size(dev,
2751                                                            ext_filter_mask));
2752         }
2753
2754         return nlmsg_total_size(min_ifinfo_dump_size);
2755 }
2756
2757 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2758 {
2759         int idx;
2760         int s_idx = cb->family;
2761
2762         if (s_idx == 0)
2763                 s_idx = 1;
2764         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2765                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2766                 if (idx < s_idx || idx == PF_PACKET)
2767                         continue;
2768                 if (rtnl_msg_handlers[idx] == NULL ||
2769                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2770                         continue;
2771                 if (idx > s_idx) {
2772                         memset(&cb->args[0], 0, sizeof(cb->args));
2773                         cb->prev_seq = 0;
2774                         cb->seq = 0;
2775                 }
2776                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2777                         break;
2778         }
2779         cb->family = idx;
2780
2781         return skb->len;
2782 }
2783
2784 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2785                                        unsigned int change, gfp_t flags)
2786 {
2787         struct net *net = dev_net(dev);
2788         struct sk_buff *skb;
2789         int err = -ENOBUFS;
2790         size_t if_info_size;
2791
2792         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2793         if (skb == NULL)
2794                 goto errout;
2795
2796         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2797         if (err < 0) {
2798                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2799                 WARN_ON(err == -EMSGSIZE);
2800                 kfree_skb(skb);
2801                 goto errout;
2802         }
2803         return skb;
2804 errout:
2805         if (err < 0)
2806                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2807         return NULL;
2808 }
2809
2810 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2811 {
2812         struct net *net = dev_net(dev);
2813
2814         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2815 }
2816
2817 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2818                   gfp_t flags)
2819 {
2820         struct sk_buff *skb;
2821
2822         if (dev->reg_state != NETREG_REGISTERED)
2823                 return;
2824
2825         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2826         if (skb)
2827                 rtmsg_ifinfo_send(skb, dev, flags);
2828 }
2829 EXPORT_SYMBOL(rtmsg_ifinfo);
2830
2831 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2832                                    struct net_device *dev,
2833                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2834                                    int type, unsigned int flags,
2835                                    int nlflags, u16 ndm_state)
2836 {
2837         struct nlmsghdr *nlh;
2838         struct ndmsg *ndm;
2839
2840         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2841         if (!nlh)
2842                 return -EMSGSIZE;
2843
2844         ndm = nlmsg_data(nlh);
2845         ndm->ndm_family  = AF_BRIDGE;
2846         ndm->ndm_pad1    = 0;
2847         ndm->ndm_pad2    = 0;
2848         ndm->ndm_flags   = flags;
2849         ndm->ndm_type    = 0;
2850         ndm->ndm_ifindex = dev->ifindex;
2851         ndm->ndm_state   = ndm_state;
2852
2853         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2854                 goto nla_put_failure;
2855         if (vid)
2856                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2857                         goto nla_put_failure;
2858
2859         nlmsg_end(skb, nlh);
2860         return 0;
2861
2862 nla_put_failure:
2863         nlmsg_cancel(skb, nlh);
2864         return -EMSGSIZE;
2865 }
2866
2867 static inline size_t rtnl_fdb_nlmsg_size(void)
2868 {
2869         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2870                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2871                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2872                0;
2873 }
2874
2875 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2876                             u16 ndm_state)
2877 {
2878         struct net *net = dev_net(dev);
2879         struct sk_buff *skb;
2880         int err = -ENOBUFS;
2881
2882         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2883         if (!skb)
2884                 goto errout;
2885
2886         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2887                                       0, 0, type, NTF_SELF, 0, ndm_state);
2888         if (err < 0) {
2889                 kfree_skb(skb);
2890                 goto errout;
2891         }
2892
2893         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2894         return;
2895 errout:
2896         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2897 }
2898
2899 /**
2900  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2901  */
2902 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2903                      struct nlattr *tb[],
2904                      struct net_device *dev,
2905                      const unsigned char *addr, u16 vid,
2906                      u16 flags)
2907 {
2908         int err = -EINVAL;
2909
2910         /* If aging addresses are supported device will need to
2911          * implement its own handler for this.
2912          */
2913         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2914                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2915                 return err;
2916         }
2917
2918         if (vid) {
2919                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2920                 return err;
2921         }
2922
2923         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2924                 err = dev_uc_add_excl(dev, addr);
2925         else if (is_multicast_ether_addr(addr))
2926                 err = dev_mc_add_excl(dev, addr);
2927
2928         /* Only return duplicate errors if NLM_F_EXCL is set */
2929         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2930                 err = 0;
2931
2932         return err;
2933 }
2934 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2935
2936 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2937 {
2938         u16 vid = 0;
2939
2940         if (vlan_attr) {
2941                 if (nla_len(vlan_attr) != sizeof(u16)) {
2942                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2943                         return -EINVAL;
2944                 }
2945
2946                 vid = nla_get_u16(vlan_attr);
2947
2948                 if (!vid || vid >= VLAN_VID_MASK) {
2949                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2950                                 vid);
2951                         return -EINVAL;
2952                 }
2953         }
2954         *p_vid = vid;
2955         return 0;
2956 }
2957
2958 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2959 {
2960         struct net *net = sock_net(skb->sk);
2961         struct ndmsg *ndm;
2962         struct nlattr *tb[NDA_MAX+1];
2963         struct net_device *dev;
2964         u8 *addr;
2965         u16 vid;
2966         int err;
2967
2968         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2969         if (err < 0)
2970                 return err;
2971
2972         ndm = nlmsg_data(nlh);
2973         if (ndm->ndm_ifindex == 0) {
2974                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2975                 return -EINVAL;
2976         }
2977
2978         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2979         if (dev == NULL) {
2980                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2981                 return -ENODEV;
2982         }
2983
2984         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2985                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2986                 return -EINVAL;
2987         }
2988
2989         addr = nla_data(tb[NDA_LLADDR]);
2990
2991         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2992         if (err)
2993                 return err;
2994
2995         err = -EOPNOTSUPP;
2996
2997         /* Support fdb on master device the net/bridge default case */
2998         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2999             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3000                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3001                 const struct net_device_ops *ops = br_dev->netdev_ops;
3002
3003                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3004                                        nlh->nlmsg_flags);
3005                 if (err)
3006                         goto out;
3007                 else
3008                         ndm->ndm_flags &= ~NTF_MASTER;
3009         }
3010
3011         /* Embedded bridge, macvlan, and any other device support */
3012         if ((ndm->ndm_flags & NTF_SELF)) {
3013                 if (dev->netdev_ops->ndo_fdb_add)
3014                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3015                                                            vid,
3016                                                            nlh->nlmsg_flags);
3017                 else
3018                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3019                                                nlh->nlmsg_flags);
3020
3021                 if (!err) {
3022                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3023                                         ndm->ndm_state);
3024                         ndm->ndm_flags &= ~NTF_SELF;
3025                 }
3026         }
3027 out:
3028         return err;
3029 }
3030
3031 /**
3032  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3033  */
3034 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3035                      struct nlattr *tb[],
3036                      struct net_device *dev,
3037                      const unsigned char *addr, u16 vid)
3038 {
3039         int err = -EINVAL;
3040
3041         /* If aging addresses are supported device will need to
3042          * implement its own handler for this.
3043          */
3044         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3045                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3046                 return err;
3047         }
3048
3049         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3050                 err = dev_uc_del(dev, addr);
3051         else if (is_multicast_ether_addr(addr))
3052                 err = dev_mc_del(dev, addr);
3053
3054         return err;
3055 }
3056 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3057
3058 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3059 {
3060         struct net *net = sock_net(skb->sk);
3061         struct ndmsg *ndm;
3062         struct nlattr *tb[NDA_MAX+1];
3063         struct net_device *dev;
3064         int err = -EINVAL;
3065         __u8 *addr;
3066         u16 vid;
3067
3068         if (!netlink_capable(skb, CAP_NET_ADMIN))
3069                 return -EPERM;
3070
3071         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3072         if (err < 0)
3073                 return err;
3074
3075         ndm = nlmsg_data(nlh);
3076         if (ndm->ndm_ifindex == 0) {
3077                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3078                 return -EINVAL;
3079         }
3080
3081         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3082         if (dev == NULL) {
3083                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3084                 return -ENODEV;
3085         }
3086
3087         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3088                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3089                 return -EINVAL;
3090         }
3091
3092         addr = nla_data(tb[NDA_LLADDR]);
3093
3094         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3095         if (err)
3096                 return err;
3097
3098         err = -EOPNOTSUPP;
3099
3100         /* Support fdb on master device the net/bridge default case */
3101         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3102             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3103                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3104                 const struct net_device_ops *ops = br_dev->netdev_ops;
3105
3106                 if (ops->ndo_fdb_del)
3107                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3108
3109                 if (err)
3110                         goto out;
3111                 else
3112                         ndm->ndm_flags &= ~NTF_MASTER;
3113         }
3114
3115         /* Embedded bridge, macvlan, and any other device support */
3116         if (ndm->ndm_flags & NTF_SELF) {
3117                 if (dev->netdev_ops->ndo_fdb_del)
3118                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3119                                                            vid);
3120                 else
3121                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3122
3123                 if (!err) {
3124                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3125                                         ndm->ndm_state);
3126                         ndm->ndm_flags &= ~NTF_SELF;
3127                 }
3128         }
3129 out:
3130         return err;
3131 }
3132
3133 static int nlmsg_populate_fdb(struct sk_buff *skb,
3134                               struct netlink_callback *cb,
3135                               struct net_device *dev,
3136                               int *idx,
3137                               struct netdev_hw_addr_list *list)
3138 {
3139         struct netdev_hw_addr *ha;
3140         int err;
3141         u32 portid, seq;
3142
3143         portid = NETLINK_CB(cb->skb).portid;
3144         seq = cb->nlh->nlmsg_seq;
3145
3146         list_for_each_entry(ha, &list->list, list) {
3147                 if (*idx < cb->args[2])
3148                         goto skip;
3149
3150                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3151                                               portid, seq,
3152                                               RTM_NEWNEIGH, NTF_SELF,
3153                                               NLM_F_MULTI, NUD_PERMANENT);
3154                 if (err < 0)
3155                         return err;
3156 skip:
3157                 *idx += 1;
3158         }
3159         return 0;
3160 }
3161
3162 /**
3163  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3164  * @nlh: netlink message header
3165  * @dev: netdevice
3166  *
3167  * Default netdevice operation to dump the existing unicast address list.
3168  * Returns number of addresses from list put in skb.
3169  */
3170 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3171                       struct netlink_callback *cb,
3172                       struct net_device *dev,
3173                       struct net_device *filter_dev,
3174                       int *idx)
3175 {
3176         int err;
3177
3178         netif_addr_lock_bh(dev);
3179         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3180         if (err)
3181                 goto out;
3182         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3183 out:
3184         netif_addr_unlock_bh(dev);
3185         return err;
3186 }
3187 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3188
3189 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3190 {
3191         struct net_device *dev;
3192         struct nlattr *tb[IFLA_MAX+1];
3193         struct net_device *br_dev = NULL;
3194         const struct net_device_ops *ops = NULL;
3195         const struct net_device_ops *cops = NULL;
3196         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3197         struct net *net = sock_net(skb->sk);
3198         struct hlist_head *head;
3199         int brport_idx = 0;
3200         int br_idx = 0;
3201         int h, s_h;
3202         int idx = 0, s_idx;
3203         int err = 0;
3204         int fidx = 0;
3205
3206         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3207                         ifla_policy) == 0) {
3208                 if (tb[IFLA_MASTER])
3209                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3210         }
3211
3212         brport_idx = ifm->ifi_index;
3213
3214         if (br_idx) {
3215                 br_dev = __dev_get_by_index(net, br_idx);
3216                 if (!br_dev)
3217                         return -ENODEV;
3218
3219                 ops = br_dev->netdev_ops;
3220         }
3221
3222         s_h = cb->args[0];
3223         s_idx = cb->args[1];
3224
3225         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3226                 idx = 0;
3227                 head = &net->dev_index_head[h];
3228                 hlist_for_each_entry(dev, head, index_hlist) {
3229
3230                         if (brport_idx && (dev->ifindex != brport_idx))
3231                                 continue;
3232
3233                         if (!br_idx) { /* user did not specify a specific bridge */
3234                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3235                                         br_dev = netdev_master_upper_dev_get(dev);
3236                                         cops = br_dev->netdev_ops;
3237                                 }
3238                         } else {
3239                                 if (dev != br_dev &&
3240                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3241                                         continue;
3242
3243                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3244                                     !(dev->priv_flags & IFF_EBRIDGE))
3245                                         continue;
3246                                 cops = ops;
3247                         }
3248
3249                         if (idx < s_idx)
3250                                 goto cont;
3251
3252                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3253                                 if (cops && cops->ndo_fdb_dump) {
3254                                         err = cops->ndo_fdb_dump(skb, cb,
3255                                                                 br_dev, dev,
3256                                                                 &fidx);
3257                                         if (err == -EMSGSIZE)
3258                                                 goto out;
3259                                 }
3260                         }
3261
3262                         if (dev->netdev_ops->ndo_fdb_dump)
3263                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3264                                                                     dev, NULL,
3265                                                                     &fidx);
3266                         else
3267                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3268                                                         &fidx);
3269                         if (err == -EMSGSIZE)
3270                                 goto out;
3271
3272                         cops = NULL;
3273
3274                         /* reset fdb offset to 0 for rest of the interfaces */
3275                         cb->args[2] = 0;
3276                         fidx = 0;
3277 cont:
3278                         idx++;
3279                 }
3280         }
3281
3282 out:
3283         cb->args[0] = h;
3284         cb->args[1] = idx;
3285         cb->args[2] = fidx;
3286
3287         return skb->len;
3288 }
3289
3290 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3291                                unsigned int attrnum, unsigned int flag)
3292 {
3293         if (mask & flag)
3294                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3295         return 0;
3296 }
3297
3298 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3299                             struct net_device *dev, u16 mode,
3300                             u32 flags, u32 mask, int nlflags,
3301                             u32 filter_mask,
3302                             int (*vlan_fill)(struct sk_buff *skb,
3303                                              struct net_device *dev,
3304                                              u32 filter_mask))
3305 {
3306         struct nlmsghdr *nlh;
3307         struct ifinfomsg *ifm;
3308         struct nlattr *br_afspec;
3309         struct nlattr *protinfo;
3310         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3311         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3312         int err = 0;
3313
3314         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3315         if (nlh == NULL)
3316                 return -EMSGSIZE;
3317
3318         ifm = nlmsg_data(nlh);
3319         ifm->ifi_family = AF_BRIDGE;
3320         ifm->__ifi_pad = 0;
3321         ifm->ifi_type = dev->type;
3322         ifm->ifi_index = dev->ifindex;
3323         ifm->ifi_flags = dev_get_flags(dev);
3324         ifm->ifi_change = 0;
3325
3326
3327         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3328             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3329             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3330             (br_dev &&
3331              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3332             (dev->addr_len &&
3333              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3334             (dev->ifindex != dev_get_iflink(dev) &&
3335              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3336                 goto nla_put_failure;
3337
3338         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3339         if (!br_afspec)
3340                 goto nla_put_failure;
3341
3342         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3343                 nla_nest_cancel(skb, br_afspec);
3344                 goto nla_put_failure;
3345         }
3346
3347         if (mode != BRIDGE_MODE_UNDEF) {
3348                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3349                         nla_nest_cancel(skb, br_afspec);
3350                         goto nla_put_failure;
3351                 }
3352         }
3353         if (vlan_fill) {
3354                 err = vlan_fill(skb, dev, filter_mask);
3355                 if (err) {
3356                         nla_nest_cancel(skb, br_afspec);
3357                         goto nla_put_failure;
3358                 }
3359         }
3360         nla_nest_end(skb, br_afspec);
3361
3362         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3363         if (!protinfo)
3364                 goto nla_put_failure;
3365
3366         if (brport_nla_put_flag(skb, flags, mask,
3367                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3368             brport_nla_put_flag(skb, flags, mask,
3369                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3370             brport_nla_put_flag(skb, flags, mask,
3371                                 IFLA_BRPORT_FAST_LEAVE,
3372                                 BR_MULTICAST_FAST_LEAVE) ||
3373             brport_nla_put_flag(skb, flags, mask,
3374                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3375             brport_nla_put_flag(skb, flags, mask,
3376                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3377             brport_nla_put_flag(skb, flags, mask,
3378                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3379             brport_nla_put_flag(skb, flags, mask,
3380                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3381             brport_nla_put_flag(skb, flags, mask,
3382                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3383                 nla_nest_cancel(skb, protinfo);
3384                 goto nla_put_failure;
3385         }
3386
3387         nla_nest_end(skb, protinfo);
3388
3389         nlmsg_end(skb, nlh);
3390         return 0;
3391 nla_put_failure:
3392         nlmsg_cancel(skb, nlh);
3393         return err ? err : -EMSGSIZE;
3394 }
3395 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3396
3397 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3398 {
3399         struct net *net = sock_net(skb->sk);
3400         struct net_device *dev;
3401         int idx = 0;
3402         u32 portid = NETLINK_CB(cb->skb).portid;
3403         u32 seq = cb->nlh->nlmsg_seq;
3404         u32 filter_mask = 0;
3405         int err;
3406
3407         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3408                 struct nlattr *extfilt;
3409
3410                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3411                                           IFLA_EXT_MASK);
3412                 if (extfilt) {
3413                         if (nla_len(extfilt) < sizeof(filter_mask))
3414                                 return -EINVAL;
3415
3416                         filter_mask = nla_get_u32(extfilt);
3417                 }
3418         }
3419
3420         rcu_read_lock();
3421         for_each_netdev_rcu(net, dev) {
3422                 const struct net_device_ops *ops = dev->netdev_ops;
3423                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3424
3425                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3426                         if (idx >= cb->args[0]) {
3427                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3428                                                 skb, portid, seq, dev,
3429                                                 filter_mask, NLM_F_MULTI);
3430                                 if (err < 0 && err != -EOPNOTSUPP)
3431                                         break;
3432                         }
3433                         idx++;
3434                 }
3435
3436                 if (ops->ndo_bridge_getlink) {
3437                         if (idx >= cb->args[0]) {
3438                                 err = ops->ndo_bridge_getlink(skb, portid,
3439                                                               seq, dev,
3440                                                               filter_mask,
3441                                                               NLM_F_MULTI);
3442                                 if (err < 0 && err != -EOPNOTSUPP)
3443                                         break;
3444                         }
3445                         idx++;
3446                 }
3447         }
3448         rcu_read_unlock();
3449         cb->args[0] = idx;
3450
3451         return skb->len;
3452 }
3453
3454 static inline size_t bridge_nlmsg_size(void)
3455 {
3456         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3457                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3458                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3459                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3460                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3461                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3462                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3463                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3464                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3465                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3466                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3467 }
3468
3469 static int rtnl_bridge_notify(struct net_device *dev)
3470 {
3471         struct net *net = dev_net(dev);
3472         struct sk_buff *skb;
3473         int err = -EOPNOTSUPP;
3474
3475         if (!dev->netdev_ops->ndo_bridge_getlink)
3476                 return 0;
3477
3478         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3479         if (!skb) {
3480                 err = -ENOMEM;
3481                 goto errout;
3482         }
3483
3484         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3485         if (err < 0)
3486                 goto errout;
3487
3488         if (!skb->len)
3489                 goto errout;
3490
3491         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3492         return 0;
3493 errout:
3494         WARN_ON(err == -EMSGSIZE);
3495         kfree_skb(skb);
3496         if (err)
3497                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3498         return err;
3499 }
3500
3501 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3502 {
3503         struct net *net = sock_net(skb->sk);
3504         struct ifinfomsg *ifm;
3505         struct net_device *dev;
3506         struct nlattr *br_spec, *attr = NULL;
3507         int rem, err = -EOPNOTSUPP;
3508         u16 flags = 0;
3509         bool have_flags = false;
3510
3511         if (nlmsg_len(nlh) < sizeof(*ifm))
3512                 return -EINVAL;
3513
3514         ifm = nlmsg_data(nlh);
3515         if (ifm->ifi_family != AF_BRIDGE)
3516                 return -EPFNOSUPPORT;
3517
3518         dev = __dev_get_by_index(net, ifm->ifi_index);
3519         if (!dev) {
3520                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3521                 return -ENODEV;
3522         }
3523
3524         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3525         if (br_spec) {
3526                 nla_for_each_nested(attr, br_spec, rem) {
3527                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3528                                 if (nla_len(attr) < sizeof(flags))
3529                                         return -EINVAL;
3530
3531                                 have_flags = true;
3532                                 flags = nla_get_u16(attr);
3533                                 break;
3534                         }
3535                 }
3536         }
3537
3538         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3539                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3540
3541                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3542                         err = -EOPNOTSUPP;
3543                         goto out;
3544                 }
3545
3546                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3547                 if (err)
3548                         goto out;
3549
3550                 flags &= ~BRIDGE_FLAGS_MASTER;
3551         }
3552
3553         if ((flags & BRIDGE_FLAGS_SELF)) {
3554                 if (!dev->netdev_ops->ndo_bridge_setlink)
3555                         err = -EOPNOTSUPP;
3556                 else
3557                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3558                                                                   flags);
3559                 if (!err) {
3560                         flags &= ~BRIDGE_FLAGS_SELF;
3561
3562                         /* Generate event to notify upper layer of bridge
3563                          * change
3564                          */
3565                         err = rtnl_bridge_notify(dev);
3566                 }
3567         }
3568
3569         if (have_flags)
3570                 memcpy(nla_data(attr), &flags, sizeof(flags));
3571 out:
3572         return err;
3573 }
3574
3575 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3576 {
3577         struct net *net = sock_net(skb->sk);
3578         struct ifinfomsg *ifm;
3579         struct net_device *dev;
3580         struct nlattr *br_spec, *attr = NULL;
3581         int rem, err = -EOPNOTSUPP;
3582         u16 flags = 0;
3583         bool have_flags = false;
3584
3585         if (nlmsg_len(nlh) < sizeof(*ifm))
3586                 return -EINVAL;
3587
3588         ifm = nlmsg_data(nlh);
3589         if (ifm->ifi_family != AF_BRIDGE)
3590                 return -EPFNOSUPPORT;
3591
3592         dev = __dev_get_by_index(net, ifm->ifi_index);
3593         if (!dev) {
3594                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3595                 return -ENODEV;
3596         }
3597
3598         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3599         if (br_spec) {
3600                 nla_for_each_nested(attr, br_spec, rem) {
3601                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3602                                 if (nla_len(attr) < sizeof(flags))
3603                                         return -EINVAL;
3604
3605                                 have_flags = true;
3606                                 flags = nla_get_u16(attr);
3607                                 break;
3608                         }
3609                 }
3610         }
3611
3612         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3613                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3614
3615                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3616                         err = -EOPNOTSUPP;
3617                         goto out;
3618                 }
3619
3620                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3621                 if (err)
3622                         goto out;
3623
3624                 flags &= ~BRIDGE_FLAGS_MASTER;
3625         }
3626
3627         if ((flags & BRIDGE_FLAGS_SELF)) {
3628                 if (!dev->netdev_ops->ndo_bridge_dellink)
3629                         err = -EOPNOTSUPP;
3630                 else
3631                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3632                                                                   flags);
3633
3634                 if (!err) {
3635                         flags &= ~BRIDGE_FLAGS_SELF;
3636
3637                         /* Generate event to notify upper layer of bridge
3638                          * change
3639                          */
3640                         err = rtnl_bridge_notify(dev);
3641                 }
3642         }
3643
3644         if (have_flags)
3645                 memcpy(nla_data(attr), &flags, sizeof(flags));
3646 out:
3647         return err;
3648 }
3649
3650 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3651 {
3652         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3653                (!idxattr || idxattr == attrid);
3654 }
3655
3656 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3657 static int rtnl_get_offload_stats_attr_size(int attr_id)
3658 {
3659         switch (attr_id) {
3660         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3661                 return sizeof(struct rtnl_link_stats64);
3662         }
3663
3664         return 0;
3665 }
3666
3667 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3668                                   int *prividx)
3669 {
3670         struct nlattr *attr = NULL;
3671         int attr_id, size;
3672         void *attr_data;
3673         int err;
3674
3675         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3676               dev->netdev_ops->ndo_get_offload_stats))
3677                 return -ENODATA;
3678
3679         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3680              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3681                 if (attr_id < *prividx)
3682                         continue;
3683
3684                 size = rtnl_get_offload_stats_attr_size(attr_id);
3685                 if (!size)
3686                         continue;
3687
3688                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3689                         continue;
3690
3691                 attr = nla_reserve_64bit(skb, attr_id, size,
3692                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3693                 if (!attr)
3694                         goto nla_put_failure;
3695
3696                 attr_data = nla_data(attr);
3697                 memset(attr_data, 0, size);
3698                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3699                                                              attr_data);
3700                 if (err)
3701                         goto get_offload_stats_failure;
3702         }
3703
3704         if (!attr)
3705                 return -ENODATA;
3706
3707         *prividx = 0;
3708         return 0;
3709
3710 nla_put_failure:
3711         err = -EMSGSIZE;
3712 get_offload_stats_failure:
3713         *prividx = attr_id;
3714         return err;
3715 }
3716
3717 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3718 {
3719         int nla_size = 0;
3720         int attr_id;
3721         int size;
3722
3723         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3724               dev->netdev_ops->ndo_get_offload_stats))
3725                 return 0;
3726
3727         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3728              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3729                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3730                         continue;
3731                 size = rtnl_get_offload_stats_attr_size(attr_id);
3732                 nla_size += nla_total_size_64bit(size);
3733         }
3734
3735         if (nla_size != 0)
3736                 nla_size += nla_total_size(0);
3737
3738         return nla_size;
3739 }
3740
3741 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3742                                int type, u32 pid, u32 seq, u32 change,
3743                                unsigned int flags, unsigned int filter_mask,
3744                                int *idxattr, int *prividx)
3745 {
3746         struct if_stats_msg *ifsm;
3747         struct nlmsghdr *nlh;
3748         struct nlattr *attr;
3749         int s_prividx = *prividx;
3750         int err;
3751
3752         ASSERT_RTNL();
3753
3754         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3755         if (!nlh)
3756                 return -EMSGSIZE;
3757
3758         ifsm = nlmsg_data(nlh);
3759         ifsm->ifindex = dev->ifindex;
3760         ifsm->filter_mask = filter_mask;
3761
3762         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3763                 struct rtnl_link_stats64 *sp;
3764
3765                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3766                                          sizeof(struct rtnl_link_stats64),
3767                                          IFLA_STATS_UNSPEC);
3768                 if (!attr)
3769                         goto nla_put_failure;
3770
3771                 sp = nla_data(attr);
3772                 dev_get_stats(dev, sp);
3773         }
3774
3775         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3776                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3777
3778                 if (ops && ops->fill_linkxstats) {
3779                         *idxattr = IFLA_STATS_LINK_XSTATS;
3780                         attr = nla_nest_start(skb,
3781                                               IFLA_STATS_LINK_XSTATS);
3782                         if (!attr)
3783                                 goto nla_put_failure;
3784
3785                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3786                         nla_nest_end(skb, attr);
3787                         if (err)
3788                                 goto nla_put_failure;
3789                         *idxattr = 0;
3790                 }
3791         }
3792
3793         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3794                              *idxattr)) {
3795                 const struct rtnl_link_ops *ops = NULL;
3796                 const struct net_device *master;
3797
3798                 master = netdev_master_upper_dev_get(dev);
3799                 if (master)
3800                         ops = master->rtnl_link_ops;
3801                 if (ops && ops->fill_linkxstats) {
3802                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3803                         attr = nla_nest_start(skb,
3804                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3805                         if (!attr)
3806                                 goto nla_put_failure;
3807
3808                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3809                         nla_nest_end(skb, attr);
3810                         if (err)
3811                                 goto nla_put_failure;
3812                         *idxattr = 0;
3813                 }
3814         }
3815
3816         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3817                              *idxattr)) {
3818                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3819                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3820                 if (!attr)
3821                         goto nla_put_failure;
3822
3823                 err = rtnl_get_offload_stats(skb, dev, prividx);
3824                 if (err == -ENODATA)
3825                         nla_nest_cancel(skb, attr);
3826                 else
3827                         nla_nest_end(skb, attr);
3828
3829                 if (err && err != -ENODATA)
3830                         goto nla_put_failure;
3831                 *idxattr = 0;
3832         }
3833
3834         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3835                 struct rtnl_af_ops *af_ops;
3836
3837                 *idxattr = IFLA_STATS_AF_SPEC;
3838                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3839                 if (!attr)
3840                         goto nla_put_failure;
3841
3842                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3843                         if (af_ops->fill_stats_af) {
3844                                 struct nlattr *af;
3845                                 int err;
3846
3847                                 af = nla_nest_start(skb, af_ops->family);
3848                                 if (!af)
3849                                         goto nla_put_failure;
3850
3851                                 err = af_ops->fill_stats_af(skb, dev);
3852
3853                                 if (err == -ENODATA)
3854                                         nla_nest_cancel(skb, af);
3855                                 else if (err < 0)
3856                                         goto nla_put_failure;
3857
3858                                 nla_nest_end(skb, af);
3859                         }
3860                 }
3861
3862                 nla_nest_end(skb, attr);
3863
3864                 *idxattr = 0;
3865         }
3866
3867         nlmsg_end(skb, nlh);
3868
3869         return 0;
3870
3871 nla_put_failure:
3872         /* not a multi message or no progress mean a real error */
3873         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3874                 nlmsg_cancel(skb, nlh);
3875         else
3876                 nlmsg_end(skb, nlh);
3877
3878         return -EMSGSIZE;
3879 }
3880
3881 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3882                                   u32 filter_mask)
3883 {
3884         size_t size = 0;
3885
3886         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3887                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3888
3889         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3890                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3891                 int attr = IFLA_STATS_LINK_XSTATS;
3892
3893                 if (ops && ops->get_linkxstats_size) {
3894                         size += nla_total_size(ops->get_linkxstats_size(dev,
3895                                                                         attr));
3896                         /* for IFLA_STATS_LINK_XSTATS */
3897                         size += nla_total_size(0);
3898                 }
3899         }
3900
3901         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3902                 struct net_device *_dev = (struct net_device *)dev;
3903                 const struct rtnl_link_ops *ops = NULL;
3904                 const struct net_device *master;
3905
3906                 /* netdev_master_upper_dev_get can't take const */
3907                 master = netdev_master_upper_dev_get(_dev);
3908                 if (master)
3909                         ops = master->rtnl_link_ops;
3910                 if (ops && ops->get_linkxstats_size) {
3911                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3912
3913                         size += nla_total_size(ops->get_linkxstats_size(dev,
3914                                                                         attr));
3915                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3916                         size += nla_total_size(0);
3917                 }
3918         }
3919
3920         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3921                 size += rtnl_get_offload_stats_size(dev);
3922
3923         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3924                 struct rtnl_af_ops *af_ops;
3925
3926                 /* for IFLA_STATS_AF_SPEC */
3927                 size += nla_total_size(0);
3928
3929                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3930                         if (af_ops->get_stats_af_size) {
3931                                 size += nla_total_size(
3932                                         af_ops->get_stats_af_size(dev));
3933
3934                                 /* for AF_* */
3935                                 size += nla_total_size(0);
3936                         }
3937                 }
3938         }
3939
3940         return size;
3941 }
3942
3943 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3944 {
3945         struct net *net = sock_net(skb->sk);
3946         struct net_device *dev = NULL;
3947         int idxattr = 0, prividx = 0;
3948         struct if_stats_msg *ifsm;
3949         struct sk_buff *nskb;
3950         u32 filter_mask;
3951         int err;
3952
3953         if (nlmsg_len(nlh) < sizeof(*ifsm))
3954                 return -EINVAL;
3955
3956         ifsm = nlmsg_data(nlh);
3957         if (ifsm->ifindex > 0)
3958                 dev = __dev_get_by_index(net, ifsm->ifindex);
3959         else
3960                 return -EINVAL;
3961
3962         if (!dev)
3963                 return -ENODEV;
3964
3965         filter_mask = ifsm->filter_mask;
3966         if (!filter_mask)
3967                 return -EINVAL;
3968
3969         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3970         if (!nskb)
3971                 return -ENOBUFS;
3972
3973         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3974                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3975                                   0, filter_mask, &idxattr, &prividx);
3976         if (err < 0) {
3977                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3978                 WARN_ON(err == -EMSGSIZE);
3979                 kfree_skb(nskb);
3980         } else {
3981                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3982         }
3983
3984         return err;
3985 }
3986
3987 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3988 {
3989         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3990         struct net *net = sock_net(skb->sk);
3991         unsigned int flags = NLM_F_MULTI;
3992         struct if_stats_msg *ifsm;
3993         struct hlist_head *head;
3994         struct net_device *dev;
3995         u32 filter_mask = 0;
3996         int idx = 0;
3997
3998         s_h = cb->args[0];
3999         s_idx = cb->args[1];
4000         s_idxattr = cb->args[2];
4001         s_prividx = cb->args[3];
4002
4003         cb->seq = net->dev_base_seq;
4004
4005         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4006                 return -EINVAL;
4007
4008         ifsm = nlmsg_data(cb->nlh);
4009         filter_mask = ifsm->filter_mask;
4010         if (!filter_mask)
4011                 return -EINVAL;
4012
4013         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4014                 idx = 0;
4015                 head = &net->dev_index_head[h];
4016                 hlist_for_each_entry(dev, head, index_hlist) {
4017                         if (idx < s_idx)
4018                                 goto cont;
4019                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4020                                                   NETLINK_CB(cb->skb).portid,
4021                                                   cb->nlh->nlmsg_seq, 0,
4022                                                   flags, filter_mask,
4023                                                   &s_idxattr, &s_prividx);
4024                         /* If we ran out of room on the first message,
4025                          * we're in trouble
4026                          */
4027                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4028
4029                         if (err < 0)
4030                                 goto out;
4031                         s_prividx = 0;
4032                         s_idxattr = 0;
4033                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4034 cont:
4035                         idx++;
4036                 }
4037         }
4038 out:
4039         cb->args[3] = s_prividx;
4040         cb->args[2] = s_idxattr;
4041         cb->args[1] = idx;
4042         cb->args[0] = h;
4043
4044         return skb->len;
4045 }
4046
4047 /* Process one rtnetlink message. */
4048
4049 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
4050 {
4051         struct net *net = sock_net(skb->sk);
4052         rtnl_doit_func doit;
4053         int kind;
4054         int family;
4055         int type;
4056         int err;
4057
4058         type = nlh->nlmsg_type;
4059         if (type > RTM_MAX)
4060                 return -EOPNOTSUPP;
4061
4062         type -= RTM_BASE;
4063
4064         /* All the messages must have at least 1 byte length */
4065         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4066                 return 0;
4067
4068         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4069         kind = type&3;
4070
4071         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4072                 return -EPERM;
4073
4074         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4075                 struct sock *rtnl;
4076                 rtnl_dumpit_func dumpit;
4077                 rtnl_calcit_func calcit;
4078                 u16 min_dump_alloc = 0;
4079
4080                 dumpit = rtnl_get_dumpit(family, type);
4081                 if (dumpit == NULL)
4082                         return -EOPNOTSUPP;
4083                 calcit = rtnl_get_calcit(family, type);
4084                 if (calcit)
4085                         min_dump_alloc = calcit(skb, nlh);
4086
4087                 __rtnl_unlock();
4088                 rtnl = net->rtnl;
4089                 {
4090                         struct netlink_dump_control c = {
4091                                 .dump           = dumpit,
4092                                 .min_dump_alloc = min_dump_alloc,
4093                         };
4094                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4095                 }
4096                 rtnl_lock();
4097                 return err;
4098         }
4099
4100         doit = rtnl_get_doit(family, type);
4101         if (doit == NULL)
4102                 return -EOPNOTSUPP;
4103
4104         return doit(skb, nlh);
4105 }
4106
4107 static void rtnetlink_rcv(struct sk_buff *skb)
4108 {
4109         rtnl_lock();
4110         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4111         rtnl_unlock();
4112 }
4113
4114 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4115 {
4116         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4117
4118         switch (event) {
4119         case NETDEV_REBOOT:
4120         case NETDEV_CHANGENAME:
4121         case NETDEV_FEAT_CHANGE:
4122         case NETDEV_BONDING_FAILOVER:
4123         case NETDEV_NOTIFY_PEERS:
4124         case NETDEV_CHANGEUPPER:
4125         case NETDEV_RESEND_IGMP:
4126         case NETDEV_CHANGEINFODATA:
4127         case NETDEV_PRECHANGEUPPER:
4128         case NETDEV_CHANGELOWERSTATE:
4129         case NETDEV_CHANGE_TX_QUEUE_LEN:
4130                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4131                 break;
4132         default:
4133                 break;
4134         }
4135         return NOTIFY_DONE;
4136 }
4137
4138 static struct notifier_block rtnetlink_dev_notifier = {
4139         .notifier_call  = rtnetlink_event,
4140 };
4141
4142
4143 static int __net_init rtnetlink_net_init(struct net *net)
4144 {
4145         struct sock *sk;
4146         struct netlink_kernel_cfg cfg = {
4147                 .groups         = RTNLGRP_MAX,
4148                 .input          = rtnetlink_rcv,
4149                 .cb_mutex       = &rtnl_mutex,
4150                 .flags          = NL_CFG_F_NONROOT_RECV,
4151         };
4152
4153         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4154         if (!sk)
4155                 return -ENOMEM;
4156         net->rtnl = sk;
4157         return 0;
4158 }
4159
4160 static void __net_exit rtnetlink_net_exit(struct net *net)
4161 {
4162         netlink_kernel_release(net->rtnl);
4163         net->rtnl = NULL;
4164 }
4165
4166 static struct pernet_operations rtnetlink_net_ops = {
4167         .init = rtnetlink_net_init,
4168         .exit = rtnetlink_net_exit,
4169 };
4170
4171 void __init rtnetlink_init(void)
4172 {
4173         if (register_pernet_subsys(&rtnetlink_net_ops))
4174                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4175
4176         register_netdevice_notifier(&rtnetlink_dev_notifier);
4177
4178         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4179                       rtnl_dump_ifinfo, rtnl_calcit);
4180         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4181         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4182         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4183
4184         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4185         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4186         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4187
4188         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4189         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4190         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4191
4192         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4193         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4194         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4195
4196         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4197                       NULL);
4198 }