net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <asm/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 static struct sk_buff *defer_kfree_skb_list;
  75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  76 {
  77         if (head && tail) {
  78                 tail->next = defer_kfree_skb_list;
  79                 defer_kfree_skb_list = head;
  80         }
  81 }
  82 EXPORT_SYMBOL(rtnl_kfree_skbs);
  83
  84 void __rtnl_unlock(void)
  85 {
  86         struct sk_buff *head = defer_kfree_skb_list;
  87
  88         defer_kfree_skb_list = NULL;
  89
  90         mutex_unlock(&rtnl_mutex);
  91
  92         while (head) {
  93                 struct sk_buff *next = head->next;
  94
  95                 kfree_skb(head);
  96                 cond_resched();
  97                 head = next;
  98         }
  99 }
 100
 101 void rtnl_unlock(void)
 102 {
 103         /* This fellow will unlock it for us. */
 104         netdev_run_todo();
 105 }
 106 EXPORT_SYMBOL(rtnl_unlock);
 107
 108 int rtnl_trylock(void)
 109 {
 110         return mutex_trylock(&rtnl_mutex);
 111 }
 112 EXPORT_SYMBOL(rtnl_trylock);
 113
 114 int rtnl_is_locked(void)
 115 {
 116         return mutex_is_locked(&rtnl_mutex);
 117 }
 118 EXPORT_SYMBOL(rtnl_is_locked);
 119
 120 #ifdef CONFIG_PROVE_LOCKING
 121 bool lockdep_rtnl_is_held(void)
 122 {
 123         return lockdep_is_held(&rtnl_mutex);
 124 }
 125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 127
 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 129
 130 static inline int rtm_msgindex(int msgtype)
 131 {
 132         int msgindex = msgtype - RTM_BASE;
 133
 134         /*
 135          * msgindex < 0 implies someone tried to register a netlink
 136          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 137          * the message type has not been added to linux/rtnetlink.h
 138          */
 139         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 140
 141         return msgindex;
 142 }
 143
 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 145 {
 146         struct rtnl_link *tab;
 147
 148         if (protocol <= RTNL_FAMILY_MAX)
 149                 tab = rtnl_msg_handlers[protocol];
 150         else
 151                 tab = NULL;
 152
 153         if (tab == NULL || tab[msgindex].doit == NULL)
 154                 tab = rtnl_msg_handlers[PF_UNSPEC];
 155
 156         return tab[msgindex].doit;
 157 }
 158
 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 160 {
 161         struct rtnl_link *tab;
 162
 163         if (protocol <= RTNL_FAMILY_MAX)
 164                 tab = rtnl_msg_handlers[protocol];
 165         else
 166                 tab = NULL;
 167
 168         if (tab == NULL || tab[msgindex].dumpit == NULL)
 169                 tab = rtnl_msg_handlers[PF_UNSPEC];
 170
 171         return tab[msgindex].dumpit;
 172 }
 173
 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 175 {
 176         struct rtnl_link *tab;
 177
 178         if (protocol <= RTNL_FAMILY_MAX)
 179                 tab = rtnl_msg_handlers[protocol];
 180         else
 181                 tab = NULL;
 182
 183         if (tab == NULL || tab[msgindex].calcit == NULL)
 184                 tab = rtnl_msg_handlers[PF_UNSPEC];
 185
 186         return tab[msgindex].calcit;
 187 }
 188
 189 /**
 190  * __rtnl_register - Register a rtnetlink message type
 191  * @protocol: Protocol family or PF_UNSPEC
 192  * @msgtype: rtnetlink message type
 193  * @doit: Function pointer called for each request message
 194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 195  * @calcit: Function pointer to calc size of dump message
 196  *
 197  * Registers the specified function pointers (at least one of them has
 198  * to be non-NULL) to be called whenever a request message for the
 199  * specified protocol family and message type is received.
 200  *
 201  * The special protocol family PF_UNSPEC may be used to define fallback
 202  * function pointers for the case when no entry for the specific protocol
 203  * family exists.
 204  *
 205  * Returns 0 on success or a negative error code.
 206  */
 207 int __rtnl_register(int protocol, int msgtype,
 208                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 209                     rtnl_calcit_func calcit)
 210 {
 211         struct rtnl_link *tab;
 212         int msgindex;
 213
 214         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 215         msgindex = rtm_msgindex(msgtype);
 216
 217         tab = rtnl_msg_handlers[protocol];
 218         if (tab == NULL) {
 219                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 220                 if (tab == NULL)
 221                         return -ENOBUFS;
 222
 223                 rtnl_msg_handlers[protocol] = tab;
 224         }
 225
 226         if (doit)
 227                 tab[msgindex].doit = doit;
 228
 229         if (dumpit)
 230                 tab[msgindex].dumpit = dumpit;
 231
 232         if (calcit)
 233                 tab[msgindex].calcit = calcit;
 234
 235         return 0;
 236 }
 237 EXPORT_SYMBOL_GPL(__rtnl_register);
 238
 239 /**
 240  * rtnl_register - Register a rtnetlink message type
 241  *
 242  * Identical to __rtnl_register() but panics on failure. This is useful
 243  * as failure of this function is very unlikely, it can only happen due
 244  * to lack of memory when allocating the chain to store all message
 245  * handlers for a protocol. Meant for use in init functions where lack
 246  * of memory implies no sense in continuing.
 247  */
 248 void rtnl_register(int protocol, int msgtype,
 249                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 250                    rtnl_calcit_func calcit)
 251 {
 252         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 253                 panic("Unable to register rtnetlink message handler, "
 254                       "protocol = %d, message type = %d\n",
 255                       protocol, msgtype);
 256 }
 257 EXPORT_SYMBOL_GPL(rtnl_register);
 258
 259 /**
 260  * rtnl_unregister - Unregister a rtnetlink message type
 261  * @protocol: Protocol family or PF_UNSPEC
 262  * @msgtype: rtnetlink message type
 263  *
 264  * Returns 0 on success or a negative error code.
 265  */
 266 int rtnl_unregister(int protocol, int msgtype)
 267 {
 268         int msgindex;
 269
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271         msgindex = rtm_msgindex(msgtype);
 272
 273         if (rtnl_msg_handlers[protocol] == NULL)
 274                 return -ENOENT;
 275
 276         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 277         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 278         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 && dev_is_pci(dev->dev.parent) &&
 841             (ext_filter_mask & RTEXT_FILTER_VF)) {
 842                 int num_vfs = dev_num_vf(dev->dev.parent);
 843                 size_t size = nla_total_size(0);
 844                 size += num_vfs *
 845                         (nla_total_size(0) +
 846                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 847                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 848                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 849                          nla_total_size(MAX_VLAN_LIST_LEN *
 850                                         sizeof(struct ifla_vf_vlan_info)) +
 851                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 852                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 853                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 854                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 855                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 856                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 857                          /* IFLA_VF_STATS_RX_PACKETS */
 858                          nla_total_size_64bit(sizeof(__u64)) +
 859                          /* IFLA_VF_STATS_TX_PACKETS */
 860                          nla_total_size_64bit(sizeof(__u64)) +
 861                          /* IFLA_VF_STATS_RX_BYTES */
 862                          nla_total_size_64bit(sizeof(__u64)) +
 863                          /* IFLA_VF_STATS_TX_BYTES */
 864                          nla_total_size_64bit(sizeof(__u64)) +
 865                          /* IFLA_VF_STATS_BROADCAST */
 866                          nla_total_size_64bit(sizeof(__u64)) +
 867                          /* IFLA_VF_STATS_MULTICAST */
 868                          nla_total_size_64bit(sizeof(__u64)) +
 869                          nla_total_size(sizeof(struct ifla_vf_trust)));
 870                 return size;
 871         } else
 872                 return 0;
 873 }
 874
 875 static size_t rtnl_port_size(const struct net_device *dev,
 876                              u32 ext_filter_mask)
 877 {
 878         size_t port_size = nla_total_size(4)            /* PORT_VF */
 879                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 880                 + nla_total_size(sizeof(struct ifla_port_vsi))
 881                                                         /* PORT_VSI_TYPE */
 882                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 883                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 884                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 885                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 886         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 887         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 888                 + port_size;
 889         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 890                 + port_size;
 891
 892         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 893             !(ext_filter_mask & RTEXT_FILTER_VF))
 894                 return 0;
 895         if (dev_num_vf(dev->dev.parent))
 896                 return port_self_size + vf_ports_size +
 897                         vf_port_size * dev_num_vf(dev->dev.parent);
 898         else
 899                 return port_self_size;
 900 }
 901
 902 static size_t rtnl_xdp_size(const struct net_device *dev)
 903 {
 904         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 905                           nla_total_size(1);    /* XDP_ATTACHED */
 906
 907         if (!dev->netdev_ops->ndo_xdp)
 908                 return 0;
 909         else
 910                 return xdp_size;
 911 }
 912
 913 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 914                                      u32 ext_filter_mask)
 915 {
 916         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 917                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 918                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 919                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 920                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 921                + nla_total_size(sizeof(struct rtnl_link_stats))
 922                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 923                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 924                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 925                + nla_total_size(4) /* IFLA_TXQLEN */
 926                + nla_total_size(4) /* IFLA_WEIGHT */
 927                + nla_total_size(4) /* IFLA_MTU */
 928                + nla_total_size(4) /* IFLA_LINK */
 929                + nla_total_size(4) /* IFLA_MASTER */
 930                + nla_total_size(1) /* IFLA_CARRIER */
 931                + nla_total_size(4) /* IFLA_PROMISCUITY */
 932                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 933                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 934                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 935                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 936                + nla_total_size(1) /* IFLA_OPERSTATE */
 937                + nla_total_size(1) /* IFLA_LINKMODE */
 938                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 939                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 940                + nla_total_size(ext_filter_mask
 941                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 942                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 943                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 944                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 945                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 946                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 947                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 948                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 949                + rtnl_xdp_size(dev) /* IFLA_XDP */
 950                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 951
 952 }
 953
 954 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 955 {
 956         struct nlattr *vf_ports;
 957         struct nlattr *vf_port;
 958         int vf;
 959         int err;
 960
 961         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 962         if (!vf_ports)
 963                 return -EMSGSIZE;
 964
 965         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 966                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 967                 if (!vf_port)
 968                         goto nla_put_failure;
 969                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 970                         goto nla_put_failure;
 971                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 972                 if (err == -EMSGSIZE)
 973                         goto nla_put_failure;
 974                 if (err) {
 975                         nla_nest_cancel(skb, vf_port);
 976                         continue;
 977                 }
 978                 nla_nest_end(skb, vf_port);
 979         }
 980
 981         nla_nest_end(skb, vf_ports);
 982
 983         return 0;
 984
 985 nla_put_failure:
 986         nla_nest_cancel(skb, vf_ports);
 987         return -EMSGSIZE;
 988 }
 989
 990 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 991 {
 992         struct nlattr *port_self;
 993         int err;
 994
 995         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 996         if (!port_self)
 997                 return -EMSGSIZE;
 998
 999         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1000         if (err) {
1001                 nla_nest_cancel(skb, port_self);
1002                 return (err == -EMSGSIZE) ? err : 0;
1003         }
1004
1005         nla_nest_end(skb, port_self);
1006
1007         return 0;
1008 }
1009
1010 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1011                           u32 ext_filter_mask)
1012 {
1013         int err;
1014
1015         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1016             !(ext_filter_mask & RTEXT_FILTER_VF))
1017                 return 0;
1018
1019         err = rtnl_port_self_fill(skb, dev);
1020         if (err)
1021                 return err;
1022
1023         if (dev_num_vf(dev->dev.parent)) {
1024                 err = rtnl_vf_ports_fill(skb, dev);
1025                 if (err)
1026                         return err;
1027         }
1028
1029         return 0;
1030 }
1031
1032 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1033 {
1034         int err;
1035         struct netdev_phys_item_id ppid;
1036
1037         err = dev_get_phys_port_id(dev, &ppid);
1038         if (err) {
1039                 if (err == -EOPNOTSUPP)
1040                         return 0;
1041                 return err;
1042         }
1043
1044         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1045                 return -EMSGSIZE;
1046
1047         return 0;
1048 }
1049
1050 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1051 {
1052         char name[IFNAMSIZ];
1053         int err;
1054
1055         err = dev_get_phys_port_name(dev, name, sizeof(name));
1056         if (err) {
1057                 if (err == -EOPNOTSUPP)
1058                         return 0;
1059                 return err;
1060         }
1061
1062         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1063                 return -EMSGSIZE;
1064
1065         return 0;
1066 }
1067
1068 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1069 {
1070         int err;
1071         struct switchdev_attr attr = {
1072                 .orig_dev = dev,
1073                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1074                 .flags = SWITCHDEV_F_NO_RECURSE,
1075         };
1076
1077         err = switchdev_port_attr_get(dev, &attr);
1078         if (err) {
1079                 if (err == -EOPNOTSUPP)
1080                         return 0;
1081                 return err;
1082         }
1083
1084         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1085                     attr.u.ppid.id))
1086                 return -EMSGSIZE;
1087
1088         return 0;
1089 }
1090
1091 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1092                                               struct net_device *dev)
1093 {
1094         struct rtnl_link_stats64 *sp;
1095         struct nlattr *attr;
1096
1097         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1098                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1099         if (!attr)
1100                 return -EMSGSIZE;
1101
1102         sp = nla_data(attr);
1103         dev_get_stats(dev, sp);
1104
1105         attr = nla_reserve(skb, IFLA_STATS,
1106                            sizeof(struct rtnl_link_stats));
1107         if (!attr)
1108                 return -EMSGSIZE;
1109
1110         copy_rtnl_link_stats(nla_data(attr), sp);
1111
1112         return 0;
1113 }
1114
1115 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1116                                                struct net_device *dev,
1117                                                int vfs_num,
1118                                                struct nlattr *vfinfo)
1119 {
1120         struct ifla_vf_rss_query_en vf_rss_query_en;
1121         struct nlattr *vf, *vfstats, *vfvlanlist;
1122         struct ifla_vf_link_state vf_linkstate;
1123         struct ifla_vf_vlan_info vf_vlan_info;
1124         struct ifla_vf_spoofchk vf_spoofchk;
1125         struct ifla_vf_tx_rate vf_tx_rate;
1126         struct ifla_vf_stats vf_stats;
1127         struct ifla_vf_trust vf_trust;
1128         struct ifla_vf_vlan vf_vlan;
1129         struct ifla_vf_rate vf_rate;
1130         struct ifla_vf_mac vf_mac;
1131         struct ifla_vf_info ivi;
1132
1133         /* Not all SR-IOV capable drivers support the
1134          * spoofcheck and "RSS query enable" query.  Preset to
1135          * -1 so the user space tool can detect that the driver
1136          * didn't report anything.
1137          */
1138         ivi.spoofchk = -1;
1139         ivi.rss_query_en = -1;
1140         ivi.trusted = -1;
1141         memset(ivi.mac, 0, sizeof(ivi.mac));
1142         /* The default value for VF link state is "auto"
1143          * IFLA_VF_LINK_STATE_AUTO which equals zero
1144          */
1145         ivi.linkstate = 0;
1146         /* VLAN Protocol by default is 802.1Q */
1147         ivi.vlan_proto = htons(ETH_P_8021Q);
1148         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1149                 return 0;
1150
1151         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1152
1153         vf_mac.vf =
1154                 vf_vlan.vf =
1155                 vf_vlan_info.vf =
1156                 vf_rate.vf =
1157                 vf_tx_rate.vf =
1158                 vf_spoofchk.vf =
1159                 vf_linkstate.vf =
1160                 vf_rss_query_en.vf =
1161                 vf_trust.vf = ivi.vf;
1162
1163         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1164         vf_vlan.vlan = ivi.vlan;
1165         vf_vlan.qos = ivi.qos;
1166         vf_vlan_info.vlan = ivi.vlan;
1167         vf_vlan_info.qos = ivi.qos;
1168         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1169         vf_tx_rate.rate = ivi.max_tx_rate;
1170         vf_rate.min_tx_rate = ivi.min_tx_rate;
1171         vf_rate.max_tx_rate = ivi.max_tx_rate;
1172         vf_spoofchk.setting = ivi.spoofchk;
1173         vf_linkstate.link_state = ivi.linkstate;
1174         vf_rss_query_en.setting = ivi.rss_query_en;
1175         vf_trust.setting = ivi.trusted;
1176         vf = nla_nest_start(skb, IFLA_VF_INFO);
1177         if (!vf)
1178                 goto nla_put_vfinfo_failure;
1179         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1182                     &vf_rate) ||
1183             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1184                     &vf_tx_rate) ||
1185             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1186                     &vf_spoofchk) ||
1187             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1188                     &vf_linkstate) ||
1189             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190                     sizeof(vf_rss_query_en),
1191                     &vf_rss_query_en) ||
1192             nla_put(skb, IFLA_VF_TRUST,
1193                     sizeof(vf_trust), &vf_trust))
1194                 goto nla_put_vf_failure;
1195         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1196         if (!vfvlanlist)
1197                 goto nla_put_vf_failure;
1198         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1199                     &vf_vlan_info)) {
1200                 nla_nest_cancel(skb, vfvlanlist);
1201                 goto nla_put_vf_failure;
1202         }
1203         nla_nest_end(skb, vfvlanlist);
1204         memset(&vf_stats, 0, sizeof(vf_stats));
1205         if (dev->netdev_ops->ndo_get_vf_stats)
1206                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1207                                                 &vf_stats);
1208         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1209         if (!vfstats)
1210                 goto nla_put_vf_failure;
1211         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1212                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1214                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1216                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1217             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1218                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1219             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1220                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1221             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1222                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1223                 nla_nest_cancel(skb, vfstats);
1224                 goto nla_put_vf_failure;
1225         }
1226         nla_nest_end(skb, vfstats);
1227         nla_nest_end(skb, vf);
1228         return 0;
1229
1230 nla_put_vf_failure:
1231         nla_nest_cancel(skb, vf);
1232 nla_put_vfinfo_failure:
1233         nla_nest_cancel(skb, vfinfo);
1234         return -EMSGSIZE;
1235 }
1236
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1238 {
1239         struct rtnl_link_ifmap map;
1240
1241         memset(&map, 0, sizeof(map));
1242         map.mem_start   = dev->mem_start;
1243         map.mem_end     = dev->mem_end;
1244         map.base_addr   = dev->base_addr;
1245         map.irq         = dev->irq;
1246         map.dma         = dev->dma;
1247         map.port        = dev->if_port;
1248
1249         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1250                 return -EMSGSIZE;
1251
1252         return 0;
1253 }
1254
1255 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1256 {
1257         struct netdev_xdp xdp_op = {};
1258         struct nlattr *xdp;
1259         int err;
1260
1261         if (!dev->netdev_ops->ndo_xdp)
1262                 return 0;
1263         xdp = nla_nest_start(skb, IFLA_XDP);
1264         if (!xdp)
1265                 return -EMSGSIZE;
1266         xdp_op.command = XDP_QUERY_PROG;
1267         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1268         if (err)
1269                 goto err_cancel;
1270         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1271         if (err)
1272                 goto err_cancel;
1273
1274         nla_nest_end(skb, xdp);
1275         return 0;
1276
1277 err_cancel:
1278         nla_nest_cancel(skb, xdp);
1279         return err;
1280 }
1281
1282 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1283                             int type, u32 pid, u32 seq, u32 change,
1284                             unsigned int flags, u32 ext_filter_mask)
1285 {
1286         struct ifinfomsg *ifm;
1287         struct nlmsghdr *nlh;
1288         struct nlattr *af_spec;
1289         struct rtnl_af_ops *af_ops;
1290         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1291
1292         ASSERT_RTNL();
1293         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1294         if (nlh == NULL)
1295                 return -EMSGSIZE;
1296
1297         ifm = nlmsg_data(nlh);
1298         ifm->ifi_family = AF_UNSPEC;
1299         ifm->__ifi_pad = 0;
1300         ifm->ifi_type = dev->type;
1301         ifm->ifi_index = dev->ifindex;
1302         ifm->ifi_flags = dev_get_flags(dev);
1303         ifm->ifi_change = change;
1304
1305         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1306             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1307             nla_put_u8(skb, IFLA_OPERSTATE,
1308                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1309             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1310             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1311             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1312             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1313             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1314             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1315             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1316 #ifdef CONFIG_RPS
1317             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1318 #endif
1319             (dev->ifindex != dev_get_iflink(dev) &&
1320              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1321             (upper_dev &&
1322              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1323             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1324             (dev->qdisc &&
1325              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1326             (dev->ifalias &&
1327              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1328             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1329                         atomic_read(&dev->carrier_changes)) ||
1330             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1331                 goto nla_put_failure;
1332
1333         if (rtnl_fill_link_ifmap(skb, dev))
1334                 goto nla_put_failure;
1335
1336         if (dev->addr_len) {
1337                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1338                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1339                         goto nla_put_failure;
1340         }
1341
1342         if (rtnl_phys_port_id_fill(skb, dev))
1343                 goto nla_put_failure;
1344
1345         if (rtnl_phys_port_name_fill(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (rtnl_phys_switch_id_fill(skb, dev))
1349                 goto nla_put_failure;
1350
1351         if (rtnl_fill_stats(skb, dev))
1352                 goto nla_put_failure;
1353
1354         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1355             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1356                 goto nla_put_failure;
1357
1358         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1359             ext_filter_mask & RTEXT_FILTER_VF) {
1360                 int i;
1361                 struct nlattr *vfinfo;
1362                 int num_vfs = dev_num_vf(dev->dev.parent);
1363
1364                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1365                 if (!vfinfo)
1366                         goto nla_put_failure;
1367                 for (i = 0; i < num_vfs; i++) {
1368                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1369                                 goto nla_put_failure;
1370                 }
1371
1372                 nla_nest_end(skb, vfinfo);
1373         }
1374
1375         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1376                 goto nla_put_failure;
1377
1378         if (rtnl_xdp_fill(skb, dev))
1379                 goto nla_put_failure;
1380
1381         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1382                 if (rtnl_link_fill(skb, dev) < 0)
1383                         goto nla_put_failure;
1384         }
1385
1386         if (dev->rtnl_link_ops &&
1387             dev->rtnl_link_ops->get_link_net) {
1388                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1389
1390                 if (!net_eq(dev_net(dev), link_net)) {
1391                         int id = peernet2id_alloc(dev_net(dev), link_net);
1392
1393                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1394                                 goto nla_put_failure;
1395                 }
1396         }
1397
1398         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1399                 goto nla_put_failure;
1400
1401         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1402                 if (af_ops->fill_link_af) {
1403                         struct nlattr *af;
1404                         int err;
1405
1406                         if (!(af = nla_nest_start(skb, af_ops->family)))
1407                                 goto nla_put_failure;
1408
1409                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1410
1411                         /*
1412                          * Caller may return ENODATA to indicate that there
1413                          * was no data to be dumped. This is not an error, it
1414                          * means we should trim the attribute header and
1415                          * continue.
1416                          */
1417                         if (err == -ENODATA)
1418                                 nla_nest_cancel(skb, af);
1419                         else if (err < 0)
1420                                 goto nla_put_failure;
1421
1422                         nla_nest_end(skb, af);
1423                 }
1424         }
1425
1426         nla_nest_end(skb, af_spec);
1427
1428         nlmsg_end(skb, nlh);
1429         return 0;
1430
1431 nla_put_failure:
1432         nlmsg_cancel(skb, nlh);
1433         return -EMSGSIZE;
1434 }
1435
1436 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1437         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1438         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1439         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1440         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1441         [IFLA_MTU]              = { .type = NLA_U32 },
1442         [IFLA_LINK]             = { .type = NLA_U32 },
1443         [IFLA_MASTER]           = { .type = NLA_U32 },
1444         [IFLA_CARRIER]          = { .type = NLA_U8 },
1445         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1446         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1447         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1448         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1449         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1450         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1451         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1452         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1453         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1454         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1455         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1456         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1457         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1458         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1459         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1460         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1461         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1462         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1463         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1464         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1465         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1466         [IFLA_XDP]              = { .type = NLA_NESTED },
1467 };
1468
1469 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1470         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1471         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1472         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1473         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1474 };
1475
1476 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1477         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1478         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1479         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1480         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1481         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1482         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1483         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1484         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1485         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1486         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1487         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1488         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1489 };
1490
1491 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1492         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1493         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1494                                     .len = PORT_PROFILE_MAX },
1495         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1496                                     .len = sizeof(struct ifla_port_vsi)},
1497         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1498                                       .len = PORT_UUID_MAX },
1499         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1500                                     .len = PORT_UUID_MAX },
1501         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1502         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1503 };
1504
1505 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1506         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1507         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1508 };
1509
1510 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1511 {
1512         const struct rtnl_link_ops *ops = NULL;
1513         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1514
1515         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1516                 return NULL;
1517
1518         if (linfo[IFLA_INFO_KIND]) {
1519                 char kind[MODULE_NAME_LEN];
1520
1521                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1522                 ops = rtnl_link_ops_get(kind);
1523         }
1524
1525         return ops;
1526 }
1527
1528 static bool link_master_filtered(struct net_device *dev, int master_idx)
1529 {
1530         struct net_device *master;
1531
1532         if (!master_idx)
1533                 return false;
1534
1535         master = netdev_master_upper_dev_get(dev);
1536         if (!master || master->ifindex != master_idx)
1537                 return true;
1538
1539         return false;
1540 }
1541
1542 static bool link_kind_filtered(const struct net_device *dev,
1543                                const struct rtnl_link_ops *kind_ops)
1544 {
1545         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1546                 return true;
1547
1548         return false;
1549 }
1550
1551 static bool link_dump_filtered(struct net_device *dev,
1552                                int master_idx,
1553                                const struct rtnl_link_ops *kind_ops)
1554 {
1555         if (link_master_filtered(dev, master_idx) ||
1556             link_kind_filtered(dev, kind_ops))
1557                 return true;
1558
1559         return false;
1560 }
1561
1562 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1563 {
1564         struct net *net = sock_net(skb->sk);
1565         int h, s_h;
1566         int idx = 0, s_idx;
1567         struct net_device *dev;
1568         struct hlist_head *head;
1569         struct nlattr *tb[IFLA_MAX+1];
1570         u32 ext_filter_mask = 0;
1571         const struct rtnl_link_ops *kind_ops = NULL;
1572         unsigned int flags = NLM_F_MULTI;
1573         int master_idx = 0;
1574         int err;
1575         int hdrlen;
1576
1577         s_h = cb->args[0];
1578         s_idx = cb->args[1];
1579
1580         cb->seq = net->dev_base_seq;
1581
1582         /* A hack to preserve kernel<->userspace interface.
1583          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1584          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1585          * what iproute2 < v3.9.0 used.
1586          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1587          * attribute, its netlink message is shorter than struct ifinfomsg.
1588          */
1589         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1590                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1591
1592         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1593
1594                 if (tb[IFLA_EXT_MASK])
1595                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1596
1597                 if (tb[IFLA_MASTER])
1598                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1599
1600                 if (tb[IFLA_LINKINFO])
1601                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1602
1603                 if (master_idx || kind_ops)
1604                         flags |= NLM_F_DUMP_FILTERED;
1605         }
1606
1607         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1608                 idx = 0;
1609                 head = &net->dev_index_head[h];
1610                 hlist_for_each_entry(dev, head, index_hlist) {
1611                         if (link_dump_filtered(dev, master_idx, kind_ops))
1612                                 goto cont;
1613                         if (idx < s_idx)
1614                                 goto cont;
1615                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1616                                                NETLINK_CB(cb->skb).portid,
1617                                                cb->nlh->nlmsg_seq, 0,
1618                                                flags,
1619                                                ext_filter_mask);
1620                         /* If we ran out of room on the first message,
1621                          * we're in trouble
1622                          */
1623                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1624
1625                         if (err < 0)
1626                                 goto out;
1627
1628                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1629 cont:
1630                         idx++;
1631                 }
1632         }
1633 out:
1634         cb->args[1] = idx;
1635         cb->args[0] = h;
1636
1637         return skb->len;
1638 }
1639
1640 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1641 {
1642         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1643 }
1644 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1645
1646 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1647 {
1648         struct net *net;
1649         /* Examine the link attributes and figure out which
1650          * network namespace we are talking about.
1651          */
1652         if (tb[IFLA_NET_NS_PID])
1653                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1654         else if (tb[IFLA_NET_NS_FD])
1655                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1656         else
1657                 net = get_net(src_net);
1658         return net;
1659 }
1660 EXPORT_SYMBOL(rtnl_link_get_net);
1661
1662 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1663 {
1664         if (dev) {
1665                 if (tb[IFLA_ADDRESS] &&
1666                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1667                         return -EINVAL;
1668
1669                 if (tb[IFLA_BROADCAST] &&
1670                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1671                         return -EINVAL;
1672         }
1673
1674         if (tb[IFLA_AF_SPEC]) {
1675                 struct nlattr *af;
1676                 int rem, err;
1677
1678                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1679                         const struct rtnl_af_ops *af_ops;
1680
1681                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1682                                 return -EAFNOSUPPORT;
1683
1684                         if (!af_ops->set_link_af)
1685                                 return -EOPNOTSUPP;
1686
1687                         if (af_ops->validate_link_af) {
1688                                 err = af_ops->validate_link_af(dev, af);
1689                                 if (err < 0)
1690                                         return err;
1691                         }
1692                 }
1693         }
1694
1695         return 0;
1696 }
1697
1698 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1699                                   int guid_type)
1700 {
1701         const struct net_device_ops *ops = dev->netdev_ops;
1702
1703         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1704 }
1705
1706 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1707 {
1708         if (dev->type != ARPHRD_INFINIBAND)
1709                 return -EOPNOTSUPP;
1710
1711         return handle_infiniband_guid(dev, ivt, guid_type);
1712 }
1713
1714 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1715 {
1716         const struct net_device_ops *ops = dev->netdev_ops;
1717         int err = -EINVAL;
1718
1719         if (tb[IFLA_VF_MAC]) {
1720                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1721
1722                 err = -EOPNOTSUPP;
1723                 if (ops->ndo_set_vf_mac)
1724                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1725                                                   ivm->mac);
1726                 if (err < 0)
1727                         return err;
1728         }
1729
1730         if (tb[IFLA_VF_VLAN]) {
1731                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1732
1733                 err = -EOPNOTSUPP;
1734                 if (ops->ndo_set_vf_vlan)
1735                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1736                                                    ivv->qos,
1737                                                    htons(ETH_P_8021Q));
1738                 if (err < 0)
1739                         return err;
1740         }
1741
1742         if (tb[IFLA_VF_VLAN_LIST]) {
1743                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1744                 struct nlattr *attr;
1745                 int rem, len = 0;
1746
1747                 err = -EOPNOTSUPP;
1748                 if (!ops->ndo_set_vf_vlan)
1749                         return err;
1750
1751                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1752                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1753                             nla_len(attr) < NLA_HDRLEN) {
1754                                 return -EINVAL;
1755                         }
1756                         if (len >= MAX_VLAN_LIST_LEN)
1757                                 return -EOPNOTSUPP;
1758                         ivvl[len] = nla_data(attr);
1759
1760                         len++;
1761                 }
1762                 if (len == 0)
1763                         return -EINVAL;
1764
1765                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1766                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1767                 if (err < 0)
1768                         return err;
1769         }
1770
1771         if (tb[IFLA_VF_TX_RATE]) {
1772                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1773                 struct ifla_vf_info ivf;
1774
1775                 err = -EOPNOTSUPP;
1776                 if (ops->ndo_get_vf_config)
1777                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1778                 if (err < 0)
1779                         return err;
1780
1781                 err = -EOPNOTSUPP;
1782                 if (ops->ndo_set_vf_rate)
1783                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1784                                                    ivf.min_tx_rate,
1785                                                    ivt->rate);
1786                 if (err < 0)
1787                         return err;
1788         }
1789
1790         if (tb[IFLA_VF_RATE]) {
1791                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1792
1793                 err = -EOPNOTSUPP;
1794                 if (ops->ndo_set_vf_rate)
1795                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1796                                                    ivt->min_tx_rate,
1797                                                    ivt->max_tx_rate);
1798                 if (err < 0)
1799                         return err;
1800         }
1801
1802         if (tb[IFLA_VF_SPOOFCHK]) {
1803                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1804
1805                 err = -EOPNOTSUPP;
1806                 if (ops->ndo_set_vf_spoofchk)
1807                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1808                                                        ivs->setting);
1809                 if (err < 0)
1810                         return err;
1811         }
1812
1813         if (tb[IFLA_VF_LINK_STATE]) {
1814                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1815
1816                 err = -EOPNOTSUPP;
1817                 if (ops->ndo_set_vf_link_state)
1818                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1819                                                          ivl->link_state);
1820                 if (err < 0)
1821                         return err;
1822         }
1823
1824         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1825                 struct ifla_vf_rss_query_en *ivrssq_en;
1826
1827                 err = -EOPNOTSUPP;
1828                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1829                 if (ops->ndo_set_vf_rss_query_en)
1830                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1831                                                            ivrssq_en->setting);
1832                 if (err < 0)
1833                         return err;
1834         }
1835
1836         if (tb[IFLA_VF_TRUST]) {
1837                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1838
1839                 err = -EOPNOTSUPP;
1840                 if (ops->ndo_set_vf_trust)
1841                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1842                 if (err < 0)
1843                         return err;
1844         }
1845
1846         if (tb[IFLA_VF_IB_NODE_GUID]) {
1847                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1848
1849                 if (!ops->ndo_set_vf_guid)
1850                         return -EOPNOTSUPP;
1851
1852                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1853         }
1854
1855         if (tb[IFLA_VF_IB_PORT_GUID]) {
1856                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1857
1858                 if (!ops->ndo_set_vf_guid)
1859                         return -EOPNOTSUPP;
1860
1861                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1862         }
1863
1864         return err;
1865 }
1866
1867 static int do_set_master(struct net_device *dev, int ifindex)
1868 {
1869         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1870         const struct net_device_ops *ops;
1871         int err;
1872
1873         if (upper_dev) {
1874                 if (upper_dev->ifindex == ifindex)
1875                         return 0;
1876                 ops = upper_dev->netdev_ops;
1877                 if (ops->ndo_del_slave) {
1878                         err = ops->ndo_del_slave(upper_dev, dev);
1879                         if (err)
1880                                 return err;
1881                 } else {
1882                         return -EOPNOTSUPP;
1883                 }
1884         }
1885
1886         if (ifindex) {
1887                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1888                 if (!upper_dev)
1889                         return -EINVAL;
1890                 ops = upper_dev->netdev_ops;
1891                 if (ops->ndo_add_slave) {
1892                         err = ops->ndo_add_slave(upper_dev, dev);
1893                         if (err)
1894                                 return err;
1895                 } else {
1896                         return -EOPNOTSUPP;
1897                 }
1898         }
1899         return 0;
1900 }
1901
1902 #define DO_SETLINK_MODIFIED     0x01
1903 /* notify flag means notify + modified. */
1904 #define DO_SETLINK_NOTIFY       0x03
1905 static int do_setlink(const struct sk_buff *skb,
1906                       struct net_device *dev, struct ifinfomsg *ifm,
1907                       struct nlattr **tb, char *ifname, int status)
1908 {
1909         const struct net_device_ops *ops = dev->netdev_ops;
1910         int err;
1911
1912         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1913                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1914                 if (IS_ERR(net)) {
1915                         err = PTR_ERR(net);
1916                         goto errout;
1917                 }
1918                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1919                         put_net(net);
1920                         err = -EPERM;
1921                         goto errout;
1922                 }
1923                 err = dev_change_net_namespace(dev, net, ifname);
1924                 put_net(net);
1925                 if (err)
1926                         goto errout;
1927                 status |= DO_SETLINK_MODIFIED;
1928         }
1929
1930         if (tb[IFLA_MAP]) {
1931                 struct rtnl_link_ifmap *u_map;
1932                 struct ifmap k_map;
1933
1934                 if (!ops->ndo_set_config) {
1935                         err = -EOPNOTSUPP;
1936                         goto errout;
1937                 }
1938
1939                 if (!netif_device_present(dev)) {
1940                         err = -ENODEV;
1941                         goto errout;
1942                 }
1943
1944                 u_map = nla_data(tb[IFLA_MAP]);
1945                 k_map.mem_start = (unsigned long) u_map->mem_start;
1946                 k_map.mem_end = (unsigned long) u_map->mem_end;
1947                 k_map.base_addr = (unsigned short) u_map->base_addr;
1948                 k_map.irq = (unsigned char) u_map->irq;
1949                 k_map.dma = (unsigned char) u_map->dma;
1950                 k_map.port = (unsigned char) u_map->port;
1951
1952                 err = ops->ndo_set_config(dev, &k_map);
1953                 if (err < 0)
1954                         goto errout;
1955
1956                 status |= DO_SETLINK_NOTIFY;
1957         }
1958
1959         if (tb[IFLA_ADDRESS]) {
1960                 struct sockaddr *sa;
1961                 int len;
1962
1963                 len = sizeof(sa_family_t) + dev->addr_len;
1964                 sa = kmalloc(len, GFP_KERNEL);
1965                 if (!sa) {
1966                         err = -ENOMEM;
1967                         goto errout;
1968                 }
1969                 sa->sa_family = dev->type;
1970                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1971                        dev->addr_len);
1972                 err = dev_set_mac_address(dev, sa);
1973                 kfree(sa);
1974                 if (err)
1975                         goto errout;
1976                 status |= DO_SETLINK_MODIFIED;
1977         }
1978
1979         if (tb[IFLA_MTU]) {
1980                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1981                 if (err < 0)
1982                         goto errout;
1983                 status |= DO_SETLINK_MODIFIED;
1984         }
1985
1986         if (tb[IFLA_GROUP]) {
1987                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1988                 status |= DO_SETLINK_NOTIFY;
1989         }
1990
1991         /*
1992          * Interface selected by interface index but interface
1993          * name provided implies that a name change has been
1994          * requested.
1995          */
1996         if (ifm->ifi_index > 0 && ifname[0]) {
1997                 err = dev_change_name(dev, ifname);
1998                 if (err < 0)
1999                         goto errout;
2000                 status |= DO_SETLINK_MODIFIED;
2001         }
2002
2003         if (tb[IFLA_IFALIAS]) {
2004                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2005                                     nla_len(tb[IFLA_IFALIAS]));
2006                 if (err < 0)
2007                         goto errout;
2008                 status |= DO_SETLINK_NOTIFY;
2009         }
2010
2011         if (tb[IFLA_BROADCAST]) {
2012                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2013                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2014         }
2015
2016         if (ifm->ifi_flags || ifm->ifi_change) {
2017                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2018                 if (err < 0)
2019                         goto errout;
2020         }
2021
2022         if (tb[IFLA_MASTER]) {
2023                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2024                 if (err)
2025                         goto errout;
2026                 status |= DO_SETLINK_MODIFIED;
2027         }
2028
2029         if (tb[IFLA_CARRIER]) {
2030                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2031                 if (err)
2032                         goto errout;
2033                 status |= DO_SETLINK_MODIFIED;
2034         }
2035
2036         if (tb[IFLA_TXQLEN]) {
2037                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2038                 unsigned long orig_len = dev->tx_queue_len;
2039
2040                 if (dev->tx_queue_len ^ value) {
2041                         dev->tx_queue_len = value;
2042                         err = call_netdevice_notifiers(
2043                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2044                         err = notifier_to_errno(err);
2045                         if (err) {
2046                                 dev->tx_queue_len = orig_len;
2047                                 goto errout;
2048                         }
2049                         status |= DO_SETLINK_NOTIFY;
2050                 }
2051         }
2052
2053         if (tb[IFLA_OPERSTATE])
2054                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2055
2056         if (tb[IFLA_LINKMODE]) {
2057                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2058
2059                 write_lock_bh(&dev_base_lock);
2060                 if (dev->link_mode ^ value)
2061                         status |= DO_SETLINK_NOTIFY;
2062                 dev->link_mode = value;
2063                 write_unlock_bh(&dev_base_lock);
2064         }
2065
2066         if (tb[IFLA_VFINFO_LIST]) {
2067                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2068                 struct nlattr *attr;
2069                 int rem;
2070
2071                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2072                         if (nla_type(attr) != IFLA_VF_INFO ||
2073                             nla_len(attr) < NLA_HDRLEN) {
2074                                 err = -EINVAL;
2075                                 goto errout;
2076                         }
2077                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2078                                                ifla_vf_policy);
2079                         if (err < 0)
2080                                 goto errout;
2081                         err = do_setvfinfo(dev, vfinfo);
2082                         if (err < 0)
2083                                 goto errout;
2084                         status |= DO_SETLINK_NOTIFY;
2085                 }
2086         }
2087         err = 0;
2088
2089         if (tb[IFLA_VF_PORTS]) {
2090                 struct nlattr *port[IFLA_PORT_MAX+1];
2091                 struct nlattr *attr;
2092                 int vf;
2093                 int rem;
2094
2095                 err = -EOPNOTSUPP;
2096                 if (!ops->ndo_set_vf_port)
2097                         goto errout;
2098
2099                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2100                         if (nla_type(attr) != IFLA_VF_PORT ||
2101                             nla_len(attr) < NLA_HDRLEN) {
2102                                 err = -EINVAL;
2103                                 goto errout;
2104                         }
2105                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2106                                                ifla_port_policy);
2107                         if (err < 0)
2108                                 goto errout;
2109                         if (!port[IFLA_PORT_VF]) {
2110                                 err = -EOPNOTSUPP;
2111                                 goto errout;
2112                         }
2113                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2114                         err = ops->ndo_set_vf_port(dev, vf, port);
2115                         if (err < 0)
2116                                 goto errout;
2117                         status |= DO_SETLINK_NOTIFY;
2118                 }
2119         }
2120         err = 0;
2121
2122         if (tb[IFLA_PORT_SELF]) {
2123                 struct nlattr *port[IFLA_PORT_MAX+1];
2124
2125                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2126                         tb[IFLA_PORT_SELF], ifla_port_policy);
2127                 if (err < 0)
2128                         goto errout;
2129
2130                 err = -EOPNOTSUPP;
2131                 if (ops->ndo_set_vf_port)
2132                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2133                 if (err < 0)
2134                         goto errout;
2135                 status |= DO_SETLINK_NOTIFY;
2136         }
2137
2138         if (tb[IFLA_AF_SPEC]) {
2139                 struct nlattr *af;
2140                 int rem;
2141
2142                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2143                         const struct rtnl_af_ops *af_ops;
2144
2145                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2146                                 BUG();
2147
2148                         err = af_ops->set_link_af(dev, af);
2149                         if (err < 0)
2150                                 goto errout;
2151
2152                         status |= DO_SETLINK_NOTIFY;
2153                 }
2154         }
2155         err = 0;
2156
2157         if (tb[IFLA_PROTO_DOWN]) {
2158                 err = dev_change_proto_down(dev,
2159                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2160                 if (err)
2161                         goto errout;
2162                 status |= DO_SETLINK_NOTIFY;
2163         }
2164
2165         if (tb[IFLA_XDP]) {
2166                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2167
2168                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2169                                        ifla_xdp_policy);
2170                 if (err < 0)
2171                         goto errout;
2172
2173                 if (xdp[IFLA_XDP_ATTACHED]) {
2174                         err = -EINVAL;
2175                         goto errout;
2176                 }
2177                 if (xdp[IFLA_XDP_FD]) {
2178                         err = dev_change_xdp_fd(dev,
2179                                                 nla_get_s32(xdp[IFLA_XDP_FD]));
2180                         if (err)
2181                                 goto errout;
2182                         status |= DO_SETLINK_NOTIFY;
2183                 }
2184         }
2185
2186 errout:
2187         if (status & DO_SETLINK_MODIFIED) {
2188                 if (status & DO_SETLINK_NOTIFY)
2189                         netdev_state_change(dev);
2190
2191                 if (err < 0)
2192                         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",
2193                                              dev->name);
2194         }
2195
2196         return err;
2197 }
2198
2199 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2200 {
2201         struct net *net = sock_net(skb->sk);
2202         struct ifinfomsg *ifm;
2203         struct net_device *dev;
2204         int err;
2205         struct nlattr *tb[IFLA_MAX+1];
2206         char ifname[IFNAMSIZ];
2207
2208         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2209         if (err < 0)
2210                 goto errout;
2211
2212         if (tb[IFLA_IFNAME])
2213                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2214         else
2215                 ifname[0] = '\0';
2216
2217         err = -EINVAL;
2218         ifm = nlmsg_data(nlh);
2219         if (ifm->ifi_index > 0)
2220                 dev = __dev_get_by_index(net, ifm->ifi_index);
2221         else if (tb[IFLA_IFNAME])
2222                 dev = __dev_get_by_name(net, ifname);
2223         else
2224                 goto errout;
2225
2226         if (dev == NULL) {
2227                 err = -ENODEV;
2228                 goto errout;
2229         }
2230
2231         err = validate_linkmsg(dev, tb);
2232         if (err < 0)
2233                 goto errout;
2234
2235         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2236 errout:
2237         return err;
2238 }
2239
2240 static int rtnl_group_dellink(const struct net *net, int group)
2241 {
2242         struct net_device *dev, *aux;
2243         LIST_HEAD(list_kill);
2244         bool found = false;
2245
2246         if (!group)
2247                 return -EPERM;
2248
2249         for_each_netdev(net, dev) {
2250                 if (dev->group == group) {
2251                         const struct rtnl_link_ops *ops;
2252
2253                         found = true;
2254                         ops = dev->rtnl_link_ops;
2255                         if (!ops || !ops->dellink)
2256                                 return -EOPNOTSUPP;
2257                 }
2258         }
2259
2260         if (!found)
2261                 return -ENODEV;
2262
2263         for_each_netdev_safe(net, dev, aux) {
2264                 if (dev->group == group) {
2265                         const struct rtnl_link_ops *ops;
2266
2267                         ops = dev->rtnl_link_ops;
2268                         ops->dellink(dev, &list_kill);
2269                 }
2270         }
2271         unregister_netdevice_many(&list_kill);
2272
2273         return 0;
2274 }
2275
2276 int rtnl_delete_link(struct net_device *dev)
2277 {
2278         const struct rtnl_link_ops *ops;
2279         LIST_HEAD(list_kill);
2280
2281         ops = dev->rtnl_link_ops;
2282         if (!ops || !ops->dellink)
2283                 return -EOPNOTSUPP;
2284
2285         ops->dellink(dev, &list_kill);
2286         unregister_netdevice_many(&list_kill);
2287
2288         return 0;
2289 }
2290 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2291
2292 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2293 {
2294         struct net *net = sock_net(skb->sk);
2295         struct net_device *dev;
2296         struct ifinfomsg *ifm;
2297         char ifname[IFNAMSIZ];
2298         struct nlattr *tb[IFLA_MAX+1];
2299         int err;
2300
2301         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2302         if (err < 0)
2303                 return err;
2304
2305         if (tb[IFLA_IFNAME])
2306                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2307
2308         ifm = nlmsg_data(nlh);
2309         if (ifm->ifi_index > 0)
2310                 dev = __dev_get_by_index(net, ifm->ifi_index);
2311         else if (tb[IFLA_IFNAME])
2312                 dev = __dev_get_by_name(net, ifname);
2313         else if (tb[IFLA_GROUP])
2314                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2315         else
2316                 return -EINVAL;
2317
2318         if (!dev)
2319                 return -ENODEV;
2320
2321         return rtnl_delete_link(dev);
2322 }
2323
2324 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2325 {
2326         unsigned int old_flags;
2327         int err;
2328
2329         old_flags = dev->flags;
2330         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2331                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2332                 if (err < 0)
2333                         return err;
2334         }
2335
2336         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2337
2338         __dev_notify_flags(dev, old_flags, ~0U);
2339         return 0;
2340 }
2341 EXPORT_SYMBOL(rtnl_configure_link);
2342
2343 struct net_device *rtnl_create_link(struct net *net,
2344         const char *ifname, unsigned char name_assign_type,
2345         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2346 {
2347         int err;
2348         struct net_device *dev;
2349         unsigned int num_tx_queues = 1;
2350         unsigned int num_rx_queues = 1;
2351
2352         if (tb[IFLA_NUM_TX_QUEUES])
2353                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2354         else if (ops->get_num_tx_queues)
2355                 num_tx_queues = ops->get_num_tx_queues();
2356
2357         if (tb[IFLA_NUM_RX_QUEUES])
2358                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2359         else if (ops->get_num_rx_queues)
2360                 num_rx_queues = ops->get_num_rx_queues();
2361
2362         err = -ENOMEM;
2363         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2364                                ops->setup, num_tx_queues, num_rx_queues);
2365         if (!dev)
2366                 goto err;
2367
2368         dev_net_set(dev, net);
2369         dev->rtnl_link_ops = ops;
2370         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2371
2372         if (tb[IFLA_MTU])
2373                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2374         if (tb[IFLA_ADDRESS]) {
2375                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2376                                 nla_len(tb[IFLA_ADDRESS]));
2377                 dev->addr_assign_type = NET_ADDR_SET;
2378         }
2379         if (tb[IFLA_BROADCAST])
2380                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2381                                 nla_len(tb[IFLA_BROADCAST]));
2382         if (tb[IFLA_TXQLEN])
2383                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2384         if (tb[IFLA_OPERSTATE])
2385                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2386         if (tb[IFLA_LINKMODE])
2387                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2388         if (tb[IFLA_GROUP])
2389                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2390
2391         return dev;
2392
2393 err:
2394         return ERR_PTR(err);
2395 }
2396 EXPORT_SYMBOL(rtnl_create_link);
2397
2398 static int rtnl_group_changelink(const struct sk_buff *skb,
2399                 struct net *net, int group,
2400                 struct ifinfomsg *ifm,
2401                 struct nlattr **tb)
2402 {
2403         struct net_device *dev, *aux;
2404         int err;
2405
2406         for_each_netdev_safe(net, dev, aux) {
2407                 if (dev->group == group) {
2408                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2409                         if (err < 0)
2410                                 return err;
2411                 }
2412         }
2413
2414         return 0;
2415 }
2416
2417 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2418 {
2419         struct net *net = sock_net(skb->sk);
2420         const struct rtnl_link_ops *ops;
2421         const struct rtnl_link_ops *m_ops = NULL;
2422         struct net_device *dev;
2423         struct net_device *master_dev = NULL;
2424         struct ifinfomsg *ifm;
2425         char kind[MODULE_NAME_LEN];
2426         char ifname[IFNAMSIZ];
2427         struct nlattr *tb[IFLA_MAX+1];
2428         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2429         unsigned char name_assign_type = NET_NAME_USER;
2430         int err;
2431
2432 #ifdef CONFIG_MODULES
2433 replay:
2434 #endif
2435         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2436         if (err < 0)
2437                 return err;
2438
2439         if (tb[IFLA_IFNAME])
2440                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2441         else
2442                 ifname[0] = '\0';
2443
2444         ifm = nlmsg_data(nlh);
2445         if (ifm->ifi_index > 0)
2446                 dev = __dev_get_by_index(net, ifm->ifi_index);
2447         else {
2448                 if (ifname[0])
2449                         dev = __dev_get_by_name(net, ifname);
2450                 else
2451                         dev = NULL;
2452         }
2453
2454         if (dev) {
2455                 master_dev = netdev_master_upper_dev_get(dev);
2456                 if (master_dev)
2457                         m_ops = master_dev->rtnl_link_ops;
2458         }
2459
2460         err = validate_linkmsg(dev, tb);
2461         if (err < 0)
2462                 return err;
2463
2464         if (tb[IFLA_LINKINFO]) {
2465                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2466                                        tb[IFLA_LINKINFO], ifla_info_policy);
2467                 if (err < 0)
2468                         return err;
2469         } else
2470                 memset(linkinfo, 0, sizeof(linkinfo));
2471
2472         if (linkinfo[IFLA_INFO_KIND]) {
2473                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2474                 ops = rtnl_link_ops_get(kind);
2475         } else {
2476                 kind[0] = '\0';
2477                 ops = NULL;
2478         }
2479
2480         if (1) {
2481                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2482                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2483                 struct nlattr **data = NULL;
2484                 struct nlattr **slave_data = NULL;
2485                 struct net *dest_net, *link_net = NULL;
2486
2487                 if (ops) {
2488                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2489                                 err = nla_parse_nested(attr, ops->maxtype,
2490                                                        linkinfo[IFLA_INFO_DATA],
2491                                                        ops->policy);
2492                                 if (err < 0)
2493                                         return err;
2494                                 data = attr;
2495                         }
2496                         if (ops->validate) {
2497                                 err = ops->validate(tb, data);
2498                                 if (err < 0)
2499                                         return err;
2500                         }
2501                 }
2502
2503                 if (m_ops) {
2504                         if (m_ops->slave_maxtype &&
2505                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2506                                 err = nla_parse_nested(slave_attr,
2507                                                        m_ops->slave_maxtype,
2508                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2509                                                        m_ops->slave_policy);
2510                                 if (err < 0)
2511                                         return err;
2512                                 slave_data = slave_attr;
2513                         }
2514                         if (m_ops->slave_validate) {
2515                                 err = m_ops->slave_validate(tb, slave_data);
2516                                 if (err < 0)
2517                                         return err;
2518                         }
2519                 }
2520
2521                 if (dev) {
2522                         int status = 0;
2523
2524                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2525                                 return -EEXIST;
2526                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2527                                 return -EOPNOTSUPP;
2528
2529                         if (linkinfo[IFLA_INFO_DATA]) {
2530                                 if (!ops || ops != dev->rtnl_link_ops ||
2531                                     !ops->changelink)
2532                                         return -EOPNOTSUPP;
2533
2534                                 err = ops->changelink(dev, tb, data);
2535                                 if (err < 0)
2536                                         return err;
2537                                 status |= DO_SETLINK_NOTIFY;
2538                         }
2539
2540                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2541                                 if (!m_ops || !m_ops->slave_changelink)
2542                                         return -EOPNOTSUPP;
2543
2544                                 err = m_ops->slave_changelink(master_dev, dev,
2545                                                               tb, slave_data);
2546                                 if (err < 0)
2547                                         return err;
2548                                 status |= DO_SETLINK_NOTIFY;
2549                         }
2550
2551                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2552                 }
2553
2554                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2555                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2556                                 return rtnl_group_changelink(skb, net,
2557                                                 nla_get_u32(tb[IFLA_GROUP]),
2558                                                 ifm, tb);
2559                         return -ENODEV;
2560                 }
2561
2562                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2563                         return -EOPNOTSUPP;
2564
2565                 if (!ops) {
2566 #ifdef CONFIG_MODULES
2567                         if (kind[0]) {
2568                                 __rtnl_unlock();
2569                                 request_module("rtnl-link-%s", kind);
2570                                 rtnl_lock();
2571                                 ops = rtnl_link_ops_get(kind);
2572                                 if (ops)
2573                                         goto replay;
2574                         }
2575 #endif
2576                         return -EOPNOTSUPP;
2577                 }
2578
2579                 if (!ops->setup)
2580                         return -EOPNOTSUPP;
2581
2582                 if (!ifname[0]) {
2583                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2584                         name_assign_type = NET_NAME_ENUM;
2585                 }
2586
2587                 dest_net = rtnl_link_get_net(net, tb);
2588                 if (IS_ERR(dest_net))
2589                         return PTR_ERR(dest_net);
2590
2591                 err = -EPERM;
2592                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2593                         goto out;
2594
2595                 if (tb[IFLA_LINK_NETNSID]) {
2596                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2597
2598                         link_net = get_net_ns_by_id(dest_net, id);
2599                         if (!link_net) {
2600                                 err =  -EINVAL;
2601                                 goto out;
2602                         }
2603                         err = -EPERM;
2604                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2605                                 goto out;
2606                 }
2607
2608                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2609                                        name_assign_type, ops, tb);
2610                 if (IS_ERR(dev)) {
2611                         err = PTR_ERR(dev);
2612                         goto out;
2613                 }
2614
2615                 dev->ifindex = ifm->ifi_index;
2616
2617                 if (ops->newlink) {
2618                         err = ops->newlink(link_net ? : net, dev, tb, data);
2619                         /* Drivers should call free_netdev() in ->destructor
2620                          * and unregister it on failure after registration
2621                          * so that device could be finally freed in rtnl_unlock.
2622                          */
2623                         if (err < 0) {
2624                                 /* If device is not registered at all, free it now */
2625                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2626                                         free_netdev(dev);
2627                                 goto out;
2628                         }
2629                 } else {
2630                         err = register_netdevice(dev);
2631                         if (err < 0) {
2632                                 free_netdev(dev);
2633                                 goto out;
2634                         }
2635                 }
2636                 err = rtnl_configure_link(dev, ifm);
2637                 if (err < 0)
2638                         goto out_unregister;
2639                 if (link_net) {
2640                         err = dev_change_net_namespace(dev, dest_net, ifname);
2641                         if (err < 0)
2642                                 goto out_unregister;
2643                 }
2644 out:
2645                 if (link_net)
2646                         put_net(link_net);
2647                 put_net(dest_net);
2648                 return err;
2649 out_unregister:
2650                 if (ops->newlink) {
2651                         LIST_HEAD(list_kill);
2652
2653                         ops->dellink(dev, &list_kill);
2654                         unregister_netdevice_many(&list_kill);
2655                 } else {
2656                         unregister_netdevice(dev);
2657                 }
2658                 goto out;
2659         }
2660 }
2661
2662 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2663 {
2664         struct net *net = sock_net(skb->sk);
2665         struct ifinfomsg *ifm;
2666         char ifname[IFNAMSIZ];
2667         struct nlattr *tb[IFLA_MAX+1];
2668         struct net_device *dev = NULL;
2669         struct sk_buff *nskb;
2670         int err;
2671         u32 ext_filter_mask = 0;
2672
2673         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2674         if (err < 0)
2675                 return err;
2676
2677         if (tb[IFLA_IFNAME])
2678                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2679
2680         if (tb[IFLA_EXT_MASK])
2681                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2682
2683         ifm = nlmsg_data(nlh);
2684         if (ifm->ifi_index > 0)
2685                 dev = __dev_get_by_index(net, ifm->ifi_index);
2686         else if (tb[IFLA_IFNAME])
2687                 dev = __dev_get_by_name(net, ifname);
2688         else
2689                 return -EINVAL;
2690
2691         if (dev == NULL)
2692                 return -ENODEV;
2693
2694         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2695         if (nskb == NULL)
2696                 return -ENOBUFS;
2697
2698         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2699                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2700         if (err < 0) {
2701                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2702                 WARN_ON(err == -EMSGSIZE);
2703                 kfree_skb(nskb);
2704         } else
2705                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2706
2707         return err;
2708 }
2709
2710 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2711 {
2712         struct net *net = sock_net(skb->sk);
2713         struct net_device *dev;
2714         struct nlattr *tb[IFLA_MAX+1];
2715         u32 ext_filter_mask = 0;
2716         u16 min_ifinfo_dump_size = 0;
2717         int hdrlen;
2718
2719         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2720         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2721                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2722
2723         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2724                 if (tb[IFLA_EXT_MASK])
2725                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2726         }
2727
2728         if (!ext_filter_mask)
2729                 return NLMSG_GOODSIZE;
2730         /*
2731          * traverse the list of net devices and compute the minimum
2732          * buffer size based upon the filter mask.
2733          */
2734         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2735                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2736                                              if_nlmsg_size(dev,
2737                                                            ext_filter_mask));
2738         }
2739
2740         return nlmsg_total_size(min_ifinfo_dump_size);
2741 }
2742
2743 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2744 {
2745         int idx;
2746         int s_idx = cb->family;
2747
2748         if (s_idx == 0)
2749                 s_idx = 1;
2750         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2751                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2752                 if (idx < s_idx || idx == PF_PACKET)
2753                         continue;
2754                 if (rtnl_msg_handlers[idx] == NULL ||
2755                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2756                         continue;
2757                 if (idx > s_idx) {
2758                         memset(&cb->args[0], 0, sizeof(cb->args));
2759                         cb->prev_seq = 0;
2760                         cb->seq = 0;
2761                 }
2762                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2763                         break;
2764         }
2765         cb->family = idx;
2766
2767         return skb->len;
2768 }
2769
2770 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2771                                        unsigned int change, gfp_t flags)
2772 {
2773         struct net *net = dev_net(dev);
2774         struct sk_buff *skb;
2775         int err = -ENOBUFS;
2776         size_t if_info_size;
2777
2778         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2779         if (skb == NULL)
2780                 goto errout;
2781
2782         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2783         if (err < 0) {
2784                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2785                 WARN_ON(err == -EMSGSIZE);
2786                 kfree_skb(skb);
2787                 goto errout;
2788         }
2789         return skb;
2790 errout:
2791         if (err < 0)
2792                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2793         return NULL;
2794 }
2795
2796 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2797 {
2798         struct net *net = dev_net(dev);
2799
2800         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2801 }
2802
2803 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2804                   gfp_t flags)
2805 {
2806         struct sk_buff *skb;
2807
2808         if (dev->reg_state != NETREG_REGISTERED)
2809                 return;
2810
2811         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2812         if (skb)
2813                 rtmsg_ifinfo_send(skb, dev, flags);
2814 }
2815 EXPORT_SYMBOL(rtmsg_ifinfo);
2816
2817 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2818                                    struct net_device *dev,
2819                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2820                                    int type, unsigned int flags,
2821                                    int nlflags, u16 ndm_state)
2822 {
2823         struct nlmsghdr *nlh;
2824         struct ndmsg *ndm;
2825
2826         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2827         if (!nlh)
2828                 return -EMSGSIZE;
2829
2830         ndm = nlmsg_data(nlh);
2831         ndm->ndm_family  = AF_BRIDGE;
2832         ndm->ndm_pad1    = 0;
2833         ndm->ndm_pad2    = 0;
2834         ndm->ndm_flags   = flags;
2835         ndm->ndm_type    = 0;
2836         ndm->ndm_ifindex = dev->ifindex;
2837         ndm->ndm_state   = ndm_state;
2838
2839         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2840                 goto nla_put_failure;
2841         if (vid)
2842                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2843                         goto nla_put_failure;
2844
2845         nlmsg_end(skb, nlh);
2846         return 0;
2847
2848 nla_put_failure:
2849         nlmsg_cancel(skb, nlh);
2850         return -EMSGSIZE;
2851 }
2852
2853 static inline size_t rtnl_fdb_nlmsg_size(void)
2854 {
2855         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2856                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2857                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2858                0;
2859 }
2860
2861 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2862                             u16 ndm_state)
2863 {
2864         struct net *net = dev_net(dev);
2865         struct sk_buff *skb;
2866         int err = -ENOBUFS;
2867
2868         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2869         if (!skb)
2870                 goto errout;
2871
2872         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2873                                       0, 0, type, NTF_SELF, 0, ndm_state);
2874         if (err < 0) {
2875                 kfree_skb(skb);
2876                 goto errout;
2877         }
2878
2879         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2880         return;
2881 errout:
2882         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2883 }
2884
2885 /**
2886  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2887  */
2888 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2889                      struct nlattr *tb[],
2890                      struct net_device *dev,
2891                      const unsigned char *addr, u16 vid,
2892                      u16 flags)
2893 {
2894         int err = -EINVAL;
2895
2896         /* If aging addresses are supported device will need to
2897          * implement its own handler for this.
2898          */
2899         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2900                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2901                 return err;
2902         }
2903
2904         if (vid) {
2905                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2906                 return err;
2907         }
2908
2909         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2910                 err = dev_uc_add_excl(dev, addr);
2911         else if (is_multicast_ether_addr(addr))
2912                 err = dev_mc_add_excl(dev, addr);
2913
2914         /* Only return duplicate errors if NLM_F_EXCL is set */
2915         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2916                 err = 0;
2917
2918         return err;
2919 }
2920 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2921
2922 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2923 {
2924         u16 vid = 0;
2925
2926         if (vlan_attr) {
2927                 if (nla_len(vlan_attr) != sizeof(u16)) {
2928                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2929                         return -EINVAL;
2930                 }
2931
2932                 vid = nla_get_u16(vlan_attr);
2933
2934                 if (!vid || vid >= VLAN_VID_MASK) {
2935                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2936                                 vid);
2937                         return -EINVAL;
2938                 }
2939         }
2940         *p_vid = vid;
2941         return 0;
2942 }
2943
2944 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2945 {
2946         struct net *net = sock_net(skb->sk);
2947         struct ndmsg *ndm;
2948         struct nlattr *tb[NDA_MAX+1];
2949         struct net_device *dev;
2950         u8 *addr;
2951         u16 vid;
2952         int err;
2953
2954         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2955         if (err < 0)
2956                 return err;
2957
2958         ndm = nlmsg_data(nlh);
2959         if (ndm->ndm_ifindex == 0) {
2960                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2961                 return -EINVAL;
2962         }
2963
2964         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2965         if (dev == NULL) {
2966                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2967                 return -ENODEV;
2968         }
2969
2970         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2971                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2972                 return -EINVAL;
2973         }
2974
2975         addr = nla_data(tb[NDA_LLADDR]);
2976
2977         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2978         if (err)
2979                 return err;
2980
2981         err = -EOPNOTSUPP;
2982
2983         /* Support fdb on master device the net/bridge default case */
2984         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2985             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2986                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2987                 const struct net_device_ops *ops = br_dev->netdev_ops;
2988
2989                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2990                                        nlh->nlmsg_flags);
2991                 if (err)
2992                         goto out;
2993                 else
2994                         ndm->ndm_flags &= ~NTF_MASTER;
2995         }
2996
2997         /* Embedded bridge, macvlan, and any other device support */
2998         if ((ndm->ndm_flags & NTF_SELF)) {
2999                 if (dev->netdev_ops->ndo_fdb_add)
3000                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3001                                                            vid,
3002                                                            nlh->nlmsg_flags);
3003                 else
3004                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3005                                                nlh->nlmsg_flags);
3006
3007                 if (!err) {
3008                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3009                                         ndm->ndm_state);
3010                         ndm->ndm_flags &= ~NTF_SELF;
3011                 }
3012         }
3013 out:
3014         return err;
3015 }
3016
3017 /**
3018  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3019  */
3020 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3021                      struct nlattr *tb[],
3022                      struct net_device *dev,
3023                      const unsigned char *addr, u16 vid)
3024 {
3025         int err = -EINVAL;
3026
3027         /* If aging addresses are supported device will need to
3028          * implement its own handler for this.
3029          */
3030         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3031                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3032                 return err;
3033         }
3034
3035         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3036                 err = dev_uc_del(dev, addr);
3037         else if (is_multicast_ether_addr(addr))
3038                 err = dev_mc_del(dev, addr);
3039
3040         return err;
3041 }
3042 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3043
3044 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3045 {
3046         struct net *net = sock_net(skb->sk);
3047         struct ndmsg *ndm;
3048         struct nlattr *tb[NDA_MAX+1];
3049         struct net_device *dev;
3050         int err = -EINVAL;
3051         __u8 *addr;
3052         u16 vid;
3053
3054         if (!netlink_capable(skb, CAP_NET_ADMIN))
3055                 return -EPERM;
3056
3057         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3058         if (err < 0)
3059                 return err;
3060
3061         ndm = nlmsg_data(nlh);
3062         if (ndm->ndm_ifindex == 0) {
3063                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3064                 return -EINVAL;
3065         }
3066
3067         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3068         if (dev == NULL) {
3069                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3070                 return -ENODEV;
3071         }
3072
3073         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3074                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3075                 return -EINVAL;
3076         }
3077
3078         addr = nla_data(tb[NDA_LLADDR]);
3079
3080         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3081         if (err)
3082                 return err;
3083
3084         err = -EOPNOTSUPP;
3085
3086         /* Support fdb on master device the net/bridge default case */
3087         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3088             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3089                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3090                 const struct net_device_ops *ops = br_dev->netdev_ops;
3091
3092                 if (ops->ndo_fdb_del)
3093                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3094
3095                 if (err)
3096                         goto out;
3097                 else
3098                         ndm->ndm_flags &= ~NTF_MASTER;
3099         }
3100
3101         /* Embedded bridge, macvlan, and any other device support */
3102         if (ndm->ndm_flags & NTF_SELF) {
3103                 if (dev->netdev_ops->ndo_fdb_del)
3104                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3105                                                            vid);
3106                 else
3107                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3108
3109                 if (!err) {
3110                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3111                                         ndm->ndm_state);
3112                         ndm->ndm_flags &= ~NTF_SELF;
3113                 }
3114         }
3115 out:
3116         return err;
3117 }
3118
3119 static int nlmsg_populate_fdb(struct sk_buff *skb,
3120                               struct netlink_callback *cb,
3121                               struct net_device *dev,
3122                               int *idx,
3123                               struct netdev_hw_addr_list *list)
3124 {
3125         struct netdev_hw_addr *ha;
3126         int err;
3127         u32 portid, seq;
3128
3129         portid = NETLINK_CB(cb->skb).portid;
3130         seq = cb->nlh->nlmsg_seq;
3131
3132         list_for_each_entry(ha, &list->list, list) {
3133                 if (*idx < cb->args[2])
3134                         goto skip;
3135
3136                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3137                                               portid, seq,
3138                                               RTM_NEWNEIGH, NTF_SELF,
3139                                               NLM_F_MULTI, NUD_PERMANENT);
3140                 if (err < 0)
3141                         return err;
3142 skip:
3143                 *idx += 1;
3144         }
3145         return 0;
3146 }
3147
3148 /**
3149  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3150  * @nlh: netlink message header
3151  * @dev: netdevice
3152  *
3153  * Default netdevice operation to dump the existing unicast address list.
3154  * Returns number of addresses from list put in skb.
3155  */
3156 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3157                       struct netlink_callback *cb,
3158                       struct net_device *dev,
3159                       struct net_device *filter_dev,
3160                       int *idx)
3161 {
3162         int err;
3163
3164         netif_addr_lock_bh(dev);
3165         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3166         if (err)
3167                 goto out;
3168         nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3169 out:
3170         netif_addr_unlock_bh(dev);
3171         return err;
3172 }
3173 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3174
3175 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3176 {
3177         struct net_device *dev;
3178         struct nlattr *tb[IFLA_MAX+1];
3179         struct net_device *br_dev = NULL;
3180         const struct net_device_ops *ops = NULL;
3181         const struct net_device_ops *cops = NULL;
3182         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3183         struct net *net = sock_net(skb->sk);
3184         struct hlist_head *head;
3185         int brport_idx = 0;
3186         int br_idx = 0;
3187         int h, s_h;
3188         int idx = 0, s_idx;
3189         int err = 0;
3190         int fidx = 0;
3191
3192         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3193                         ifla_policy) == 0) {
3194                 if (tb[IFLA_MASTER])
3195                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3196         }
3197
3198         brport_idx = ifm->ifi_index;
3199
3200         if (br_idx) {
3201                 br_dev = __dev_get_by_index(net, br_idx);
3202                 if (!br_dev)
3203                         return -ENODEV;
3204
3205                 ops = br_dev->netdev_ops;
3206         }
3207
3208         s_h = cb->args[0];
3209         s_idx = cb->args[1];
3210
3211         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3212                 idx = 0;
3213                 head = &net->dev_index_head[h];
3214                 hlist_for_each_entry(dev, head, index_hlist) {
3215
3216                         if (brport_idx && (dev->ifindex != brport_idx))
3217                                 continue;
3218
3219                         if (!br_idx) { /* user did not specify a specific bridge */
3220                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3221                                         br_dev = netdev_master_upper_dev_get(dev);
3222                                         cops = br_dev->netdev_ops;
3223                                 }
3224                         } else {
3225                                 if (dev != br_dev &&
3226                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3227                                         continue;
3228
3229                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3230                                     !(dev->priv_flags & IFF_EBRIDGE))
3231                                         continue;
3232                                 cops = ops;
3233                         }
3234
3235                         if (idx < s_idx)
3236                                 goto cont;
3237
3238                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3239                                 if (cops && cops->ndo_fdb_dump) {
3240                                         err = cops->ndo_fdb_dump(skb, cb,
3241                                                                 br_dev, dev,
3242                                                                 &fidx);
3243                                         if (err == -EMSGSIZE)
3244                                                 goto out;
3245                                 }
3246                         }
3247
3248                         if (dev->netdev_ops->ndo_fdb_dump)
3249                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3250                                                                     dev, NULL,
3251                                                                     &fidx);
3252                         else
3253                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3254                                                         &fidx);
3255                         if (err == -EMSGSIZE)
3256                                 goto out;
3257
3258                         cops = NULL;
3259
3260                         /* reset fdb offset to 0 for rest of the interfaces */
3261                         cb->args[2] = 0;
3262                         fidx = 0;
3263 cont:
3264                         idx++;
3265                 }
3266         }
3267
3268 out:
3269         cb->args[0] = h;
3270         cb->args[1] = idx;
3271         cb->args[2] = fidx;
3272
3273         return skb->len;
3274 }
3275
3276 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3277                                unsigned int attrnum, unsigned int flag)
3278 {
3279         if (mask & flag)
3280                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3281         return 0;
3282 }
3283
3284 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3285                             struct net_device *dev, u16 mode,
3286                             u32 flags, u32 mask, int nlflags,
3287                             u32 filter_mask,
3288                             int (*vlan_fill)(struct sk_buff *skb,
3289                                              struct net_device *dev,
3290                                              u32 filter_mask))
3291 {
3292         struct nlmsghdr *nlh;
3293         struct ifinfomsg *ifm;
3294         struct nlattr *br_afspec;
3295         struct nlattr *protinfo;
3296         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3297         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3298         int err = 0;
3299
3300         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3301         if (nlh == NULL)
3302                 return -EMSGSIZE;
3303
3304         ifm = nlmsg_data(nlh);
3305         ifm->ifi_family = AF_BRIDGE;
3306         ifm->__ifi_pad = 0;
3307         ifm->ifi_type = dev->type;
3308         ifm->ifi_index = dev->ifindex;
3309         ifm->ifi_flags = dev_get_flags(dev);
3310         ifm->ifi_change = 0;
3311
3312
3313         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3314             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3315             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3316             (br_dev &&
3317              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3318             (dev->addr_len &&
3319              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3320             (dev->ifindex != dev_get_iflink(dev) &&
3321              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3322                 goto nla_put_failure;
3323
3324         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3325         if (!br_afspec)
3326                 goto nla_put_failure;
3327
3328         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3329                 nla_nest_cancel(skb, br_afspec);
3330                 goto nla_put_failure;
3331         }
3332
3333         if (mode != BRIDGE_MODE_UNDEF) {
3334                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3335                         nla_nest_cancel(skb, br_afspec);
3336                         goto nla_put_failure;
3337                 }
3338         }
3339         if (vlan_fill) {
3340                 err = vlan_fill(skb, dev, filter_mask);
3341                 if (err) {
3342                         nla_nest_cancel(skb, br_afspec);
3343                         goto nla_put_failure;
3344                 }
3345         }
3346         nla_nest_end(skb, br_afspec);
3347
3348         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3349         if (!protinfo)
3350                 goto nla_put_failure;
3351
3352         if (brport_nla_put_flag(skb, flags, mask,
3353                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3354             brport_nla_put_flag(skb, flags, mask,
3355                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3356             brport_nla_put_flag(skb, flags, mask,
3357                                 IFLA_BRPORT_FAST_LEAVE,
3358                                 BR_MULTICAST_FAST_LEAVE) ||
3359             brport_nla_put_flag(skb, flags, mask,
3360                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3361             brport_nla_put_flag(skb, flags, mask,
3362                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3363             brport_nla_put_flag(skb, flags, mask,
3364                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3365             brport_nla_put_flag(skb, flags, mask,
3366                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3367             brport_nla_put_flag(skb, flags, mask,
3368                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3369                 nla_nest_cancel(skb, protinfo);
3370                 goto nla_put_failure;
3371         }
3372
3373         nla_nest_end(skb, protinfo);
3374
3375         nlmsg_end(skb, nlh);
3376         return 0;
3377 nla_put_failure:
3378         nlmsg_cancel(skb, nlh);
3379         return err ? err : -EMSGSIZE;
3380 }
3381 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3382
3383 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3384 {
3385         struct net *net = sock_net(skb->sk);
3386         struct net_device *dev;
3387         int idx = 0;
3388         u32 portid = NETLINK_CB(cb->skb).portid;
3389         u32 seq = cb->nlh->nlmsg_seq;
3390         u32 filter_mask = 0;
3391         int err;
3392
3393         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3394                 struct nlattr *extfilt;
3395
3396                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3397                                           IFLA_EXT_MASK);
3398                 if (extfilt) {
3399                         if (nla_len(extfilt) < sizeof(filter_mask))
3400                                 return -EINVAL;
3401
3402                         filter_mask = nla_get_u32(extfilt);
3403                 }
3404         }
3405
3406         rcu_read_lock();
3407         for_each_netdev_rcu(net, dev) {
3408                 const struct net_device_ops *ops = dev->netdev_ops;
3409                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3410
3411                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3412                         if (idx >= cb->args[0]) {
3413                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3414                                                 skb, portid, seq, dev,
3415                                                 filter_mask, NLM_F_MULTI);
3416                                 if (err < 0 && err != -EOPNOTSUPP)
3417                                         break;
3418                         }
3419                         idx++;
3420                 }
3421
3422                 if (ops->ndo_bridge_getlink) {
3423                         if (idx >= cb->args[0]) {
3424                                 err = ops->ndo_bridge_getlink(skb, portid,
3425                                                               seq, dev,
3426                                                               filter_mask,
3427                                                               NLM_F_MULTI);
3428                                 if (err < 0 && err != -EOPNOTSUPP)
3429                                         break;
3430                         }
3431                         idx++;
3432                 }
3433         }
3434         rcu_read_unlock();
3435         cb->args[0] = idx;
3436
3437         return skb->len;
3438 }
3439
3440 static inline size_t bridge_nlmsg_size(void)
3441 {
3442         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3443                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3444                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3445                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3446                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3447                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3448                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3449                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3450                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3451                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3452                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3453 }
3454
3455 static int rtnl_bridge_notify(struct net_device *dev)
3456 {
3457         struct net *net = dev_net(dev);
3458         struct sk_buff *skb;
3459         int err = -EOPNOTSUPP;
3460
3461         if (!dev->netdev_ops->ndo_bridge_getlink)
3462                 return 0;
3463
3464         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3465         if (!skb) {
3466                 err = -ENOMEM;
3467                 goto errout;
3468         }
3469
3470         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3471         if (err < 0)
3472                 goto errout;
3473
3474         if (!skb->len)
3475                 goto errout;
3476
3477         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3478         return 0;
3479 errout:
3480         WARN_ON(err == -EMSGSIZE);
3481         kfree_skb(skb);
3482         if (err)
3483                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3484         return err;
3485 }
3486
3487 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3488 {
3489         struct net *net = sock_net(skb->sk);
3490         struct ifinfomsg *ifm;
3491         struct net_device *dev;
3492         struct nlattr *br_spec, *attr = NULL;
3493         int rem, err = -EOPNOTSUPP;
3494         u16 flags = 0;
3495         bool have_flags = false;
3496
3497         if (nlmsg_len(nlh) < sizeof(*ifm))
3498                 return -EINVAL;
3499
3500         ifm = nlmsg_data(nlh);
3501         if (ifm->ifi_family != AF_BRIDGE)
3502                 return -EPFNOSUPPORT;
3503
3504         dev = __dev_get_by_index(net, ifm->ifi_index);
3505         if (!dev) {
3506                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3507                 return -ENODEV;
3508         }
3509
3510         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3511         if (br_spec) {
3512                 nla_for_each_nested(attr, br_spec, rem) {
3513                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3514                                 if (nla_len(attr) < sizeof(flags))
3515                                         return -EINVAL;
3516
3517                                 have_flags = true;
3518                                 flags = nla_get_u16(attr);
3519                                 break;
3520                         }
3521                 }
3522         }
3523
3524         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3525                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3526
3527                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3528                         err = -EOPNOTSUPP;
3529                         goto out;
3530                 }
3531
3532                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3533                 if (err)
3534                         goto out;
3535
3536                 flags &= ~BRIDGE_FLAGS_MASTER;
3537         }
3538
3539         if ((flags & BRIDGE_FLAGS_SELF)) {
3540                 if (!dev->netdev_ops->ndo_bridge_setlink)
3541                         err = -EOPNOTSUPP;
3542                 else
3543                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3544                                                                   flags);
3545                 if (!err) {
3546                         flags &= ~BRIDGE_FLAGS_SELF;
3547
3548                         /* Generate event to notify upper layer of bridge
3549                          * change
3550                          */
3551                         err = rtnl_bridge_notify(dev);
3552                 }
3553         }
3554
3555         if (have_flags)
3556                 memcpy(nla_data(attr), &flags, sizeof(flags));
3557 out:
3558         return err;
3559 }
3560
3561 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3562 {
3563         struct net *net = sock_net(skb->sk);
3564         struct ifinfomsg *ifm;
3565         struct net_device *dev;
3566         struct nlattr *br_spec, *attr = NULL;
3567         int rem, err = -EOPNOTSUPP;
3568         u16 flags = 0;
3569         bool have_flags = false;
3570
3571         if (nlmsg_len(nlh) < sizeof(*ifm))
3572                 return -EINVAL;
3573
3574         ifm = nlmsg_data(nlh);
3575         if (ifm->ifi_family != AF_BRIDGE)
3576                 return -EPFNOSUPPORT;
3577
3578         dev = __dev_get_by_index(net, ifm->ifi_index);
3579         if (!dev) {
3580                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3581                 return -ENODEV;
3582         }
3583
3584         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3585         if (br_spec) {
3586                 nla_for_each_nested(attr, br_spec, rem) {
3587                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3588                                 if (nla_len(attr) < sizeof(flags))
3589                                         return -EINVAL;
3590
3591                                 have_flags = true;
3592                                 flags = nla_get_u16(attr);
3593                                 break;
3594                         }
3595                 }
3596         }
3597
3598         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3599                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3600
3601                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3602                         err = -EOPNOTSUPP;
3603                         goto out;
3604                 }
3605
3606                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3607                 if (err)
3608                         goto out;
3609
3610                 flags &= ~BRIDGE_FLAGS_MASTER;
3611         }
3612
3613         if ((flags & BRIDGE_FLAGS_SELF)) {
3614                 if (!dev->netdev_ops->ndo_bridge_dellink)
3615                         err = -EOPNOTSUPP;
3616                 else
3617                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3618                                                                   flags);
3619
3620                 if (!err) {
3621                         flags &= ~BRIDGE_FLAGS_SELF;
3622
3623                         /* Generate event to notify upper layer of bridge
3624                          * change
3625                          */
3626                         err = rtnl_bridge_notify(dev);
3627                 }
3628         }
3629
3630         if (have_flags)
3631                 memcpy(nla_data(attr), &flags, sizeof(flags));
3632 out:
3633         return err;
3634 }
3635
3636 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3637 {
3638         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3639                (!idxattr || idxattr == attrid);
3640 }
3641
3642 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3643 static int rtnl_get_offload_stats_attr_size(int attr_id)
3644 {
3645         switch (attr_id) {
3646         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3647                 return sizeof(struct rtnl_link_stats64);
3648         }
3649
3650         return 0;
3651 }
3652
3653 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3654                                   int *prividx)
3655 {
3656         struct nlattr *attr = NULL;
3657         int attr_id, size;
3658         void *attr_data;
3659         int err;
3660
3661         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3662               dev->netdev_ops->ndo_get_offload_stats))
3663                 return -ENODATA;
3664
3665         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3666              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3667                 if (attr_id < *prividx)
3668                         continue;
3669
3670                 size = rtnl_get_offload_stats_attr_size(attr_id);
3671                 if (!size)
3672                         continue;
3673
3674                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3675                         continue;
3676
3677                 attr = nla_reserve_64bit(skb, attr_id, size,
3678                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3679                 if (!attr)
3680                         goto nla_put_failure;
3681
3682                 attr_data = nla_data(attr);
3683                 memset(attr_data, 0, size);
3684                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3685                                                              attr_data);
3686                 if (err)
3687                         goto get_offload_stats_failure;
3688         }
3689
3690         if (!attr)
3691                 return -ENODATA;
3692
3693         *prividx = 0;
3694         return 0;
3695
3696 nla_put_failure:
3697         err = -EMSGSIZE;
3698 get_offload_stats_failure:
3699         *prividx = attr_id;
3700         return err;
3701 }
3702
3703 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3704 {
3705         int nla_size = 0;
3706         int attr_id;
3707         int size;
3708
3709         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3710               dev->netdev_ops->ndo_get_offload_stats))
3711                 return 0;
3712
3713         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3714              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3715                 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3716                         continue;
3717                 size = rtnl_get_offload_stats_attr_size(attr_id);
3718                 nla_size += nla_total_size_64bit(size);
3719         }
3720
3721         if (nla_size != 0)
3722                 nla_size += nla_total_size(0);
3723
3724         return nla_size;
3725 }
3726
3727 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3728                                int type, u32 pid, u32 seq, u32 change,
3729                                unsigned int flags, unsigned int filter_mask,
3730                                int *idxattr, int *prividx)
3731 {
3732         struct if_stats_msg *ifsm;
3733         struct nlmsghdr *nlh;
3734         struct nlattr *attr;
3735         int s_prividx = *prividx;
3736         int err;
3737
3738         ASSERT_RTNL();
3739
3740         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3741         if (!nlh)
3742                 return -EMSGSIZE;
3743
3744         ifsm = nlmsg_data(nlh);
3745         ifsm->ifindex = dev->ifindex;
3746         ifsm->filter_mask = filter_mask;
3747
3748         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3749                 struct rtnl_link_stats64 *sp;
3750
3751                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3752                                          sizeof(struct rtnl_link_stats64),
3753                                          IFLA_STATS_UNSPEC);
3754                 if (!attr)
3755                         goto nla_put_failure;
3756
3757                 sp = nla_data(attr);
3758                 dev_get_stats(dev, sp);
3759         }
3760
3761         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3762                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3763
3764                 if (ops && ops->fill_linkxstats) {
3765                         *idxattr = IFLA_STATS_LINK_XSTATS;
3766                         attr = nla_nest_start(skb,
3767                                               IFLA_STATS_LINK_XSTATS);
3768                         if (!attr)
3769                                 goto nla_put_failure;
3770
3771                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3772                         nla_nest_end(skb, attr);
3773                         if (err)
3774                                 goto nla_put_failure;
3775                         *idxattr = 0;
3776                 }
3777         }
3778
3779         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3780                              *idxattr)) {
3781                 const struct rtnl_link_ops *ops = NULL;
3782                 const struct net_device *master;
3783
3784                 master = netdev_master_upper_dev_get(dev);
3785                 if (master)
3786                         ops = master->rtnl_link_ops;
3787                 if (ops && ops->fill_linkxstats) {
3788                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3789                         attr = nla_nest_start(skb,
3790                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3791                         if (!attr)
3792                                 goto nla_put_failure;
3793
3794                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3795                         nla_nest_end(skb, attr);
3796                         if (err)
3797                                 goto nla_put_failure;
3798                         *idxattr = 0;
3799                 }
3800         }
3801
3802         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3803                              *idxattr)) {
3804                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3805                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3806                 if (!attr)
3807                         goto nla_put_failure;
3808
3809                 err = rtnl_get_offload_stats(skb, dev, prividx);
3810                 if (err == -ENODATA)
3811                         nla_nest_cancel(skb, attr);
3812                 else
3813                         nla_nest_end(skb, attr);
3814
3815                 if (err && err != -ENODATA)
3816                         goto nla_put_failure;
3817                 *idxattr = 0;
3818         }
3819
3820         nlmsg_end(skb, nlh);
3821
3822         return 0;
3823
3824 nla_put_failure:
3825         /* not a multi message or no progress mean a real error */
3826         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3827                 nlmsg_cancel(skb, nlh);
3828         else
3829                 nlmsg_end(skb, nlh);
3830
3831         return -EMSGSIZE;
3832 }
3833
3834 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3835                                   u32 filter_mask)
3836 {
3837         size_t size = 0;
3838
3839         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3840                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3841
3842         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3843                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3844                 int attr = IFLA_STATS_LINK_XSTATS;
3845
3846                 if (ops && ops->get_linkxstats_size) {
3847                         size += nla_total_size(ops->get_linkxstats_size(dev,
3848                                                                         attr));
3849                         /* for IFLA_STATS_LINK_XSTATS */
3850                         size += nla_total_size(0);
3851                 }
3852         }
3853
3854         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3855                 struct net_device *_dev = (struct net_device *)dev;
3856                 const struct rtnl_link_ops *ops = NULL;
3857                 const struct net_device *master;
3858
3859                 /* netdev_master_upper_dev_get can't take const */
3860                 master = netdev_master_upper_dev_get(_dev);
3861                 if (master)
3862                         ops = master->rtnl_link_ops;
3863                 if (ops && ops->get_linkxstats_size) {
3864                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3865
3866                         size += nla_total_size(ops->get_linkxstats_size(dev,
3867                                                                         attr));
3868                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3869                         size += nla_total_size(0);
3870                 }
3871         }
3872
3873         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3874                 size += rtnl_get_offload_stats_size(dev);
3875
3876         return size;
3877 }
3878
3879 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3880 {
3881         struct net *net = sock_net(skb->sk);
3882         struct net_device *dev = NULL;
3883         int idxattr = 0, prividx = 0;
3884         struct if_stats_msg *ifsm;
3885         struct sk_buff *nskb;
3886         u32 filter_mask;
3887         int err;
3888
3889         ifsm = nlmsg_data(nlh);
3890         if (ifsm->ifindex > 0)
3891                 dev = __dev_get_by_index(net, ifsm->ifindex);
3892         else
3893                 return -EINVAL;
3894
3895         if (!dev)
3896                 return -ENODEV;
3897
3898         filter_mask = ifsm->filter_mask;
3899         if (!filter_mask)
3900                 return -EINVAL;
3901
3902         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3903         if (!nskb)
3904                 return -ENOBUFS;
3905
3906         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3907                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3908                                   0, filter_mask, &idxattr, &prividx);
3909         if (err < 0) {
3910                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3911                 WARN_ON(err == -EMSGSIZE);
3912                 kfree_skb(nskb);
3913         } else {
3914                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3915         }
3916
3917         return err;
3918 }
3919
3920 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3921 {
3922         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3923         struct net *net = sock_net(skb->sk);
3924         unsigned int flags = NLM_F_MULTI;
3925         struct if_stats_msg *ifsm;
3926         struct hlist_head *head;
3927         struct net_device *dev;
3928         u32 filter_mask = 0;
3929         int idx = 0;
3930
3931         s_h = cb->args[0];
3932         s_idx = cb->args[1];
3933         s_idxattr = cb->args[2];
3934         s_prividx = cb->args[3];
3935
3936         cb->seq = net->dev_base_seq;
3937
3938         ifsm = nlmsg_data(cb->nlh);
3939         filter_mask = ifsm->filter_mask;
3940         if (!filter_mask)
3941                 return -EINVAL;
3942
3943         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3944                 idx = 0;
3945                 head = &net->dev_index_head[h];
3946                 hlist_for_each_entry(dev, head, index_hlist) {
3947                         if (idx < s_idx)
3948                                 goto cont;
3949                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3950                                                   NETLINK_CB(cb->skb).portid,
3951                                                   cb->nlh->nlmsg_seq, 0,
3952                                                   flags, filter_mask,
3953                                                   &s_idxattr, &s_prividx);
3954                         /* If we ran out of room on the first message,
3955                          * we're in trouble
3956                          */
3957                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3958
3959                         if (err < 0)
3960                                 goto out;
3961                         s_prividx = 0;
3962                         s_idxattr = 0;
3963                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3964 cont:
3965                         idx++;
3966                 }
3967         }
3968 out:
3969         cb->args[3] = s_prividx;
3970         cb->args[2] = s_idxattr;
3971         cb->args[1] = idx;
3972         cb->args[0] = h;
3973
3974         return skb->len;
3975 }
3976
3977 /* Process one rtnetlink message. */
3978
3979 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3980 {
3981         struct net *net = sock_net(skb->sk);
3982         rtnl_doit_func doit;
3983         int kind;
3984         int family;
3985         int type;
3986         int err;
3987
3988         type = nlh->nlmsg_type;
3989         if (type > RTM_MAX)
3990                 return -EOPNOTSUPP;
3991
3992         type -= RTM_BASE;
3993
3994         /* All the messages must have at least 1 byte length */
3995         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3996                 return 0;
3997
3998         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3999         kind = type&3;
4000
4001         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4002                 return -EPERM;
4003
4004         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4005                 struct sock *rtnl;
4006                 rtnl_dumpit_func dumpit;
4007                 rtnl_calcit_func calcit;
4008                 u16 min_dump_alloc = 0;
4009
4010                 dumpit = rtnl_get_dumpit(family, type);
4011                 if (dumpit == NULL)
4012                         return -EOPNOTSUPP;
4013                 calcit = rtnl_get_calcit(family, type);
4014                 if (calcit)
4015                         min_dump_alloc = calcit(skb, nlh);
4016
4017                 __rtnl_unlock();
4018                 rtnl = net->rtnl;
4019                 {
4020                         struct netlink_dump_control c = {
4021                                 .dump           = dumpit,
4022                                 .min_dump_alloc = min_dump_alloc,
4023                         };
4024                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4025                 }
4026                 rtnl_lock();
4027                 return err;
4028         }
4029
4030         doit = rtnl_get_doit(family, type);
4031         if (doit == NULL)
4032                 return -EOPNOTSUPP;
4033
4034         return doit(skb, nlh);
4035 }
4036
4037 static void rtnetlink_rcv(struct sk_buff *skb)
4038 {
4039         rtnl_lock();
4040         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4041         rtnl_unlock();
4042 }
4043
4044 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4045 {
4046         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4047
4048         switch (event) {
4049         case NETDEV_UP:
4050         case NETDEV_DOWN:
4051         case NETDEV_PRE_UP:
4052         case NETDEV_POST_INIT:
4053         case NETDEV_REGISTER:
4054         case NETDEV_CHANGE:
4055         case NETDEV_PRE_TYPE_CHANGE:
4056         case NETDEV_GOING_DOWN:
4057         case NETDEV_UNREGISTER:
4058         case NETDEV_UNREGISTER_FINAL:
4059         case NETDEV_RELEASE:
4060         case NETDEV_JOIN:
4061         case NETDEV_BONDING_INFO:
4062                 break;
4063         default:
4064                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4065                 break;
4066         }
4067         return NOTIFY_DONE;
4068 }
4069
4070 static struct notifier_block rtnetlink_dev_notifier = {
4071         .notifier_call  = rtnetlink_event,
4072 };
4073
4074
4075 static int __net_init rtnetlink_net_init(struct net *net)
4076 {
4077         struct sock *sk;
4078         struct netlink_kernel_cfg cfg = {
4079                 .groups         = RTNLGRP_MAX,
4080                 .input          = rtnetlink_rcv,
4081                 .cb_mutex       = &rtnl_mutex,
4082                 .flags          = NL_CFG_F_NONROOT_RECV,
4083         };
4084
4085         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4086         if (!sk)
4087                 return -ENOMEM;
4088         net->rtnl = sk;
4089         return 0;
4090 }
4091
4092 static void __net_exit rtnetlink_net_exit(struct net *net)
4093 {
4094         netlink_kernel_release(net->rtnl);
4095         net->rtnl = NULL;
4096 }
4097
4098 static struct pernet_operations rtnetlink_net_ops = {
4099         .init = rtnetlink_net_init,
4100         .exit = rtnetlink_net_exit,
4101 };
4102
4103 void __init rtnetlink_init(void)
4104 {
4105         if (register_pernet_subsys(&rtnetlink_net_ops))
4106                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4107
4108         register_netdevice_notifier(&rtnetlink_dev_notifier);
4109
4110         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4111                       rtnl_dump_ifinfo, rtnl_calcit);
4112         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4113         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4114         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4115
4116         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4117         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4118
4119         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4120         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4121         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4122
4123         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4124         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4125         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4126
4127         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4128                       NULL);
4129 }