net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <linux/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 static struct sk_buff *defer_kfree_skb_list;
  75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  76 {
  77         if (head && tail) {
  78                 tail->next = defer_kfree_skb_list;
  79                 defer_kfree_skb_list = head;
  80         }
  81 }
  82 EXPORT_SYMBOL(rtnl_kfree_skbs);
  83
  84 void __rtnl_unlock(void)
  85 {
  86         struct sk_buff *head = defer_kfree_skb_list;
  87
  88         defer_kfree_skb_list = NULL;
  89
  90         mutex_unlock(&rtnl_mutex);
  91
  92         while (head) {
  93                 struct sk_buff *next = head->next;
  94
  95                 kfree_skb(head);
  96                 cond_resched();
  97                 head = next;
  98         }
  99 }
 100
 101 void rtnl_unlock(void)
 102 {
 103         /* This fellow will unlock it for us. */
 104         netdev_run_todo();
 105 }
 106 EXPORT_SYMBOL(rtnl_unlock);
 107
 108 int rtnl_trylock(void)
 109 {
 110         return mutex_trylock(&rtnl_mutex);
 111 }
 112 EXPORT_SYMBOL(rtnl_trylock);
 113
 114 int rtnl_is_locked(void)
 115 {
 116         return mutex_is_locked(&rtnl_mutex);
 117 }
 118 EXPORT_SYMBOL(rtnl_is_locked);
 119
 120 #ifdef CONFIG_PROVE_LOCKING
 121 bool lockdep_rtnl_is_held(void)
 122 {
 123         return lockdep_is_held(&rtnl_mutex);
 124 }
 125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 127
 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 129
 130 static inline int rtm_msgindex(int msgtype)
 131 {
 132         int msgindex = msgtype - RTM_BASE;
 133
 134         /*
 135          * msgindex < 0 implies someone tried to register a netlink
 136          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 137          * the message type has not been added to linux/rtnetlink.h
 138          */
 139         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 140
 141         return msgindex;
 142 }
 143
 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 145 {
 146         struct rtnl_link *tab;
 147
 148         if (protocol <= RTNL_FAMILY_MAX)
 149                 tab = rtnl_msg_handlers[protocol];
 150         else
 151                 tab = NULL;
 152
 153         if (tab == NULL || tab[msgindex].doit == NULL)
 154                 tab = rtnl_msg_handlers[PF_UNSPEC];
 155
 156         return tab[msgindex].doit;
 157 }
 158
 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 160 {
 161         struct rtnl_link *tab;
 162
 163         if (protocol <= RTNL_FAMILY_MAX)
 164                 tab = rtnl_msg_handlers[protocol];
 165         else
 166                 tab = NULL;
 167
 168         if (tab == NULL || tab[msgindex].dumpit == NULL)
 169                 tab = rtnl_msg_handlers[PF_UNSPEC];
 170
 171         return tab[msgindex].dumpit;
 172 }
 173
 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 175 {
 176         struct rtnl_link *tab;
 177
 178         if (protocol <= RTNL_FAMILY_MAX)
 179                 tab = rtnl_msg_handlers[protocol];
 180         else
 181                 tab = NULL;
 182
 183         if (tab == NULL || tab[msgindex].calcit == NULL)
 184                 tab = rtnl_msg_handlers[PF_UNSPEC];
 185
 186         return tab[msgindex].calcit;
 187 }
 188
 189 /**
 190  * __rtnl_register - Register a rtnetlink message type
 191  * @protocol: Protocol family or PF_UNSPEC
 192  * @msgtype: rtnetlink message type
 193  * @doit: Function pointer called for each request message
 194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 195  * @calcit: Function pointer to calc size of dump message
 196  *
 197  * Registers the specified function pointers (at least one of them has
 198  * to be non-NULL) to be called whenever a request message for the
 199  * specified protocol family and message type is received.
 200  *
 201  * The special protocol family PF_UNSPEC may be used to define fallback
 202  * function pointers for the case when no entry for the specific protocol
 203  * family exists.
 204  *
 205  * Returns 0 on success or a negative error code.
 206  */
 207 int __rtnl_register(int protocol, int msgtype,
 208                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 209                     rtnl_calcit_func calcit)
 210 {
 211         struct rtnl_link *tab;
 212         int msgindex;
 213
 214         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 215         msgindex = rtm_msgindex(msgtype);
 216
 217         tab = rtnl_msg_handlers[protocol];
 218         if (tab == NULL) {
 219                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 220                 if (tab == NULL)
 221                         return -ENOBUFS;
 222
 223                 rtnl_msg_handlers[protocol] = tab;
 224         }
 225
 226         if (doit)
 227                 tab[msgindex].doit = doit;
 228
 229         if (dumpit)
 230                 tab[msgindex].dumpit = dumpit;
 231
 232         if (calcit)
 233                 tab[msgindex].calcit = calcit;
 234
 235         return 0;
 236 }
 237 EXPORT_SYMBOL_GPL(__rtnl_register);
 238
 239 /**
 240  * rtnl_register - Register a rtnetlink message type
 241  *
 242  * Identical to __rtnl_register() but panics on failure. This is useful
 243  * as failure of this function is very unlikely, it can only happen due
 244  * to lack of memory when allocating the chain to store all message
 245  * handlers for a protocol. Meant for use in init functions where lack
 246  * of memory implies no sense in continuing.
 247  */
 248 void rtnl_register(int protocol, int msgtype,
 249                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 250                    rtnl_calcit_func calcit)
 251 {
 252         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 253                 panic("Unable to register rtnetlink message handler, "
 254                       "protocol = %d, message type = %d\n",
 255                       protocol, msgtype);
 256 }
 257 EXPORT_SYMBOL_GPL(rtnl_register);
 258
 259 /**
 260  * rtnl_unregister - Unregister a rtnetlink message type
 261  * @protocol: Protocol family or PF_UNSPEC
 262  * @msgtype: rtnetlink message type
 263  *
 264  * Returns 0 on success or a negative error code.
 265  */
 266 int rtnl_unregister(int protocol, int msgtype)
 267 {
 268         int msgindex;
 269
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271         msgindex = rtm_msgindex(msgtype);
 272
 273         if (rtnl_msg_handlers[protocol] == NULL)
 274                 return -ENOENT;
 275
 276         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 277         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 278         rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
 279
 280         return 0;
 281 }
 282 EXPORT_SYMBOL_GPL(rtnl_unregister);
 283
 284 /**
 285  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 286  * @protocol : Protocol family or PF_UNSPEC
 287  *
 288  * Identical to calling rtnl_unregster() for all registered message types
 289  * of a certain protocol family.
 290  */
 291 void rtnl_unregister_all(int protocol)
 292 {
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294
 295         kfree(rtnl_msg_handlers[protocol]);
 296         rtnl_msg_handlers[protocol] = NULL;
 297 }
 298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 299
 300 static LIST_HEAD(link_ops);
 301
 302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 303 {
 304         const struct rtnl_link_ops *ops;
 305
 306         list_for_each_entry(ops, &link_ops, list) {
 307                 if (!strcmp(ops->kind, kind))
 308                         return ops;
 309         }
 310         return NULL;
 311 }
 312
 313 /**
 314  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 315  * @ops: struct rtnl_link_ops * to register
 316  *
 317  * The caller must hold the rtnl_mutex. This function should be used
 318  * by drivers that create devices during module initialization. It
 319  * must be called before registering the devices.
 320  *
 321  * Returns 0 on success or a negative error code.
 322  */
 323 int __rtnl_link_register(struct rtnl_link_ops *ops)
 324 {
 325         if (rtnl_link_ops_get(ops->kind))
 326                 return -EEXIST;
 327
 328         /* The check for setup is here because if ops
 329          * does not have that filled up, it is not possible
 330          * to use the ops for creating device. So do not
 331          * fill up dellink as well. That disables rtnl_dellink.
 332          */
 333         if (ops->setup && !ops->dellink)
 334                 ops->dellink = unregister_netdevice_queue;
 335
 336         list_add_tail(&ops->list, &link_ops);
 337         return 0;
 338 }
 339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 340
 341 /**
 342  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 343  * @ops: struct rtnl_link_ops * to register
 344  *
 345  * Returns 0 on success or a negative error code.
 346  */
 347 int rtnl_link_register(struct rtnl_link_ops *ops)
 348 {
 349         int err;
 350
 351         rtnl_lock();
 352         err = __rtnl_link_register(ops);
 353         rtnl_unlock();
 354         return err;
 355 }
 356 EXPORT_SYMBOL_GPL(rtnl_link_register);
 357
 358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 359 {
 360         struct net_device *dev;
 361         LIST_HEAD(list_kill);
 362
 363         for_each_netdev(net, dev) {
 364                 if (dev->rtnl_link_ops == ops)
 365                         ops->dellink(dev, &list_kill);
 366         }
 367         unregister_netdevice_many(&list_kill);
 368 }
 369
 370 /**
 371  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 372  * @ops: struct rtnl_link_ops * to unregister
 373  *
 374  * The caller must hold the rtnl_mutex.
 375  */
 376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 377 {
 378         struct net *net;
 379
 380         for_each_net(net) {
 381                 __rtnl_kill_links(net, ops);
 382         }
 383         list_del(&ops->list);
 384 }
 385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 386
 387 /* Return with the rtnl_lock held when there are no network
 388  * devices unregistering in any network namespace.
 389  */
 390 static void rtnl_lock_unregistering_all(void)
 391 {
 392         struct net *net;
 393         bool unregistering;
 394         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 395
 396         add_wait_queue(&netdev_unregistering_wq, &wait);
 397         for (;;) {
 398                 unregistering = false;
 399                 rtnl_lock();
 400                 for_each_net(net) {
 401                         if (net->dev_unreg_count > 0) {
 402                                 unregistering = true;
 403                                 break;
 404                         }
 405                 }
 406                 if (!unregistering)
 407                         break;
 408                 __rtnl_unlock();
 409
 410                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 411         }
 412         remove_wait_queue(&netdev_unregistering_wq, &wait);
 413 }
 414
 415 /**
 416  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 417  * @ops: struct rtnl_link_ops * to unregister
 418  */
 419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 420 {
 421         /* Close the race with cleanup_net() */
 422         mutex_lock(&net_mutex);
 423         rtnl_lock_unregistering_all();
 424         __rtnl_link_unregister(ops);
 425         rtnl_unlock();
 426         mutex_unlock(&net_mutex);
 427 }
 428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 429
 430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 431 {
 432         struct net_device *master_dev;
 433         const struct rtnl_link_ops *ops;
 434
 435         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 436         if (!master_dev)
 437                 return 0;
 438         ops = master_dev->rtnl_link_ops;
 439         if (!ops || !ops->get_slave_size)
 440                 return 0;
 441         /* IFLA_INFO_SLAVE_DATA + nested data */
 442         return nla_total_size(sizeof(struct nlattr)) +
 443                ops->get_slave_size(master_dev, dev);
 444 }
 445
 446 static size_t rtnl_link_get_size(const struct net_device *dev)
 447 {
 448         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 449         size_t size;
 450
 451         if (!ops)
 452                 return 0;
 453
 454         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 455                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 456
 457         if (ops->get_size)
 458                 /* IFLA_INFO_DATA + nested data */
 459                 size += nla_total_size(sizeof(struct nlattr)) +
 460                         ops->get_size(dev);
 461
 462         if (ops->get_xstats_size)
 463                 /* IFLA_INFO_XSTATS */
 464                 size += nla_total_size(ops->get_xstats_size(dev));
 465
 466         size += rtnl_link_get_slave_info_data_size(dev);
 467
 468         return size;
 469 }
 470
 471 static LIST_HEAD(rtnl_af_ops);
 472
 473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 474 {
 475         const struct rtnl_af_ops *ops;
 476
 477         list_for_each_entry(ops, &rtnl_af_ops, list) {
 478                 if (ops->family == family)
 479                         return ops;
 480         }
 481
 482         return NULL;
 483 }
 484
 485 /**
 486  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 487  * @ops: struct rtnl_af_ops * to register
 488  *
 489  * Returns 0 on success or a negative error code.
 490  */
 491 void rtnl_af_register(struct rtnl_af_ops *ops)
 492 {
 493         rtnl_lock();
 494         list_add_tail(&ops->list, &rtnl_af_ops);
 495         rtnl_unlock();
 496 }
 497 EXPORT_SYMBOL_GPL(rtnl_af_register);
 498
 499 /**
 500  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 501  * @ops: struct rtnl_af_ops * to unregister
 502  *
 503  * The caller must hold the rtnl_mutex.
 504  */
 505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 506 {
 507         list_del(&ops->list);
 508 }
 509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 510
 511 /**
 512  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 513  * @ops: struct rtnl_af_ops * to unregister
 514  */
 515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 516 {
 517         rtnl_lock();
 518         __rtnl_af_unregister(ops);
 519         rtnl_unlock();
 520 }
 521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 522
 523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 524                                     u32 ext_filter_mask)
 525 {
 526         struct rtnl_af_ops *af_ops;
 527         size_t size;
 528
 529         /* IFLA_AF_SPEC */
 530         size = nla_total_size(sizeof(struct nlattr));
 531
 532         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 533                 if (af_ops->get_link_af_size) {
 534                         /* AF_* + nested data */
 535                         size += nla_total_size(sizeof(struct nlattr)) +
 536                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 537                 }
 538         }
 539
 540         return size;
 541 }
 542
 543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 544 {
 545         struct net_device *master_dev;
 546
 547         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 548         if (master_dev && master_dev->rtnl_link_ops)
 549                 return true;
 550         return false;
 551 }
 552
 553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 554                                      const struct net_device *dev)
 555 {
 556         struct net_device *master_dev;
 557         const struct rtnl_link_ops *ops;
 558         struct nlattr *slave_data;
 559         int err;
 560
 561         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 562         if (!master_dev)
 563                 return 0;
 564         ops = master_dev->rtnl_link_ops;
 565         if (!ops)
 566                 return 0;
 567         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 568                 return -EMSGSIZE;
 569         if (ops->fill_slave_info) {
 570                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 571                 if (!slave_data)
 572                         return -EMSGSIZE;
 573                 err = ops->fill_slave_info(skb, master_dev, dev);
 574                 if (err < 0)
 575                         goto err_cancel_slave_data;
 576                 nla_nest_end(skb, slave_data);
 577         }
 578         return 0;
 579
 580 err_cancel_slave_data:
 581         nla_nest_cancel(skb, slave_data);
 582         return err;
 583 }
 584
 585 static int rtnl_link_info_fill(struct sk_buff *skb,
 586                                const struct net_device *dev)
 587 {
 588         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 589         struct nlattr *data;
 590         int err;
 591
 592         if (!ops)
 593                 return 0;
 594         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 595                 return -EMSGSIZE;
 596         if (ops->fill_xstats) {
 597                 err = ops->fill_xstats(skb, dev);
 598                 if (err < 0)
 599                         return err;
 600         }
 601         if (ops->fill_info) {
 602                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 603                 if (data == NULL)
 604                         return -EMSGSIZE;
 605                 err = ops->fill_info(skb, dev);
 606                 if (err < 0)
 607                         goto err_cancel_data;
 608                 nla_nest_end(skb, data);
 609         }
 610         return 0;
 611
 612 err_cancel_data:
 613         nla_nest_cancel(skb, data);
 614         return err;
 615 }
 616
 617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 618 {
 619         struct nlattr *linkinfo;
 620         int err = -EMSGSIZE;
 621
 622         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 623         if (linkinfo == NULL)
 624                 goto out;
 625
 626         err = rtnl_link_info_fill(skb, dev);
 627         if (err < 0)
 628                 goto err_cancel_link;
 629
 630         err = rtnl_link_slave_info_fill(skb, dev);
 631         if (err < 0)
 632                 goto err_cancel_link;
 633
 634         nla_nest_end(skb, linkinfo);
 635         return 0;
 636
 637 err_cancel_link:
 638         nla_nest_cancel(skb, linkinfo);
 639 out:
 640         return err;
 641 }
 642
 643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 644 {
 645         struct sock *rtnl = net->rtnl;
 646         int err = 0;
 647
 648         NETLINK_CB(skb).dst_group = group;
 649         if (echo)
 650                 atomic_inc(&skb->users);
 651         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 652         if (echo)
 653                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 654         return err;
 655 }
 656
 657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 658 {
 659         struct sock *rtnl = net->rtnl;
 660
 661         return nlmsg_unicast(rtnl, skb, pid);
 662 }
 663 EXPORT_SYMBOL(rtnl_unicast);
 664
 665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 666                  struct nlmsghdr *nlh, gfp_t flags)
 667 {
 668         struct sock *rtnl = net->rtnl;
 669         int report = 0;
 670
 671         if (nlh)
 672                 report = nlmsg_report(nlh);
 673
 674         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 675 }
 676 EXPORT_SYMBOL(rtnl_notify);
 677
 678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 679 {
 680         struct sock *rtnl = net->rtnl;
 681
 682         netlink_set_err(rtnl, 0, group, error);
 683 }
 684 EXPORT_SYMBOL(rtnl_set_sk_err);
 685
 686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 687 {
 688         struct nlattr *mx;
 689         int i, valid = 0;
 690
 691         mx = nla_nest_start(skb, RTA_METRICS);
 692         if (mx == NULL)
 693                 return -ENOBUFS;
 694
 695         for (i = 0; i < RTAX_MAX; i++) {
 696                 if (metrics[i]) {
 697                         if (i == RTAX_CC_ALGO - 1) {
 698                                 char tmp[TCP_CA_NAME_MAX], *name;
 699
 700                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 701                                 if (!name)
 702                                         continue;
 703                                 if (nla_put_string(skb, i + 1, name))
 704                                         goto nla_put_failure;
 705                         } else if (i == RTAX_FEATURES - 1) {
 706                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 707
 708                                 if (!user_features)
 709                                         continue;
 710                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 711                                 if (nla_put_u32(skb, i + 1, user_features))
 712                                         goto nla_put_failure;
 713                         } else {
 714                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 715                                         goto nla_put_failure;
 716                         }
 717                         valid++;
 718                 }
 719         }
 720
 721         if (!valid) {
 722                 nla_nest_cancel(skb, mx);
 723                 return 0;
 724         }
 725
 726         return nla_nest_end(skb, mx);
 727
 728 nla_put_failure:
 729         nla_nest_cancel(skb, mx);
 730         return -EMSGSIZE;
 731 }
 732 EXPORT_SYMBOL(rtnetlink_put_metrics);
 733
 734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 735                        long expires, u32 error)
 736 {
 737         struct rta_cacheinfo ci = {
 738                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 739                 .rta_used = dst->__use,
 740                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 741                 .rta_error = error,
 742                 .rta_id =  id,
 743         };
 744
 745         if (expires) {
 746                 unsigned long clock;
 747
 748                 clock = jiffies_to_clock_t(abs(expires));
 749                 clock = min_t(unsigned long, clock, INT_MAX);
 750                 ci.rta_expires = (expires > 0) ? clock : -clock;
 751         }
 752         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 753 }
 754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 755
 756 static void set_operstate(struct net_device *dev, unsigned char transition)
 757 {
 758         unsigned char operstate = dev->operstate;
 759
 760         switch (transition) {
 761         case IF_OPER_UP:
 762                 if ((operstate == IF_OPER_DORMANT ||
 763                      operstate == IF_OPER_UNKNOWN) &&
 764                     !netif_dormant(dev))
 765                         operstate = IF_OPER_UP;
 766                 break;
 767
 768         case IF_OPER_DORMANT:
 769                 if (operstate == IF_OPER_UP ||
 770                     operstate == IF_OPER_UNKNOWN)
 771                         operstate = IF_OPER_DORMANT;
 772                 break;
 773         }
 774
 775         if (dev->operstate != operstate) {
 776                 write_lock_bh(&dev_base_lock);
 777                 dev->operstate = operstate;
 778                 write_unlock_bh(&dev_base_lock);
 779                 netdev_state_change(dev);
 780         }
 781 }
 782
 783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 784 {
 785         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 786                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 787 }
 788
 789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 790                                            const struct ifinfomsg *ifm)
 791 {
 792         unsigned int flags = ifm->ifi_flags;
 793
 794         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 795         if (ifm->ifi_change)
 796                 flags = (flags & ifm->ifi_change) |
 797                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 798
 799         return flags;
 800 }
 801
 802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 803                                  const struct rtnl_link_stats64 *b)
 804 {
 805         a->rx_packets = b->rx_packets;
 806         a->tx_packets = b->tx_packets;
 807         a->rx_bytes = b->rx_bytes;
 808         a->tx_bytes = b->tx_bytes;
 809         a->rx_errors = b->rx_errors;
 810         a->tx_errors = b->tx_errors;
 811         a->rx_dropped = b->rx_dropped;
 812         a->tx_dropped = b->tx_dropped;
 813
 814         a->multicast = b->multicast;
 815         a->collisions = b->collisions;
 816
 817         a->rx_length_errors = b->rx_length_errors;
 818         a->rx_over_errors = b->rx_over_errors;
 819         a->rx_crc_errors = b->rx_crc_errors;
 820         a->rx_frame_errors = b->rx_frame_errors;
 821         a->rx_fifo_errors = b->rx_fifo_errors;
 822         a->rx_missed_errors = b->rx_missed_errors;
 823
 824         a->tx_aborted_errors = b->tx_aborted_errors;
 825         a->tx_carrier_errors = b->tx_carrier_errors;
 826         a->tx_fifo_errors = b->tx_fifo_errors;
 827         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 828         a->tx_window_errors = b->tx_window_errors;
 829
 830         a->rx_compressed = b->rx_compressed;
 831         a->tx_compressed = b->tx_compressed;
 832
 833         a->rx_nohandler = b->rx_nohandler;
 834 }
 835
 836 /* All VF info */
 837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 838                                    u32 ext_filter_mask)
 839 {
 840         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 841                 int num_vfs = dev_num_vf(dev->dev.parent);
 842                 size_t size = nla_total_size(0);
 843                 size += num_vfs *
 844                         (nla_total_size(0) +
 845                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 846                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 847                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 848                          nla_total_size(MAX_VLAN_LIST_LEN *
 849                                         sizeof(struct ifla_vf_vlan_info)) +
 850                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 851                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 852                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 853                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 854                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 855                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 856                          /* IFLA_VF_STATS_RX_PACKETS */
 857                          nla_total_size_64bit(sizeof(__u64)) +
 858                          /* IFLA_VF_STATS_TX_PACKETS */
 859                          nla_total_size_64bit(sizeof(__u64)) +
 860                          /* IFLA_VF_STATS_RX_BYTES */
 861                          nla_total_size_64bit(sizeof(__u64)) +
 862                          /* IFLA_VF_STATS_TX_BYTES */
 863                          nla_total_size_64bit(sizeof(__u64)) +
 864                          /* IFLA_VF_STATS_BROADCAST */
 865                          nla_total_size_64bit(sizeof(__u64)) +
 866                          /* IFLA_VF_STATS_MULTICAST */
 867                          nla_total_size_64bit(sizeof(__u64)) +
 868                          nla_total_size(sizeof(struct ifla_vf_trust)));
 869                 return size;
 870         } else
 871                 return 0;
 872 }
 873
 874 static size_t rtnl_port_size(const struct net_device *dev,
 875                              u32 ext_filter_mask)
 876 {
 877         size_t port_size = nla_total_size(4)            /* PORT_VF */
 878                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 879                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 880                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 881                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 882                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 883         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 884         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 885                 + port_size;
 886         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 887                 + port_size;
 888
 889         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 890             !(ext_filter_mask & RTEXT_FILTER_VF))
 891                 return 0;
 892         if (dev_num_vf(dev->dev.parent))
 893                 return port_self_size + vf_ports_size +
 894                         vf_port_size * dev_num_vf(dev->dev.parent);
 895         else
 896                 return port_self_size;
 897 }
 898
 899 static size_t rtnl_xdp_size(const struct net_device *dev)
 900 {
 901         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 902                           nla_total_size(1);    /* XDP_ATTACHED */
 903
 904         if (!dev->netdev_ops->ndo_xdp)
 905                 return 0;
 906         else
 907                 return xdp_size;
 908 }
 909
 910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 911                                      u32 ext_filter_mask)
 912 {
 913         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 914                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 915                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 916                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 917                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 918                + nla_total_size(sizeof(struct rtnl_link_stats))
 919                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 920                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 921                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 922                + nla_total_size(4) /* IFLA_TXQLEN */
 923                + nla_total_size(4) /* IFLA_WEIGHT */
 924                + nla_total_size(4) /* IFLA_MTU */
 925                + nla_total_size(4) /* IFLA_LINK */
 926                + nla_total_size(4) /* IFLA_MASTER */
 927                + nla_total_size(1) /* IFLA_CARRIER */
 928                + nla_total_size(4) /* IFLA_PROMISCUITY */
 929                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 930                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 931                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 932                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 933                + nla_total_size(1) /* IFLA_OPERSTATE */
 934                + nla_total_size(1) /* IFLA_LINKMODE */
 935                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 936                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 937                + nla_total_size(ext_filter_mask
 938                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 939                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 940                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 941                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 942                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 943                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 944                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 945                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 946                + rtnl_xdp_size(dev) /* IFLA_XDP */
 947                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 948
 949 }
 950
 951 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 952 {
 953         struct nlattr *vf_ports;
 954         struct nlattr *vf_port;
 955         int vf;
 956         int err;
 957
 958         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 959         if (!vf_ports)
 960                 return -EMSGSIZE;
 961
 962         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 963                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 964                 if (!vf_port)
 965                         goto nla_put_failure;
 966                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 967                         goto nla_put_failure;
 968                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 969                 if (err == -EMSGSIZE)
 970                         goto nla_put_failure;
 971                 if (err) {
 972                         nla_nest_cancel(skb, vf_port);
 973                         continue;
 974                 }
 975                 nla_nest_end(skb, vf_port);
 976         }
 977
 978         nla_nest_end(skb, vf_ports);
 979
 980         return 0;
 981
 982 nla_put_failure:
 983         nla_nest_cancel(skb, vf_ports);
 984         return -EMSGSIZE;
 985 }
 986
 987 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 988 {
 989         struct nlattr *port_self;
 990         int err;
 991
 992         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 993         if (!port_self)
 994                 return -EMSGSIZE;
 995
 996         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 997         if (err) {
 998                 nla_nest_cancel(skb, port_self);
 999                 return (err == -EMSGSIZE) ? err : 0;
1000         }
1001
1002         nla_nest_end(skb, port_self);
1003
1004         return 0;
1005 }
1006
1007 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1008                           u32 ext_filter_mask)
1009 {
1010         int err;
1011
1012         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1013             !(ext_filter_mask & RTEXT_FILTER_VF))
1014                 return 0;
1015
1016         err = rtnl_port_self_fill(skb, dev);
1017         if (err)
1018                 return err;
1019
1020         if (dev_num_vf(dev->dev.parent)) {
1021                 err = rtnl_vf_ports_fill(skb, dev);
1022                 if (err)
1023                         return err;
1024         }
1025
1026         return 0;
1027 }
1028
1029 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1030 {
1031         int err;
1032         struct netdev_phys_item_id ppid;
1033
1034         err = dev_get_phys_port_id(dev, &ppid);
1035         if (err) {
1036                 if (err == -EOPNOTSUPP)
1037                         return 0;
1038                 return err;
1039         }
1040
1041         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1042                 return -EMSGSIZE;
1043
1044         return 0;
1045 }
1046
1047 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1048 {
1049         char name[IFNAMSIZ];
1050         int err;
1051
1052         err = dev_get_phys_port_name(dev, name, sizeof(name));
1053         if (err) {
1054                 if (err == -EOPNOTSUPP)
1055                         return 0;
1056                 return err;
1057         }
1058
1059         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1060                 return -EMSGSIZE;
1061
1062         return 0;
1063 }
1064
1065 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1066 {
1067         int err;
1068         struct switchdev_attr attr = {
1069                 .orig_dev = dev,
1070                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1071                 .flags = SWITCHDEV_F_NO_RECURSE,
1072         };
1073
1074         err = switchdev_port_attr_get(dev, &attr);
1075         if (err) {
1076                 if (err == -EOPNOTSUPP)
1077                         return 0;
1078                 return err;
1079         }
1080
1081         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1082                     attr.u.ppid.id))
1083                 return -EMSGSIZE;
1084
1085         return 0;
1086 }
1087
1088 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1089                                               struct net_device *dev)
1090 {
1091         struct rtnl_link_stats64 *sp;
1092         struct nlattr *attr;
1093
1094         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1095                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1096         if (!attr)
1097                 return -EMSGSIZE;
1098
1099         sp = nla_data(attr);
1100         dev_get_stats(dev, sp);
1101
1102         attr = nla_reserve(skb, IFLA_STATS,
1103                            sizeof(struct rtnl_link_stats));
1104         if (!attr)
1105                 return -EMSGSIZE;
1106
1107         copy_rtnl_link_stats(nla_data(attr), sp);
1108
1109         return 0;
1110 }
1111
1112 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1113                                                struct net_device *dev,
1114                                                int vfs_num,
1115                                                struct nlattr *vfinfo)
1116 {
1117         struct ifla_vf_rss_query_en vf_rss_query_en;
1118         struct nlattr *vf, *vfstats, *vfvlanlist;
1119         struct ifla_vf_link_state vf_linkstate;
1120         struct ifla_vf_vlan_info vf_vlan_info;
1121         struct ifla_vf_spoofchk vf_spoofchk;
1122         struct ifla_vf_tx_rate vf_tx_rate;
1123         struct ifla_vf_stats vf_stats;
1124         struct ifla_vf_trust vf_trust;
1125         struct ifla_vf_vlan vf_vlan;
1126         struct ifla_vf_rate vf_rate;
1127         struct ifla_vf_mac vf_mac;
1128         struct ifla_vf_info ivi;
1129
1130         /* Not all SR-IOV capable drivers support the
1131          * spoofcheck and "RSS query enable" query.  Preset to
1132          * -1 so the user space tool can detect that the driver
1133          * didn't report anything.
1134          */
1135         ivi.spoofchk = -1;
1136         ivi.rss_query_en = -1;
1137         ivi.trusted = -1;
1138         memset(ivi.mac, 0, sizeof(ivi.mac));
1139         /* The default value for VF link state is "auto"
1140          * IFLA_VF_LINK_STATE_AUTO which equals zero
1141          */
1142         ivi.linkstate = 0;
1143         /* VLAN Protocol by default is 802.1Q */
1144         ivi.vlan_proto = htons(ETH_P_8021Q);
1145         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1146                 return 0;
1147
1148         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1149
1150         vf_mac.vf =
1151                 vf_vlan.vf =
1152                 vf_vlan_info.vf =
1153                 vf_rate.vf =
1154                 vf_tx_rate.vf =
1155                 vf_spoofchk.vf =
1156                 vf_linkstate.vf =
1157                 vf_rss_query_en.vf =
1158                 vf_trust.vf = ivi.vf;
1159
1160         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1161         vf_vlan.vlan = ivi.vlan;
1162         vf_vlan.qos = ivi.qos;
1163         vf_vlan_info.vlan = ivi.vlan;
1164         vf_vlan_info.qos = ivi.qos;
1165         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1166         vf_tx_rate.rate = ivi.max_tx_rate;
1167         vf_rate.min_tx_rate = ivi.min_tx_rate;
1168         vf_rate.max_tx_rate = ivi.max_tx_rate;
1169         vf_spoofchk.setting = ivi.spoofchk;
1170         vf_linkstate.link_state = ivi.linkstate;
1171         vf_rss_query_en.setting = ivi.rss_query_en;
1172         vf_trust.setting = ivi.trusted;
1173         vf = nla_nest_start(skb, IFLA_VF_INFO);
1174         if (!vf)
1175                 goto nla_put_vfinfo_failure;
1176         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1177             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1178             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1179                     &vf_rate) ||
1180             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1181                     &vf_tx_rate) ||
1182             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1183                     &vf_spoofchk) ||
1184             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1185                     &vf_linkstate) ||
1186             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1187                     sizeof(vf_rss_query_en),
1188                     &vf_rss_query_en) ||
1189             nla_put(skb, IFLA_VF_TRUST,
1190                     sizeof(vf_trust), &vf_trust))
1191                 goto nla_put_vf_failure;
1192         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1193         if (!vfvlanlist)
1194                 goto nla_put_vf_failure;
1195         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1196                     &vf_vlan_info)) {
1197                 nla_nest_cancel(skb, vfvlanlist);
1198                 goto nla_put_vf_failure;
1199         }
1200         nla_nest_end(skb, vfvlanlist);
1201         memset(&vf_stats, 0, sizeof(vf_stats));
1202         if (dev->netdev_ops->ndo_get_vf_stats)
1203                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1204                                                 &vf_stats);
1205         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1206         if (!vfstats)
1207                 goto nla_put_vf_failure;
1208         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1209                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1210             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1211                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1212             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1213                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1214             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1215                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1216             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1217                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1218             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1219                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1220                 nla_nest_cancel(skb, vfstats);
1221                 goto nla_put_vf_failure;
1222         }
1223         nla_nest_end(skb, vfstats);
1224         nla_nest_end(skb, vf);
1225         return 0;
1226
1227 nla_put_vf_failure:
1228         nla_nest_cancel(skb, vf);
1229 nla_put_vfinfo_failure:
1230         nla_nest_cancel(skb, vfinfo);
1231         return -EMSGSIZE;
1232 }
1233
1234 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1235 {
1236         struct rtnl_link_ifmap map;
1237
1238         memset(&map, 0, sizeof(map));
1239         map.mem_start   = dev->mem_start;
1240         map.mem_end     = dev->mem_end;
1241         map.base_addr   = dev->base_addr;
1242         map.irq         = dev->irq;
1243         map.dma         = dev->dma;
1244         map.port        = dev->if_port;
1245
1246         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1247                 return -EMSGSIZE;
1248
1249         return 0;
1250 }
1251
1252 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1253 {
1254         struct netdev_xdp xdp_op = {};
1255         struct nlattr *xdp;
1256         int err;
1257
1258         if (!dev->netdev_ops->ndo_xdp)
1259                 return 0;
1260         xdp = nla_nest_start(skb, IFLA_XDP);
1261         if (!xdp)
1262                 return -EMSGSIZE;
1263         xdp_op.command = XDP_QUERY_PROG;
1264         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1265         if (err)
1266                 goto err_cancel;
1267         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1268         if (err)
1269                 goto err_cancel;
1270
1271         nla_nest_end(skb, xdp);
1272         return 0;
1273
1274 err_cancel:
1275         nla_nest_cancel(skb, xdp);
1276         return err;
1277 }
1278
1279 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1280                             int type, u32 pid, u32 seq, u32 change,
1281                             unsigned int flags, u32 ext_filter_mask)
1282 {
1283         struct ifinfomsg *ifm;
1284         struct nlmsghdr *nlh;
1285         struct nlattr *af_spec;
1286         struct rtnl_af_ops *af_ops;
1287         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1288
1289         ASSERT_RTNL();
1290         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1291         if (nlh == NULL)
1292                 return -EMSGSIZE;
1293
1294         ifm = nlmsg_data(nlh);
1295         ifm->ifi_family = AF_UNSPEC;
1296         ifm->__ifi_pad = 0;
1297         ifm->ifi_type = dev->type;
1298         ifm->ifi_index = dev->ifindex;
1299         ifm->ifi_flags = dev_get_flags(dev);
1300         ifm->ifi_change = change;
1301
1302         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1303             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1304             nla_put_u8(skb, IFLA_OPERSTATE,
1305                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1306             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1307             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1308             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1309             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1310             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1311             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1312             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1313 #ifdef CONFIG_RPS
1314             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1315 #endif
1316             (dev->ifindex != dev_get_iflink(dev) &&
1317              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1318             (upper_dev &&
1319              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1320             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1321             (dev->qdisc &&
1322              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1323             (dev->ifalias &&
1324              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1325             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1326                         atomic_read(&dev->carrier_changes)) ||
1327             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1328                 goto nla_put_failure;
1329
1330         if (rtnl_fill_link_ifmap(skb, dev))
1331                 goto nla_put_failure;
1332
1333         if (dev->addr_len) {
1334                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1335                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1336                         goto nla_put_failure;
1337         }
1338
1339         if (rtnl_phys_port_id_fill(skb, dev))
1340                 goto nla_put_failure;
1341
1342         if (rtnl_phys_port_name_fill(skb, dev))
1343                 goto nla_put_failure;
1344
1345         if (rtnl_phys_switch_id_fill(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (rtnl_fill_stats(skb, dev))
1349                 goto nla_put_failure;
1350
1351         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1352             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1353                 goto nla_put_failure;
1354
1355         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1356             ext_filter_mask & RTEXT_FILTER_VF) {
1357                 int i;
1358                 struct nlattr *vfinfo;
1359                 int num_vfs = dev_num_vf(dev->dev.parent);
1360
1361                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1362                 if (!vfinfo)
1363                         goto nla_put_failure;
1364                 for (i = 0; i < num_vfs; i++) {
1365                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1366                                 goto nla_put_failure;
1367                 }
1368
1369                 nla_nest_end(skb, vfinfo);
1370         }
1371
1372         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1373                 goto nla_put_failure;
1374
1375         if (rtnl_xdp_fill(skb, dev))
1376                 goto nla_put_failure;
1377
1378         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1379                 if (rtnl_link_fill(skb, dev) < 0)
1380                         goto nla_put_failure;
1381         }
1382
1383         if (dev->rtnl_link_ops &&
1384             dev->rtnl_link_ops->get_link_net) {
1385                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1386
1387                 if (!net_eq(dev_net(dev), link_net)) {
1388                         int id = peernet2id_alloc(dev_net(dev), link_net);
1389
1390                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1391                                 goto nla_put_failure;
1392                 }
1393         }
1394
1395         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1396                 goto nla_put_failure;
1397
1398         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1399                 if (af_ops->fill_link_af) {
1400                         struct nlattr *af;
1401                         int err;
1402
1403                         if (!(af = nla_nest_start(skb, af_ops->family)))
1404                                 goto nla_put_failure;
1405
1406                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1407
1408                         /*
1409                          * Caller may return ENODATA to indicate that there
1410                          * was no data to be dumped. This is not an error, it
1411                          * means we should trim the attribute header and
1412                          * continue.
1413                          */
1414                         if (err == -ENODATA)
1415                                 nla_nest_cancel(skb, af);
1416                         else if (err < 0)
1417                                 goto nla_put_failure;
1418
1419                         nla_nest_end(skb, af);
1420                 }
1421         }
1422
1423         nla_nest_end(skb, af_spec);
1424
1425         nlmsg_end(skb, nlh);
1426         return 0;
1427
1428 nla_put_failure:
1429         nlmsg_cancel(skb, nlh);
1430         return -EMSGSIZE;
1431 }
1432
1433 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1434         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1435         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1436         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1437         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1438         [IFLA_MTU]              = { .type = NLA_U32 },
1439         [IFLA_LINK]             = { .type = NLA_U32 },
1440         [IFLA_MASTER]           = { .type = NLA_U32 },
1441         [IFLA_CARRIER]          = { .type = NLA_U8 },
1442         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1443         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1444         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1445         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1446         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1447         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1448         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1449         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1450         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1451         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1452         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1453         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1454         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1455         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1456         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1457         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1458         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1459         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1460         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1461         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1462         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1463         [IFLA_XDP]              = { .type = NLA_NESTED },
1464 };
1465
1466 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1467         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1468         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1469         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1470         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1471 };
1472
1473 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1474         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1475         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1476         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1477         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1478         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1479         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1480         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1481         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1482         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1483         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1484         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1485         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1486 };
1487
1488 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1489         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1490         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1491                                     .len = PORT_PROFILE_MAX },
1492         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1493                                       .len = PORT_UUID_MAX },
1494         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1495                                     .len = PORT_UUID_MAX },
1496         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1497         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1498
1499         /* Unused, but we need to keep it here since user space could
1500          * fill it. It's also broken with regard to NLA_BINARY use in
1501          * combination with structs.
1502          */
1503         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1504                                     .len = sizeof(struct ifla_port_vsi) },
1505 };
1506
1507 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1508         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1509         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1510         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1511 };
1512
1513 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1514 {
1515         const struct rtnl_link_ops *ops = NULL;
1516         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1517
1518         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1519                              ifla_info_policy, NULL) < 0)
1520                 return NULL;
1521
1522         if (linfo[IFLA_INFO_KIND]) {
1523                 char kind[MODULE_NAME_LEN];
1524
1525                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1526                 ops = rtnl_link_ops_get(kind);
1527         }
1528
1529         return ops;
1530 }
1531
1532 static bool link_master_filtered(struct net_device *dev, int master_idx)
1533 {
1534         struct net_device *master;
1535
1536         if (!master_idx)
1537                 return false;
1538
1539         master = netdev_master_upper_dev_get(dev);
1540         if (!master || master->ifindex != master_idx)
1541                 return true;
1542
1543         return false;
1544 }
1545
1546 static bool link_kind_filtered(const struct net_device *dev,
1547                                const struct rtnl_link_ops *kind_ops)
1548 {
1549         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1550                 return true;
1551
1552         return false;
1553 }
1554
1555 static bool link_dump_filtered(struct net_device *dev,
1556                                int master_idx,
1557                                const struct rtnl_link_ops *kind_ops)
1558 {
1559         if (link_master_filtered(dev, master_idx) ||
1560             link_kind_filtered(dev, kind_ops))
1561                 return true;
1562
1563         return false;
1564 }
1565
1566 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1567 {
1568         struct net *net = sock_net(skb->sk);
1569         int h, s_h;
1570         int idx = 0, s_idx;
1571         struct net_device *dev;
1572         struct hlist_head *head;
1573         struct nlattr *tb[IFLA_MAX+1];
1574         u32 ext_filter_mask = 0;
1575         const struct rtnl_link_ops *kind_ops = NULL;
1576         unsigned int flags = NLM_F_MULTI;
1577         int master_idx = 0;
1578         int err;
1579         int hdrlen;
1580
1581         s_h = cb->args[0];
1582         s_idx = cb->args[1];
1583
1584         cb->seq = net->dev_base_seq;
1585
1586         /* A hack to preserve kernel<->userspace interface.
1587          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1588          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1589          * what iproute2 < v3.9.0 used.
1590          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1591          * attribute, its netlink message is shorter than struct ifinfomsg.
1592          */
1593         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1594                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1595
1596         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1597                         ifla_policy, NULL) >= 0) {
1598                 if (tb[IFLA_EXT_MASK])
1599                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1600
1601                 if (tb[IFLA_MASTER])
1602                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1603
1604                 if (tb[IFLA_LINKINFO])
1605                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1606
1607                 if (master_idx || kind_ops)
1608                         flags |= NLM_F_DUMP_FILTERED;
1609         }
1610
1611         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1612                 idx = 0;
1613                 head = &net->dev_index_head[h];
1614                 hlist_for_each_entry(dev, head, index_hlist) {
1615                         if (link_dump_filtered(dev, master_idx, kind_ops))
1616                                 goto cont;
1617                         if (idx < s_idx)
1618                                 goto cont;
1619                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1620                                                NETLINK_CB(cb->skb).portid,
1621                                                cb->nlh->nlmsg_seq, 0,
1622                                                flags,
1623                                                ext_filter_mask);
1624                         /* If we ran out of room on the first message,
1625                          * we're in trouble
1626                          */
1627                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1628
1629                         if (err < 0)
1630                                 goto out;
1631
1632                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1633 cont:
1634                         idx++;
1635                 }
1636         }
1637 out:
1638         cb->args[1] = idx;
1639         cb->args[0] = h;
1640
1641         return skb->len;
1642 }
1643
1644 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1645                         struct netlink_ext_ack *exterr)
1646 {
1647         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1648 }
1649 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1650
1651 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1652 {
1653         struct net *net;
1654         /* Examine the link attributes and figure out which
1655          * network namespace we are talking about.
1656          */
1657         if (tb[IFLA_NET_NS_PID])
1658                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1659         else if (tb[IFLA_NET_NS_FD])
1660                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1661         else
1662                 net = get_net(src_net);
1663         return net;
1664 }
1665 EXPORT_SYMBOL(rtnl_link_get_net);
1666
1667 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1668 {
1669         if (dev) {
1670                 if (tb[IFLA_ADDRESS] &&
1671                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1672                         return -EINVAL;
1673
1674                 if (tb[IFLA_BROADCAST] &&
1675                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1676                         return -EINVAL;
1677         }
1678
1679         if (tb[IFLA_AF_SPEC]) {
1680                 struct nlattr *af;
1681                 int rem, err;
1682
1683                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1684                         const struct rtnl_af_ops *af_ops;
1685
1686                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1687                                 return -EAFNOSUPPORT;
1688
1689                         if (!af_ops->set_link_af)
1690                                 return -EOPNOTSUPP;
1691
1692                         if (af_ops->validate_link_af) {
1693                                 err = af_ops->validate_link_af(dev, af);
1694                                 if (err < 0)
1695                                         return err;
1696                         }
1697                 }
1698         }
1699
1700         return 0;
1701 }
1702
1703 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1704                                   int guid_type)
1705 {
1706         const struct net_device_ops *ops = dev->netdev_ops;
1707
1708         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1709 }
1710
1711 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1712 {
1713         if (dev->type != ARPHRD_INFINIBAND)
1714                 return -EOPNOTSUPP;
1715
1716         return handle_infiniband_guid(dev, ivt, guid_type);
1717 }
1718
1719 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1720 {
1721         const struct net_device_ops *ops = dev->netdev_ops;
1722         int err = -EINVAL;
1723
1724         if (tb[IFLA_VF_MAC]) {
1725                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1726
1727                 err = -EOPNOTSUPP;
1728                 if (ops->ndo_set_vf_mac)
1729                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1730                                                   ivm->mac);
1731                 if (err < 0)
1732                         return err;
1733         }
1734
1735         if (tb[IFLA_VF_VLAN]) {
1736                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1737
1738                 err = -EOPNOTSUPP;
1739                 if (ops->ndo_set_vf_vlan)
1740                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1741                                                    ivv->qos,
1742                                                    htons(ETH_P_8021Q));
1743                 if (err < 0)
1744                         return err;
1745         }
1746
1747         if (tb[IFLA_VF_VLAN_LIST]) {
1748                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1749                 struct nlattr *attr;
1750                 int rem, len = 0;
1751
1752                 err = -EOPNOTSUPP;
1753                 if (!ops->ndo_set_vf_vlan)
1754                         return err;
1755
1756                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1757                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1758                             nla_len(attr) < NLA_HDRLEN) {
1759                                 return -EINVAL;
1760                         }
1761                         if (len >= MAX_VLAN_LIST_LEN)
1762                                 return -EOPNOTSUPP;
1763                         ivvl[len] = nla_data(attr);
1764
1765                         len++;
1766                 }
1767                 if (len == 0)
1768                         return -EINVAL;
1769
1770                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1771                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1772                 if (err < 0)
1773                         return err;
1774         }
1775
1776         if (tb[IFLA_VF_TX_RATE]) {
1777                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1778                 struct ifla_vf_info ivf;
1779
1780                 err = -EOPNOTSUPP;
1781                 if (ops->ndo_get_vf_config)
1782                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1783                 if (err < 0)
1784                         return err;
1785
1786                 err = -EOPNOTSUPP;
1787                 if (ops->ndo_set_vf_rate)
1788                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1789                                                    ivf.min_tx_rate,
1790                                                    ivt->rate);
1791                 if (err < 0)
1792                         return err;
1793         }
1794
1795         if (tb[IFLA_VF_RATE]) {
1796                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1797
1798                 err = -EOPNOTSUPP;
1799                 if (ops->ndo_set_vf_rate)
1800                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1801                                                    ivt->min_tx_rate,
1802                                                    ivt->max_tx_rate);
1803                 if (err < 0)
1804                         return err;
1805         }
1806
1807         if (tb[IFLA_VF_SPOOFCHK]) {
1808                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1809
1810                 err = -EOPNOTSUPP;
1811                 if (ops->ndo_set_vf_spoofchk)
1812                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1813                                                        ivs->setting);
1814                 if (err < 0)
1815                         return err;
1816         }
1817
1818         if (tb[IFLA_VF_LINK_STATE]) {
1819                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1820
1821                 err = -EOPNOTSUPP;
1822                 if (ops->ndo_set_vf_link_state)
1823                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1824                                                          ivl->link_state);
1825                 if (err < 0)
1826                         return err;
1827         }
1828
1829         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1830                 struct ifla_vf_rss_query_en *ivrssq_en;
1831
1832                 err = -EOPNOTSUPP;
1833                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1834                 if (ops->ndo_set_vf_rss_query_en)
1835                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1836                                                            ivrssq_en->setting);
1837                 if (err < 0)
1838                         return err;
1839         }
1840
1841         if (tb[IFLA_VF_TRUST]) {
1842                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1843
1844                 err = -EOPNOTSUPP;
1845                 if (ops->ndo_set_vf_trust)
1846                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1847                 if (err < 0)
1848                         return err;
1849         }
1850
1851         if (tb[IFLA_VF_IB_NODE_GUID]) {
1852                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1853
1854                 if (!ops->ndo_set_vf_guid)
1855                         return -EOPNOTSUPP;
1856
1857                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1858         }
1859
1860         if (tb[IFLA_VF_IB_PORT_GUID]) {
1861                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1862
1863                 if (!ops->ndo_set_vf_guid)
1864                         return -EOPNOTSUPP;
1865
1866                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1867         }
1868
1869         return err;
1870 }
1871
1872 static int do_set_master(struct net_device *dev, int ifindex)
1873 {
1874         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1875         const struct net_device_ops *ops;
1876         int err;
1877
1878         if (upper_dev) {
1879                 if (upper_dev->ifindex == ifindex)
1880                         return 0;
1881                 ops = upper_dev->netdev_ops;
1882                 if (ops->ndo_del_slave) {
1883                         err = ops->ndo_del_slave(upper_dev, dev);
1884                         if (err)
1885                                 return err;
1886                 } else {
1887                         return -EOPNOTSUPP;
1888                 }
1889         }
1890
1891         if (ifindex) {
1892                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1893                 if (!upper_dev)
1894                         return -EINVAL;
1895                 ops = upper_dev->netdev_ops;
1896                 if (ops->ndo_add_slave) {
1897                         err = ops->ndo_add_slave(upper_dev, dev);
1898                         if (err)
1899                                 return err;
1900                 } else {
1901                         return -EOPNOTSUPP;
1902                 }
1903         }
1904         return 0;
1905 }
1906
1907 #define DO_SETLINK_MODIFIED     0x01
1908 /* notify flag means notify + modified. */
1909 #define DO_SETLINK_NOTIFY       0x03
1910 static int do_setlink(const struct sk_buff *skb,
1911                       struct net_device *dev, struct ifinfomsg *ifm,
1912                       struct nlattr **tb, char *ifname, int status)
1913 {
1914         const struct net_device_ops *ops = dev->netdev_ops;
1915         int err;
1916
1917         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1918                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1919                 if (IS_ERR(net)) {
1920                         err = PTR_ERR(net);
1921                         goto errout;
1922                 }
1923                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1924                         put_net(net);
1925                         err = -EPERM;
1926                         goto errout;
1927                 }
1928                 err = dev_change_net_namespace(dev, net, ifname);
1929                 put_net(net);
1930                 if (err)
1931                         goto errout;
1932                 status |= DO_SETLINK_MODIFIED;
1933         }
1934
1935         if (tb[IFLA_MAP]) {
1936                 struct rtnl_link_ifmap *u_map;
1937                 struct ifmap k_map;
1938
1939                 if (!ops->ndo_set_config) {
1940                         err = -EOPNOTSUPP;
1941                         goto errout;
1942                 }
1943
1944                 if (!netif_device_present(dev)) {
1945                         err = -ENODEV;
1946                         goto errout;
1947                 }
1948
1949                 u_map = nla_data(tb[IFLA_MAP]);
1950                 k_map.mem_start = (unsigned long) u_map->mem_start;
1951                 k_map.mem_end = (unsigned long) u_map->mem_end;
1952                 k_map.base_addr = (unsigned short) u_map->base_addr;
1953                 k_map.irq = (unsigned char) u_map->irq;
1954                 k_map.dma = (unsigned char) u_map->dma;
1955                 k_map.port = (unsigned char) u_map->port;
1956
1957                 err = ops->ndo_set_config(dev, &k_map);
1958                 if (err < 0)
1959                         goto errout;
1960
1961                 status |= DO_SETLINK_NOTIFY;
1962         }
1963
1964         if (tb[IFLA_ADDRESS]) {
1965                 struct sockaddr *sa;
1966                 int len;
1967
1968                 len = sizeof(sa_family_t) + dev->addr_len;
1969                 sa = kmalloc(len, GFP_KERNEL);
1970                 if (!sa) {
1971                         err = -ENOMEM;
1972                         goto errout;
1973                 }
1974                 sa->sa_family = dev->type;
1975                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1976                        dev->addr_len);
1977                 err = dev_set_mac_address(dev, sa);
1978                 kfree(sa);
1979                 if (err)
1980                         goto errout;
1981                 status |= DO_SETLINK_MODIFIED;
1982         }
1983
1984         if (tb[IFLA_MTU]) {
1985                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1986                 if (err < 0)
1987                         goto errout;
1988                 status |= DO_SETLINK_MODIFIED;
1989         }
1990
1991         if (tb[IFLA_GROUP]) {
1992                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1993                 status |= DO_SETLINK_NOTIFY;
1994         }
1995
1996         /*
1997          * Interface selected by interface index but interface
1998          * name provided implies that a name change has been
1999          * requested.
2000          */
2001         if (ifm->ifi_index > 0 && ifname[0]) {
2002                 err = dev_change_name(dev, ifname);
2003                 if (err < 0)
2004                         goto errout;
2005                 status |= DO_SETLINK_MODIFIED;
2006         }
2007
2008         if (tb[IFLA_IFALIAS]) {
2009                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2010                                     nla_len(tb[IFLA_IFALIAS]));
2011                 if (err < 0)
2012                         goto errout;
2013                 status |= DO_SETLINK_NOTIFY;
2014         }
2015
2016         if (tb[IFLA_BROADCAST]) {
2017                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2018                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2019         }
2020
2021         if (ifm->ifi_flags || ifm->ifi_change) {
2022                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2023                 if (err < 0)
2024                         goto errout;
2025         }
2026
2027         if (tb[IFLA_MASTER]) {
2028                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2029                 if (err)
2030                         goto errout;
2031                 status |= DO_SETLINK_MODIFIED;
2032         }
2033
2034         if (tb[IFLA_CARRIER]) {
2035                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2036                 if (err)
2037                         goto errout;
2038                 status |= DO_SETLINK_MODIFIED;
2039         }
2040
2041         if (tb[IFLA_TXQLEN]) {
2042                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2043                 unsigned long orig_len = dev->tx_queue_len;
2044
2045                 if (dev->tx_queue_len ^ value) {
2046                         dev->tx_queue_len = value;
2047                         err = call_netdevice_notifiers(
2048                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2049                         err = notifier_to_errno(err);
2050                         if (err) {
2051                                 dev->tx_queue_len = orig_len;
2052                                 goto errout;
2053                         }
2054                         status |= DO_SETLINK_NOTIFY;
2055                 }
2056         }
2057
2058         if (tb[IFLA_OPERSTATE])
2059                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2060
2061         if (tb[IFLA_LINKMODE]) {
2062                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2063
2064                 write_lock_bh(&dev_base_lock);
2065                 if (dev->link_mode ^ value)
2066                         status |= DO_SETLINK_NOTIFY;
2067                 dev->link_mode = value;
2068                 write_unlock_bh(&dev_base_lock);
2069         }
2070
2071         if (tb[IFLA_VFINFO_LIST]) {
2072                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2073                 struct nlattr *attr;
2074                 int rem;
2075
2076                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2077                         if (nla_type(attr) != IFLA_VF_INFO ||
2078                             nla_len(attr) < NLA_HDRLEN) {
2079                                 err = -EINVAL;
2080                                 goto errout;
2081                         }
2082                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2083                                                ifla_vf_policy, NULL);
2084                         if (err < 0)
2085                                 goto errout;
2086                         err = do_setvfinfo(dev, vfinfo);
2087                         if (err < 0)
2088                                 goto errout;
2089                         status |= DO_SETLINK_NOTIFY;
2090                 }
2091         }
2092         err = 0;
2093
2094         if (tb[IFLA_VF_PORTS]) {
2095                 struct nlattr *port[IFLA_PORT_MAX+1];
2096                 struct nlattr *attr;
2097                 int vf;
2098                 int rem;
2099
2100                 err = -EOPNOTSUPP;
2101                 if (!ops->ndo_set_vf_port)
2102                         goto errout;
2103
2104                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2105                         if (nla_type(attr) != IFLA_VF_PORT ||
2106                             nla_len(attr) < NLA_HDRLEN) {
2107                                 err = -EINVAL;
2108                                 goto errout;
2109                         }
2110                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2111                                                ifla_port_policy, NULL);
2112                         if (err < 0)
2113                                 goto errout;
2114                         if (!port[IFLA_PORT_VF]) {
2115                                 err = -EOPNOTSUPP;
2116                                 goto errout;
2117                         }
2118                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2119                         err = ops->ndo_set_vf_port(dev, vf, port);
2120                         if (err < 0)
2121                                 goto errout;
2122                         status |= DO_SETLINK_NOTIFY;
2123                 }
2124         }
2125         err = 0;
2126
2127         if (tb[IFLA_PORT_SELF]) {
2128                 struct nlattr *port[IFLA_PORT_MAX+1];
2129
2130                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2131                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2132                                        NULL);
2133                 if (err < 0)
2134                         goto errout;
2135
2136                 err = -EOPNOTSUPP;
2137                 if (ops->ndo_set_vf_port)
2138                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2139                 if (err < 0)
2140                         goto errout;
2141                 status |= DO_SETLINK_NOTIFY;
2142         }
2143
2144         if (tb[IFLA_AF_SPEC]) {
2145                 struct nlattr *af;
2146                 int rem;
2147
2148                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2149                         const struct rtnl_af_ops *af_ops;
2150
2151                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2152                                 BUG();
2153
2154                         err = af_ops->set_link_af(dev, af);
2155                         if (err < 0)
2156                                 goto errout;
2157
2158                         status |= DO_SETLINK_NOTIFY;
2159                 }
2160         }
2161         err = 0;
2162
2163         if (tb[IFLA_PROTO_DOWN]) {
2164                 err = dev_change_proto_down(dev,
2165                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2166                 if (err)
2167                         goto errout;
2168                 status |= DO_SETLINK_NOTIFY;
2169         }
2170
2171         if (tb[IFLA_XDP]) {
2172                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2173                 u32 xdp_flags = 0;
2174
2175                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2176                                        ifla_xdp_policy, NULL);
2177                 if (err < 0)
2178                         goto errout;
2179
2180                 if (xdp[IFLA_XDP_ATTACHED]) {
2181                         err = -EINVAL;
2182                         goto errout;
2183                 }
2184
2185                 if (xdp[IFLA_XDP_FLAGS]) {
2186                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2187                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2188                                 err = -EINVAL;
2189                                 goto errout;
2190                         }
2191                 }
2192
2193                 if (xdp[IFLA_XDP_FD]) {
2194                         err = dev_change_xdp_fd(dev,
2195                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2196                                                 xdp_flags);
2197                         if (err)
2198                                 goto errout;
2199                         status |= DO_SETLINK_NOTIFY;
2200                 }
2201         }
2202
2203 errout:
2204         if (status & DO_SETLINK_MODIFIED) {
2205                 if (status & DO_SETLINK_NOTIFY)
2206                         netdev_state_change(dev);
2207
2208                 if (err < 0)
2209                         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",
2210                                              dev->name);
2211         }
2212
2213         return err;
2214 }
2215
2216 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2217 {
2218         struct net *net = sock_net(skb->sk);
2219         struct ifinfomsg *ifm;
2220         struct net_device *dev;
2221         int err;
2222         struct nlattr *tb[IFLA_MAX+1];
2223         char ifname[IFNAMSIZ];
2224
2225         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, NULL);
2226         if (err < 0)
2227                 goto errout;
2228
2229         if (tb[IFLA_IFNAME])
2230                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2231         else
2232                 ifname[0] = '\0';
2233
2234         err = -EINVAL;
2235         ifm = nlmsg_data(nlh);
2236         if (ifm->ifi_index > 0)
2237                 dev = __dev_get_by_index(net, ifm->ifi_index);
2238         else if (tb[IFLA_IFNAME])
2239                 dev = __dev_get_by_name(net, ifname);
2240         else
2241                 goto errout;
2242
2243         if (dev == NULL) {
2244                 err = -ENODEV;
2245                 goto errout;
2246         }
2247
2248         err = validate_linkmsg(dev, tb);
2249         if (err < 0)
2250                 goto errout;
2251
2252         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2253 errout:
2254         return err;
2255 }
2256
2257 static int rtnl_group_dellink(const struct net *net, int group)
2258 {
2259         struct net_device *dev, *aux;
2260         LIST_HEAD(list_kill);
2261         bool found = false;
2262
2263         if (!group)
2264                 return -EPERM;
2265
2266         for_each_netdev(net, dev) {
2267                 if (dev->group == group) {
2268                         const struct rtnl_link_ops *ops;
2269
2270                         found = true;
2271                         ops = dev->rtnl_link_ops;
2272                         if (!ops || !ops->dellink)
2273                                 return -EOPNOTSUPP;
2274                 }
2275         }
2276
2277         if (!found)
2278                 return -ENODEV;
2279
2280         for_each_netdev_safe(net, dev, aux) {
2281                 if (dev->group == group) {
2282                         const struct rtnl_link_ops *ops;
2283
2284                         ops = dev->rtnl_link_ops;
2285                         ops->dellink(dev, &list_kill);
2286                 }
2287         }
2288         unregister_netdevice_many(&list_kill);
2289
2290         return 0;
2291 }
2292
2293 int rtnl_delete_link(struct net_device *dev)
2294 {
2295         const struct rtnl_link_ops *ops;
2296         LIST_HEAD(list_kill);
2297
2298         ops = dev->rtnl_link_ops;
2299         if (!ops || !ops->dellink)
2300                 return -EOPNOTSUPP;
2301
2302         ops->dellink(dev, &list_kill);
2303         unregister_netdevice_many(&list_kill);
2304
2305         return 0;
2306 }
2307 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2308
2309 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2310 {
2311         struct net *net = sock_net(skb->sk);
2312         struct net_device *dev;
2313         struct ifinfomsg *ifm;
2314         char ifname[IFNAMSIZ];
2315         struct nlattr *tb[IFLA_MAX+1];
2316         int err;
2317
2318         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, NULL);
2319         if (err < 0)
2320                 return err;
2321
2322         if (tb[IFLA_IFNAME])
2323                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2324
2325         ifm = nlmsg_data(nlh);
2326         if (ifm->ifi_index > 0)
2327                 dev = __dev_get_by_index(net, ifm->ifi_index);
2328         else if (tb[IFLA_IFNAME])
2329                 dev = __dev_get_by_name(net, ifname);
2330         else if (tb[IFLA_GROUP])
2331                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2332         else
2333                 return -EINVAL;
2334
2335         if (!dev)
2336                 return -ENODEV;
2337
2338         return rtnl_delete_link(dev);
2339 }
2340
2341 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2342 {
2343         unsigned int old_flags;
2344         int err;
2345
2346         old_flags = dev->flags;
2347         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2348                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2349                 if (err < 0)
2350                         return err;
2351         }
2352
2353         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2354
2355         __dev_notify_flags(dev, old_flags, ~0U);
2356         return 0;
2357 }
2358 EXPORT_SYMBOL(rtnl_configure_link);
2359
2360 struct net_device *rtnl_create_link(struct net *net,
2361         const char *ifname, unsigned char name_assign_type,
2362         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2363 {
2364         struct net_device *dev;
2365         unsigned int num_tx_queues = 1;
2366         unsigned int num_rx_queues = 1;
2367
2368         if (tb[IFLA_NUM_TX_QUEUES])
2369                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2370         else if (ops->get_num_tx_queues)
2371                 num_tx_queues = ops->get_num_tx_queues();
2372
2373         if (tb[IFLA_NUM_RX_QUEUES])
2374                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2375         else if (ops->get_num_rx_queues)
2376                 num_rx_queues = ops->get_num_rx_queues();
2377
2378         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2379                                ops->setup, num_tx_queues, num_rx_queues);
2380         if (!dev)
2381                 return ERR_PTR(-ENOMEM);
2382
2383         dev_net_set(dev, net);
2384         dev->rtnl_link_ops = ops;
2385         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2386
2387         if (tb[IFLA_MTU])
2388                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2389         if (tb[IFLA_ADDRESS]) {
2390                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2391                                 nla_len(tb[IFLA_ADDRESS]));
2392                 dev->addr_assign_type = NET_ADDR_SET;
2393         }
2394         if (tb[IFLA_BROADCAST])
2395                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2396                                 nla_len(tb[IFLA_BROADCAST]));
2397         if (tb[IFLA_TXQLEN])
2398                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2399         if (tb[IFLA_OPERSTATE])
2400                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2401         if (tb[IFLA_LINKMODE])
2402                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2403         if (tb[IFLA_GROUP])
2404                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2405
2406         return dev;
2407 }
2408 EXPORT_SYMBOL(rtnl_create_link);
2409
2410 static int rtnl_group_changelink(const struct sk_buff *skb,
2411                 struct net *net, int group,
2412                 struct ifinfomsg *ifm,
2413                 struct nlattr **tb)
2414 {
2415         struct net_device *dev, *aux;
2416         int err;
2417
2418         for_each_netdev_safe(net, dev, aux) {
2419                 if (dev->group == group) {
2420                         err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2421                         if (err < 0)
2422                                 return err;
2423                 }
2424         }
2425
2426         return 0;
2427 }
2428
2429 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2430 {
2431         struct net *net = sock_net(skb->sk);
2432         const struct rtnl_link_ops *ops;
2433         const struct rtnl_link_ops *m_ops = NULL;
2434         struct net_device *dev;
2435         struct net_device *master_dev = NULL;
2436         struct ifinfomsg *ifm;
2437         char kind[MODULE_NAME_LEN];
2438         char ifname[IFNAMSIZ];
2439         struct nlattr *tb[IFLA_MAX+1];
2440         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2441         unsigned char name_assign_type = NET_NAME_USER;
2442         int err;
2443
2444 #ifdef CONFIG_MODULES
2445 replay:
2446 #endif
2447         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, NULL);
2448         if (err < 0)
2449                 return err;
2450
2451         if (tb[IFLA_IFNAME])
2452                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2453         else
2454                 ifname[0] = '\0';
2455
2456         ifm = nlmsg_data(nlh);
2457         if (ifm->ifi_index > 0)
2458                 dev = __dev_get_by_index(net, ifm->ifi_index);
2459         else {
2460                 if (ifname[0])
2461                         dev = __dev_get_by_name(net, ifname);
2462                 else
2463                         dev = NULL;
2464         }
2465
2466         if (dev) {
2467                 master_dev = netdev_master_upper_dev_get(dev);
2468                 if (master_dev)
2469                         m_ops = master_dev->rtnl_link_ops;
2470         }
2471
2472         err = validate_linkmsg(dev, tb);
2473         if (err < 0)
2474                 return err;
2475
2476         if (tb[IFLA_LINKINFO]) {
2477                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2478                                        tb[IFLA_LINKINFO], ifla_info_policy,
2479                                        NULL);
2480                 if (err < 0)
2481                         return err;
2482         } else
2483                 memset(linkinfo, 0, sizeof(linkinfo));
2484
2485         if (linkinfo[IFLA_INFO_KIND]) {
2486                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2487                 ops = rtnl_link_ops_get(kind);
2488         } else {
2489                 kind[0] = '\0';
2490                 ops = NULL;
2491         }
2492
2493         if (1) {
2494                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2495                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2496                 struct nlattr **data = NULL;
2497                 struct nlattr **slave_data = NULL;
2498                 struct net *dest_net, *link_net = NULL;
2499
2500                 if (ops) {
2501                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2502                                 err = nla_parse_nested(attr, ops->maxtype,
2503                                                        linkinfo[IFLA_INFO_DATA],
2504                                                        ops->policy, NULL);
2505                                 if (err < 0)
2506                                         return err;
2507                                 data = attr;
2508                         }
2509                         if (ops->validate) {
2510                                 err = ops->validate(tb, data);
2511                                 if (err < 0)
2512                                         return err;
2513                         }
2514                 }
2515
2516                 if (m_ops) {
2517                         if (m_ops->slave_maxtype &&
2518                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2519                                 err = nla_parse_nested(slave_attr,
2520                                                        m_ops->slave_maxtype,
2521                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2522                                                        m_ops->slave_policy,
2523                                                        NULL);
2524                                 if (err < 0)
2525                                         return err;
2526                                 slave_data = slave_attr;
2527                         }
2528                         if (m_ops->slave_validate) {
2529                                 err = m_ops->slave_validate(tb, slave_data);
2530                                 if (err < 0)
2531                                         return err;
2532                         }
2533                 }
2534
2535                 if (dev) {
2536                         int status = 0;
2537
2538                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2539                                 return -EEXIST;
2540                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2541                                 return -EOPNOTSUPP;
2542
2543                         if (linkinfo[IFLA_INFO_DATA]) {
2544                                 if (!ops || ops != dev->rtnl_link_ops ||
2545                                     !ops->changelink)
2546                                         return -EOPNOTSUPP;
2547
2548                                 err = ops->changelink(dev, tb, data);
2549                                 if (err < 0)
2550                                         return err;
2551                                 status |= DO_SETLINK_NOTIFY;
2552                         }
2553
2554                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2555                                 if (!m_ops || !m_ops->slave_changelink)
2556                                         return -EOPNOTSUPP;
2557
2558                                 err = m_ops->slave_changelink(master_dev, dev,
2559                                                               tb, slave_data);
2560                                 if (err < 0)
2561                                         return err;
2562                                 status |= DO_SETLINK_NOTIFY;
2563                         }
2564
2565                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2566                 }
2567
2568                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2569                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2570                                 return rtnl_group_changelink(skb, net,
2571                                                 nla_get_u32(tb[IFLA_GROUP]),
2572                                                 ifm, tb);
2573                         return -ENODEV;
2574                 }
2575
2576                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2577                         return -EOPNOTSUPP;
2578
2579                 if (!ops) {
2580 #ifdef CONFIG_MODULES
2581                         if (kind[0]) {
2582                                 __rtnl_unlock();
2583                                 request_module("rtnl-link-%s", kind);
2584                                 rtnl_lock();
2585                                 ops = rtnl_link_ops_get(kind);
2586                                 if (ops)
2587                                         goto replay;
2588                         }
2589 #endif
2590                         return -EOPNOTSUPP;
2591                 }
2592
2593                 if (!ops->setup)
2594                         return -EOPNOTSUPP;
2595
2596                 if (!ifname[0]) {
2597                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2598                         name_assign_type = NET_NAME_ENUM;
2599                 }
2600
2601                 dest_net = rtnl_link_get_net(net, tb);
2602                 if (IS_ERR(dest_net))
2603                         return PTR_ERR(dest_net);
2604
2605                 err = -EPERM;
2606                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2607                         goto out;
2608
2609                 if (tb[IFLA_LINK_NETNSID]) {
2610                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2611
2612                         link_net = get_net_ns_by_id(dest_net, id);
2613                         if (!link_net) {
2614                                 err =  -EINVAL;
2615                                 goto out;
2616                         }
2617                         err = -EPERM;
2618                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2619                                 goto out;
2620                 }
2621
2622                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2623                                        name_assign_type, ops, tb);
2624                 if (IS_ERR(dev)) {
2625                         err = PTR_ERR(dev);
2626                         goto out;
2627                 }
2628
2629                 dev->ifindex = ifm->ifi_index;
2630
2631                 if (ops->newlink) {
2632                         err = ops->newlink(link_net ? : net, dev, tb, data);
2633                         /* Drivers should call free_netdev() in ->destructor
2634                          * and unregister it on failure after registration
2635                          * so that device could be finally freed in rtnl_unlock.
2636                          */
2637                         if (err < 0) {
2638                                 /* If device is not registered at all, free it now */
2639                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2640                                         free_netdev(dev);
2641                                 goto out;
2642                         }
2643                 } else {
2644                         err = register_netdevice(dev);
2645                         if (err < 0) {
2646                                 free_netdev(dev);
2647                                 goto out;
2648                         }
2649                 }
2650                 err = rtnl_configure_link(dev, ifm);
2651                 if (err < 0)
2652                         goto out_unregister;
2653                 if (link_net) {
2654                         err = dev_change_net_namespace(dev, dest_net, ifname);
2655                         if (err < 0)
2656                                 goto out_unregister;
2657                 }
2658                 if (tb[IFLA_MASTER]) {
2659                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2660                         if (err)
2661                                 goto out_unregister;
2662                 }
2663 out:
2664                 if (link_net)
2665                         put_net(link_net);
2666                 put_net(dest_net);
2667                 return err;
2668 out_unregister:
2669                 if (ops->newlink) {
2670                         LIST_HEAD(list_kill);
2671
2672                         ops->dellink(dev, &list_kill);
2673                         unregister_netdevice_many(&list_kill);
2674                 } else {
2675                         unregister_netdevice(dev);
2676                 }
2677                 goto out;
2678         }
2679 }
2680
2681 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2682 {
2683         struct net *net = sock_net(skb->sk);
2684         struct ifinfomsg *ifm;
2685         char ifname[IFNAMSIZ];
2686         struct nlattr *tb[IFLA_MAX+1];
2687         struct net_device *dev = NULL;
2688         struct sk_buff *nskb;
2689         int err;
2690         u32 ext_filter_mask = 0;
2691
2692         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, NULL);
2693         if (err < 0)
2694                 return err;
2695
2696         if (tb[IFLA_IFNAME])
2697                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2698
2699         if (tb[IFLA_EXT_MASK])
2700                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2701
2702         ifm = nlmsg_data(nlh);
2703         if (ifm->ifi_index > 0)
2704                 dev = __dev_get_by_index(net, ifm->ifi_index);
2705         else if (tb[IFLA_IFNAME])
2706                 dev = __dev_get_by_name(net, ifname);
2707         else
2708                 return -EINVAL;
2709
2710         if (dev == NULL)
2711                 return -ENODEV;
2712
2713         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2714         if (nskb == NULL)
2715                 return -ENOBUFS;
2716
2717         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2718                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2719         if (err < 0) {
2720                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2721                 WARN_ON(err == -EMSGSIZE);
2722                 kfree_skb(nskb);
2723         } else
2724                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2725
2726         return err;
2727 }
2728
2729 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2730 {
2731         struct net *net = sock_net(skb->sk);
2732         struct net_device *dev;
2733         struct nlattr *tb[IFLA_MAX+1];
2734         u32 ext_filter_mask = 0;
2735         u16 min_ifinfo_dump_size = 0;
2736         int hdrlen;
2737
2738         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2739         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2740                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2741
2742         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2743                 if (tb[IFLA_EXT_MASK])
2744                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2745         }
2746
2747         if (!ext_filter_mask)
2748                 return NLMSG_GOODSIZE;
2749         /*
2750          * traverse the list of net devices and compute the minimum
2751          * buffer size based upon the filter mask.
2752          */
2753         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2754                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2755                                              if_nlmsg_size(dev,
2756                                                            ext_filter_mask));
2757         }
2758
2759         return nlmsg_total_size(min_ifinfo_dump_size);
2760 }
2761
2762 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2763 {
2764         int idx;
2765         int s_idx = cb->family;
2766
2767         if (s_idx == 0)
2768                 s_idx = 1;
2769         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2770                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2771                 if (idx < s_idx || idx == PF_PACKET)
2772                         continue;
2773                 if (rtnl_msg_handlers[idx] == NULL ||
2774                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2775                         continue;
2776                 if (idx > s_idx) {
2777                         memset(&cb->args[0], 0, sizeof(cb->args));
2778                         cb->prev_seq = 0;
2779                         cb->seq = 0;
2780                 }
2781                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2782                         break;
2783         }
2784         cb->family = idx;
2785
2786         return skb->len;
2787 }
2788
2789 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2790                                        unsigned int change, gfp_t flags)
2791 {
2792         struct net *net = dev_net(dev);
2793         struct sk_buff *skb;
2794         int err = -ENOBUFS;
2795         size_t if_info_size;
2796
2797         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2798         if (skb == NULL)
2799                 goto errout;
2800
2801         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2802         if (err < 0) {
2803                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2804                 WARN_ON(err == -EMSGSIZE);
2805                 kfree_skb(skb);
2806                 goto errout;
2807         }
2808         return skb;
2809 errout:
2810         if (err < 0)
2811                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2812         return NULL;
2813 }
2814
2815 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2816 {
2817         struct net *net = dev_net(dev);
2818
2819         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2820 }
2821
2822 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2823                   gfp_t flags)
2824 {
2825         struct sk_buff *skb;
2826
2827         if (dev->reg_state != NETREG_REGISTERED)
2828                 return;
2829
2830         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2831         if (skb)
2832                 rtmsg_ifinfo_send(skb, dev, flags);
2833 }
2834 EXPORT_SYMBOL(rtmsg_ifinfo);
2835
2836 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2837                                    struct net_device *dev,
2838                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2839                                    int type, unsigned int flags,
2840                                    int nlflags, u16 ndm_state)
2841 {
2842         struct nlmsghdr *nlh;
2843         struct ndmsg *ndm;
2844
2845         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2846         if (!nlh)
2847                 return -EMSGSIZE;
2848
2849         ndm = nlmsg_data(nlh);
2850         ndm->ndm_family  = AF_BRIDGE;
2851         ndm->ndm_pad1    = 0;
2852         ndm->ndm_pad2    = 0;
2853         ndm->ndm_flags   = flags;
2854         ndm->ndm_type    = 0;
2855         ndm->ndm_ifindex = dev->ifindex;
2856         ndm->ndm_state   = ndm_state;
2857
2858         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2859                 goto nla_put_failure;
2860         if (vid)
2861                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2862                         goto nla_put_failure;
2863
2864         nlmsg_end(skb, nlh);
2865         return 0;
2866
2867 nla_put_failure:
2868         nlmsg_cancel(skb, nlh);
2869         return -EMSGSIZE;
2870 }
2871
2872 static inline size_t rtnl_fdb_nlmsg_size(void)
2873 {
2874         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2875                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2876                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2877                0;
2878 }
2879
2880 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2881                             u16 ndm_state)
2882 {
2883         struct net *net = dev_net(dev);
2884         struct sk_buff *skb;
2885         int err = -ENOBUFS;
2886
2887         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2888         if (!skb)
2889                 goto errout;
2890
2891         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2892                                       0, 0, type, NTF_SELF, 0, ndm_state);
2893         if (err < 0) {
2894                 kfree_skb(skb);
2895                 goto errout;
2896         }
2897
2898         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2899         return;
2900 errout:
2901         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2902 }
2903
2904 /**
2905  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2906  */
2907 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2908                      struct nlattr *tb[],
2909                      struct net_device *dev,
2910                      const unsigned char *addr, u16 vid,
2911                      u16 flags)
2912 {
2913         int err = -EINVAL;
2914
2915         /* If aging addresses are supported device will need to
2916          * implement its own handler for this.
2917          */
2918         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2919                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2920                 return err;
2921         }
2922
2923         if (vid) {
2924                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2925                 return err;
2926         }
2927
2928         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2929                 err = dev_uc_add_excl(dev, addr);
2930         else if (is_multicast_ether_addr(addr))
2931                 err = dev_mc_add_excl(dev, addr);
2932
2933         /* Only return duplicate errors if NLM_F_EXCL is set */
2934         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2935                 err = 0;
2936
2937         return err;
2938 }
2939 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2940
2941 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2942 {
2943         u16 vid = 0;
2944
2945         if (vlan_attr) {
2946                 if (nla_len(vlan_attr) != sizeof(u16)) {
2947                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2948                         return -EINVAL;
2949                 }
2950
2951                 vid = nla_get_u16(vlan_attr);
2952
2953                 if (!vid || vid >= VLAN_VID_MASK) {
2954                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2955                                 vid);
2956                         return -EINVAL;
2957                 }
2958         }
2959         *p_vid = vid;
2960         return 0;
2961 }
2962
2963 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2964 {
2965         struct net *net = sock_net(skb->sk);
2966         struct ndmsg *ndm;
2967         struct nlattr *tb[NDA_MAX+1];
2968         struct net_device *dev;
2969         u8 *addr;
2970         u16 vid;
2971         int err;
2972
2973         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, NULL);
2974         if (err < 0)
2975                 return err;
2976
2977         ndm = nlmsg_data(nlh);
2978         if (ndm->ndm_ifindex == 0) {
2979                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2980                 return -EINVAL;
2981         }
2982
2983         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2984         if (dev == NULL) {
2985                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2986                 return -ENODEV;
2987         }
2988
2989         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2990                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2991                 return -EINVAL;
2992         }
2993
2994         addr = nla_data(tb[NDA_LLADDR]);
2995
2996         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2997         if (err)
2998                 return err;
2999
3000         err = -EOPNOTSUPP;
3001
3002         /* Support fdb on master device the net/bridge default case */
3003         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3004             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3005                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3006                 const struct net_device_ops *ops = br_dev->netdev_ops;
3007
3008                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3009                                        nlh->nlmsg_flags);
3010                 if (err)
3011                         goto out;
3012                 else
3013                         ndm->ndm_flags &= ~NTF_MASTER;
3014         }
3015
3016         /* Embedded bridge, macvlan, and any other device support */
3017         if ((ndm->ndm_flags & NTF_SELF)) {
3018                 if (dev->netdev_ops->ndo_fdb_add)
3019                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3020                                                            vid,
3021                                                            nlh->nlmsg_flags);
3022                 else
3023                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3024                                                nlh->nlmsg_flags);
3025
3026                 if (!err) {
3027                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3028                                         ndm->ndm_state);
3029                         ndm->ndm_flags &= ~NTF_SELF;
3030                 }
3031         }
3032 out:
3033         return err;
3034 }
3035
3036 /**
3037  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3038  */
3039 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3040                      struct nlattr *tb[],
3041                      struct net_device *dev,
3042                      const unsigned char *addr, u16 vid)
3043 {
3044         int err = -EINVAL;
3045
3046         /* If aging addresses are supported device will need to
3047          * implement its own handler for this.
3048          */
3049         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3050                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3051                 return err;
3052         }
3053
3054         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3055                 err = dev_uc_del(dev, addr);
3056         else if (is_multicast_ether_addr(addr))
3057                 err = dev_mc_del(dev, addr);
3058
3059         return err;
3060 }
3061 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3062
3063 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3064 {
3065         struct net *net = sock_net(skb->sk);
3066         struct ndmsg *ndm;
3067         struct nlattr *tb[NDA_MAX+1];
3068         struct net_device *dev;
3069         int err = -EINVAL;
3070         __u8 *addr;
3071         u16 vid;
3072
3073         if (!netlink_capable(skb, CAP_NET_ADMIN))
3074                 return -EPERM;
3075
3076         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, NULL);
3077         if (err < 0)
3078                 return err;
3079
3080         ndm = nlmsg_data(nlh);
3081         if (ndm->ndm_ifindex == 0) {
3082                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3083                 return -EINVAL;
3084         }
3085
3086         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3087         if (dev == NULL) {
3088                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3089                 return -ENODEV;
3090         }
3091
3092         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3093                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3094                 return -EINVAL;
3095         }
3096
3097         addr = nla_data(tb[NDA_LLADDR]);
3098
3099         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3100         if (err)
3101                 return err;
3102
3103         err = -EOPNOTSUPP;
3104
3105         /* Support fdb on master device the net/bridge default case */
3106         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3107             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3108                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3109                 const struct net_device_ops *ops = br_dev->netdev_ops;
3110
3111                 if (ops->ndo_fdb_del)
3112                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3113
3114                 if (err)
3115                         goto out;
3116                 else
3117                         ndm->ndm_flags &= ~NTF_MASTER;
3118         }
3119
3120         /* Embedded bridge, macvlan, and any other device support */
3121         if (ndm->ndm_flags & NTF_SELF) {
3122                 if (dev->netdev_ops->ndo_fdb_del)
3123                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3124                                                            vid);
3125                 else
3126                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3127
3128                 if (!err) {
3129                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3130                                         ndm->ndm_state);
3131                         ndm->ndm_flags &= ~NTF_SELF;
3132                 }
3133         }
3134 out:
3135         return err;
3136 }
3137
3138 static int nlmsg_populate_fdb(struct sk_buff *skb,
3139                               struct netlink_callback *cb,
3140                               struct net_device *dev,
3141                               int *idx,
3142                               struct netdev_hw_addr_list *list)
3143 {
3144         struct netdev_hw_addr *ha;
3145         int err;
3146         u32 portid, seq;
3147
3148         portid = NETLINK_CB(cb->skb).portid;
3149         seq = cb->nlh->nlmsg_seq;
3150
3151         list_for_each_entry(ha, &list->list, list) {
3152                 if (*idx < cb->args[2])
3153                         goto skip;
3154
3155                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3156                                               portid, seq,
3157                                               RTM_NEWNEIGH, NTF_SELF,
3158                                               NLM_F_MULTI, NUD_PERMANENT);
3159                 if (err < 0)
3160                         return err;
3161 skip:
3162                 *idx += 1;
3163         }
3164         return 0;
3165 }
3166
3167 /**
3168  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3169  * @nlh: netlink message header
3170  * @dev: netdevice
3171  *
3172  * Default netdevice operation to dump the existing unicast address list.
3173  * Returns number of addresses from list put in skb.
3174  */
3175 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3176                       struct netlink_callback *cb,
3177                       struct net_device *dev,
3178                       struct net_device *filter_dev,
3179                       int *idx)
3180 {
3181         int err;
3182
3183         netif_addr_lock_bh(dev);
3184         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3185         if (err)
3186                 goto out;
3187         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3188 out:
3189         netif_addr_unlock_bh(dev);
3190         return err;
3191 }
3192 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3193
3194 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3195 {
3196         struct net_device *dev;
3197         struct nlattr *tb[IFLA_MAX+1];
3198         struct net_device *br_dev = NULL;
3199         const struct net_device_ops *ops = NULL;
3200         const struct net_device_ops *cops = NULL;
3201         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3202         struct net *net = sock_net(skb->sk);
3203         struct hlist_head *head;
3204         int brport_idx = 0;
3205         int br_idx = 0;
3206         int h, s_h;
3207         int idx = 0, s_idx;
3208         int err = 0;
3209         int fidx = 0;
3210
3211         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3212                         IFLA_MAX, ifla_policy, NULL) == 0) {
3213                 if (tb[IFLA_MASTER])
3214                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3215         }
3216
3217         brport_idx = ifm->ifi_index;
3218
3219         if (br_idx) {
3220                 br_dev = __dev_get_by_index(net, br_idx);
3221                 if (!br_dev)
3222                         return -ENODEV;
3223
3224                 ops = br_dev->netdev_ops;
3225         }
3226
3227         s_h = cb->args[0];
3228         s_idx = cb->args[1];
3229
3230         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3231                 idx = 0;
3232                 head = &net->dev_index_head[h];
3233                 hlist_for_each_entry(dev, head, index_hlist) {
3234
3235                         if (brport_idx && (dev->ifindex != brport_idx))
3236                                 continue;
3237
3238                         if (!br_idx) { /* user did not specify a specific bridge */
3239                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3240                                         br_dev = netdev_master_upper_dev_get(dev);
3241                                         cops = br_dev->netdev_ops;
3242                                 }
3243                         } else {
3244                                 if (dev != br_dev &&
3245                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3246                                         continue;
3247
3248                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3249                                     !(dev->priv_flags & IFF_EBRIDGE))
3250                                         continue;
3251                                 cops = ops;
3252                         }
3253
3254                         if (idx < s_idx)
3255                                 goto cont;
3256
3257                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3258                                 if (cops && cops->ndo_fdb_dump) {
3259                                         err = cops->ndo_fdb_dump(skb, cb,
3260                                                                 br_dev, dev,
3261                                                                 &fidx);
3262                                         if (err == -EMSGSIZE)
3263                                                 goto out;
3264                                 }
3265                         }
3266
3267                         if (dev->netdev_ops->ndo_fdb_dump)
3268                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3269                                                                     dev, NULL,
3270                                                                     &fidx);
3271                         else
3272                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3273                                                         &fidx);
3274                         if (err == -EMSGSIZE)
3275                                 goto out;
3276
3277                         cops = NULL;
3278
3279                         /* reset fdb offset to 0 for rest of the interfaces */
3280                         cb->args[2] = 0;
3281                         fidx = 0;
3282 cont:
3283                         idx++;
3284                 }
3285         }
3286
3287 out:
3288         cb->args[0] = h;
3289         cb->args[1] = idx;
3290         cb->args[2] = fidx;
3291
3292         return skb->len;
3293 }
3294
3295 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3296                                unsigned int attrnum, unsigned int flag)
3297 {
3298         if (mask & flag)
3299                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3300         return 0;
3301 }
3302
3303 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3304                             struct net_device *dev, u16 mode,
3305                             u32 flags, u32 mask, int nlflags,
3306                             u32 filter_mask,
3307                             int (*vlan_fill)(struct sk_buff *skb,
3308                                              struct net_device *dev,
3309                                              u32 filter_mask))
3310 {
3311         struct nlmsghdr *nlh;
3312         struct ifinfomsg *ifm;
3313         struct nlattr *br_afspec;
3314         struct nlattr *protinfo;
3315         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3316         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3317         int err = 0;
3318
3319         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3320         if (nlh == NULL)
3321                 return -EMSGSIZE;
3322
3323         ifm = nlmsg_data(nlh);
3324         ifm->ifi_family = AF_BRIDGE;
3325         ifm->__ifi_pad = 0;
3326         ifm->ifi_type = dev->type;
3327         ifm->ifi_index = dev->ifindex;
3328         ifm->ifi_flags = dev_get_flags(dev);
3329         ifm->ifi_change = 0;
3330
3331
3332         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3333             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3334             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3335             (br_dev &&
3336              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3337             (dev->addr_len &&
3338              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3339             (dev->ifindex != dev_get_iflink(dev) &&
3340              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3341                 goto nla_put_failure;
3342
3343         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3344         if (!br_afspec)
3345                 goto nla_put_failure;
3346
3347         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3348                 nla_nest_cancel(skb, br_afspec);
3349                 goto nla_put_failure;
3350         }
3351
3352         if (mode != BRIDGE_MODE_UNDEF) {
3353                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3354                         nla_nest_cancel(skb, br_afspec);
3355                         goto nla_put_failure;
3356                 }
3357         }
3358         if (vlan_fill) {
3359                 err = vlan_fill(skb, dev, filter_mask);
3360                 if (err) {
3361                         nla_nest_cancel(skb, br_afspec);
3362                         goto nla_put_failure;
3363                 }
3364         }
3365         nla_nest_end(skb, br_afspec);
3366
3367         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3368         if (!protinfo)
3369                 goto nla_put_failure;
3370
3371         if (brport_nla_put_flag(skb, flags, mask,
3372                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3373             brport_nla_put_flag(skb, flags, mask,
3374                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3375             brport_nla_put_flag(skb, flags, mask,
3376                                 IFLA_BRPORT_FAST_LEAVE,
3377                                 BR_MULTICAST_FAST_LEAVE) ||
3378             brport_nla_put_flag(skb, flags, mask,
3379                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3380             brport_nla_put_flag(skb, flags, mask,
3381                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3382             brport_nla_put_flag(skb, flags, mask,
3383                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3384             brport_nla_put_flag(skb, flags, mask,
3385                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3386             brport_nla_put_flag(skb, flags, mask,
3387                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3388                 nla_nest_cancel(skb, protinfo);
3389                 goto nla_put_failure;
3390         }
3391
3392         nla_nest_end(skb, protinfo);
3393
3394         nlmsg_end(skb, nlh);
3395         return 0;
3396 nla_put_failure:
3397         nlmsg_cancel(skb, nlh);
3398         return err ? err : -EMSGSIZE;
3399 }
3400 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3401
3402 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3403 {
3404         struct net *net = sock_net(skb->sk);
3405         struct net_device *dev;
3406         int idx = 0;
3407         u32 portid = NETLINK_CB(cb->skb).portid;
3408         u32 seq = cb->nlh->nlmsg_seq;
3409         u32 filter_mask = 0;
3410         int err;
3411
3412         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3413                 struct nlattr *extfilt;
3414
3415                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3416                                           IFLA_EXT_MASK);
3417                 if (extfilt) {
3418                         if (nla_len(extfilt) < sizeof(filter_mask))
3419                                 return -EINVAL;
3420
3421                         filter_mask = nla_get_u32(extfilt);
3422                 }
3423         }
3424
3425         rcu_read_lock();
3426         for_each_netdev_rcu(net, dev) {
3427                 const struct net_device_ops *ops = dev->netdev_ops;
3428                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3429
3430                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3431                         if (idx >= cb->args[0]) {
3432                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3433                                                 skb, portid, seq, dev,
3434                                                 filter_mask, NLM_F_MULTI);
3435                                 if (err < 0 && err != -EOPNOTSUPP)
3436                                         break;
3437                         }
3438                         idx++;
3439                 }
3440
3441                 if (ops->ndo_bridge_getlink) {
3442                         if (idx >= cb->args[0]) {
3443                                 err = ops->ndo_bridge_getlink(skb, portid,
3444                                                               seq, dev,
3445                                                               filter_mask,
3446                                                               NLM_F_MULTI);
3447                                 if (err < 0 && err != -EOPNOTSUPP)
3448                                         break;
3449                         }
3450                         idx++;
3451                 }
3452         }
3453         rcu_read_unlock();
3454         cb->args[0] = idx;
3455
3456         return skb->len;
3457 }
3458
3459 static inline size_t bridge_nlmsg_size(void)
3460 {
3461         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3462                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3463                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3464                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3465                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3466                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3467                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3468                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3469                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3470                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3471                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3472 }
3473
3474 static int rtnl_bridge_notify(struct net_device *dev)
3475 {
3476         struct net *net = dev_net(dev);
3477         struct sk_buff *skb;
3478         int err = -EOPNOTSUPP;
3479
3480         if (!dev->netdev_ops->ndo_bridge_getlink)
3481                 return 0;
3482
3483         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3484         if (!skb) {
3485                 err = -ENOMEM;
3486                 goto errout;
3487         }
3488
3489         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3490         if (err < 0)
3491                 goto errout;
3492
3493         if (!skb->len)
3494                 goto errout;
3495
3496         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3497         return 0;
3498 errout:
3499         WARN_ON(err == -EMSGSIZE);
3500         kfree_skb(skb);
3501         if (err)
3502                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3503         return err;
3504 }
3505
3506 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3507 {
3508         struct net *net = sock_net(skb->sk);
3509         struct ifinfomsg *ifm;
3510         struct net_device *dev;
3511         struct nlattr *br_spec, *attr = NULL;
3512         int rem, err = -EOPNOTSUPP;
3513         u16 flags = 0;
3514         bool have_flags = false;
3515
3516         if (nlmsg_len(nlh) < sizeof(*ifm))
3517                 return -EINVAL;
3518
3519         ifm = nlmsg_data(nlh);
3520         if (ifm->ifi_family != AF_BRIDGE)
3521                 return -EPFNOSUPPORT;
3522
3523         dev = __dev_get_by_index(net, ifm->ifi_index);
3524         if (!dev) {
3525                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3526                 return -ENODEV;
3527         }
3528
3529         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3530         if (br_spec) {
3531                 nla_for_each_nested(attr, br_spec, rem) {
3532                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3533                                 if (nla_len(attr) < sizeof(flags))
3534                                         return -EINVAL;
3535
3536                                 have_flags = true;
3537                                 flags = nla_get_u16(attr);
3538                                 break;
3539                         }
3540                 }
3541         }
3542
3543         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3544                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3545
3546                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3547                         err = -EOPNOTSUPP;
3548                         goto out;
3549                 }
3550
3551                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3552                 if (err)
3553                         goto out;
3554
3555                 flags &= ~BRIDGE_FLAGS_MASTER;
3556         }
3557
3558         if ((flags & BRIDGE_FLAGS_SELF)) {
3559                 if (!dev->netdev_ops->ndo_bridge_setlink)
3560                         err = -EOPNOTSUPP;
3561                 else
3562                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3563                                                                   flags);
3564                 if (!err) {
3565                         flags &= ~BRIDGE_FLAGS_SELF;
3566
3567                         /* Generate event to notify upper layer of bridge
3568                          * change
3569                          */
3570                         err = rtnl_bridge_notify(dev);
3571                 }
3572         }
3573
3574         if (have_flags)
3575                 memcpy(nla_data(attr), &flags, sizeof(flags));
3576 out:
3577         return err;
3578 }
3579
3580 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3581 {
3582         struct net *net = sock_net(skb->sk);
3583         struct ifinfomsg *ifm;
3584         struct net_device *dev;
3585         struct nlattr *br_spec, *attr = NULL;
3586         int rem, err = -EOPNOTSUPP;
3587         u16 flags = 0;
3588         bool have_flags = false;
3589
3590         if (nlmsg_len(nlh) < sizeof(*ifm))
3591                 return -EINVAL;
3592
3593         ifm = nlmsg_data(nlh);
3594         if (ifm->ifi_family != AF_BRIDGE)
3595                 return -EPFNOSUPPORT;
3596
3597         dev = __dev_get_by_index(net, ifm->ifi_index);
3598         if (!dev) {
3599                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3600                 return -ENODEV;
3601         }
3602
3603         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3604         if (br_spec) {
3605                 nla_for_each_nested(attr, br_spec, rem) {
3606                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3607                                 if (nla_len(attr) < sizeof(flags))
3608                                         return -EINVAL;
3609
3610                                 have_flags = true;
3611                                 flags = nla_get_u16(attr);
3612                                 break;
3613                         }
3614                 }
3615         }
3616
3617         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3618                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3619
3620                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3621                         err = -EOPNOTSUPP;
3622                         goto out;
3623                 }
3624
3625                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3626                 if (err)
3627                         goto out;
3628
3629                 flags &= ~BRIDGE_FLAGS_MASTER;
3630         }
3631
3632         if ((flags & BRIDGE_FLAGS_SELF)) {
3633                 if (!dev->netdev_ops->ndo_bridge_dellink)
3634                         err = -EOPNOTSUPP;
3635                 else
3636                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3637                                                                   flags);
3638
3639                 if (!err) {
3640                         flags &= ~BRIDGE_FLAGS_SELF;
3641
3642                         /* Generate event to notify upper layer of bridge
3643                          * change
3644                          */
3645                         err = rtnl_bridge_notify(dev);
3646                 }
3647         }
3648
3649         if (have_flags)
3650                 memcpy(nla_data(attr), &flags, sizeof(flags));
3651 out:
3652         return err;
3653 }
3654
3655 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3656 {
3657         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3658                (!idxattr || idxattr == attrid);
3659 }
3660
3661 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3662 static int rtnl_get_offload_stats_attr_size(int attr_id)
3663 {
3664         switch (attr_id) {
3665         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3666                 return sizeof(struct rtnl_link_stats64);
3667         }
3668
3669         return 0;
3670 }
3671
3672 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3673                                   int *prividx)
3674 {
3675         struct nlattr *attr = NULL;
3676         int attr_id, size;
3677         void *attr_data;
3678         int err;
3679
3680         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3681               dev->netdev_ops->ndo_get_offload_stats))
3682                 return -ENODATA;
3683
3684         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3685              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3686                 if (attr_id < *prividx)
3687                         continue;
3688
3689                 size = rtnl_get_offload_stats_attr_size(attr_id);
3690                 if (!size)
3691                         continue;
3692
3693                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3694                         continue;
3695
3696                 attr = nla_reserve_64bit(skb, attr_id, size,
3697                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3698                 if (!attr)
3699                         goto nla_put_failure;
3700
3701                 attr_data = nla_data(attr);
3702                 memset(attr_data, 0, size);
3703                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3704                                                              attr_data);
3705                 if (err)
3706                         goto get_offload_stats_failure;
3707         }
3708
3709         if (!attr)
3710                 return -ENODATA;
3711
3712         *prividx = 0;
3713         return 0;
3714
3715 nla_put_failure:
3716         err = -EMSGSIZE;
3717 get_offload_stats_failure:
3718         *prividx = attr_id;
3719         return err;
3720 }
3721
3722 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3723 {
3724         int nla_size = 0;
3725         int attr_id;
3726         int size;
3727
3728         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3729               dev->netdev_ops->ndo_get_offload_stats))
3730                 return 0;
3731
3732         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3733              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3734                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3735                         continue;
3736                 size = rtnl_get_offload_stats_attr_size(attr_id);
3737                 nla_size += nla_total_size_64bit(size);
3738         }
3739
3740         if (nla_size != 0)
3741                 nla_size += nla_total_size(0);
3742
3743         return nla_size;
3744 }
3745
3746 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3747                                int type, u32 pid, u32 seq, u32 change,
3748                                unsigned int flags, unsigned int filter_mask,
3749                                int *idxattr, int *prividx)
3750 {
3751         struct if_stats_msg *ifsm;
3752         struct nlmsghdr *nlh;
3753         struct nlattr *attr;
3754         int s_prividx = *prividx;
3755         int err;
3756
3757         ASSERT_RTNL();
3758
3759         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3760         if (!nlh)
3761                 return -EMSGSIZE;
3762
3763         ifsm = nlmsg_data(nlh);
3764         ifsm->ifindex = dev->ifindex;
3765         ifsm->filter_mask = filter_mask;
3766
3767         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3768                 struct rtnl_link_stats64 *sp;
3769
3770                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3771                                          sizeof(struct rtnl_link_stats64),
3772                                          IFLA_STATS_UNSPEC);
3773                 if (!attr)
3774                         goto nla_put_failure;
3775
3776                 sp = nla_data(attr);
3777                 dev_get_stats(dev, sp);
3778         }
3779
3780         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3781                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3782
3783                 if (ops && ops->fill_linkxstats) {
3784                         *idxattr = IFLA_STATS_LINK_XSTATS;
3785                         attr = nla_nest_start(skb,
3786                                               IFLA_STATS_LINK_XSTATS);
3787                         if (!attr)
3788                                 goto nla_put_failure;
3789
3790                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3791                         nla_nest_end(skb, attr);
3792                         if (err)
3793                                 goto nla_put_failure;
3794                         *idxattr = 0;
3795                 }
3796         }
3797
3798         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3799                              *idxattr)) {
3800                 const struct rtnl_link_ops *ops = NULL;
3801                 const struct net_device *master;
3802
3803                 master = netdev_master_upper_dev_get(dev);
3804                 if (master)
3805                         ops = master->rtnl_link_ops;
3806                 if (ops && ops->fill_linkxstats) {
3807                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3808                         attr = nla_nest_start(skb,
3809                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3810                         if (!attr)
3811                                 goto nla_put_failure;
3812
3813                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3814                         nla_nest_end(skb, attr);
3815                         if (err)
3816                                 goto nla_put_failure;
3817                         *idxattr = 0;
3818                 }
3819         }
3820
3821         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3822                              *idxattr)) {
3823                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3824                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3825                 if (!attr)
3826                         goto nla_put_failure;
3827
3828                 err = rtnl_get_offload_stats(skb, dev, prividx);
3829                 if (err == -ENODATA)
3830                         nla_nest_cancel(skb, attr);
3831                 else
3832                         nla_nest_end(skb, attr);
3833
3834                 if (err && err != -ENODATA)
3835                         goto nla_put_failure;
3836                 *idxattr = 0;
3837         }
3838
3839         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3840                 struct rtnl_af_ops *af_ops;
3841
3842                 *idxattr = IFLA_STATS_AF_SPEC;
3843                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3844                 if (!attr)
3845                         goto nla_put_failure;
3846
3847                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3848                         if (af_ops->fill_stats_af) {
3849                                 struct nlattr *af;
3850                                 int err;
3851
3852                                 af = nla_nest_start(skb, af_ops->family);
3853                                 if (!af)
3854                                         goto nla_put_failure;
3855
3856                                 err = af_ops->fill_stats_af(skb, dev);
3857
3858                                 if (err == -ENODATA)
3859                                         nla_nest_cancel(skb, af);
3860                                 else if (err < 0)
3861                                         goto nla_put_failure;
3862
3863                                 nla_nest_end(skb, af);
3864                         }
3865                 }
3866
3867                 nla_nest_end(skb, attr);
3868
3869                 *idxattr = 0;
3870         }
3871
3872         nlmsg_end(skb, nlh);
3873
3874         return 0;
3875
3876 nla_put_failure:
3877         /* not a multi message or no progress mean a real error */
3878         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3879                 nlmsg_cancel(skb, nlh);
3880         else
3881                 nlmsg_end(skb, nlh);
3882
3883         return -EMSGSIZE;
3884 }
3885
3886 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3887                                   u32 filter_mask)
3888 {
3889         size_t size = 0;
3890
3891         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3892                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3893
3894         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3895                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3896                 int attr = IFLA_STATS_LINK_XSTATS;
3897
3898                 if (ops && ops->get_linkxstats_size) {
3899                         size += nla_total_size(ops->get_linkxstats_size(dev,
3900                                                                         attr));
3901                         /* for IFLA_STATS_LINK_XSTATS */
3902                         size += nla_total_size(0);
3903                 }
3904         }
3905
3906         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3907                 struct net_device *_dev = (struct net_device *)dev;
3908                 const struct rtnl_link_ops *ops = NULL;
3909                 const struct net_device *master;
3910
3911                 /* netdev_master_upper_dev_get can't take const */
3912                 master = netdev_master_upper_dev_get(_dev);
3913                 if (master)
3914                         ops = master->rtnl_link_ops;
3915                 if (ops && ops->get_linkxstats_size) {
3916                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3917
3918                         size += nla_total_size(ops->get_linkxstats_size(dev,
3919                                                                         attr));
3920                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3921                         size += nla_total_size(0);
3922                 }
3923         }
3924
3925         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3926                 size += rtnl_get_offload_stats_size(dev);
3927
3928         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3929                 struct rtnl_af_ops *af_ops;
3930
3931                 /* for IFLA_STATS_AF_SPEC */
3932                 size += nla_total_size(0);
3933
3934                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3935                         if (af_ops->get_stats_af_size) {
3936                                 size += nla_total_size(
3937                                         af_ops->get_stats_af_size(dev));
3938
3939                                 /* for AF_* */
3940                                 size += nla_total_size(0);
3941                         }
3942                 }
3943         }
3944
3945         return size;
3946 }
3947
3948 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3949 {
3950         struct net *net = sock_net(skb->sk);
3951         struct net_device *dev = NULL;
3952         int idxattr = 0, prividx = 0;
3953         struct if_stats_msg *ifsm;
3954         struct sk_buff *nskb;
3955         u32 filter_mask;
3956         int err;
3957
3958         if (nlmsg_len(nlh) < sizeof(*ifsm))
3959                 return -EINVAL;
3960
3961         ifsm = nlmsg_data(nlh);
3962         if (ifsm->ifindex > 0)
3963                 dev = __dev_get_by_index(net, ifsm->ifindex);
3964         else
3965                 return -EINVAL;
3966
3967         if (!dev)
3968                 return -ENODEV;
3969
3970         filter_mask = ifsm->filter_mask;
3971         if (!filter_mask)
3972                 return -EINVAL;
3973
3974         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3975         if (!nskb)
3976                 return -ENOBUFS;
3977
3978         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3979                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3980                                   0, filter_mask, &idxattr, &prividx);
3981         if (err < 0) {
3982                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3983                 WARN_ON(err == -EMSGSIZE);
3984                 kfree_skb(nskb);
3985         } else {
3986                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3987         }
3988
3989         return err;
3990 }
3991
3992 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3993 {
3994         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3995         struct net *net = sock_net(skb->sk);
3996         unsigned int flags = NLM_F_MULTI;
3997         struct if_stats_msg *ifsm;
3998         struct hlist_head *head;
3999         struct net_device *dev;
4000         u32 filter_mask = 0;
4001         int idx = 0;
4002
4003         s_h = cb->args[0];
4004         s_idx = cb->args[1];
4005         s_idxattr = cb->args[2];
4006         s_prividx = cb->args[3];
4007
4008         cb->seq = net->dev_base_seq;
4009
4010         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4011                 return -EINVAL;
4012
4013         ifsm = nlmsg_data(cb->nlh);
4014         filter_mask = ifsm->filter_mask;
4015         if (!filter_mask)
4016                 return -EINVAL;
4017
4018         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4019                 idx = 0;
4020                 head = &net->dev_index_head[h];
4021                 hlist_for_each_entry(dev, head, index_hlist) {
4022                         if (idx < s_idx)
4023                                 goto cont;
4024                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4025                                                   NETLINK_CB(cb->skb).portid,
4026                                                   cb->nlh->nlmsg_seq, 0,
4027                                                   flags, filter_mask,
4028                                                   &s_idxattr, &s_prividx);
4029                         /* If we ran out of room on the first message,
4030                          * we're in trouble
4031                          */
4032                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4033
4034                         if (err < 0)
4035                                 goto out;
4036                         s_prividx = 0;
4037                         s_idxattr = 0;
4038                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4039 cont:
4040                         idx++;
4041                 }
4042         }
4043 out:
4044         cb->args[3] = s_prividx;
4045         cb->args[2] = s_idxattr;
4046         cb->args[1] = idx;
4047         cb->args[0] = h;
4048
4049         return skb->len;
4050 }
4051
4052 /* Process one rtnetlink message. */
4053
4054 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4055                              struct netlink_ext_ack *extack)
4056 {
4057         struct net *net = sock_net(skb->sk);
4058         rtnl_doit_func doit;
4059         int kind;
4060         int family;
4061         int type;
4062         int err;
4063
4064         type = nlh->nlmsg_type;
4065         if (type > RTM_MAX)
4066                 return -EOPNOTSUPP;
4067
4068         type -= RTM_BASE;
4069
4070         /* All the messages must have at least 1 byte length */
4071         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4072                 return 0;
4073
4074         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4075         kind = type&3;
4076
4077         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4078                 return -EPERM;
4079
4080         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4081                 struct sock *rtnl;
4082                 rtnl_dumpit_func dumpit;
4083                 rtnl_calcit_func calcit;
4084                 u16 min_dump_alloc = 0;
4085
4086                 dumpit = rtnl_get_dumpit(family, type);
4087                 if (dumpit == NULL)
4088                         return -EOPNOTSUPP;
4089                 calcit = rtnl_get_calcit(family, type);
4090                 if (calcit)
4091                         min_dump_alloc = calcit(skb, nlh);
4092
4093                 __rtnl_unlock();
4094                 rtnl = net->rtnl;
4095                 {
4096                         struct netlink_dump_control c = {
4097                                 .dump           = dumpit,
4098                                 .min_dump_alloc = min_dump_alloc,
4099                         };
4100                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4101                 }
4102                 rtnl_lock();
4103                 return err;
4104         }
4105
4106         doit = rtnl_get_doit(family, type);
4107         if (doit == NULL)
4108                 return -EOPNOTSUPP;
4109
4110         return doit(skb, nlh);
4111 }
4112
4113 static void rtnetlink_rcv(struct sk_buff *skb)
4114 {
4115         rtnl_lock();
4116         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4117         rtnl_unlock();
4118 }
4119
4120 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4121 {
4122         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4123
4124         switch (event) {
4125         case NETDEV_REBOOT:
4126         case NETDEV_CHANGENAME:
4127         case NETDEV_FEAT_CHANGE:
4128         case NETDEV_BONDING_FAILOVER:
4129         case NETDEV_NOTIFY_PEERS:
4130         case NETDEV_RESEND_IGMP:
4131         case NETDEV_CHANGEINFODATA:
4132                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4133                 break;
4134         default:
4135                 break;
4136         }
4137         return NOTIFY_DONE;
4138 }
4139
4140 static struct notifier_block rtnetlink_dev_notifier = {
4141         .notifier_call  = rtnetlink_event,
4142 };
4143
4144
4145 static int __net_init rtnetlink_net_init(struct net *net)
4146 {
4147         struct sock *sk;
4148         struct netlink_kernel_cfg cfg = {
4149                 .groups         = RTNLGRP_MAX,
4150                 .input          = rtnetlink_rcv,
4151                 .cb_mutex       = &rtnl_mutex,
4152                 .flags          = NL_CFG_F_NONROOT_RECV,
4153         };
4154
4155         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4156         if (!sk)
4157                 return -ENOMEM;
4158         net->rtnl = sk;
4159         return 0;
4160 }
4161
4162 static void __net_exit rtnetlink_net_exit(struct net *net)
4163 {
4164         netlink_kernel_release(net->rtnl);
4165         net->rtnl = NULL;
4166 }
4167
4168 static struct pernet_operations rtnetlink_net_ops = {
4169         .init = rtnetlink_net_init,
4170         .exit = rtnetlink_net_exit,
4171 };
4172
4173 void __init rtnetlink_init(void)
4174 {
4175         if (register_pernet_subsys(&rtnetlink_net_ops))
4176                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4177
4178         register_netdevice_notifier(&rtnetlink_dev_notifier);
4179
4180         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4181                       rtnl_dump_ifinfo, rtnl_calcit);
4182         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4183         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4184         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4185
4186         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4187         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4188         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4189
4190         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4191         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4192         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4193
4194         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4195         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4196         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4197
4198         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4199                       NULL);
4200 }