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 && dev_is_pci(dev->dev.parent) &&
 841             (ext_filter_mask & RTEXT_FILTER_VF)) {
 842                 int num_vfs = dev_num_vf(dev->dev.parent);
 843                 size_t size = nla_total_size(0);
 844                 size += num_vfs *
 845                         (nla_total_size(0) +
 846                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 847                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 848                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 849                          nla_total_size(MAX_VLAN_LIST_LEN *
 850                                         sizeof(struct ifla_vf_vlan_info)) +
 851                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 852                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 853                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 854                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 855                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 856                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 857                          /* IFLA_VF_STATS_RX_PACKETS */
 858                          nla_total_size_64bit(sizeof(__u64)) +
 859                          /* IFLA_VF_STATS_TX_PACKETS */
 860                          nla_total_size_64bit(sizeof(__u64)) +
 861                          /* IFLA_VF_STATS_RX_BYTES */
 862                          nla_total_size_64bit(sizeof(__u64)) +
 863                          /* IFLA_VF_STATS_TX_BYTES */
 864                          nla_total_size_64bit(sizeof(__u64)) +
 865                          /* IFLA_VF_STATS_BROADCAST */
 866                          nla_total_size_64bit(sizeof(__u64)) +
 867                          /* IFLA_VF_STATS_MULTICAST */
 868                          nla_total_size_64bit(sizeof(__u64)) +
 869                          nla_total_size(sizeof(struct ifla_vf_trust)));
 870                 return size;
 871         } else
 872                 return 0;
 873 }
 874
 875 static size_t rtnl_port_size(const struct net_device *dev,
 876                              u32 ext_filter_mask)
 877 {
 878         size_t port_size = nla_total_size(4)            /* PORT_VF */
 879                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 880                 + nla_total_size(sizeof(struct ifla_port_vsi))
 881                                                         /* PORT_VSI_TYPE */
 882                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 883                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 884                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 885                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 886         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 887         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 888                 + port_size;
 889         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 890                 + port_size;
 891
 892         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 893             !(ext_filter_mask & RTEXT_FILTER_VF))
 894                 return 0;
 895         if (dev_num_vf(dev->dev.parent))
 896                 return port_self_size + vf_ports_size +
 897                         vf_port_size * dev_num_vf(dev->dev.parent);
 898         else
 899                 return port_self_size;
 900 }
 901
 902 static size_t rtnl_xdp_size(const struct net_device *dev)
 903 {
 904         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 905                           nla_total_size(1);    /* XDP_ATTACHED */
 906
 907         if (!dev->netdev_ops->ndo_xdp)
 908                 return 0;
 909         else
 910                 return xdp_size;
 911 }
 912
 913 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 914                                      u32 ext_filter_mask)
 915 {
 916         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 917                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 918                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 919                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 920                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 921                + nla_total_size(sizeof(struct rtnl_link_stats))
 922                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 923                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 924                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 925                + nla_total_size(4) /* IFLA_TXQLEN */
 926                + nla_total_size(4) /* IFLA_WEIGHT */
 927                + nla_total_size(4) /* IFLA_MTU */
 928                + nla_total_size(4) /* IFLA_LINK */
 929                + nla_total_size(4) /* IFLA_MASTER */
 930                + nla_total_size(1) /* IFLA_CARRIER */
 931                + nla_total_size(4) /* IFLA_PROMISCUITY */
 932                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 933                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 934                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 935                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 936                + nla_total_size(1) /* IFLA_OPERSTATE */
 937                + nla_total_size(1) /* IFLA_LINKMODE */
 938                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 939                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 940                + nla_total_size(ext_filter_mask
 941                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 942                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 943                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 944                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 945                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 946                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 947                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 948                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 949                + rtnl_xdp_size(dev) /* IFLA_XDP */
 950                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 951
 952 }
 953
 954 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 955 {
 956         struct nlattr *vf_ports;
 957         struct nlattr *vf_port;
 958         int vf;
 959         int err;
 960
 961         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 962         if (!vf_ports)
 963                 return -EMSGSIZE;
 964
 965         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 966                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 967                 if (!vf_port)
 968                         goto nla_put_failure;
 969                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 970                         goto nla_put_failure;
 971                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 972                 if (err == -EMSGSIZE)
 973                         goto nla_put_failure;
 974                 if (err) {
 975                         nla_nest_cancel(skb, vf_port);
 976                         continue;
 977                 }
 978                 nla_nest_end(skb, vf_port);
 979         }
 980
 981         nla_nest_end(skb, vf_ports);
 982
 983         return 0;
 984
 985 nla_put_failure:
 986         nla_nest_cancel(skb, vf_ports);
 987         return -EMSGSIZE;
 988 }
 989
 990 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 991 {
 992         struct nlattr *port_self;
 993         int err;
 994
 995         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 996         if (!port_self)
 997                 return -EMSGSIZE;
 998
 999         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1000         if (err) {
1001                 nla_nest_cancel(skb, port_self);
1002                 return (err == -EMSGSIZE) ? err : 0;
1003         }
1004
1005         nla_nest_end(skb, port_self);
1006
1007         return 0;
1008 }
1009
1010 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1011                           u32 ext_filter_mask)
1012 {
1013         int err;
1014
1015         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1016             !(ext_filter_mask & RTEXT_FILTER_VF))
1017                 return 0;
1018
1019         err = rtnl_port_self_fill(skb, dev);
1020         if (err)
1021                 return err;
1022
1023         if (dev_num_vf(dev->dev.parent)) {
1024                 err = rtnl_vf_ports_fill(skb, dev);
1025                 if (err)
1026                         return err;
1027         }
1028
1029         return 0;
1030 }
1031
1032 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1033 {
1034         int err;
1035         struct netdev_phys_item_id ppid;
1036
1037         err = dev_get_phys_port_id(dev, &ppid);
1038         if (err) {
1039                 if (err == -EOPNOTSUPP)
1040                         return 0;
1041                 return err;
1042         }
1043
1044         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1045                 return -EMSGSIZE;
1046
1047         return 0;
1048 }
1049
1050 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1051 {
1052         char name[IFNAMSIZ];
1053         int err;
1054
1055         err = dev_get_phys_port_name(dev, name, sizeof(name));
1056         if (err) {
1057                 if (err == -EOPNOTSUPP)
1058                         return 0;
1059                 return err;
1060         }
1061
1062         if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1063                 return -EMSGSIZE;
1064
1065         return 0;
1066 }
1067
1068 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1069 {
1070         int err;
1071         struct switchdev_attr attr = {
1072                 .orig_dev = dev,
1073                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1074                 .flags = SWITCHDEV_F_NO_RECURSE,
1075         };
1076
1077         err = switchdev_port_attr_get(dev, &attr);
1078         if (err) {
1079                 if (err == -EOPNOTSUPP)
1080                         return 0;
1081                 return err;
1082         }
1083
1084         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1085                     attr.u.ppid.id))
1086                 return -EMSGSIZE;
1087
1088         return 0;
1089 }
1090
1091 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1092                                               struct net_device *dev)
1093 {
1094         struct rtnl_link_stats64 *sp;
1095         struct nlattr *attr;
1096
1097         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1098                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1099         if (!attr)
1100                 return -EMSGSIZE;
1101
1102         sp = nla_data(attr);
1103         dev_get_stats(dev, sp);
1104
1105         attr = nla_reserve(skb, IFLA_STATS,
1106                            sizeof(struct rtnl_link_stats));
1107         if (!attr)
1108                 return -EMSGSIZE;
1109
1110         copy_rtnl_link_stats(nla_data(attr), sp);
1111
1112         return 0;
1113 }
1114
1115 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1116                                                struct net_device *dev,
1117                                                int vfs_num,
1118                                                struct nlattr *vfinfo)
1119 {
1120         struct ifla_vf_rss_query_en vf_rss_query_en;
1121         struct nlattr *vf, *vfstats, *vfvlanlist;
1122         struct ifla_vf_link_state vf_linkstate;
1123         struct ifla_vf_vlan_info vf_vlan_info;
1124         struct ifla_vf_spoofchk vf_spoofchk;
1125         struct ifla_vf_tx_rate vf_tx_rate;
1126         struct ifla_vf_stats vf_stats;
1127         struct ifla_vf_trust vf_trust;
1128         struct ifla_vf_vlan vf_vlan;
1129         struct ifla_vf_rate vf_rate;
1130         struct ifla_vf_mac vf_mac;
1131         struct ifla_vf_info ivi;
1132
1133         /* Not all SR-IOV capable drivers support the
1134          * spoofcheck and "RSS query enable" query.  Preset to
1135          * -1 so the user space tool can detect that the driver
1136          * didn't report anything.
1137          */
1138         ivi.spoofchk = -1;
1139         ivi.rss_query_en = -1;
1140         ivi.trusted = -1;
1141         memset(ivi.mac, 0, sizeof(ivi.mac));
1142         /* The default value for VF link state is "auto"
1143          * IFLA_VF_LINK_STATE_AUTO which equals zero
1144          */
1145         ivi.linkstate = 0;
1146         /* VLAN Protocol by default is 802.1Q */
1147         ivi.vlan_proto = htons(ETH_P_8021Q);
1148         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1149                 return 0;
1150
1151         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1152
1153         vf_mac.vf =
1154                 vf_vlan.vf =
1155                 vf_vlan_info.vf =
1156                 vf_rate.vf =
1157                 vf_tx_rate.vf =
1158                 vf_spoofchk.vf =
1159                 vf_linkstate.vf =
1160                 vf_rss_query_en.vf =
1161                 vf_trust.vf = ivi.vf;
1162
1163         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1164         vf_vlan.vlan = ivi.vlan;
1165         vf_vlan.qos = ivi.qos;
1166         vf_vlan_info.vlan = ivi.vlan;
1167         vf_vlan_info.qos = ivi.qos;
1168         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1169         vf_tx_rate.rate = ivi.max_tx_rate;
1170         vf_rate.min_tx_rate = ivi.min_tx_rate;
1171         vf_rate.max_tx_rate = ivi.max_tx_rate;
1172         vf_spoofchk.setting = ivi.spoofchk;
1173         vf_linkstate.link_state = ivi.linkstate;
1174         vf_rss_query_en.setting = ivi.rss_query_en;
1175         vf_trust.setting = ivi.trusted;
1176         vf = nla_nest_start(skb, IFLA_VF_INFO);
1177         if (!vf)
1178                 goto nla_put_vfinfo_failure;
1179         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1182                     &vf_rate) ||
1183             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1184                     &vf_tx_rate) ||
1185             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1186                     &vf_spoofchk) ||
1187             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1188                     &vf_linkstate) ||
1189             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190                     sizeof(vf_rss_query_en),
1191                     &vf_rss_query_en) ||
1192             nla_put(skb, IFLA_VF_TRUST,
1193                     sizeof(vf_trust), &vf_trust))
1194                 goto nla_put_vf_failure;
1195         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1196         if (!vfvlanlist)
1197                 goto nla_put_vf_failure;
1198         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1199                     &vf_vlan_info)) {
1200                 nla_nest_cancel(skb, vfvlanlist);
1201                 goto nla_put_vf_failure;
1202         }
1203         nla_nest_end(skb, vfvlanlist);
1204         memset(&vf_stats, 0, sizeof(vf_stats));
1205         if (dev->netdev_ops->ndo_get_vf_stats)
1206                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1207                                                 &vf_stats);
1208         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1209         if (!vfstats)
1210                 goto nla_put_vf_failure;
1211         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1212                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1214                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1216                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1217             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1218                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1219             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1220                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1221             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1222                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1223                 nla_nest_cancel(skb, vfstats);
1224                 goto nla_put_vf_failure;
1225         }
1226         nla_nest_end(skb, vfstats);
1227         nla_nest_end(skb, vf);
1228         return 0;
1229
1230 nla_put_vf_failure:
1231         nla_nest_cancel(skb, vf);
1232 nla_put_vfinfo_failure:
1233         nla_nest_cancel(skb, vfinfo);
1234         return -EMSGSIZE;
1235 }
1236
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1238 {
1239         struct rtnl_link_ifmap map;
1240
1241         memset(&map, 0, sizeof(map));
1242         map.mem_start   = dev->mem_start;
1243         map.mem_end     = dev->mem_end;
1244         map.base_addr   = dev->base_addr;
1245         map.irq         = dev->irq;
1246         map.dma         = dev->dma;
1247         map.port        = dev->if_port;
1248
1249         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1250                 return -EMSGSIZE;
1251
1252         return 0;
1253 }
1254
1255 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1256 {
1257         struct netdev_xdp xdp_op = {};
1258         struct nlattr *xdp;
1259         int err;
1260
1261         if (!dev->netdev_ops->ndo_xdp)
1262                 return 0;
1263         xdp = nla_nest_start(skb, IFLA_XDP);
1264         if (!xdp)
1265                 return -EMSGSIZE;
1266         xdp_op.command = XDP_QUERY_PROG;
1267         err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1268         if (err)
1269                 goto err_cancel;
1270         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1271         if (err)
1272                 goto err_cancel;
1273
1274         nla_nest_end(skb, xdp);
1275         return 0;
1276
1277 err_cancel:
1278         nla_nest_cancel(skb, xdp);
1279         return err;
1280 }
1281
1282 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1283                             int type, u32 pid, u32 seq, u32 change,
1284                             unsigned int flags, u32 ext_filter_mask)
1285 {
1286         struct ifinfomsg *ifm;
1287         struct nlmsghdr *nlh;
1288         struct nlattr *af_spec;
1289         struct rtnl_af_ops *af_ops;
1290         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1291
1292         ASSERT_RTNL();
1293         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1294         if (nlh == NULL)
1295                 return -EMSGSIZE;
1296
1297         ifm = nlmsg_data(nlh);
1298         ifm->ifi_family = AF_UNSPEC;
1299         ifm->__ifi_pad = 0;
1300         ifm->ifi_type = dev->type;
1301         ifm->ifi_index = dev->ifindex;
1302         ifm->ifi_flags = dev_get_flags(dev);
1303         ifm->ifi_change = change;
1304
1305         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1306             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1307             nla_put_u8(skb, IFLA_OPERSTATE,
1308                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1309             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1310             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1311             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1312             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1313             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1314             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1315             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1316 #ifdef CONFIG_RPS
1317             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1318 #endif
1319             (dev->ifindex != dev_get_iflink(dev) &&
1320              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1321             (upper_dev &&
1322              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1323             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1324             (dev->qdisc &&
1325              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1326             (dev->ifalias &&
1327              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1328             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1329                         atomic_read(&dev->carrier_changes)) ||
1330             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1331                 goto nla_put_failure;
1332
1333         if (rtnl_fill_link_ifmap(skb, dev))
1334                 goto nla_put_failure;
1335
1336         if (dev->addr_len) {
1337                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1338                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1339                         goto nla_put_failure;
1340         }
1341
1342         if (rtnl_phys_port_id_fill(skb, dev))
1343                 goto nla_put_failure;
1344
1345         if (rtnl_phys_port_name_fill(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (rtnl_phys_switch_id_fill(skb, dev))
1349                 goto nla_put_failure;
1350
1351         if (rtnl_fill_stats(skb, dev))
1352                 goto nla_put_failure;
1353
1354         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1355             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1356                 goto nla_put_failure;
1357
1358         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1359             ext_filter_mask & RTEXT_FILTER_VF) {
1360                 int i;
1361                 struct nlattr *vfinfo;
1362                 int num_vfs = dev_num_vf(dev->dev.parent);
1363
1364                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1365                 if (!vfinfo)
1366                         goto nla_put_failure;
1367                 for (i = 0; i < num_vfs; i++) {
1368                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1369                                 goto nla_put_failure;
1370                 }
1371
1372                 nla_nest_end(skb, vfinfo);
1373         }
1374
1375         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1376                 goto nla_put_failure;
1377
1378         if (rtnl_xdp_fill(skb, dev))
1379                 goto nla_put_failure;
1380
1381         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1382                 if (rtnl_link_fill(skb, dev) < 0)
1383                         goto nla_put_failure;
1384         }
1385
1386         if (dev->rtnl_link_ops &&
1387             dev->rtnl_link_ops->get_link_net) {
1388                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1389
1390                 if (!net_eq(dev_net(dev), link_net)) {
1391                         int id = peernet2id_alloc(dev_net(dev), link_net);
1392
1393                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1394                                 goto nla_put_failure;
1395                 }
1396         }
1397
1398         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1399                 goto nla_put_failure;
1400
1401         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1402                 if (af_ops->fill_link_af) {
1403                         struct nlattr *af;
1404                         int err;
1405
1406                         if (!(af = nla_nest_start(skb, af_ops->family)))
1407                                 goto nla_put_failure;
1408
1409                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1410
1411                         /*
1412                          * Caller may return ENODATA to indicate that there
1413                          * was no data to be dumped. This is not an error, it
1414                          * means we should trim the attribute header and
1415                          * continue.
1416                          */
1417                         if (err == -ENODATA)
1418                                 nla_nest_cancel(skb, af);
1419                         else if (err < 0)
1420                                 goto nla_put_failure;
1421
1422                         nla_nest_end(skb, af);
1423                 }
1424         }
1425
1426         nla_nest_end(skb, af_spec);
1427
1428         nlmsg_end(skb, nlh);
1429         return 0;
1430
1431 nla_put_failure:
1432         nlmsg_cancel(skb, nlh);
1433         return -EMSGSIZE;
1434 }
1435
1436 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1437         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1438         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1439         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1440         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1441         [IFLA_MTU]              = { .type = NLA_U32 },
1442         [IFLA_LINK]             = { .type = NLA_U32 },
1443         [IFLA_MASTER]           = { .type = NLA_U32 },
1444         [IFLA_CARRIER]          = { .type = NLA_U8 },
1445         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1446         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1447         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1448         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1449         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1450         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1451         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1452         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1453         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1454         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1455         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1456         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1457         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1458         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1459         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1460         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1461         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1462         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1463         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1464         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1465         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1466         [IFLA_XDP]              = { .type = NLA_NESTED },
1467 };
1468
1469 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1470         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1471         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1472         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1473         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1474 };
1475
1476 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1477         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1478         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1479         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1480         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1481         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1482         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1483         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1484         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1485         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1486         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1487         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1488         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1489 };
1490
1491 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1492         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1493         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1494                                     .len = PORT_PROFILE_MAX },
1495         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1496                                     .len = sizeof(struct ifla_port_vsi)},
1497         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1498                                       .len = PORT_UUID_MAX },
1499         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1500                                     .len = PORT_UUID_MAX },
1501         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1502         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1503 };
1504
1505 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1506         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1507         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1508         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1509 };
1510
1511 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1512 {
1513         const struct rtnl_link_ops *ops = NULL;
1514         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1515
1516         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1517                 return NULL;
1518
1519         if (linfo[IFLA_INFO_KIND]) {
1520                 char kind[MODULE_NAME_LEN];
1521
1522                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1523                 ops = rtnl_link_ops_get(kind);
1524         }
1525
1526         return ops;
1527 }
1528
1529 static bool link_master_filtered(struct net_device *dev, int master_idx)
1530 {
1531         struct net_device *master;
1532
1533         if (!master_idx)
1534                 return false;
1535
1536         master = netdev_master_upper_dev_get(dev);
1537         if (!master || master->ifindex != master_idx)
1538                 return true;
1539
1540         return false;
1541 }
1542
1543 static bool link_kind_filtered(const struct net_device *dev,
1544                                const struct rtnl_link_ops *kind_ops)
1545 {
1546         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1547                 return true;
1548
1549         return false;
1550 }
1551
1552 static bool link_dump_filtered(struct net_device *dev,
1553                                int master_idx,
1554                                const struct rtnl_link_ops *kind_ops)
1555 {
1556         if (link_master_filtered(dev, master_idx) ||
1557             link_kind_filtered(dev, kind_ops))
1558                 return true;
1559
1560         return false;
1561 }
1562
1563 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1564 {
1565         struct net *net = sock_net(skb->sk);
1566         int h, s_h;
1567         int idx = 0, s_idx;
1568         struct net_device *dev;
1569         struct hlist_head *head;
1570         struct nlattr *tb[IFLA_MAX+1];
1571         u32 ext_filter_mask = 0;
1572         const struct rtnl_link_ops *kind_ops = NULL;
1573         unsigned int flags = NLM_F_MULTI;
1574         int master_idx = 0;
1575         int err;
1576         int hdrlen;
1577
1578         s_h = cb->args[0];
1579         s_idx = cb->args[1];
1580
1581         cb->seq = net->dev_base_seq;
1582
1583         /* A hack to preserve kernel<->userspace interface.
1584          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1585          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1586          * what iproute2 < v3.9.0 used.
1587          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1588          * attribute, its netlink message is shorter than struct ifinfomsg.
1589          */
1590         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1591                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1592
1593         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1594
1595                 if (tb[IFLA_EXT_MASK])
1596                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1597
1598                 if (tb[IFLA_MASTER])
1599                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1600
1601                 if (tb[IFLA_LINKINFO])
1602                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1603
1604                 if (master_idx || kind_ops)
1605                         flags |= NLM_F_DUMP_FILTERED;
1606         }
1607
1608         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1609                 idx = 0;
1610                 head = &net->dev_index_head[h];
1611                 hlist_for_each_entry(dev, head, index_hlist) {
1612                         if (link_dump_filtered(dev, master_idx, kind_ops))
1613                                 goto cont;
1614                         if (idx < s_idx)
1615                                 goto cont;
1616                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1617                                                NETLINK_CB(cb->skb).portid,
1618                                                cb->nlh->nlmsg_seq, 0,
1619                                                flags,
1620                                                ext_filter_mask);
1621                         /* If we ran out of room on the first message,
1622                          * we're in trouble
1623                          */
1624                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1625
1626                         if (err < 0)
1627                                 goto out;
1628
1629                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1630 cont:
1631                         idx++;
1632                 }
1633         }
1634 out:
1635         cb->args[1] = idx;
1636         cb->args[0] = h;
1637
1638         return skb->len;
1639 }
1640
1641 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1642 {
1643         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1644 }
1645 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1646
1647 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1648 {
1649         struct net *net;
1650         /* Examine the link attributes and figure out which
1651          * network namespace we are talking about.
1652          */
1653         if (tb[IFLA_NET_NS_PID])
1654                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1655         else if (tb[IFLA_NET_NS_FD])
1656                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1657         else
1658                 net = get_net(src_net);
1659         return net;
1660 }
1661 EXPORT_SYMBOL(rtnl_link_get_net);
1662
1663 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1664 {
1665         if (dev) {
1666                 if (tb[IFLA_ADDRESS] &&
1667                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1668                         return -EINVAL;
1669
1670                 if (tb[IFLA_BROADCAST] &&
1671                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1672                         return -EINVAL;
1673         }
1674
1675         if (tb[IFLA_AF_SPEC]) {
1676                 struct nlattr *af;
1677                 int rem, err;
1678
1679                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1680                         const struct rtnl_af_ops *af_ops;
1681
1682                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1683                                 return -EAFNOSUPPORT;
1684
1685                         if (!af_ops->set_link_af)
1686                                 return -EOPNOTSUPP;
1687
1688                         if (af_ops->validate_link_af) {
1689                                 err = af_ops->validate_link_af(dev, af);
1690                                 if (err < 0)
1691                                         return err;
1692                         }
1693                 }
1694         }
1695
1696         return 0;
1697 }
1698
1699 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1700                                   int guid_type)
1701 {
1702         const struct net_device_ops *ops = dev->netdev_ops;
1703
1704         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1705 }
1706
1707 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1708 {
1709         if (dev->type != ARPHRD_INFINIBAND)
1710                 return -EOPNOTSUPP;
1711
1712         return handle_infiniband_guid(dev, ivt, guid_type);
1713 }
1714
1715 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1716 {
1717         const struct net_device_ops *ops = dev->netdev_ops;
1718         int err = -EINVAL;
1719
1720         if (tb[IFLA_VF_MAC]) {
1721                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1722
1723                 err = -EOPNOTSUPP;
1724                 if (ops->ndo_set_vf_mac)
1725                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1726                                                   ivm->mac);
1727                 if (err < 0)
1728                         return err;
1729         }
1730
1731         if (tb[IFLA_VF_VLAN]) {
1732                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1733
1734                 err = -EOPNOTSUPP;
1735                 if (ops->ndo_set_vf_vlan)
1736                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1737                                                    ivv->qos,
1738                                                    htons(ETH_P_8021Q));
1739                 if (err < 0)
1740                         return err;
1741         }
1742
1743         if (tb[IFLA_VF_VLAN_LIST]) {
1744                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1745                 struct nlattr *attr;
1746                 int rem, len = 0;
1747
1748                 err = -EOPNOTSUPP;
1749                 if (!ops->ndo_set_vf_vlan)
1750                         return err;
1751
1752                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1753                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1754                             nla_len(attr) < NLA_HDRLEN) {
1755                                 return -EINVAL;
1756                         }
1757                         if (len >= MAX_VLAN_LIST_LEN)
1758                                 return -EOPNOTSUPP;
1759                         ivvl[len] = nla_data(attr);
1760
1761                         len++;
1762                 }
1763                 if (len == 0)
1764                         return -EINVAL;
1765
1766                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1767                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1768                 if (err < 0)
1769                         return err;
1770         }
1771
1772         if (tb[IFLA_VF_TX_RATE]) {
1773                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1774                 struct ifla_vf_info ivf;
1775
1776                 err = -EOPNOTSUPP;
1777                 if (ops->ndo_get_vf_config)
1778                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1779                 if (err < 0)
1780                         return err;
1781
1782                 err = -EOPNOTSUPP;
1783                 if (ops->ndo_set_vf_rate)
1784                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1785                                                    ivf.min_tx_rate,
1786                                                    ivt->rate);
1787                 if (err < 0)
1788                         return err;
1789         }
1790
1791         if (tb[IFLA_VF_RATE]) {
1792                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1793
1794                 err = -EOPNOTSUPP;
1795                 if (ops->ndo_set_vf_rate)
1796                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1797                                                    ivt->min_tx_rate,
1798                                                    ivt->max_tx_rate);
1799                 if (err < 0)
1800                         return err;
1801         }
1802
1803         if (tb[IFLA_VF_SPOOFCHK]) {
1804                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1805
1806                 err = -EOPNOTSUPP;
1807                 if (ops->ndo_set_vf_spoofchk)
1808                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1809                                                        ivs->setting);
1810                 if (err < 0)
1811                         return err;
1812         }
1813
1814         if (tb[IFLA_VF_LINK_STATE]) {
1815                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1816
1817                 err = -EOPNOTSUPP;
1818                 if (ops->ndo_set_vf_link_state)
1819                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1820                                                          ivl->link_state);
1821                 if (err < 0)
1822                         return err;
1823         }
1824
1825         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1826                 struct ifla_vf_rss_query_en *ivrssq_en;
1827
1828                 err = -EOPNOTSUPP;
1829                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1830                 if (ops->ndo_set_vf_rss_query_en)
1831                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1832                                                            ivrssq_en->setting);
1833                 if (err < 0)
1834                         return err;
1835         }
1836
1837         if (tb[IFLA_VF_TRUST]) {
1838                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1839
1840                 err = -EOPNOTSUPP;
1841                 if (ops->ndo_set_vf_trust)
1842                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1843                 if (err < 0)
1844                         return err;
1845         }
1846
1847         if (tb[IFLA_VF_IB_NODE_GUID]) {
1848                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1849
1850                 if (!ops->ndo_set_vf_guid)
1851                         return -EOPNOTSUPP;
1852
1853                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1854         }
1855
1856         if (tb[IFLA_VF_IB_PORT_GUID]) {
1857                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1858
1859                 if (!ops->ndo_set_vf_guid)
1860                         return -EOPNOTSUPP;
1861
1862                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1863         }
1864
1865         return err;
1866 }
1867
1868 static int do_set_master(struct net_device *dev, int ifindex)
1869 {
1870         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1871         const struct net_device_ops *ops;
1872         int err;
1873
1874         if (upper_dev) {
1875                 if (upper_dev->ifindex == ifindex)
1876                         return 0;
1877                 ops = upper_dev->netdev_ops;
1878                 if (ops->ndo_del_slave) {
1879                         err = ops->ndo_del_slave(upper_dev, dev);
1880                         if (err)
1881                                 return err;
1882                 } else {
1883                         return -EOPNOTSUPP;
1884                 }
1885         }
1886
1887         if (ifindex) {
1888                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1889                 if (!upper_dev)
1890                         return -EINVAL;
1891                 ops = upper_dev->netdev_ops;
1892                 if (ops->ndo_add_slave) {
1893                         err = ops->ndo_add_slave(upper_dev, dev);
1894                         if (err)
1895                                 return err;
1896                 } else {
1897                         return -EOPNOTSUPP;
1898                 }
1899         }
1900         return 0;
1901 }
1902
1903 #define DO_SETLINK_MODIFIED     0x01
1904 /* notify flag means notify + modified. */
1905 #define DO_SETLINK_NOTIFY       0x03
1906 static int do_setlink(const struct sk_buff *skb,
1907                       struct net_device *dev, struct ifinfomsg *ifm,
1908                       struct nlattr **tb, char *ifname, int status)
1909 {
1910         const struct net_device_ops *ops = dev->netdev_ops;
1911         int err;
1912
1913         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1914                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1915                 if (IS_ERR(net)) {
1916                         err = PTR_ERR(net);
1917                         goto errout;
1918                 }
1919                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1920                         put_net(net);
1921                         err = -EPERM;
1922                         goto errout;
1923                 }
1924                 err = dev_change_net_namespace(dev, net, ifname);
1925                 put_net(net);
1926                 if (err)
1927                         goto errout;
1928                 status |= DO_SETLINK_MODIFIED;
1929         }
1930
1931         if (tb[IFLA_MAP]) {
1932                 struct rtnl_link_ifmap *u_map;
1933                 struct ifmap k_map;
1934
1935                 if (!ops->ndo_set_config) {
1936                         err = -EOPNOTSUPP;
1937                         goto errout;
1938                 }
1939
1940                 if (!netif_device_present(dev)) {
1941                         err = -ENODEV;
1942                         goto errout;
1943                 }
1944
1945                 u_map = nla_data(tb[IFLA_MAP]);
1946                 k_map.mem_start = (unsigned long) u_map->mem_start;
1947                 k_map.mem_end = (unsigned long) u_map->mem_end;
1948                 k_map.base_addr = (unsigned short) u_map->base_addr;
1949                 k_map.irq = (unsigned char) u_map->irq;
1950                 k_map.dma = (unsigned char) u_map->dma;
1951                 k_map.port = (unsigned char) u_map->port;
1952
1953                 err = ops->ndo_set_config(dev, &k_map);
1954                 if (err < 0)
1955                         goto errout;
1956
1957                 status |= DO_SETLINK_NOTIFY;
1958         }
1959
1960         if (tb[IFLA_ADDRESS]) {
1961                 struct sockaddr *sa;
1962                 int len;
1963
1964                 len = sizeof(sa_family_t) + dev->addr_len;
1965                 sa = kmalloc(len, GFP_KERNEL);
1966                 if (!sa) {
1967                         err = -ENOMEM;
1968                         goto errout;
1969                 }
1970                 sa->sa_family = dev->type;
1971                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1972                        dev->addr_len);
1973                 err = dev_set_mac_address(dev, sa);
1974                 kfree(sa);
1975                 if (err)
1976                         goto errout;
1977                 status |= DO_SETLINK_MODIFIED;
1978         }
1979
1980         if (tb[IFLA_MTU]) {
1981                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1982                 if (err < 0)
1983                         goto errout;
1984                 status |= DO_SETLINK_MODIFIED;
1985         }
1986
1987         if (tb[IFLA_GROUP]) {
1988                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1989                 status |= DO_SETLINK_NOTIFY;
1990         }
1991
1992         /*
1993          * Interface selected by interface index but interface
1994          * name provided implies that a name change has been
1995          * requested.
1996          */
1997         if (ifm->ifi_index > 0 && ifname[0]) {
1998                 err = dev_change_name(dev, ifname);
1999                 if (err < 0)
2000                         goto errout;
2001                 status |= DO_SETLINK_MODIFIED;
2002         }
2003
2004         if (tb[IFLA_IFALIAS]) {
2005                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2006                                     nla_len(tb[IFLA_IFALIAS]));
2007                 if (err < 0)
2008                         goto errout;
2009                 status |= DO_SETLINK_NOTIFY;
2010         }
2011
2012         if (tb[IFLA_BROADCAST]) {
2013                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2014                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2015         }
2016
2017         if (ifm->ifi_flags || ifm->ifi_change) {
2018                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2019                 if (err < 0)
2020                         goto errout;
2021         }
2022
2023         if (tb[IFLA_MASTER]) {
2024                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2025                 if (err)
2026                         goto errout;
2027                 status |= DO_SETLINK_MODIFIED;
2028         }
2029
2030         if (tb[IFLA_CARRIER]) {
2031                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2032                 if (err)
2033                         goto errout;
2034                 status |= DO_SETLINK_MODIFIED;
2035         }
2036
2037         if (tb[IFLA_TXQLEN]) {
2038                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2039                 unsigned long orig_len = dev->tx_queue_len;
2040
2041                 if (dev->tx_queue_len ^ value) {
2042                         dev->tx_queue_len = value;
2043                         err = call_netdevice_notifiers(
2044                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2045                         err = notifier_to_errno(err);
2046                         if (err) {
2047                                 dev->tx_queue_len = orig_len;
2048                                 goto errout;
2049                         }
2050                         status |= DO_SETLINK_NOTIFY;
2051                 }
2052         }
2053
2054         if (tb[IFLA_OPERSTATE])
2055                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2056
2057         if (tb[IFLA_LINKMODE]) {
2058                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2059
2060                 write_lock_bh(&dev_base_lock);
2061                 if (dev->link_mode ^ value)
2062                         status |= DO_SETLINK_NOTIFY;
2063                 dev->link_mode = value;
2064                 write_unlock_bh(&dev_base_lock);
2065         }
2066
2067         if (tb[IFLA_VFINFO_LIST]) {
2068                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2069                 struct nlattr *attr;
2070                 int rem;
2071
2072                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2073                         if (nla_type(attr) != IFLA_VF_INFO ||
2074                             nla_len(attr) < NLA_HDRLEN) {
2075                                 err = -EINVAL;
2076                                 goto errout;
2077                         }
2078                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2079                                                ifla_vf_policy);
2080                         if (err < 0)
2081                                 goto errout;
2082                         err = do_setvfinfo(dev, vfinfo);
2083                         if (err < 0)
2084                                 goto errout;
2085                         status |= DO_SETLINK_NOTIFY;
2086                 }
2087         }
2088         err = 0;
2089
2090         if (tb[IFLA_VF_PORTS]) {
2091                 struct nlattr *port[IFLA_PORT_MAX+1];
2092                 struct nlattr *attr;
2093                 int vf;
2094                 int rem;
2095
2096                 err = -EOPNOTSUPP;
2097                 if (!ops->ndo_set_vf_port)
2098                         goto errout;
2099
2100                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2101                         if (nla_type(attr) != IFLA_VF_PORT ||
2102                             nla_len(attr) < NLA_HDRLEN) {
2103                                 err = -EINVAL;
2104                                 goto errout;
2105                         }
2106                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2107                                                ifla_port_policy);
2108                         if (err < 0)
2109                                 goto errout;
2110                         if (!port[IFLA_PORT_VF]) {
2111                                 err = -EOPNOTSUPP;
2112                                 goto errout;
2113                         }
2114                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2115                         err = ops->ndo_set_vf_port(dev, vf, port);
2116                         if (err < 0)
2117                                 goto errout;
2118                         status |= DO_SETLINK_NOTIFY;
2119                 }
2120         }
2121         err = 0;
2122
2123         if (tb[IFLA_PORT_SELF]) {
2124                 struct nlattr *port[IFLA_PORT_MAX+1];
2125
2126                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2127                         tb[IFLA_PORT_SELF], ifla_port_policy);
2128                 if (err < 0)
2129                         goto errout;
2130
2131                 err = -EOPNOTSUPP;
2132                 if (ops->ndo_set_vf_port)
2133                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2134                 if (err < 0)
2135                         goto errout;
2136                 status |= DO_SETLINK_NOTIFY;
2137         }
2138
2139         if (tb[IFLA_AF_SPEC]) {
2140                 struct nlattr *af;
2141                 int rem;
2142
2143                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2144                         const struct rtnl_af_ops *af_ops;
2145
2146                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2147                                 BUG();
2148
2149                         err = af_ops->set_link_af(dev, af);
2150                         if (err < 0)
2151                                 goto errout;
2152
2153                         status |= DO_SETLINK_NOTIFY;
2154                 }
2155         }
2156         err = 0;
2157
2158         if (tb[IFLA_PROTO_DOWN]) {
2159                 err = dev_change_proto_down(dev,
2160                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2161                 if (err)
2162                         goto errout;
2163                 status |= DO_SETLINK_NOTIFY;
2164         }
2165
2166         if (tb[IFLA_XDP]) {
2167                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2168                 u32 xdp_flags = 0;
2169
2170                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2171                                        ifla_xdp_policy);
2172                 if (err < 0)
2173                         goto errout;
2174
2175                 if (xdp[IFLA_XDP_ATTACHED]) {
2176                         err = -EINVAL;
2177                         goto errout;
2178                 }
2179
2180                 if (xdp[IFLA_XDP_FLAGS]) {
2181                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2182                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2183                                 err = -EINVAL;
2184                                 goto errout;
2185                         }
2186                 }
2187
2188                 if (xdp[IFLA_XDP_FD]) {
2189                         err = dev_change_xdp_fd(dev,
2190                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2191                                                 xdp_flags);
2192                         if (err)
2193                                 goto errout;
2194                         status |= DO_SETLINK_NOTIFY;
2195                 }
2196         }
2197
2198 errout:
2199         if (status & DO_SETLINK_MODIFIED) {
2200                 if (status & DO_SETLINK_NOTIFY)
2201                         netdev_state_change(dev);
2202
2203                 if (err < 0)
2204                         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",
2205                                              dev->name);
2206         }
2207
2208         return err;
2209 }
2210
2211 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2212 {
2213         struct net *net = sock_net(skb->sk);
2214         struct ifinfomsg *ifm;
2215         struct net_device *dev;
2216         int err;
2217         struct nlattr *tb[IFLA_MAX+1];
2218         char ifname[IFNAMSIZ];
2219
2220         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2221         if (err < 0)
2222                 goto errout;
2223
2224         if (tb[IFLA_IFNAME])
2225                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2226         else
2227                 ifname[0] = '\0';
2228
2229         err = -EINVAL;
2230         ifm = nlmsg_data(nlh);
2231         if (ifm->ifi_index > 0)
2232                 dev = __dev_get_by_index(net, ifm->ifi_index);
2233         else if (tb[IFLA_IFNAME])
2234                 dev = __dev_get_by_name(net, ifname);
2235         else
2236                 goto errout;
2237
2238         if (dev == NULL) {
2239                 err = -ENODEV;
2240                 goto errout;
2241         }
2242
2243         err = validate_linkmsg(dev, tb);
2244         if (err < 0)
2245                 goto errout;
2246
2247         err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2248 errout:
2249         return err;
2250 }
2251
2252 static int rtnl_group_dellink(const struct net *net, int group)
2253 {
2254         struct net_device *dev, *aux;
2255         LIST_HEAD(list_kill);
2256         bool found = false;
2257
2258         if (!group)
2259                 return -EPERM;
2260
2261         for_each_netdev(net, dev) {
2262                 if (dev->group == group) {
2263                         const struct rtnl_link_ops *ops;
2264
2265                         found = true;
2266                         ops = dev->rtnl_link_ops;
2267                         if (!ops || !ops->dellink)
2268                                 return -EOPNOTSUPP;
2269                 }
2270         }
2271
2272         if (!found)
2273                 return -ENODEV;
2274
2275         for_each_netdev_safe(net, dev, aux) {
2276                 if (dev->group == group) {
2277                         const struct rtnl_link_ops *ops;
2278
2279                         ops = dev->rtnl_link_ops;
2280                         ops->dellink(dev, &list_kill);
2281                 }
2282         }
2283         unregister_netdevice_many(&list_kill);
2284
2285         return 0;
2286 }
2287
2288 int rtnl_delete_link(struct net_device *dev)
2289 {
2290         const struct rtnl_link_ops *ops;
2291         LIST_HEAD(list_kill);
2292
2293         ops = dev->rtnl_link_ops;
2294         if (!ops || !ops->dellink)
2295                 return -EOPNOTSUPP;
2296
2297         ops->dellink(dev, &list_kill);
2298         unregister_netdevice_many(&list_kill);
2299
2300         return 0;
2301 }
2302 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2303
2304 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2305 {
2306         struct net *net = sock_net(skb->sk);
2307         struct net_device *dev;
2308         struct ifinfomsg *ifm;
2309         char ifname[IFNAMSIZ];
2310         struct nlattr *tb[IFLA_MAX+1];
2311         int err;
2312
2313         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2314         if (err < 0)
2315                 return err;
2316
2317         if (tb[IFLA_IFNAME])
2318                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2319
2320         ifm = nlmsg_data(nlh);
2321         if (ifm->ifi_index > 0)
2322                 dev = __dev_get_by_index(net, ifm->ifi_index);
2323         else if (tb[IFLA_IFNAME])
2324                 dev = __dev_get_by_name(net, ifname);
2325         else if (tb[IFLA_GROUP])
2326                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2327         else
2328                 return -EINVAL;
2329
2330         if (!dev)
2331                 return -ENODEV;
2332
2333         return rtnl_delete_link(dev);
2334 }
2335
2336 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2337 {
2338         unsigned int old_flags;
2339         int err;
2340
2341         old_flags = dev->flags;
2342         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2343                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2344                 if (err < 0)
2345                         return err;
2346         }
2347
2348         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2349
2350         __dev_notify_flags(dev, old_flags, ~0U);
2351         return 0;
2352 }
2353 EXPORT_SYMBOL(rtnl_configure_link);
2354
2355 struct net_device *rtnl_create_link(struct net *net,
2356         const char *ifname, unsigned char name_assign_type,
2357         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2358 {
2359         int err;
2360         struct net_device *dev;
2361         unsigned int num_tx_queues = 1;
2362         unsigned int num_rx_queues = 1;
2363
2364         if (tb[IFLA_NUM_TX_QUEUES])
2365                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2366         else if (ops->get_num_tx_queues)
2367                 num_tx_queues = ops->get_num_tx_queues();
2368
2369         if (tb[IFLA_NUM_RX_QUEUES])
2370                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2371         else if (ops->get_num_rx_queues)
2372                 num_rx_queues = ops->get_num_rx_queues();
2373
2374         err = -ENOMEM;
2375         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2376                                ops->setup, num_tx_queues, num_rx_queues);
2377         if (!dev)
2378                 goto err;
2379
2380         dev_net_set(dev, net);
2381         dev->rtnl_link_ops = ops;
2382         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2383
2384         if (tb[IFLA_MTU])
2385                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2386         if (tb[IFLA_ADDRESS]) {
2387                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2388                                 nla_len(tb[IFLA_ADDRESS]));
2389                 dev->addr_assign_type = NET_ADDR_SET;
2390         }
2391         if (tb[IFLA_BROADCAST])
2392                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2393                                 nla_len(tb[IFLA_BROADCAST]));
2394         if (tb[IFLA_TXQLEN])
2395                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2396         if (tb[IFLA_OPERSTATE])
2397                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2398         if (tb[IFLA_LINKMODE])
2399                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2400         if (tb[IFLA_GROUP])
2401                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2402
2403         return dev;
2404
2405 err:
2406         return ERR_PTR(err);
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);
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                 if (err < 0)
2480                         return err;
2481         } else
2482                 memset(linkinfo, 0, sizeof(linkinfo));
2483
2484         if (linkinfo[IFLA_INFO_KIND]) {
2485                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2486                 ops = rtnl_link_ops_get(kind);
2487         } else {
2488                 kind[0] = '\0';
2489                 ops = NULL;
2490         }
2491
2492         if (1) {
2493                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2494                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2495                 struct nlattr **data = NULL;
2496                 struct nlattr **slave_data = NULL;
2497                 struct net *dest_net, *link_net = NULL;
2498
2499                 if (ops) {
2500                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2501                                 err = nla_parse_nested(attr, ops->maxtype,
2502                                                        linkinfo[IFLA_INFO_DATA],
2503                                                        ops->policy);
2504                                 if (err < 0)
2505                                         return err;
2506                                 data = attr;
2507                         }
2508                         if (ops->validate) {
2509                                 err = ops->validate(tb, data);
2510                                 if (err < 0)
2511                                         return err;
2512                         }
2513                 }
2514
2515                 if (m_ops) {
2516                         if (m_ops->slave_maxtype &&
2517                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2518                                 err = nla_parse_nested(slave_attr,
2519                                                        m_ops->slave_maxtype,
2520                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2521                                                        m_ops->slave_policy);
2522                                 if (err < 0)
2523                                         return err;
2524                                 slave_data = slave_attr;
2525                         }
2526                         if (m_ops->slave_validate) {
2527                                 err = m_ops->slave_validate(tb, slave_data);
2528                                 if (err < 0)
2529                                         return err;
2530                         }
2531                 }
2532
2533                 if (dev) {
2534                         int status = 0;
2535
2536                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2537                                 return -EEXIST;
2538                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2539                                 return -EOPNOTSUPP;
2540
2541                         if (linkinfo[IFLA_INFO_DATA]) {
2542                                 if (!ops || ops != dev->rtnl_link_ops ||
2543                                     !ops->changelink)
2544                                         return -EOPNOTSUPP;
2545
2546                                 err = ops->changelink(dev, tb, data);
2547                                 if (err < 0)
2548                                         return err;
2549                                 status |= DO_SETLINK_NOTIFY;
2550                         }
2551
2552                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2553                                 if (!m_ops || !m_ops->slave_changelink)
2554                                         return -EOPNOTSUPP;
2555
2556                                 err = m_ops->slave_changelink(master_dev, dev,
2557                                                               tb, slave_data);
2558                                 if (err < 0)
2559                                         return err;
2560                                 status |= DO_SETLINK_NOTIFY;
2561                         }
2562
2563                         return do_setlink(skb, dev, ifm, tb, ifname, status);
2564                 }
2565
2566                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2567                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2568                                 return rtnl_group_changelink(skb, net,
2569                                                 nla_get_u32(tb[IFLA_GROUP]),
2570                                                 ifm, tb);
2571                         return -ENODEV;
2572                 }
2573
2574                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2575                         return -EOPNOTSUPP;
2576
2577                 if (!ops) {
2578 #ifdef CONFIG_MODULES
2579                         if (kind[0]) {
2580                                 __rtnl_unlock();
2581                                 request_module("rtnl-link-%s", kind);
2582                                 rtnl_lock();
2583                                 ops = rtnl_link_ops_get(kind);
2584                                 if (ops)
2585                                         goto replay;
2586                         }
2587 #endif
2588                         return -EOPNOTSUPP;
2589                 }
2590
2591                 if (!ops->setup)
2592                         return -EOPNOTSUPP;
2593
2594                 if (!ifname[0]) {
2595                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2596                         name_assign_type = NET_NAME_ENUM;
2597                 }
2598
2599                 dest_net = rtnl_link_get_net(net, tb);
2600                 if (IS_ERR(dest_net))
2601                         return PTR_ERR(dest_net);
2602
2603                 err = -EPERM;
2604                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2605                         goto out;
2606
2607                 if (tb[IFLA_LINK_NETNSID]) {
2608                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2609
2610                         link_net = get_net_ns_by_id(dest_net, id);
2611                         if (!link_net) {
2612                                 err =  -EINVAL;
2613                                 goto out;
2614                         }
2615                         err = -EPERM;
2616                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2617                                 goto out;
2618                 }
2619
2620                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2621                                        name_assign_type, ops, tb);
2622                 if (IS_ERR(dev)) {
2623                         err = PTR_ERR(dev);
2624                         goto out;
2625                 }
2626
2627                 dev->ifindex = ifm->ifi_index;
2628
2629                 if (ops->newlink) {
2630                         err = ops->newlink(link_net ? : net, dev, tb, data);
2631                         /* Drivers should call free_netdev() in ->destructor
2632                          * and unregister it on failure after registration
2633                          * so that device could be finally freed in rtnl_unlock.
2634                          */
2635                         if (err < 0) {
2636                                 /* If device is not registered at all, free it now */
2637                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2638                                         free_netdev(dev);
2639                                 goto out;
2640                         }
2641                 } else {
2642                         err = register_netdevice(dev);
2643                         if (err < 0) {
2644                                 free_netdev(dev);
2645                                 goto out;
2646                         }
2647                 }
2648                 err = rtnl_configure_link(dev, ifm);
2649                 if (err < 0)
2650                         goto out_unregister;
2651                 if (link_net) {
2652                         err = dev_change_net_namespace(dev, dest_net, ifname);
2653                         if (err < 0)
2654                                 goto out_unregister;
2655                 }
2656 out:
2657                 if (link_net)
2658                         put_net(link_net);
2659                 put_net(dest_net);
2660                 return err;
2661 out_unregister:
2662                 if (ops->newlink) {
2663                         LIST_HEAD(list_kill);
2664
2665                         ops->dellink(dev, &list_kill);
2666                         unregister_netdevice_many(&list_kill);
2667                 } else {
2668                         unregister_netdevice(dev);
2669                 }
2670                 goto out;
2671         }
2672 }
2673
2674 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2675 {
2676         struct net *net = sock_net(skb->sk);
2677         struct ifinfomsg *ifm;
2678         char ifname[IFNAMSIZ];
2679         struct nlattr *tb[IFLA_MAX+1];
2680         struct net_device *dev = NULL;
2681         struct sk_buff *nskb;
2682         int err;
2683         u32 ext_filter_mask = 0;
2684
2685         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2686         if (err < 0)
2687                 return err;
2688
2689         if (tb[IFLA_IFNAME])
2690                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2691
2692         if (tb[IFLA_EXT_MASK])
2693                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2694
2695         ifm = nlmsg_data(nlh);
2696         if (ifm->ifi_index > 0)
2697                 dev = __dev_get_by_index(net, ifm->ifi_index);
2698         else if (tb[IFLA_IFNAME])
2699                 dev = __dev_get_by_name(net, ifname);
2700         else
2701                 return -EINVAL;
2702
2703         if (dev == NULL)
2704                 return -ENODEV;
2705
2706         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2707         if (nskb == NULL)
2708                 return -ENOBUFS;
2709
2710         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2711                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2712         if (err < 0) {
2713                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2714                 WARN_ON(err == -EMSGSIZE);
2715                 kfree_skb(nskb);
2716         } else
2717                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2718
2719         return err;
2720 }
2721
2722 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2723 {
2724         struct net *net = sock_net(skb->sk);
2725         struct net_device *dev;
2726         struct nlattr *tb[IFLA_MAX+1];
2727         u32 ext_filter_mask = 0;
2728         u16 min_ifinfo_dump_size = 0;
2729         int hdrlen;
2730
2731         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2732         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2733                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2734
2735         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2736                 if (tb[IFLA_EXT_MASK])
2737                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2738         }
2739
2740         if (!ext_filter_mask)
2741                 return NLMSG_GOODSIZE;
2742         /*
2743          * traverse the list of net devices and compute the minimum
2744          * buffer size based upon the filter mask.
2745          */
2746         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2747                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2748                                              if_nlmsg_size(dev,
2749                                                            ext_filter_mask));
2750         }
2751
2752         return nlmsg_total_size(min_ifinfo_dump_size);
2753 }
2754
2755 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2756 {
2757         int idx;
2758         int s_idx = cb->family;
2759
2760         if (s_idx == 0)
2761                 s_idx = 1;
2762         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2763                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2764                 if (idx < s_idx || idx == PF_PACKET)
2765                         continue;
2766                 if (rtnl_msg_handlers[idx] == NULL ||
2767                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2768                         continue;
2769                 if (idx > s_idx) {
2770                         memset(&cb->args[0], 0, sizeof(cb->args));
2771                         cb->prev_seq = 0;
2772                         cb->seq = 0;
2773                 }
2774                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2775                         break;
2776         }
2777         cb->family = idx;
2778
2779         return skb->len;
2780 }
2781
2782 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2783                                        unsigned int change, gfp_t flags)
2784 {
2785         struct net *net = dev_net(dev);
2786         struct sk_buff *skb;
2787         int err = -ENOBUFS;
2788         size_t if_info_size;
2789
2790         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2791         if (skb == NULL)
2792                 goto errout;
2793
2794         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2795         if (err < 0) {
2796                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2797                 WARN_ON(err == -EMSGSIZE);
2798                 kfree_skb(skb);
2799                 goto errout;
2800         }
2801         return skb;
2802 errout:
2803         if (err < 0)
2804                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2805         return NULL;
2806 }
2807
2808 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2809 {
2810         struct net *net = dev_net(dev);
2811
2812         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2813 }
2814
2815 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2816                   gfp_t flags)
2817 {
2818         struct sk_buff *skb;
2819
2820         if (dev->reg_state != NETREG_REGISTERED)
2821                 return;
2822
2823         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2824         if (skb)
2825                 rtmsg_ifinfo_send(skb, dev, flags);
2826 }
2827 EXPORT_SYMBOL(rtmsg_ifinfo);
2828
2829 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2830                                    struct net_device *dev,
2831                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2832                                    int type, unsigned int flags,
2833                                    int nlflags, u16 ndm_state)
2834 {
2835         struct nlmsghdr *nlh;
2836         struct ndmsg *ndm;
2837
2838         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2839         if (!nlh)
2840                 return -EMSGSIZE;
2841
2842         ndm = nlmsg_data(nlh);
2843         ndm->ndm_family  = AF_BRIDGE;
2844         ndm->ndm_pad1    = 0;
2845         ndm->ndm_pad2    = 0;
2846         ndm->ndm_flags   = flags;
2847         ndm->ndm_type    = 0;
2848         ndm->ndm_ifindex = dev->ifindex;
2849         ndm->ndm_state   = ndm_state;
2850
2851         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2852                 goto nla_put_failure;
2853         if (vid)
2854                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2855                         goto nla_put_failure;
2856
2857         nlmsg_end(skb, nlh);
2858         return 0;
2859
2860 nla_put_failure:
2861         nlmsg_cancel(skb, nlh);
2862         return -EMSGSIZE;
2863 }
2864
2865 static inline size_t rtnl_fdb_nlmsg_size(void)
2866 {
2867         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2868                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2869                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2870                0;
2871 }
2872
2873 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2874                             u16 ndm_state)
2875 {
2876         struct net *net = dev_net(dev);
2877         struct sk_buff *skb;
2878         int err = -ENOBUFS;
2879
2880         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2881         if (!skb)
2882                 goto errout;
2883
2884         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2885                                       0, 0, type, NTF_SELF, 0, ndm_state);
2886         if (err < 0) {
2887                 kfree_skb(skb);
2888                 goto errout;
2889         }
2890
2891         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2892         return;
2893 errout:
2894         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2895 }
2896
2897 /**
2898  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2899  */
2900 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2901                      struct nlattr *tb[],
2902                      struct net_device *dev,
2903                      const unsigned char *addr, u16 vid,
2904                      u16 flags)
2905 {
2906         int err = -EINVAL;
2907
2908         /* If aging addresses are supported device will need to
2909          * implement its own handler for this.
2910          */
2911         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2912                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2913                 return err;
2914         }
2915
2916         if (vid) {
2917                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2918                 return err;
2919         }
2920
2921         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2922                 err = dev_uc_add_excl(dev, addr);
2923         else if (is_multicast_ether_addr(addr))
2924                 err = dev_mc_add_excl(dev, addr);
2925
2926         /* Only return duplicate errors if NLM_F_EXCL is set */
2927         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2928                 err = 0;
2929
2930         return err;
2931 }
2932 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2933
2934 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2935 {
2936         u16 vid = 0;
2937
2938         if (vlan_attr) {
2939                 if (nla_len(vlan_attr) != sizeof(u16)) {
2940                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2941                         return -EINVAL;
2942                 }
2943
2944                 vid = nla_get_u16(vlan_attr);
2945
2946                 if (!vid || vid >= VLAN_VID_MASK) {
2947                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2948                                 vid);
2949                         return -EINVAL;
2950                 }
2951         }
2952         *p_vid = vid;
2953         return 0;
2954 }
2955
2956 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2957 {
2958         struct net *net = sock_net(skb->sk);
2959         struct ndmsg *ndm;
2960         struct nlattr *tb[NDA_MAX+1];
2961         struct net_device *dev;
2962         u8 *addr;
2963         u16 vid;
2964         int err;
2965
2966         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2967         if (err < 0)
2968                 return err;
2969
2970         ndm = nlmsg_data(nlh);
2971         if (ndm->ndm_ifindex == 0) {
2972                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2973                 return -EINVAL;
2974         }
2975
2976         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2977         if (dev == NULL) {
2978                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2979                 return -ENODEV;
2980         }
2981
2982         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2983                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2984                 return -EINVAL;
2985         }
2986
2987         addr = nla_data(tb[NDA_LLADDR]);
2988
2989         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2990         if (err)
2991                 return err;
2992
2993         err = -EOPNOTSUPP;
2994
2995         /* Support fdb on master device the net/bridge default case */
2996         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2997             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2998                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2999                 const struct net_device_ops *ops = br_dev->netdev_ops;
3000
3001                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3002                                        nlh->nlmsg_flags);
3003                 if (err)
3004                         goto out;
3005                 else
3006                         ndm->ndm_flags &= ~NTF_MASTER;
3007         }
3008
3009         /* Embedded bridge, macvlan, and any other device support */
3010         if ((ndm->ndm_flags & NTF_SELF)) {
3011                 if (dev->netdev_ops->ndo_fdb_add)
3012                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3013                                                            vid,
3014                                                            nlh->nlmsg_flags);
3015                 else
3016                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3017                                                nlh->nlmsg_flags);
3018
3019                 if (!err) {
3020                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3021                                         ndm->ndm_state);
3022                         ndm->ndm_flags &= ~NTF_SELF;
3023                 }
3024         }
3025 out:
3026         return err;
3027 }
3028
3029 /**
3030  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3031  */
3032 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3033                      struct nlattr *tb[],
3034                      struct net_device *dev,
3035                      const unsigned char *addr, u16 vid)
3036 {
3037         int err = -EINVAL;
3038
3039         /* If aging addresses are supported device will need to
3040          * implement its own handler for this.
3041          */
3042         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3043                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3044                 return err;
3045         }
3046
3047         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3048                 err = dev_uc_del(dev, addr);
3049         else if (is_multicast_ether_addr(addr))
3050                 err = dev_mc_del(dev, addr);
3051
3052         return err;
3053 }
3054 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3055
3056 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3057 {
3058         struct net *net = sock_net(skb->sk);
3059         struct ndmsg *ndm;
3060         struct nlattr *tb[NDA_MAX+1];
3061         struct net_device *dev;
3062         int err = -EINVAL;
3063         __u8 *addr;
3064         u16 vid;
3065
3066         if (!netlink_capable(skb, CAP_NET_ADMIN))
3067                 return -EPERM;
3068
3069         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3070         if (err < 0)
3071                 return err;
3072
3073         ndm = nlmsg_data(nlh);
3074         if (ndm->ndm_ifindex == 0) {
3075                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3076                 return -EINVAL;
3077         }
3078
3079         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3080         if (dev == NULL) {
3081                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3082                 return -ENODEV;
3083         }
3084
3085         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3086                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3087                 return -EINVAL;
3088         }
3089
3090         addr = nla_data(tb[NDA_LLADDR]);
3091
3092         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3093         if (err)
3094                 return err;
3095
3096         err = -EOPNOTSUPP;
3097
3098         /* Support fdb on master device the net/bridge default case */
3099         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3100             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3101                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3102                 const struct net_device_ops *ops = br_dev->netdev_ops;
3103
3104                 if (ops->ndo_fdb_del)
3105                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3106
3107                 if (err)
3108                         goto out;
3109                 else
3110                         ndm->ndm_flags &= ~NTF_MASTER;
3111         }
3112
3113         /* Embedded bridge, macvlan, and any other device support */
3114         if (ndm->ndm_flags & NTF_SELF) {
3115                 if (dev->netdev_ops->ndo_fdb_del)
3116                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3117                                                            vid);
3118                 else
3119                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3120
3121                 if (!err) {
3122                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3123                                         ndm->ndm_state);
3124                         ndm->ndm_flags &= ~NTF_SELF;
3125                 }
3126         }
3127 out:
3128         return err;
3129 }
3130
3131 static int nlmsg_populate_fdb(struct sk_buff *skb,
3132                               struct netlink_callback *cb,
3133                               struct net_device *dev,
3134                               int *idx,
3135                               struct netdev_hw_addr_list *list)
3136 {
3137         struct netdev_hw_addr *ha;
3138         int err;
3139         u32 portid, seq;
3140
3141         portid = NETLINK_CB(cb->skb).portid;
3142         seq = cb->nlh->nlmsg_seq;
3143
3144         list_for_each_entry(ha, &list->list, list) {
3145                 if (*idx < cb->args[2])
3146                         goto skip;
3147
3148                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3149                                               portid, seq,
3150                                               RTM_NEWNEIGH, NTF_SELF,
3151                                               NLM_F_MULTI, NUD_PERMANENT);
3152                 if (err < 0)
3153                         return err;
3154 skip:
3155                 *idx += 1;
3156         }
3157         return 0;
3158 }
3159
3160 /**
3161  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3162  * @nlh: netlink message header
3163  * @dev: netdevice
3164  *
3165  * Default netdevice operation to dump the existing unicast address list.
3166  * Returns number of addresses from list put in skb.
3167  */
3168 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3169                       struct netlink_callback *cb,
3170                       struct net_device *dev,
3171                       struct net_device *filter_dev,
3172                       int *idx)
3173 {
3174         int err;
3175
3176         netif_addr_lock_bh(dev);
3177         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3178         if (err)
3179                 goto out;
3180         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3181 out:
3182         netif_addr_unlock_bh(dev);
3183         return err;
3184 }
3185 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3186
3187 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3188 {
3189         struct net_device *dev;
3190         struct nlattr *tb[IFLA_MAX+1];
3191         struct net_device *br_dev = NULL;
3192         const struct net_device_ops *ops = NULL;
3193         const struct net_device_ops *cops = NULL;
3194         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3195         struct net *net = sock_net(skb->sk);
3196         struct hlist_head *head;
3197         int brport_idx = 0;
3198         int br_idx = 0;
3199         int h, s_h;
3200         int idx = 0, s_idx;
3201         int err = 0;
3202         int fidx = 0;
3203
3204         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3205                         ifla_policy) == 0) {
3206                 if (tb[IFLA_MASTER])
3207                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3208         }
3209
3210         brport_idx = ifm->ifi_index;
3211
3212         if (br_idx) {
3213                 br_dev = __dev_get_by_index(net, br_idx);
3214                 if (!br_dev)
3215                         return -ENODEV;
3216
3217                 ops = br_dev->netdev_ops;
3218         }
3219
3220         s_h = cb->args[0];
3221         s_idx = cb->args[1];
3222
3223         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3224                 idx = 0;
3225                 head = &net->dev_index_head[h];
3226                 hlist_for_each_entry(dev, head, index_hlist) {
3227
3228                         if (brport_idx && (dev->ifindex != brport_idx))
3229                                 continue;
3230
3231                         if (!br_idx) { /* user did not specify a specific bridge */
3232                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3233                                         br_dev = netdev_master_upper_dev_get(dev);
3234                                         cops = br_dev->netdev_ops;
3235                                 }
3236                         } else {
3237                                 if (dev != br_dev &&
3238                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3239                                         continue;
3240
3241                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3242                                     !(dev->priv_flags & IFF_EBRIDGE))
3243                                         continue;
3244                                 cops = ops;
3245                         }
3246
3247                         if (idx < s_idx)
3248                                 goto cont;
3249
3250                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3251                                 if (cops && cops->ndo_fdb_dump) {
3252                                         err = cops->ndo_fdb_dump(skb, cb,
3253                                                                 br_dev, dev,
3254                                                                 &fidx);
3255                                         if (err == -EMSGSIZE)
3256                                                 goto out;
3257                                 }
3258                         }
3259
3260                         if (dev->netdev_ops->ndo_fdb_dump)
3261                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3262                                                                     dev, NULL,
3263                                                                     &fidx);
3264                         else
3265                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3266                                                         &fidx);
3267                         if (err == -EMSGSIZE)
3268                                 goto out;
3269
3270                         cops = NULL;
3271
3272                         /* reset fdb offset to 0 for rest of the interfaces */
3273                         cb->args[2] = 0;
3274                         fidx = 0;
3275 cont:
3276                         idx++;
3277                 }
3278         }
3279
3280 out:
3281         cb->args[0] = h;
3282         cb->args[1] = idx;
3283         cb->args[2] = fidx;
3284
3285         return skb->len;
3286 }
3287
3288 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3289                                unsigned int attrnum, unsigned int flag)
3290 {
3291         if (mask & flag)
3292                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3293         return 0;
3294 }
3295
3296 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3297                             struct net_device *dev, u16 mode,
3298                             u32 flags, u32 mask, int nlflags,
3299                             u32 filter_mask,
3300                             int (*vlan_fill)(struct sk_buff *skb,
3301                                              struct net_device *dev,
3302                                              u32 filter_mask))
3303 {
3304         struct nlmsghdr *nlh;
3305         struct ifinfomsg *ifm;
3306         struct nlattr *br_afspec;
3307         struct nlattr *protinfo;
3308         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3309         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3310         int err = 0;
3311
3312         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3313         if (nlh == NULL)
3314                 return -EMSGSIZE;
3315
3316         ifm = nlmsg_data(nlh);
3317         ifm->ifi_family = AF_BRIDGE;
3318         ifm->__ifi_pad = 0;
3319         ifm->ifi_type = dev->type;
3320         ifm->ifi_index = dev->ifindex;
3321         ifm->ifi_flags = dev_get_flags(dev);
3322         ifm->ifi_change = 0;
3323
3324
3325         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3326             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3327             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3328             (br_dev &&
3329              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3330             (dev->addr_len &&
3331              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3332             (dev->ifindex != dev_get_iflink(dev) &&
3333              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3334                 goto nla_put_failure;
3335
3336         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3337         if (!br_afspec)
3338                 goto nla_put_failure;
3339
3340         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3341                 nla_nest_cancel(skb, br_afspec);
3342                 goto nla_put_failure;
3343         }
3344
3345         if (mode != BRIDGE_MODE_UNDEF) {
3346                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3347                         nla_nest_cancel(skb, br_afspec);
3348                         goto nla_put_failure;
3349                 }
3350         }
3351         if (vlan_fill) {
3352                 err = vlan_fill(skb, dev, filter_mask);
3353                 if (err) {
3354                         nla_nest_cancel(skb, br_afspec);
3355                         goto nla_put_failure;
3356                 }
3357         }
3358         nla_nest_end(skb, br_afspec);
3359
3360         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3361         if (!protinfo)
3362                 goto nla_put_failure;
3363
3364         if (brport_nla_put_flag(skb, flags, mask,
3365                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3366             brport_nla_put_flag(skb, flags, mask,
3367                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3368             brport_nla_put_flag(skb, flags, mask,
3369                                 IFLA_BRPORT_FAST_LEAVE,
3370                                 BR_MULTICAST_FAST_LEAVE) ||
3371             brport_nla_put_flag(skb, flags, mask,
3372                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3373             brport_nla_put_flag(skb, flags, mask,
3374                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3375             brport_nla_put_flag(skb, flags, mask,
3376                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3377             brport_nla_put_flag(skb, flags, mask,
3378                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3379             brport_nla_put_flag(skb, flags, mask,
3380                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3381                 nla_nest_cancel(skb, protinfo);
3382                 goto nla_put_failure;
3383         }
3384
3385         nla_nest_end(skb, protinfo);
3386
3387         nlmsg_end(skb, nlh);
3388         return 0;
3389 nla_put_failure:
3390         nlmsg_cancel(skb, nlh);
3391         return err ? err : -EMSGSIZE;
3392 }
3393 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3394
3395 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3396 {
3397         struct net *net = sock_net(skb->sk);
3398         struct net_device *dev;
3399         int idx = 0;
3400         u32 portid = NETLINK_CB(cb->skb).portid;
3401         u32 seq = cb->nlh->nlmsg_seq;
3402         u32 filter_mask = 0;
3403         int err;
3404
3405         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3406                 struct nlattr *extfilt;
3407
3408                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3409                                           IFLA_EXT_MASK);
3410                 if (extfilt) {
3411                         if (nla_len(extfilt) < sizeof(filter_mask))
3412                                 return -EINVAL;
3413
3414                         filter_mask = nla_get_u32(extfilt);
3415                 }
3416         }
3417
3418         rcu_read_lock();
3419         for_each_netdev_rcu(net, dev) {
3420                 const struct net_device_ops *ops = dev->netdev_ops;
3421                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3422
3423                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3424                         if (idx >= cb->args[0]) {
3425                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3426                                                 skb, portid, seq, dev,
3427                                                 filter_mask, NLM_F_MULTI);
3428                                 if (err < 0 && err != -EOPNOTSUPP)
3429                                         break;
3430                         }
3431                         idx++;
3432                 }
3433
3434                 if (ops->ndo_bridge_getlink) {
3435                         if (idx >= cb->args[0]) {
3436                                 err = ops->ndo_bridge_getlink(skb, portid,
3437                                                               seq, dev,
3438                                                               filter_mask,
3439                                                               NLM_F_MULTI);
3440                                 if (err < 0 && err != -EOPNOTSUPP)
3441                                         break;
3442                         }
3443                         idx++;
3444                 }
3445         }
3446         rcu_read_unlock();
3447         cb->args[0] = idx;
3448
3449         return skb->len;
3450 }
3451
3452 static inline size_t bridge_nlmsg_size(void)
3453 {
3454         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3455                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3456                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3457                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3458                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3459                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3460                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3461                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3462                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3463                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3464                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3465 }
3466
3467 static int rtnl_bridge_notify(struct net_device *dev)
3468 {
3469         struct net *net = dev_net(dev);
3470         struct sk_buff *skb;
3471         int err = -EOPNOTSUPP;
3472
3473         if (!dev->netdev_ops->ndo_bridge_getlink)
3474                 return 0;
3475
3476         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3477         if (!skb) {
3478                 err = -ENOMEM;
3479                 goto errout;
3480         }
3481
3482         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3483         if (err < 0)
3484                 goto errout;
3485
3486         if (!skb->len)
3487                 goto errout;
3488
3489         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3490         return 0;
3491 errout:
3492         WARN_ON(err == -EMSGSIZE);
3493         kfree_skb(skb);
3494         if (err)
3495                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3496         return err;
3497 }
3498
3499 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3500 {
3501         struct net *net = sock_net(skb->sk);
3502         struct ifinfomsg *ifm;
3503         struct net_device *dev;
3504         struct nlattr *br_spec, *attr = NULL;
3505         int rem, err = -EOPNOTSUPP;
3506         u16 flags = 0;
3507         bool have_flags = false;
3508
3509         if (nlmsg_len(nlh) < sizeof(*ifm))
3510                 return -EINVAL;
3511
3512         ifm = nlmsg_data(nlh);
3513         if (ifm->ifi_family != AF_BRIDGE)
3514                 return -EPFNOSUPPORT;
3515
3516         dev = __dev_get_by_index(net, ifm->ifi_index);
3517         if (!dev) {
3518                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3519                 return -ENODEV;
3520         }
3521
3522         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3523         if (br_spec) {
3524                 nla_for_each_nested(attr, br_spec, rem) {
3525                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3526                                 if (nla_len(attr) < sizeof(flags))
3527                                         return -EINVAL;
3528
3529                                 have_flags = true;
3530                                 flags = nla_get_u16(attr);
3531                                 break;
3532                         }
3533                 }
3534         }
3535
3536         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3537                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3538
3539                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3540                         err = -EOPNOTSUPP;
3541                         goto out;
3542                 }
3543
3544                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3545                 if (err)
3546                         goto out;
3547
3548                 flags &= ~BRIDGE_FLAGS_MASTER;
3549         }
3550
3551         if ((flags & BRIDGE_FLAGS_SELF)) {
3552                 if (!dev->netdev_ops->ndo_bridge_setlink)
3553                         err = -EOPNOTSUPP;
3554                 else
3555                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3556                                                                   flags);
3557                 if (!err) {
3558                         flags &= ~BRIDGE_FLAGS_SELF;
3559
3560                         /* Generate event to notify upper layer of bridge
3561                          * change
3562                          */
3563                         err = rtnl_bridge_notify(dev);
3564                 }
3565         }
3566
3567         if (have_flags)
3568                 memcpy(nla_data(attr), &flags, sizeof(flags));
3569 out:
3570         return err;
3571 }
3572
3573 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3574 {
3575         struct net *net = sock_net(skb->sk);
3576         struct ifinfomsg *ifm;
3577         struct net_device *dev;
3578         struct nlattr *br_spec, *attr = NULL;
3579         int rem, err = -EOPNOTSUPP;
3580         u16 flags = 0;
3581         bool have_flags = false;
3582
3583         if (nlmsg_len(nlh) < sizeof(*ifm))
3584                 return -EINVAL;
3585
3586         ifm = nlmsg_data(nlh);
3587         if (ifm->ifi_family != AF_BRIDGE)
3588                 return -EPFNOSUPPORT;
3589
3590         dev = __dev_get_by_index(net, ifm->ifi_index);
3591         if (!dev) {
3592                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3593                 return -ENODEV;
3594         }
3595
3596         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3597         if (br_spec) {
3598                 nla_for_each_nested(attr, br_spec, rem) {
3599                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3600                                 if (nla_len(attr) < sizeof(flags))
3601                                         return -EINVAL;
3602
3603                                 have_flags = true;
3604                                 flags = nla_get_u16(attr);
3605                                 break;
3606                         }
3607                 }
3608         }
3609
3610         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3611                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3612
3613                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3614                         err = -EOPNOTSUPP;
3615                         goto out;
3616                 }
3617
3618                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3619                 if (err)
3620                         goto out;
3621
3622                 flags &= ~BRIDGE_FLAGS_MASTER;
3623         }
3624
3625         if ((flags & BRIDGE_FLAGS_SELF)) {
3626                 if (!dev->netdev_ops->ndo_bridge_dellink)
3627                         err = -EOPNOTSUPP;
3628                 else
3629                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3630                                                                   flags);
3631
3632                 if (!err) {
3633                         flags &= ~BRIDGE_FLAGS_SELF;
3634
3635                         /* Generate event to notify upper layer of bridge
3636                          * change
3637                          */
3638                         err = rtnl_bridge_notify(dev);
3639                 }
3640         }
3641
3642         if (have_flags)
3643                 memcpy(nla_data(attr), &flags, sizeof(flags));
3644 out:
3645         return err;
3646 }
3647
3648 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3649 {
3650         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3651                (!idxattr || idxattr == attrid);
3652 }
3653
3654 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3655 static int rtnl_get_offload_stats_attr_size(int attr_id)
3656 {
3657         switch (attr_id) {
3658         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3659                 return sizeof(struct rtnl_link_stats64);
3660         }
3661
3662         return 0;
3663 }
3664
3665 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3666                                   int *prividx)
3667 {
3668         struct nlattr *attr = NULL;
3669         int attr_id, size;
3670         void *attr_data;
3671         int err;
3672
3673         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3674               dev->netdev_ops->ndo_get_offload_stats))
3675                 return -ENODATA;
3676
3677         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3678              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3679                 if (attr_id < *prividx)
3680                         continue;
3681
3682                 size = rtnl_get_offload_stats_attr_size(attr_id);
3683                 if (!size)
3684                         continue;
3685
3686                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3687                         continue;
3688
3689                 attr = nla_reserve_64bit(skb, attr_id, size,
3690                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3691                 if (!attr)
3692                         goto nla_put_failure;
3693
3694                 attr_data = nla_data(attr);
3695                 memset(attr_data, 0, size);
3696                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3697                                                              attr_data);
3698                 if (err)
3699                         goto get_offload_stats_failure;
3700         }
3701
3702         if (!attr)
3703                 return -ENODATA;
3704
3705         *prividx = 0;
3706         return 0;
3707
3708 nla_put_failure:
3709         err = -EMSGSIZE;
3710 get_offload_stats_failure:
3711         *prividx = attr_id;
3712         return err;
3713 }
3714
3715 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3716 {
3717         int nla_size = 0;
3718         int attr_id;
3719         int size;
3720
3721         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3722               dev->netdev_ops->ndo_get_offload_stats))
3723                 return 0;
3724
3725         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3726              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3727                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3728                         continue;
3729                 size = rtnl_get_offload_stats_attr_size(attr_id);
3730                 nla_size += nla_total_size_64bit(size);
3731         }
3732
3733         if (nla_size != 0)
3734                 nla_size += nla_total_size(0);
3735
3736         return nla_size;
3737 }
3738
3739 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3740                                int type, u32 pid, u32 seq, u32 change,
3741                                unsigned int flags, unsigned int filter_mask,
3742                                int *idxattr, int *prividx)
3743 {
3744         struct if_stats_msg *ifsm;
3745         struct nlmsghdr *nlh;
3746         struct nlattr *attr;
3747         int s_prividx = *prividx;
3748         int err;
3749
3750         ASSERT_RTNL();
3751
3752         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3753         if (!nlh)
3754                 return -EMSGSIZE;
3755
3756         ifsm = nlmsg_data(nlh);
3757         ifsm->ifindex = dev->ifindex;
3758         ifsm->filter_mask = filter_mask;
3759
3760         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3761                 struct rtnl_link_stats64 *sp;
3762
3763                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3764                                          sizeof(struct rtnl_link_stats64),
3765                                          IFLA_STATS_UNSPEC);
3766                 if (!attr)
3767                         goto nla_put_failure;
3768
3769                 sp = nla_data(attr);
3770                 dev_get_stats(dev, sp);
3771         }
3772
3773         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3774                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3775
3776                 if (ops && ops->fill_linkxstats) {
3777                         *idxattr = IFLA_STATS_LINK_XSTATS;
3778                         attr = nla_nest_start(skb,
3779                                               IFLA_STATS_LINK_XSTATS);
3780                         if (!attr)
3781                                 goto nla_put_failure;
3782
3783                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3784                         nla_nest_end(skb, attr);
3785                         if (err)
3786                                 goto nla_put_failure;
3787                         *idxattr = 0;
3788                 }
3789         }
3790
3791         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3792                              *idxattr)) {
3793                 const struct rtnl_link_ops *ops = NULL;
3794                 const struct net_device *master;
3795
3796                 master = netdev_master_upper_dev_get(dev);
3797                 if (master)
3798                         ops = master->rtnl_link_ops;
3799                 if (ops && ops->fill_linkxstats) {
3800                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3801                         attr = nla_nest_start(skb,
3802                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3803                         if (!attr)
3804                                 goto nla_put_failure;
3805
3806                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3807                         nla_nest_end(skb, attr);
3808                         if (err)
3809                                 goto nla_put_failure;
3810                         *idxattr = 0;
3811                 }
3812         }
3813
3814         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3815                              *idxattr)) {
3816                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3817                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3818                 if (!attr)
3819                         goto nla_put_failure;
3820
3821                 err = rtnl_get_offload_stats(skb, dev, prividx);
3822                 if (err == -ENODATA)
3823                         nla_nest_cancel(skb, attr);
3824                 else
3825                         nla_nest_end(skb, attr);
3826
3827                 if (err && err != -ENODATA)
3828                         goto nla_put_failure;
3829                 *idxattr = 0;
3830         }
3831
3832         nlmsg_end(skb, nlh);
3833
3834         return 0;
3835
3836 nla_put_failure:
3837         /* not a multi message or no progress mean a real error */
3838         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3839                 nlmsg_cancel(skb, nlh);
3840         else
3841                 nlmsg_end(skb, nlh);
3842
3843         return -EMSGSIZE;
3844 }
3845
3846 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3847                                   u32 filter_mask)
3848 {
3849         size_t size = 0;
3850
3851         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3852                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3853
3854         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3855                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3856                 int attr = IFLA_STATS_LINK_XSTATS;
3857
3858                 if (ops && ops->get_linkxstats_size) {
3859                         size += nla_total_size(ops->get_linkxstats_size(dev,
3860                                                                         attr));
3861                         /* for IFLA_STATS_LINK_XSTATS */
3862                         size += nla_total_size(0);
3863                 }
3864         }
3865
3866         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3867                 struct net_device *_dev = (struct net_device *)dev;
3868                 const struct rtnl_link_ops *ops = NULL;
3869                 const struct net_device *master;
3870
3871                 /* netdev_master_upper_dev_get can't take const */
3872                 master = netdev_master_upper_dev_get(_dev);
3873                 if (master)
3874                         ops = master->rtnl_link_ops;
3875                 if (ops && ops->get_linkxstats_size) {
3876                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3877
3878                         size += nla_total_size(ops->get_linkxstats_size(dev,
3879                                                                         attr));
3880                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3881                         size += nla_total_size(0);
3882                 }
3883         }
3884
3885         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3886                 size += rtnl_get_offload_stats_size(dev);
3887
3888         return size;
3889 }
3890
3891 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3892 {
3893         struct net *net = sock_net(skb->sk);
3894         struct net_device *dev = NULL;
3895         int idxattr = 0, prividx = 0;
3896         struct if_stats_msg *ifsm;
3897         struct sk_buff *nskb;
3898         u32 filter_mask;
3899         int err;
3900
3901         if (nlmsg_len(nlh) < sizeof(*ifsm))
3902                 return -EINVAL;
3903
3904         ifsm = nlmsg_data(nlh);
3905         if (ifsm->ifindex > 0)
3906                 dev = __dev_get_by_index(net, ifsm->ifindex);
3907         else
3908                 return -EINVAL;
3909
3910         if (!dev)
3911                 return -ENODEV;
3912
3913         filter_mask = ifsm->filter_mask;
3914         if (!filter_mask)
3915                 return -EINVAL;
3916
3917         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3918         if (!nskb)
3919                 return -ENOBUFS;
3920
3921         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3922                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3923                                   0, filter_mask, &idxattr, &prividx);
3924         if (err < 0) {
3925                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3926                 WARN_ON(err == -EMSGSIZE);
3927                 kfree_skb(nskb);
3928         } else {
3929                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3930         }
3931
3932         return err;
3933 }
3934
3935 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3936 {
3937         int h, s_h, err, s_idx, s_idxattr, s_prividx;
3938         struct net *net = sock_net(skb->sk);
3939         unsigned int flags = NLM_F_MULTI;
3940         struct if_stats_msg *ifsm;
3941         struct hlist_head *head;
3942         struct net_device *dev;
3943         u32 filter_mask = 0;
3944         int idx = 0;
3945
3946         s_h = cb->args[0];
3947         s_idx = cb->args[1];
3948         s_idxattr = cb->args[2];
3949         s_prividx = cb->args[3];
3950
3951         cb->seq = net->dev_base_seq;
3952
3953         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
3954                 return -EINVAL;
3955
3956         ifsm = nlmsg_data(cb->nlh);
3957         filter_mask = ifsm->filter_mask;
3958         if (!filter_mask)
3959                 return -EINVAL;
3960
3961         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3962                 idx = 0;
3963                 head = &net->dev_index_head[h];
3964                 hlist_for_each_entry(dev, head, index_hlist) {
3965                         if (idx < s_idx)
3966                                 goto cont;
3967                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3968                                                   NETLINK_CB(cb->skb).portid,
3969                                                   cb->nlh->nlmsg_seq, 0,
3970                                                   flags, filter_mask,
3971                                                   &s_idxattr, &s_prividx);
3972                         /* If we ran out of room on the first message,
3973                          * we're in trouble
3974                          */
3975                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3976
3977                         if (err < 0)
3978                                 goto out;
3979                         s_prividx = 0;
3980                         s_idxattr = 0;
3981                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3982 cont:
3983                         idx++;
3984                 }
3985         }
3986 out:
3987         cb->args[3] = s_prividx;
3988         cb->args[2] = s_idxattr;
3989         cb->args[1] = idx;
3990         cb->args[0] = h;
3991
3992         return skb->len;
3993 }
3994
3995 /* Process one rtnetlink message. */
3996
3997 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3998 {
3999         struct net *net = sock_net(skb->sk);
4000         rtnl_doit_func doit;
4001         int kind;
4002         int family;
4003         int type;
4004         int err;
4005
4006         type = nlh->nlmsg_type;
4007         if (type > RTM_MAX)
4008                 return -EOPNOTSUPP;
4009
4010         type -= RTM_BASE;
4011
4012         /* All the messages must have at least 1 byte length */
4013         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4014                 return 0;
4015
4016         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4017         kind = type&3;
4018
4019         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4020                 return -EPERM;
4021
4022         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4023                 struct sock *rtnl;
4024                 rtnl_dumpit_func dumpit;
4025                 rtnl_calcit_func calcit;
4026                 u16 min_dump_alloc = 0;
4027
4028                 dumpit = rtnl_get_dumpit(family, type);
4029                 if (dumpit == NULL)
4030                         return -EOPNOTSUPP;
4031                 calcit = rtnl_get_calcit(family, type);
4032                 if (calcit)
4033                         min_dump_alloc = calcit(skb, nlh);
4034
4035                 __rtnl_unlock();
4036                 rtnl = net->rtnl;
4037                 {
4038                         struct netlink_dump_control c = {
4039                                 .dump           = dumpit,
4040                                 .min_dump_alloc = min_dump_alloc,
4041                         };
4042                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4043                 }
4044                 rtnl_lock();
4045                 return err;
4046         }
4047
4048         doit = rtnl_get_doit(family, type);
4049         if (doit == NULL)
4050                 return -EOPNOTSUPP;
4051
4052         return doit(skb, nlh);
4053 }
4054
4055 static void rtnetlink_rcv(struct sk_buff *skb)
4056 {
4057         rtnl_lock();
4058         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4059         rtnl_unlock();
4060 }
4061
4062 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4063 {
4064         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4065
4066         switch (event) {
4067         case NETDEV_UP:
4068         case NETDEV_DOWN:
4069         case NETDEV_PRE_UP:
4070         case NETDEV_POST_INIT:
4071         case NETDEV_REGISTER:
4072         case NETDEV_CHANGE:
4073         case NETDEV_PRE_TYPE_CHANGE:
4074         case NETDEV_GOING_DOWN:
4075         case NETDEV_UNREGISTER:
4076         case NETDEV_UNREGISTER_FINAL:
4077         case NETDEV_RELEASE:
4078         case NETDEV_JOIN:
4079         case NETDEV_BONDING_INFO:
4080                 break;
4081         default:
4082                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4083                 break;
4084         }
4085         return NOTIFY_DONE;
4086 }
4087
4088 static struct notifier_block rtnetlink_dev_notifier = {
4089         .notifier_call  = rtnetlink_event,
4090 };
4091
4092
4093 static int __net_init rtnetlink_net_init(struct net *net)
4094 {
4095         struct sock *sk;
4096         struct netlink_kernel_cfg cfg = {
4097                 .groups         = RTNLGRP_MAX,
4098                 .input          = rtnetlink_rcv,
4099                 .cb_mutex       = &rtnl_mutex,
4100                 .flags          = NL_CFG_F_NONROOT_RECV,
4101         };
4102
4103         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4104         if (!sk)
4105                 return -ENOMEM;
4106         net->rtnl = sk;
4107         return 0;
4108 }
4109
4110 static void __net_exit rtnetlink_net_exit(struct net *net)
4111 {
4112         netlink_kernel_release(net->rtnl);
4113         net->rtnl = NULL;
4114 }
4115
4116 static struct pernet_operations rtnetlink_net_ops = {
4117         .init = rtnetlink_net_init,
4118         .exit = rtnetlink_net_exit,
4119 };
4120
4121 void __init rtnetlink_init(void)
4122 {
4123         if (register_pernet_subsys(&rtnetlink_net_ops))
4124                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4125
4126         register_netdevice_notifier(&rtnetlink_dev_notifier);
4127
4128         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4129                       rtnl_dump_ifinfo, rtnl_calcit);
4130         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4131         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4132         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4133
4134         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4135         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4136
4137         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4138         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4139         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4140
4141         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4142         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4143         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4144
4145         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4146                       NULL);
4147 }