net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/errno.h>
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/kernel.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/mm.h>
  30 #include <linux/slab.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/capability.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/init.h>
  35 #include <linux/security.h>
  36 #include <linux/mutex.h>
  37 #include <linux/if_addr.h>
  38 #include <linux/if_bridge.h>
  39 #include <linux/if_vlan.h>
  40 #include <linux/pci.h>
  41 #include <linux/etherdevice.h>
  42
  43 #include <linux/uaccess.h>
  44
  45 #include <linux/inet.h>
  46 #include <linux/netdevice.h>
  47 #include <net/switchdev.h>
  48 #include <net/ip.h>
  49 #include <net/protocol.h>
  50 #include <net/arp.h>
  51 #include <net/route.h>
  52 #include <net/udp.h>
  53 #include <net/tcp.h>
  54 #include <net/sock.h>
  55 #include <net/pkt_sched.h>
  56 #include <net/fib_rules.h>
  57 #include <net/rtnetlink.h>
  58 #include <net/net_namespace.h>
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         rtnl_calcit_func        calcit;
  64 };
  65
  66 static DEFINE_MUTEX(rtnl_mutex);
  67
  68 void rtnl_lock(void)
  69 {
  70         mutex_lock(&rtnl_mutex);
  71 }
  72 EXPORT_SYMBOL(rtnl_lock);
  73
  74 static struct sk_buff *defer_kfree_skb_list;
  75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  76 {
  77         if (head && tail) {
  78                 tail->next = defer_kfree_skb_list;
  79                 defer_kfree_skb_list = head;
  80         }
  81 }
  82 EXPORT_SYMBOL(rtnl_kfree_skbs);
  83
  84 void __rtnl_unlock(void)
  85 {
  86         struct sk_buff *head = defer_kfree_skb_list;
  87
  88         defer_kfree_skb_list = NULL;
  89
  90         mutex_unlock(&rtnl_mutex);
  91
  92         while (head) {
  93                 struct sk_buff *next = head->next;
  94
  95                 kfree_skb(head);
  96                 cond_resched();
  97                 head = next;
  98         }
  99 }
 100
 101 void rtnl_unlock(void)
 102 {
 103         /* This fellow will unlock it for us. */
 104         netdev_run_todo();
 105 }
 106 EXPORT_SYMBOL(rtnl_unlock);
 107
 108 int rtnl_trylock(void)
 109 {
 110         return mutex_trylock(&rtnl_mutex);
 111 }
 112 EXPORT_SYMBOL(rtnl_trylock);
 113
 114 int rtnl_is_locked(void)
 115 {
 116         return mutex_is_locked(&rtnl_mutex);
 117 }
 118 EXPORT_SYMBOL(rtnl_is_locked);
 119
 120 #ifdef CONFIG_PROVE_LOCKING
 121 bool lockdep_rtnl_is_held(void)
 122 {
 123         return lockdep_is_held(&rtnl_mutex);
 124 }
 125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 127
 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 129
 130 static inline int rtm_msgindex(int msgtype)
 131 {
 132         int msgindex = msgtype - RTM_BASE;
 133
 134         /*
 135          * msgindex < 0 implies someone tried to register a netlink
 136          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 137          * the message type has not been added to linux/rtnetlink.h
 138          */
 139         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 140
 141         return msgindex;
 142 }
 143
 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 145 {
 146         struct rtnl_link *tab;
 147
 148         if (protocol <= RTNL_FAMILY_MAX)
 149                 tab = rtnl_msg_handlers[protocol];
 150         else
 151                 tab = NULL;
 152
 153         if (tab == NULL || tab[msgindex].doit == NULL)
 154                 tab = rtnl_msg_handlers[PF_UNSPEC];
 155
 156         return tab[msgindex].doit;
 157 }
 158
 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 160 {
 161         struct rtnl_link *tab;
 162
 163         if (protocol <= RTNL_FAMILY_MAX)
 164                 tab = rtnl_msg_handlers[protocol];
 165         else
 166                 tab = NULL;
 167
 168         if (tab == NULL || tab[msgindex].dumpit == NULL)
 169                 tab = rtnl_msg_handlers[PF_UNSPEC];
 170
 171         return tab[msgindex].dumpit;
 172 }
 173
 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 175 {
 176         struct rtnl_link *tab;
 177
 178         if (protocol <= RTNL_FAMILY_MAX)
 179                 tab = rtnl_msg_handlers[protocol];
 180         else
 181                 tab = NULL;
 182
 183         if (tab == NULL || tab[msgindex].calcit == NULL)
 184                 tab = rtnl_msg_handlers[PF_UNSPEC];
 185
 186         return tab[msgindex].calcit;
 187 }
 188
 189 /**
 190  * __rtnl_register - Register a rtnetlink message type
 191  * @protocol: Protocol family or PF_UNSPEC
 192  * @msgtype: rtnetlink message type
 193  * @doit: Function pointer called for each request message
 194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 195  * @calcit: Function pointer to calc size of dump message
 196  *
 197  * Registers the specified function pointers (at least one of them has
 198  * to be non-NULL) to be called whenever a request message for the
 199  * specified protocol family and message type is received.
 200  *
 201  * The special protocol family PF_UNSPEC may be used to define fallback
 202  * function pointers for the case when no entry for the specific protocol
 203  * family exists.
 204  *
 205  * Returns 0 on success or a negative error code.
 206  */
 207 int __rtnl_register(int protocol, int msgtype,
 208                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 209                     rtnl_calcit_func calcit)
 210 {
 211         struct rtnl_link *tab;
 212         int msgindex;
 213
 214         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 215         msgindex = rtm_msgindex(msgtype);
 216
 217         tab = rtnl_msg_handlers[protocol];
 218         if (tab == NULL) {
 219                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 220                 if (tab == NULL)
 221                         return -ENOBUFS;
 222
 223                 rtnl_msg_handlers[protocol] = tab;
 224         }
 225
 226         if (doit)
 227                 tab[msgindex].doit = doit;
 228
 229         if (dumpit)
 230                 tab[msgindex].dumpit = dumpit;
 231
 232         if (calcit)
 233                 tab[msgindex].calcit = calcit;
 234
 235         return 0;
 236 }
 237 EXPORT_SYMBOL_GPL(__rtnl_register);
 238
 239 /**
 240  * rtnl_register - Register a rtnetlink message type
 241  *
 242  * Identical to __rtnl_register() but panics on failure. This is useful
 243  * as failure of this function is very unlikely, it can only happen due
 244  * to lack of memory when allocating the chain to store all message
 245  * handlers for a protocol. Meant for use in init functions where lack
 246  * of memory implies no sense in continuing.
 247  */
 248 void rtnl_register(int protocol, int msgtype,
 249                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 250                    rtnl_calcit_func calcit)
 251 {
 252         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 253                 panic("Unable to register rtnetlink message handler, "
 254                       "protocol = %d, message type = %d\n",
 255                       protocol, msgtype);
 256 }
 257 EXPORT_SYMBOL_GPL(rtnl_register);
 258
 259 /**
 260  * rtnl_unregister - Unregister a rtnetlink message type
 261  * @protocol: Protocol family or PF_UNSPEC
 262  * @msgtype: rtnetlink message type
 263  *
 264  * Returns 0 on success or a negative error code.
 265  */
 266 int rtnl_unregister(int protocol, int msgtype)
 267 {
 268         int msgindex;
 269
 270         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 271         msgindex = rtm_msgindex(msgtype);
 272
 273         if (rtnl_msg_handlers[protocol] == NULL)
 274                 return -ENOENT;
 275
 276         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 277         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 278         rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
 279
 280         return 0;
 281 }
 282 EXPORT_SYMBOL_GPL(rtnl_unregister);
 283
 284 /**
 285  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 286  * @protocol : Protocol family or PF_UNSPEC
 287  *
 288  * Identical to calling rtnl_unregster() for all registered message types
 289  * of a certain protocol family.
 290  */
 291 void rtnl_unregister_all(int protocol)
 292 {
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294
 295         kfree(rtnl_msg_handlers[protocol]);
 296         rtnl_msg_handlers[protocol] = NULL;
 297 }
 298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 299
 300 static LIST_HEAD(link_ops);
 301
 302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 303 {
 304         const struct rtnl_link_ops *ops;
 305
 306         list_for_each_entry(ops, &link_ops, list) {
 307                 if (!strcmp(ops->kind, kind))
 308                         return ops;
 309         }
 310         return NULL;
 311 }
 312
 313 /**
 314  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 315  * @ops: struct rtnl_link_ops * to register
 316  *
 317  * The caller must hold the rtnl_mutex. This function should be used
 318  * by drivers that create devices during module initialization. It
 319  * must be called before registering the devices.
 320  *
 321  * Returns 0 on success or a negative error code.
 322  */
 323 int __rtnl_link_register(struct rtnl_link_ops *ops)
 324 {
 325         if (rtnl_link_ops_get(ops->kind))
 326                 return -EEXIST;
 327
 328         /* The check for setup is here because if ops
 329          * does not have that filled up, it is not possible
 330          * to use the ops for creating device. So do not
 331          * fill up dellink as well. That disables rtnl_dellink.
 332          */
 333         if (ops->setup && !ops->dellink)
 334                 ops->dellink = unregister_netdevice_queue;
 335
 336         list_add_tail(&ops->list, &link_ops);
 337         return 0;
 338 }
 339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 340
 341 /**
 342  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 343  * @ops: struct rtnl_link_ops * to register
 344  *
 345  * Returns 0 on success or a negative error code.
 346  */
 347 int rtnl_link_register(struct rtnl_link_ops *ops)
 348 {
 349         int err;
 350
 351         rtnl_lock();
 352         err = __rtnl_link_register(ops);
 353         rtnl_unlock();
 354         return err;
 355 }
 356 EXPORT_SYMBOL_GPL(rtnl_link_register);
 357
 358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 359 {
 360         struct net_device *dev;
 361         LIST_HEAD(list_kill);
 362
 363         for_each_netdev(net, dev) {
 364                 if (dev->rtnl_link_ops == ops)
 365                         ops->dellink(dev, &list_kill);
 366         }
 367         unregister_netdevice_many(&list_kill);
 368 }
 369
 370 /**
 371  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 372  * @ops: struct rtnl_link_ops * to unregister
 373  *
 374  * The caller must hold the rtnl_mutex.
 375  */
 376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 377 {
 378         struct net *net;
 379
 380         for_each_net(net) {
 381                 __rtnl_kill_links(net, ops);
 382         }
 383         list_del(&ops->list);
 384 }
 385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 386
 387 /* Return with the rtnl_lock held when there are no network
 388  * devices unregistering in any network namespace.
 389  */
 390 static void rtnl_lock_unregistering_all(void)
 391 {
 392         struct net *net;
 393         bool unregistering;
 394         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 395
 396         add_wait_queue(&netdev_unregistering_wq, &wait);
 397         for (;;) {
 398                 unregistering = false;
 399                 rtnl_lock();
 400                 for_each_net(net) {
 401                         if (net->dev_unreg_count > 0) {
 402                                 unregistering = true;
 403                                 break;
 404                         }
 405                 }
 406                 if (!unregistering)
 407                         break;
 408                 __rtnl_unlock();
 409
 410                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 411         }
 412         remove_wait_queue(&netdev_unregistering_wq, &wait);
 413 }
 414
 415 /**
 416  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 417  * @ops: struct rtnl_link_ops * to unregister
 418  */
 419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 420 {
 421         /* Close the race with cleanup_net() */
 422         mutex_lock(&net_mutex);
 423         rtnl_lock_unregistering_all();
 424         __rtnl_link_unregister(ops);
 425         rtnl_unlock();
 426         mutex_unlock(&net_mutex);
 427 }
 428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 429
 430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 431 {
 432         struct net_device *master_dev;
 433         const struct rtnl_link_ops *ops;
 434
 435         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 436         if (!master_dev)
 437                 return 0;
 438         ops = master_dev->rtnl_link_ops;
 439         if (!ops || !ops->get_slave_size)
 440                 return 0;
 441         /* IFLA_INFO_SLAVE_DATA + nested data */
 442         return nla_total_size(sizeof(struct nlattr)) +
 443                ops->get_slave_size(master_dev, dev);
 444 }
 445
 446 static size_t rtnl_link_get_size(const struct net_device *dev)
 447 {
 448         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 449         size_t size;
 450
 451         if (!ops)
 452                 return 0;
 453
 454         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 455                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 456
 457         if (ops->get_size)
 458                 /* IFLA_INFO_DATA + nested data */
 459                 size += nla_total_size(sizeof(struct nlattr)) +
 460                         ops->get_size(dev);
 461
 462         if (ops->get_xstats_size)
 463                 /* IFLA_INFO_XSTATS */
 464                 size += nla_total_size(ops->get_xstats_size(dev));
 465
 466         size += rtnl_link_get_slave_info_data_size(dev);
 467
 468         return size;
 469 }
 470
 471 static LIST_HEAD(rtnl_af_ops);
 472
 473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 474 {
 475         const struct rtnl_af_ops *ops;
 476
 477         list_for_each_entry(ops, &rtnl_af_ops, list) {
 478                 if (ops->family == family)
 479                         return ops;
 480         }
 481
 482         return NULL;
 483 }
 484
 485 /**
 486  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 487  * @ops: struct rtnl_af_ops * to register
 488  *
 489  * Returns 0 on success or a negative error code.
 490  */
 491 void rtnl_af_register(struct rtnl_af_ops *ops)
 492 {
 493         rtnl_lock();
 494         list_add_tail(&ops->list, &rtnl_af_ops);
 495         rtnl_unlock();
 496 }
 497 EXPORT_SYMBOL_GPL(rtnl_af_register);
 498
 499 /**
 500  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 501  * @ops: struct rtnl_af_ops * to unregister
 502  *
 503  * The caller must hold the rtnl_mutex.
 504  */
 505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 506 {
 507         list_del(&ops->list);
 508 }
 509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 510
 511 /**
 512  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 513  * @ops: struct rtnl_af_ops * to unregister
 514  */
 515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 516 {
 517         rtnl_lock();
 518         __rtnl_af_unregister(ops);
 519         rtnl_unlock();
 520 }
 521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 522
 523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 524                                     u32 ext_filter_mask)
 525 {
 526         struct rtnl_af_ops *af_ops;
 527         size_t size;
 528
 529         /* IFLA_AF_SPEC */
 530         size = nla_total_size(sizeof(struct nlattr));
 531
 532         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 533                 if (af_ops->get_link_af_size) {
 534                         /* AF_* + nested data */
 535                         size += nla_total_size(sizeof(struct nlattr)) +
 536                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 537                 }
 538         }
 539
 540         return size;
 541 }
 542
 543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 544 {
 545         struct net_device *master_dev;
 546
 547         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 548         if (master_dev && master_dev->rtnl_link_ops)
 549                 return true;
 550         return false;
 551 }
 552
 553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 554                                      const struct net_device *dev)
 555 {
 556         struct net_device *master_dev;
 557         const struct rtnl_link_ops *ops;
 558         struct nlattr *slave_data;
 559         int err;
 560
 561         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 562         if (!master_dev)
 563                 return 0;
 564         ops = master_dev->rtnl_link_ops;
 565         if (!ops)
 566                 return 0;
 567         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 568                 return -EMSGSIZE;
 569         if (ops->fill_slave_info) {
 570                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 571                 if (!slave_data)
 572                         return -EMSGSIZE;
 573                 err = ops->fill_slave_info(skb, master_dev, dev);
 574                 if (err < 0)
 575                         goto err_cancel_slave_data;
 576                 nla_nest_end(skb, slave_data);
 577         }
 578         return 0;
 579
 580 err_cancel_slave_data:
 581         nla_nest_cancel(skb, slave_data);
 582         return err;
 583 }
 584
 585 static int rtnl_link_info_fill(struct sk_buff *skb,
 586                                const struct net_device *dev)
 587 {
 588         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 589         struct nlattr *data;
 590         int err;
 591
 592         if (!ops)
 593                 return 0;
 594         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 595                 return -EMSGSIZE;
 596         if (ops->fill_xstats) {
 597                 err = ops->fill_xstats(skb, dev);
 598                 if (err < 0)
 599                         return err;
 600         }
 601         if (ops->fill_info) {
 602                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 603                 if (data == NULL)
 604                         return -EMSGSIZE;
 605                 err = ops->fill_info(skb, dev);
 606                 if (err < 0)
 607                         goto err_cancel_data;
 608                 nla_nest_end(skb, data);
 609         }
 610         return 0;
 611
 612 err_cancel_data:
 613         nla_nest_cancel(skb, data);
 614         return err;
 615 }
 616
 617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 618 {
 619         struct nlattr *linkinfo;
 620         int err = -EMSGSIZE;
 621
 622         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 623         if (linkinfo == NULL)
 624                 goto out;
 625
 626         err = rtnl_link_info_fill(skb, dev);
 627         if (err < 0)
 628                 goto err_cancel_link;
 629
 630         err = rtnl_link_slave_info_fill(skb, dev);
 631         if (err < 0)
 632                 goto err_cancel_link;
 633
 634         nla_nest_end(skb, linkinfo);
 635         return 0;
 636
 637 err_cancel_link:
 638         nla_nest_cancel(skb, linkinfo);
 639 out:
 640         return err;
 641 }
 642
 643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 644 {
 645         struct sock *rtnl = net->rtnl;
 646         int err = 0;
 647
 648         NETLINK_CB(skb).dst_group = group;
 649         if (echo)
 650                 atomic_inc(&skb->users);
 651         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 652         if (echo)
 653                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 654         return err;
 655 }
 656
 657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 658 {
 659         struct sock *rtnl = net->rtnl;
 660
 661         return nlmsg_unicast(rtnl, skb, pid);
 662 }
 663 EXPORT_SYMBOL(rtnl_unicast);
 664
 665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 666                  struct nlmsghdr *nlh, gfp_t flags)
 667 {
 668         struct sock *rtnl = net->rtnl;
 669         int report = 0;
 670
 671         if (nlh)
 672                 report = nlmsg_report(nlh);
 673
 674         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 675 }
 676 EXPORT_SYMBOL(rtnl_notify);
 677
 678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 679 {
 680         struct sock *rtnl = net->rtnl;
 681
 682         netlink_set_err(rtnl, 0, group, error);
 683 }
 684 EXPORT_SYMBOL(rtnl_set_sk_err);
 685
 686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 687 {
 688         struct nlattr *mx;
 689         int i, valid = 0;
 690
 691         mx = nla_nest_start(skb, RTA_METRICS);
 692         if (mx == NULL)
 693                 return -ENOBUFS;
 694
 695         for (i = 0; i < RTAX_MAX; i++) {
 696                 if (metrics[i]) {
 697                         if (i == RTAX_CC_ALGO - 1) {
 698                                 char tmp[TCP_CA_NAME_MAX], *name;
 699
 700                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 701                                 if (!name)
 702                                         continue;
 703                                 if (nla_put_string(skb, i + 1, name))
 704                                         goto nla_put_failure;
 705                         } else if (i == RTAX_FEATURES - 1) {
 706                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 707
 708                                 if (!user_features)
 709                                         continue;
 710                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 711                                 if (nla_put_u32(skb, i + 1, user_features))
 712                                         goto nla_put_failure;
 713                         } else {
 714                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 715                                         goto nla_put_failure;
 716                         }
 717                         valid++;
 718                 }
 719         }
 720
 721         if (!valid) {
 722                 nla_nest_cancel(skb, mx);
 723                 return 0;
 724         }
 725
 726         return nla_nest_end(skb, mx);
 727
 728 nla_put_failure:
 729         nla_nest_cancel(skb, mx);
 730         return -EMSGSIZE;
 731 }
 732 EXPORT_SYMBOL(rtnetlink_put_metrics);
 733
 734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 735                        long expires, u32 error)
 736 {
 737         struct rta_cacheinfo ci = {
 738                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 739                 .rta_used = dst->__use,
 740                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 741                 .rta_error = error,
 742                 .rta_id =  id,
 743         };
 744
 745         if (expires) {
 746                 unsigned long clock;
 747
 748                 clock = jiffies_to_clock_t(abs(expires));
 749                 clock = min_t(unsigned long, clock, INT_MAX);
 750                 ci.rta_expires = (expires > 0) ? clock : -clock;
 751         }
 752         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 753 }
 754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 755
 756 static void set_operstate(struct net_device *dev, unsigned char transition)
 757 {
 758         unsigned char operstate = dev->operstate;
 759
 760         switch (transition) {
 761         case IF_OPER_UP:
 762                 if ((operstate == IF_OPER_DORMANT ||
 763                      operstate == IF_OPER_UNKNOWN) &&
 764                     !netif_dormant(dev))
 765                         operstate = IF_OPER_UP;
 766                 break;
 767
 768         case IF_OPER_DORMANT:
 769                 if (operstate == IF_OPER_UP ||
 770                     operstate == IF_OPER_UNKNOWN)
 771                         operstate = IF_OPER_DORMANT;
 772                 break;
 773         }
 774
 775         if (dev->operstate != operstate) {
 776                 write_lock_bh(&dev_base_lock);
 777                 dev->operstate = operstate;
 778                 write_unlock_bh(&dev_base_lock);
 779                 netdev_state_change(dev);
 780         }
 781 }
 782
 783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 784 {
 785         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 786                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 787 }
 788
 789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 790                                            const struct ifinfomsg *ifm)
 791 {
 792         unsigned int flags = ifm->ifi_flags;
 793
 794         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 795         if (ifm->ifi_change)
 796                 flags = (flags & ifm->ifi_change) |
 797                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 798
 799         return flags;
 800 }
 801
 802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 803                                  const struct rtnl_link_stats64 *b)
 804 {
 805         a->rx_packets = b->rx_packets;
 806         a->tx_packets = b->tx_packets;
 807         a->rx_bytes = b->rx_bytes;
 808         a->tx_bytes = b->tx_bytes;
 809         a->rx_errors = b->rx_errors;
 810         a->tx_errors = b->tx_errors;
 811         a->rx_dropped = b->rx_dropped;
 812         a->tx_dropped = b->tx_dropped;
 813
 814         a->multicast = b->multicast;
 815         a->collisions = b->collisions;
 816
 817         a->rx_length_errors = b->rx_length_errors;
 818         a->rx_over_errors = b->rx_over_errors;
 819         a->rx_crc_errors = b->rx_crc_errors;
 820         a->rx_frame_errors = b->rx_frame_errors;
 821         a->rx_fifo_errors = b->rx_fifo_errors;
 822         a->rx_missed_errors = b->rx_missed_errors;
 823
 824         a->tx_aborted_errors = b->tx_aborted_errors;
 825         a->tx_carrier_errors = b->tx_carrier_errors;
 826         a->tx_fifo_errors = b->tx_fifo_errors;
 827         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 828         a->tx_window_errors = b->tx_window_errors;
 829
 830         a->rx_compressed = b->rx_compressed;
 831         a->tx_compressed = b->tx_compressed;
 832
 833         a->rx_nohandler = b->rx_nohandler;
 834 }
 835
 836 /* All VF info */
 837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 838                                    u32 ext_filter_mask)
 839 {
 840         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 841                 int num_vfs = dev_num_vf(dev->dev.parent);
 842                 size_t size = nla_total_size(0);
 843                 size += num_vfs *
 844                         (nla_total_size(0) +
 845                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 846                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 847                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 848                          nla_total_size(MAX_VLAN_LIST_LEN *
 849                                         sizeof(struct ifla_vf_vlan_info)) +
 850                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 851                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 852                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 853                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 854                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 855                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 856                          /* IFLA_VF_STATS_RX_PACKETS */
 857                          nla_total_size_64bit(sizeof(__u64)) +
 858                          /* IFLA_VF_STATS_TX_PACKETS */
 859                          nla_total_size_64bit(sizeof(__u64)) +
 860                          /* IFLA_VF_STATS_RX_BYTES */
 861                          nla_total_size_64bit(sizeof(__u64)) +
 862                          /* IFLA_VF_STATS_TX_BYTES */
 863                          nla_total_size_64bit(sizeof(__u64)) +
 864                          /* IFLA_VF_STATS_BROADCAST */
 865                          nla_total_size_64bit(sizeof(__u64)) +
 866                          /* IFLA_VF_STATS_MULTICAST */
 867                          nla_total_size_64bit(sizeof(__u64)) +
 868                          nla_total_size(sizeof(struct ifla_vf_trust)));
 869                 return size;
 870         } else
 871                 return 0;
 872 }
 873
 874 static size_t rtnl_port_size(const struct net_device *dev,
 875                              u32 ext_filter_mask)
 876 {
 877         size_t port_size = nla_total_size(4)            /* PORT_VF */
 878                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 879                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 880                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 881                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 882                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 883         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 884         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 885                 + port_size;
 886         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 887                 + port_size;
 888
 889         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 890             !(ext_filter_mask & RTEXT_FILTER_VF))
 891                 return 0;
 892         if (dev_num_vf(dev->dev.parent))
 893                 return port_self_size + vf_ports_size +
 894                         vf_port_size * dev_num_vf(dev->dev.parent);
 895         else
 896                 return port_self_size;
 897 }
 898
 899 static size_t rtnl_xdp_size(void)
 900 {
 901         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 902                           nla_total_size(1) +   /* XDP_ATTACHED */
 903                           nla_total_size(4);    /* XDP_FLAGS */
 904
 905         return xdp_size;
 906 }
 907
 908 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 909                                      u32 ext_filter_mask)
 910 {
 911         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 912                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 913                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 914                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 915                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 916                + nla_total_size(sizeof(struct rtnl_link_stats))
 917                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 918                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 919                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 920                + nla_total_size(4) /* IFLA_TXQLEN */
 921                + nla_total_size(4) /* IFLA_WEIGHT */
 922                + nla_total_size(4) /* IFLA_MTU */
 923                + nla_total_size(4) /* IFLA_LINK */
 924                + nla_total_size(4) /* IFLA_MASTER */
 925                + nla_total_size(1) /* IFLA_CARRIER */
 926                + nla_total_size(4) /* IFLA_PROMISCUITY */
 927                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 928                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 929                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 930                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 931                + nla_total_size(1) /* IFLA_OPERSTATE */
 932                + nla_total_size(1) /* IFLA_LINKMODE */
 933                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 934                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 935                + nla_total_size(ext_filter_mask
 936                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 937                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 938                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 939                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 940                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 941                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 942                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 943                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 944                + rtnl_xdp_size() /* IFLA_XDP */
 945                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 946
 947 }
 948
 949 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 950 {
 951         struct nlattr *vf_ports;
 952         struct nlattr *vf_port;
 953         int vf;
 954         int err;
 955
 956         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 957         if (!vf_ports)
 958                 return -EMSGSIZE;
 959
 960         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 961                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 962                 if (!vf_port)
 963                         goto nla_put_failure;
 964                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 965                         goto nla_put_failure;
 966                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 967                 if (err == -EMSGSIZE)
 968                         goto nla_put_failure;
 969                 if (err) {
 970                         nla_nest_cancel(skb, vf_port);
 971                         continue;
 972                 }
 973                 nla_nest_end(skb, vf_port);
 974         }
 975
 976         nla_nest_end(skb, vf_ports);
 977
 978         return 0;
 979
 980 nla_put_failure:
 981         nla_nest_cancel(skb, vf_ports);
 982         return -EMSGSIZE;
 983 }
 984
 985 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 986 {
 987         struct nlattr *port_self;
 988         int err;
 989
 990         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 991         if (!port_self)
 992                 return -EMSGSIZE;
 993
 994         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 995         if (err) {
 996                 nla_nest_cancel(skb, port_self);
 997                 return (err == -EMSGSIZE) ? err : 0;
 998         }
 999
1000         nla_nest_end(skb, port_self);
1001
1002         return 0;
1003 }
1004
1005 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1006                           u32 ext_filter_mask)
1007 {
1008         int err;
1009
1010         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1011             !(ext_filter_mask & RTEXT_FILTER_VF))
1012                 return 0;
1013
1014         err = rtnl_port_self_fill(skb, dev);
1015         if (err)
1016                 return err;
1017
1018         if (dev_num_vf(dev->dev.parent)) {
1019                 err = rtnl_vf_ports_fill(skb, dev);
1020                 if (err)
1021                         return err;
1022         }
1023
1024         return 0;
1025 }
1026
1027 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1028 {
1029         int err;
1030         struct netdev_phys_item_id ppid;
1031
1032         err = dev_get_phys_port_id(dev, &ppid);
1033         if (err) {
1034                 if (err == -EOPNOTSUPP)
1035                         return 0;
1036                 return err;
1037         }
1038
1039         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1040                 return -EMSGSIZE;
1041
1042         return 0;
1043 }
1044
1045 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1046 {
1047         char name[IFNAMSIZ];
1048         int err;
1049
1050         err = dev_get_phys_port_name(dev, name, sizeof(name));
1051         if (err) {
1052                 if (err == -EOPNOTSUPP)
1053                         return 0;
1054                 return err;
1055         }
1056
1057         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1058                 return -EMSGSIZE;
1059
1060         return 0;
1061 }
1062
1063 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1064 {
1065         int err;
1066         struct switchdev_attr attr = {
1067                 .orig_dev = dev,
1068                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1069                 .flags = SWITCHDEV_F_NO_RECURSE,
1070         };
1071
1072         err = switchdev_port_attr_get(dev, &attr);
1073         if (err) {
1074                 if (err == -EOPNOTSUPP)
1075                         return 0;
1076                 return err;
1077         }
1078
1079         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1080                     attr.u.ppid.id))
1081                 return -EMSGSIZE;
1082
1083         return 0;
1084 }
1085
1086 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1087                                               struct net_device *dev)
1088 {
1089         struct rtnl_link_stats64 *sp;
1090         struct nlattr *attr;
1091
1092         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1093                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1094         if (!attr)
1095                 return -EMSGSIZE;
1096
1097         sp = nla_data(attr);
1098         dev_get_stats(dev, sp);
1099
1100         attr = nla_reserve(skb, IFLA_STATS,
1101                            sizeof(struct rtnl_link_stats));
1102         if (!attr)
1103                 return -EMSGSIZE;
1104
1105         copy_rtnl_link_stats(nla_data(attr), sp);
1106
1107         return 0;
1108 }
1109
1110 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1111                                                struct net_device *dev,
1112                                                int vfs_num,
1113                                                struct nlattr *vfinfo)
1114 {
1115         struct ifla_vf_rss_query_en vf_rss_query_en;
1116         struct nlattr *vf, *vfstats, *vfvlanlist;
1117         struct ifla_vf_link_state vf_linkstate;
1118         struct ifla_vf_vlan_info vf_vlan_info;
1119         struct ifla_vf_spoofchk vf_spoofchk;
1120         struct ifla_vf_tx_rate vf_tx_rate;
1121         struct ifla_vf_stats vf_stats;
1122         struct ifla_vf_trust vf_trust;
1123         struct ifla_vf_vlan vf_vlan;
1124         struct ifla_vf_rate vf_rate;
1125         struct ifla_vf_mac vf_mac;
1126         struct ifla_vf_info ivi;
1127
1128         /* Not all SR-IOV capable drivers support the
1129          * spoofcheck and "RSS query enable" query.  Preset to
1130          * -1 so the user space tool can detect that the driver
1131          * didn't report anything.
1132          */
1133         ivi.spoofchk = -1;
1134         ivi.rss_query_en = -1;
1135         ivi.trusted = -1;
1136         memset(ivi.mac, 0, sizeof(ivi.mac));
1137         /* The default value for VF link state is "auto"
1138          * IFLA_VF_LINK_STATE_AUTO which equals zero
1139          */
1140         ivi.linkstate = 0;
1141         /* VLAN Protocol by default is 802.1Q */
1142         ivi.vlan_proto = htons(ETH_P_8021Q);
1143         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1144                 return 0;
1145
1146         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1147
1148         vf_mac.vf =
1149                 vf_vlan.vf =
1150                 vf_vlan_info.vf =
1151                 vf_rate.vf =
1152                 vf_tx_rate.vf =
1153                 vf_spoofchk.vf =
1154                 vf_linkstate.vf =
1155                 vf_rss_query_en.vf =
1156                 vf_trust.vf = ivi.vf;
1157
1158         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1159         vf_vlan.vlan = ivi.vlan;
1160         vf_vlan.qos = ivi.qos;
1161         vf_vlan_info.vlan = ivi.vlan;
1162         vf_vlan_info.qos = ivi.qos;
1163         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1164         vf_tx_rate.rate = ivi.max_tx_rate;
1165         vf_rate.min_tx_rate = ivi.min_tx_rate;
1166         vf_rate.max_tx_rate = ivi.max_tx_rate;
1167         vf_spoofchk.setting = ivi.spoofchk;
1168         vf_linkstate.link_state = ivi.linkstate;
1169         vf_rss_query_en.setting = ivi.rss_query_en;
1170         vf_trust.setting = ivi.trusted;
1171         vf = nla_nest_start(skb, IFLA_VF_INFO);
1172         if (!vf)
1173                 goto nla_put_vfinfo_failure;
1174         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1175             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1176             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1177                     &vf_rate) ||
1178             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1179                     &vf_tx_rate) ||
1180             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1181                     &vf_spoofchk) ||
1182             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1183                     &vf_linkstate) ||
1184             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1185                     sizeof(vf_rss_query_en),
1186                     &vf_rss_query_en) ||
1187             nla_put(skb, IFLA_VF_TRUST,
1188                     sizeof(vf_trust), &vf_trust))
1189                 goto nla_put_vf_failure;
1190         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1191         if (!vfvlanlist)
1192                 goto nla_put_vf_failure;
1193         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1194                     &vf_vlan_info)) {
1195                 nla_nest_cancel(skb, vfvlanlist);
1196                 goto nla_put_vf_failure;
1197         }
1198         nla_nest_end(skb, vfvlanlist);
1199         memset(&vf_stats, 0, sizeof(vf_stats));
1200         if (dev->netdev_ops->ndo_get_vf_stats)
1201                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1202                                                 &vf_stats);
1203         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1204         if (!vfstats)
1205                 goto nla_put_vf_failure;
1206         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1207                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1208             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1209                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1210             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1211                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1212             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1213                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1214             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1215                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1216             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1217                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1218                 nla_nest_cancel(skb, vfstats);
1219                 goto nla_put_vf_failure;
1220         }
1221         nla_nest_end(skb, vfstats);
1222         nla_nest_end(skb, vf);
1223         return 0;
1224
1225 nla_put_vf_failure:
1226         nla_nest_cancel(skb, vf);
1227 nla_put_vfinfo_failure:
1228         nla_nest_cancel(skb, vfinfo);
1229         return -EMSGSIZE;
1230 }
1231
1232 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1233 {
1234         struct rtnl_link_ifmap map;
1235
1236         memset(&map, 0, sizeof(map));
1237         map.mem_start   = dev->mem_start;
1238         map.mem_end     = dev->mem_end;
1239         map.base_addr   = dev->base_addr;
1240         map.irq         = dev->irq;
1241         map.dma         = dev->dma;
1242         map.port        = dev->if_port;
1243
1244         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1245                 return -EMSGSIZE;
1246
1247         return 0;
1248 }
1249
1250 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1251 {
1252         struct nlattr *xdp;
1253         u32 xdp_flags = 0;
1254         u8 val = 0;
1255         int err;
1256
1257         xdp = nla_nest_start(skb, IFLA_XDP);
1258         if (!xdp)
1259                 return -EMSGSIZE;
1260         if (rcu_access_pointer(dev->xdp_prog)) {
1261                 xdp_flags = XDP_FLAGS_SKB_MODE;
1262                 val = 1;
1263         } else if (dev->netdev_ops->ndo_xdp) {
1264                 struct netdev_xdp xdp_op = {};
1265
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                 val = xdp_op.prog_attached;
1271         }
1272         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, val);
1273         if (err)
1274                 goto err_cancel;
1275
1276         if (xdp_flags) {
1277                 err = nla_put_u32(skb, IFLA_XDP_FLAGS, xdp_flags);
1278                 if (err)
1279                         goto err_cancel;
1280         }
1281         nla_nest_end(skb, xdp);
1282         return 0;
1283
1284 err_cancel:
1285         nla_nest_cancel(skb, xdp);
1286         return err;
1287 }
1288
1289 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1290                             int type, u32 pid, u32 seq, u32 change,
1291                             unsigned int flags, u32 ext_filter_mask)
1292 {
1293         struct ifinfomsg *ifm;
1294         struct nlmsghdr *nlh;
1295         struct nlattr *af_spec;
1296         struct rtnl_af_ops *af_ops;
1297         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1298
1299         ASSERT_RTNL();
1300         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1301         if (nlh == NULL)
1302                 return -EMSGSIZE;
1303
1304         ifm = nlmsg_data(nlh);
1305         ifm->ifi_family = AF_UNSPEC;
1306         ifm->__ifi_pad = 0;
1307         ifm->ifi_type = dev->type;
1308         ifm->ifi_index = dev->ifindex;
1309         ifm->ifi_flags = dev_get_flags(dev);
1310         ifm->ifi_change = change;
1311
1312         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1313             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1314             nla_put_u8(skb, IFLA_OPERSTATE,
1315                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1316             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1317             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1318             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1319             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1320             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1321             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1322             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1323 #ifdef CONFIG_RPS
1324             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1325 #endif
1326             (dev->ifindex != dev_get_iflink(dev) &&
1327              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1328             (upper_dev &&
1329              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1330             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1331             (dev->qdisc &&
1332              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1333             (dev->ifalias &&
1334              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1335             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1336                         atomic_read(&dev->carrier_changes)) ||
1337             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1338                 goto nla_put_failure;
1339
1340         if (rtnl_fill_link_ifmap(skb, dev))
1341                 goto nla_put_failure;
1342
1343         if (dev->addr_len) {
1344                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1345                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1346                         goto nla_put_failure;
1347         }
1348
1349         if (rtnl_phys_port_id_fill(skb, dev))
1350                 goto nla_put_failure;
1351
1352         if (rtnl_phys_port_name_fill(skb, dev))
1353                 goto nla_put_failure;
1354
1355         if (rtnl_phys_switch_id_fill(skb, dev))
1356                 goto nla_put_failure;
1357
1358         if (rtnl_fill_stats(skb, dev))
1359                 goto nla_put_failure;
1360
1361         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1362             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1363                 goto nla_put_failure;
1364
1365         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1366             ext_filter_mask & RTEXT_FILTER_VF) {
1367                 int i;
1368                 struct nlattr *vfinfo;
1369                 int num_vfs = dev_num_vf(dev->dev.parent);
1370
1371                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1372                 if (!vfinfo)
1373                         goto nla_put_failure;
1374                 for (i = 0; i < num_vfs; i++) {
1375                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1376                                 goto nla_put_failure;
1377                 }
1378
1379                 nla_nest_end(skb, vfinfo);
1380         }
1381
1382         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1383                 goto nla_put_failure;
1384
1385         if (rtnl_xdp_fill(skb, dev))
1386                 goto nla_put_failure;
1387
1388         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1389                 if (rtnl_link_fill(skb, dev) < 0)
1390                         goto nla_put_failure;
1391         }
1392
1393         if (dev->rtnl_link_ops &&
1394             dev->rtnl_link_ops->get_link_net) {
1395                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1396
1397                 if (!net_eq(dev_net(dev), link_net)) {
1398                         int id = peernet2id_alloc(dev_net(dev), link_net);
1399
1400                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1401                                 goto nla_put_failure;
1402                 }
1403         }
1404
1405         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1406                 goto nla_put_failure;
1407
1408         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1409                 if (af_ops->fill_link_af) {
1410                         struct nlattr *af;
1411                         int err;
1412
1413                         if (!(af = nla_nest_start(skb, af_ops->family)))
1414                                 goto nla_put_failure;
1415
1416                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1417
1418                         /*
1419                          * Caller may return ENODATA to indicate that there
1420                          * was no data to be dumped. This is not an error, it
1421                          * means we should trim the attribute header and
1422                          * continue.
1423                          */
1424                         if (err == -ENODATA)
1425                                 nla_nest_cancel(skb, af);
1426                         else if (err < 0)
1427                                 goto nla_put_failure;
1428
1429                         nla_nest_end(skb, af);
1430                 }
1431         }
1432
1433         nla_nest_end(skb, af_spec);
1434
1435         nlmsg_end(skb, nlh);
1436         return 0;
1437
1438 nla_put_failure:
1439         nlmsg_cancel(skb, nlh);
1440         return -EMSGSIZE;
1441 }
1442
1443 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1444         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1445         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1446         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1447         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1448         [IFLA_MTU]              = { .type = NLA_U32 },
1449         [IFLA_LINK]             = { .type = NLA_U32 },
1450         [IFLA_MASTER]           = { .type = NLA_U32 },
1451         [IFLA_CARRIER]          = { .type = NLA_U8 },
1452         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1453         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1454         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1455         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1456         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1457         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1458         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1459         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1460         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1461         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1462         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1463         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1464         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1465         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1466         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1467         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1468         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1469         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1470         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1471         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1472         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1473         [IFLA_XDP]              = { .type = NLA_NESTED },
1474 };
1475
1476 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1477         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1478         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1479         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1480         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1481 };
1482
1483 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1484         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1485         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1486         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1487         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1488         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1489         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1490         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1491         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1492         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1493         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1494         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1495         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1496 };
1497
1498 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1499         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1500         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1501                                     .len = PORT_PROFILE_MAX },
1502         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1503                                       .len = PORT_UUID_MAX },
1504         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1505                                     .len = PORT_UUID_MAX },
1506         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1507         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1508
1509         /* Unused, but we need to keep it here since user space could
1510          * fill it. It's also broken with regard to NLA_BINARY use in
1511          * combination with structs.
1512          */
1513         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1514                                     .len = sizeof(struct ifla_port_vsi) },
1515 };
1516
1517 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1518         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1519         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1520         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1521 };
1522
1523 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1524 {
1525         const struct rtnl_link_ops *ops = NULL;
1526         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1527
1528         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1529                              ifla_info_policy, NULL) < 0)
1530                 return NULL;
1531
1532         if (linfo[IFLA_INFO_KIND]) {
1533                 char kind[MODULE_NAME_LEN];
1534
1535                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1536                 ops = rtnl_link_ops_get(kind);
1537         }
1538
1539         return ops;
1540 }
1541
1542 static bool link_master_filtered(struct net_device *dev, int master_idx)
1543 {
1544         struct net_device *master;
1545
1546         if (!master_idx)
1547                 return false;
1548
1549         master = netdev_master_upper_dev_get(dev);
1550         if (!master || master->ifindex != master_idx)
1551                 return true;
1552
1553         return false;
1554 }
1555
1556 static bool link_kind_filtered(const struct net_device *dev,
1557                                const struct rtnl_link_ops *kind_ops)
1558 {
1559         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1560                 return true;
1561
1562         return false;
1563 }
1564
1565 static bool link_dump_filtered(struct net_device *dev,
1566                                int master_idx,
1567                                const struct rtnl_link_ops *kind_ops)
1568 {
1569         if (link_master_filtered(dev, master_idx) ||
1570             link_kind_filtered(dev, kind_ops))
1571                 return true;
1572
1573         return false;
1574 }
1575
1576 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1577 {
1578         struct net *net = sock_net(skb->sk);
1579         int h, s_h;
1580         int idx = 0, s_idx;
1581         struct net_device *dev;
1582         struct hlist_head *head;
1583         struct nlattr *tb[IFLA_MAX+1];
1584         u32 ext_filter_mask = 0;
1585         const struct rtnl_link_ops *kind_ops = NULL;
1586         unsigned int flags = NLM_F_MULTI;
1587         int master_idx = 0;
1588         int err;
1589         int hdrlen;
1590
1591         s_h = cb->args[0];
1592         s_idx = cb->args[1];
1593
1594         cb->seq = net->dev_base_seq;
1595
1596         /* A hack to preserve kernel<->userspace interface.
1597          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1598          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1599          * what iproute2 < v3.9.0 used.
1600          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1601          * attribute, its netlink message is shorter than struct ifinfomsg.
1602          */
1603         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1604                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1605
1606         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1607                         ifla_policy, NULL) >= 0) {
1608                 if (tb[IFLA_EXT_MASK])
1609                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1610
1611                 if (tb[IFLA_MASTER])
1612                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1613
1614                 if (tb[IFLA_LINKINFO])
1615                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1616
1617                 if (master_idx || kind_ops)
1618                         flags |= NLM_F_DUMP_FILTERED;
1619         }
1620
1621         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1622                 idx = 0;
1623                 head = &net->dev_index_head[h];
1624                 hlist_for_each_entry(dev, head, index_hlist) {
1625                         if (link_dump_filtered(dev, master_idx, kind_ops))
1626                                 goto cont;
1627                         if (idx < s_idx)
1628                                 goto cont;
1629                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1630                                                NETLINK_CB(cb->skb).portid,
1631                                                cb->nlh->nlmsg_seq, 0,
1632                                                flags,
1633                                                ext_filter_mask);
1634                         /* If we ran out of room on the first message,
1635                          * we're in trouble
1636                          */
1637                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1638
1639                         if (err < 0)
1640                                 goto out;
1641
1642                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1643 cont:
1644                         idx++;
1645                 }
1646         }
1647 out:
1648         cb->args[1] = idx;
1649         cb->args[0] = h;
1650
1651         return skb->len;
1652 }
1653
1654 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1655                         struct netlink_ext_ack *exterr)
1656 {
1657         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1658 }
1659 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1660
1661 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1662 {
1663         struct net *net;
1664         /* Examine the link attributes and figure out which
1665          * network namespace we are talking about.
1666          */
1667         if (tb[IFLA_NET_NS_PID])
1668                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1669         else if (tb[IFLA_NET_NS_FD])
1670                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1671         else
1672                 net = get_net(src_net);
1673         return net;
1674 }
1675 EXPORT_SYMBOL(rtnl_link_get_net);
1676
1677 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1678 {
1679         if (dev) {
1680                 if (tb[IFLA_ADDRESS] &&
1681                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1682                         return -EINVAL;
1683
1684                 if (tb[IFLA_BROADCAST] &&
1685                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1686                         return -EINVAL;
1687         }
1688
1689         if (tb[IFLA_AF_SPEC]) {
1690                 struct nlattr *af;
1691                 int rem, err;
1692
1693                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1694                         const struct rtnl_af_ops *af_ops;
1695
1696                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1697                                 return -EAFNOSUPPORT;
1698
1699                         if (!af_ops->set_link_af)
1700                                 return -EOPNOTSUPP;
1701
1702                         if (af_ops->validate_link_af) {
1703                                 err = af_ops->validate_link_af(dev, af);
1704                                 if (err < 0)
1705                                         return err;
1706                         }
1707                 }
1708         }
1709
1710         return 0;
1711 }
1712
1713 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1714                                   int guid_type)
1715 {
1716         const struct net_device_ops *ops = dev->netdev_ops;
1717
1718         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1719 }
1720
1721 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1722 {
1723         if (dev->type != ARPHRD_INFINIBAND)
1724                 return -EOPNOTSUPP;
1725
1726         return handle_infiniband_guid(dev, ivt, guid_type);
1727 }
1728
1729 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1730 {
1731         const struct net_device_ops *ops = dev->netdev_ops;
1732         int err = -EINVAL;
1733
1734         if (tb[IFLA_VF_MAC]) {
1735                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1736
1737                 err = -EOPNOTSUPP;
1738                 if (ops->ndo_set_vf_mac)
1739                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1740                                                   ivm->mac);
1741                 if (err < 0)
1742                         return err;
1743         }
1744
1745         if (tb[IFLA_VF_VLAN]) {
1746                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1747
1748                 err = -EOPNOTSUPP;
1749                 if (ops->ndo_set_vf_vlan)
1750                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1751                                                    ivv->qos,
1752                                                    htons(ETH_P_8021Q));
1753                 if (err < 0)
1754                         return err;
1755         }
1756
1757         if (tb[IFLA_VF_VLAN_LIST]) {
1758                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1759                 struct nlattr *attr;
1760                 int rem, len = 0;
1761
1762                 err = -EOPNOTSUPP;
1763                 if (!ops->ndo_set_vf_vlan)
1764                         return err;
1765
1766                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1767                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1768                             nla_len(attr) < NLA_HDRLEN) {
1769                                 return -EINVAL;
1770                         }
1771                         if (len >= MAX_VLAN_LIST_LEN)
1772                                 return -EOPNOTSUPP;
1773                         ivvl[len] = nla_data(attr);
1774
1775                         len++;
1776                 }
1777                 if (len == 0)
1778                         return -EINVAL;
1779
1780                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1781                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1782                 if (err < 0)
1783                         return err;
1784         }
1785
1786         if (tb[IFLA_VF_TX_RATE]) {
1787                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1788                 struct ifla_vf_info ivf;
1789
1790                 err = -EOPNOTSUPP;
1791                 if (ops->ndo_get_vf_config)
1792                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1793                 if (err < 0)
1794                         return err;
1795
1796                 err = -EOPNOTSUPP;
1797                 if (ops->ndo_set_vf_rate)
1798                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1799                                                    ivf.min_tx_rate,
1800                                                    ivt->rate);
1801                 if (err < 0)
1802                         return err;
1803         }
1804
1805         if (tb[IFLA_VF_RATE]) {
1806                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1807
1808                 err = -EOPNOTSUPP;
1809                 if (ops->ndo_set_vf_rate)
1810                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1811                                                    ivt->min_tx_rate,
1812                                                    ivt->max_tx_rate);
1813                 if (err < 0)
1814                         return err;
1815         }
1816
1817         if (tb[IFLA_VF_SPOOFCHK]) {
1818                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1819
1820                 err = -EOPNOTSUPP;
1821                 if (ops->ndo_set_vf_spoofchk)
1822                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1823                                                        ivs->setting);
1824                 if (err < 0)
1825                         return err;
1826         }
1827
1828         if (tb[IFLA_VF_LINK_STATE]) {
1829                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1830
1831                 err = -EOPNOTSUPP;
1832                 if (ops->ndo_set_vf_link_state)
1833                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1834                                                          ivl->link_state);
1835                 if (err < 0)
1836                         return err;
1837         }
1838
1839         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1840                 struct ifla_vf_rss_query_en *ivrssq_en;
1841
1842                 err = -EOPNOTSUPP;
1843                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1844                 if (ops->ndo_set_vf_rss_query_en)
1845                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1846                                                            ivrssq_en->setting);
1847                 if (err < 0)
1848                         return err;
1849         }
1850
1851         if (tb[IFLA_VF_TRUST]) {
1852                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1853
1854                 err = -EOPNOTSUPP;
1855                 if (ops->ndo_set_vf_trust)
1856                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1857                 if (err < 0)
1858                         return err;
1859         }
1860
1861         if (tb[IFLA_VF_IB_NODE_GUID]) {
1862                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1863
1864                 if (!ops->ndo_set_vf_guid)
1865                         return -EOPNOTSUPP;
1866
1867                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1868         }
1869
1870         if (tb[IFLA_VF_IB_PORT_GUID]) {
1871                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1872
1873                 if (!ops->ndo_set_vf_guid)
1874                         return -EOPNOTSUPP;
1875
1876                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1877         }
1878
1879         return err;
1880 }
1881
1882 static int do_set_master(struct net_device *dev, int ifindex)
1883 {
1884         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1885         const struct net_device_ops *ops;
1886         int err;
1887
1888         if (upper_dev) {
1889                 if (upper_dev->ifindex == ifindex)
1890                         return 0;
1891                 ops = upper_dev->netdev_ops;
1892                 if (ops->ndo_del_slave) {
1893                         err = ops->ndo_del_slave(upper_dev, dev);
1894                         if (err)
1895                                 return err;
1896                 } else {
1897                         return -EOPNOTSUPP;
1898                 }
1899         }
1900
1901         if (ifindex) {
1902                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1903                 if (!upper_dev)
1904                         return -EINVAL;
1905                 ops = upper_dev->netdev_ops;
1906                 if (ops->ndo_add_slave) {
1907                         err = ops->ndo_add_slave(upper_dev, dev);
1908                         if (err)
1909                                 return err;
1910                 } else {
1911                         return -EOPNOTSUPP;
1912                 }
1913         }
1914         return 0;
1915 }
1916
1917 #define DO_SETLINK_MODIFIED     0x01
1918 /* notify flag means notify + modified. */
1919 #define DO_SETLINK_NOTIFY       0x03
1920 static int do_setlink(const struct sk_buff *skb,
1921                       struct net_device *dev, struct ifinfomsg *ifm,
1922                       struct netlink_ext_ack *extack,
1923                       struct nlattr **tb, char *ifname, int status)
1924 {
1925         const struct net_device_ops *ops = dev->netdev_ops;
1926         int err;
1927
1928         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1929                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1930                 if (IS_ERR(net)) {
1931                         err = PTR_ERR(net);
1932                         goto errout;
1933                 }
1934                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1935                         put_net(net);
1936                         err = -EPERM;
1937                         goto errout;
1938                 }
1939                 err = dev_change_net_namespace(dev, net, ifname);
1940                 put_net(net);
1941                 if (err)
1942                         goto errout;
1943                 status |= DO_SETLINK_MODIFIED;
1944         }
1945
1946         if (tb[IFLA_MAP]) {
1947                 struct rtnl_link_ifmap *u_map;
1948                 struct ifmap k_map;
1949
1950                 if (!ops->ndo_set_config) {
1951                         err = -EOPNOTSUPP;
1952                         goto errout;
1953                 }
1954
1955                 if (!netif_device_present(dev)) {
1956                         err = -ENODEV;
1957                         goto errout;
1958                 }
1959
1960                 u_map = nla_data(tb[IFLA_MAP]);
1961                 k_map.mem_start = (unsigned long) u_map->mem_start;
1962                 k_map.mem_end = (unsigned long) u_map->mem_end;
1963                 k_map.base_addr = (unsigned short) u_map->base_addr;
1964                 k_map.irq = (unsigned char) u_map->irq;
1965                 k_map.dma = (unsigned char) u_map->dma;
1966                 k_map.port = (unsigned char) u_map->port;
1967
1968                 err = ops->ndo_set_config(dev, &k_map);
1969                 if (err < 0)
1970                         goto errout;
1971
1972                 status |= DO_SETLINK_NOTIFY;
1973         }
1974
1975         if (tb[IFLA_ADDRESS]) {
1976                 struct sockaddr *sa;
1977                 int len;
1978
1979                 len = sizeof(sa_family_t) + dev->addr_len;
1980                 sa = kmalloc(len, GFP_KERNEL);
1981                 if (!sa) {
1982                         err = -ENOMEM;
1983                         goto errout;
1984                 }
1985                 sa->sa_family = dev->type;
1986                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1987                        dev->addr_len);
1988                 err = dev_set_mac_address(dev, sa);
1989                 kfree(sa);
1990                 if (err)
1991                         goto errout;
1992                 status |= DO_SETLINK_MODIFIED;
1993         }
1994
1995         if (tb[IFLA_MTU]) {
1996                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1997                 if (err < 0)
1998                         goto errout;
1999                 status |= DO_SETLINK_MODIFIED;
2000         }
2001
2002         if (tb[IFLA_GROUP]) {
2003                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2004                 status |= DO_SETLINK_NOTIFY;
2005         }
2006
2007         /*
2008          * Interface selected by interface index but interface
2009          * name provided implies that a name change has been
2010          * requested.
2011          */
2012         if (ifm->ifi_index > 0 && ifname[0]) {
2013                 err = dev_change_name(dev, ifname);
2014                 if (err < 0)
2015                         goto errout;
2016                 status |= DO_SETLINK_MODIFIED;
2017         }
2018
2019         if (tb[IFLA_IFALIAS]) {
2020                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2021                                     nla_len(tb[IFLA_IFALIAS]));
2022                 if (err < 0)
2023                         goto errout;
2024                 status |= DO_SETLINK_NOTIFY;
2025         }
2026
2027         if (tb[IFLA_BROADCAST]) {
2028                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2029                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2030         }
2031
2032         if (ifm->ifi_flags || ifm->ifi_change) {
2033                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2034                 if (err < 0)
2035                         goto errout;
2036         }
2037
2038         if (tb[IFLA_MASTER]) {
2039                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2040                 if (err)
2041                         goto errout;
2042                 status |= DO_SETLINK_MODIFIED;
2043         }
2044
2045         if (tb[IFLA_CARRIER]) {
2046                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2047                 if (err)
2048                         goto errout;
2049                 status |= DO_SETLINK_MODIFIED;
2050         }
2051
2052         if (tb[IFLA_TXQLEN]) {
2053                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2054                 unsigned long orig_len = dev->tx_queue_len;
2055
2056                 if (dev->tx_queue_len ^ value) {
2057                         dev->tx_queue_len = value;
2058                         err = call_netdevice_notifiers(
2059                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2060                         err = notifier_to_errno(err);
2061                         if (err) {
2062                                 dev->tx_queue_len = orig_len;
2063                                 goto errout;
2064                         }
2065                         status |= DO_SETLINK_NOTIFY;
2066                 }
2067         }
2068
2069         if (tb[IFLA_OPERSTATE])
2070                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2071
2072         if (tb[IFLA_LINKMODE]) {
2073                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2074
2075                 write_lock_bh(&dev_base_lock);
2076                 if (dev->link_mode ^ value)
2077                         status |= DO_SETLINK_NOTIFY;
2078                 dev->link_mode = value;
2079                 write_unlock_bh(&dev_base_lock);
2080         }
2081
2082         if (tb[IFLA_VFINFO_LIST]) {
2083                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2084                 struct nlattr *attr;
2085                 int rem;
2086
2087                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2088                         if (nla_type(attr) != IFLA_VF_INFO ||
2089                             nla_len(attr) < NLA_HDRLEN) {
2090                                 err = -EINVAL;
2091                                 goto errout;
2092                         }
2093                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2094                                                ifla_vf_policy, NULL);
2095                         if (err < 0)
2096                                 goto errout;
2097                         err = do_setvfinfo(dev, vfinfo);
2098                         if (err < 0)
2099                                 goto errout;
2100                         status |= DO_SETLINK_NOTIFY;
2101                 }
2102         }
2103         err = 0;
2104
2105         if (tb[IFLA_VF_PORTS]) {
2106                 struct nlattr *port[IFLA_PORT_MAX+1];
2107                 struct nlattr *attr;
2108                 int vf;
2109                 int rem;
2110
2111                 err = -EOPNOTSUPP;
2112                 if (!ops->ndo_set_vf_port)
2113                         goto errout;
2114
2115                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2116                         if (nla_type(attr) != IFLA_VF_PORT ||
2117                             nla_len(attr) < NLA_HDRLEN) {
2118                                 err = -EINVAL;
2119                                 goto errout;
2120                         }
2121                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2122                                                ifla_port_policy, NULL);
2123                         if (err < 0)
2124                                 goto errout;
2125                         if (!port[IFLA_PORT_VF]) {
2126                                 err = -EOPNOTSUPP;
2127                                 goto errout;
2128                         }
2129                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2130                         err = ops->ndo_set_vf_port(dev, vf, port);
2131                         if (err < 0)
2132                                 goto errout;
2133                         status |= DO_SETLINK_NOTIFY;
2134                 }
2135         }
2136         err = 0;
2137
2138         if (tb[IFLA_PORT_SELF]) {
2139                 struct nlattr *port[IFLA_PORT_MAX+1];
2140
2141                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2142                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2143                                        NULL);
2144                 if (err < 0)
2145                         goto errout;
2146
2147                 err = -EOPNOTSUPP;
2148                 if (ops->ndo_set_vf_port)
2149                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2150                 if (err < 0)
2151                         goto errout;
2152                 status |= DO_SETLINK_NOTIFY;
2153         }
2154
2155         if (tb[IFLA_AF_SPEC]) {
2156                 struct nlattr *af;
2157                 int rem;
2158
2159                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2160                         const struct rtnl_af_ops *af_ops;
2161
2162                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2163                                 BUG();
2164
2165                         err = af_ops->set_link_af(dev, af);
2166                         if (err < 0)
2167                                 goto errout;
2168
2169                         status |= DO_SETLINK_NOTIFY;
2170                 }
2171         }
2172         err = 0;
2173
2174         if (tb[IFLA_PROTO_DOWN]) {
2175                 err = dev_change_proto_down(dev,
2176                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2177                 if (err)
2178                         goto errout;
2179                 status |= DO_SETLINK_NOTIFY;
2180         }
2181
2182         if (tb[IFLA_XDP]) {
2183                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2184                 u32 xdp_flags = 0;
2185
2186                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2187                                        ifla_xdp_policy, NULL);
2188                 if (err < 0)
2189                         goto errout;
2190
2191                 if (xdp[IFLA_XDP_ATTACHED]) {
2192                         err = -EINVAL;
2193                         goto errout;
2194                 }
2195
2196                 if (xdp[IFLA_XDP_FLAGS]) {
2197                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2198                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2199                                 err = -EINVAL;
2200                                 goto errout;
2201                         }
2202                 }
2203
2204                 if (xdp[IFLA_XDP_FD]) {
2205                         err = dev_change_xdp_fd(dev, extack,
2206                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2207                                                 xdp_flags);
2208                         if (err)
2209                                 goto errout;
2210                         status |= DO_SETLINK_NOTIFY;
2211                 }
2212         }
2213
2214 errout:
2215         if (status & DO_SETLINK_MODIFIED) {
2216                 if (status & DO_SETLINK_NOTIFY)
2217                         netdev_state_change(dev);
2218
2219                 if (err < 0)
2220                         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",
2221                                              dev->name);
2222         }
2223
2224         return err;
2225 }
2226
2227 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2228                         struct netlink_ext_ack *extack)
2229 {
2230         struct net *net = sock_net(skb->sk);
2231         struct ifinfomsg *ifm;
2232         struct net_device *dev;
2233         int err;
2234         struct nlattr *tb[IFLA_MAX+1];
2235         char ifname[IFNAMSIZ];
2236
2237         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2238                           extack);
2239         if (err < 0)
2240                 goto errout;
2241
2242         if (tb[IFLA_IFNAME])
2243                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2244         else
2245                 ifname[0] = '\0';
2246
2247         err = -EINVAL;
2248         ifm = nlmsg_data(nlh);
2249         if (ifm->ifi_index > 0)
2250                 dev = __dev_get_by_index(net, ifm->ifi_index);
2251         else if (tb[IFLA_IFNAME])
2252                 dev = __dev_get_by_name(net, ifname);
2253         else
2254                 goto errout;
2255
2256         if (dev == NULL) {
2257                 err = -ENODEV;
2258                 goto errout;
2259         }
2260
2261         err = validate_linkmsg(dev, tb);
2262         if (err < 0)
2263                 goto errout;
2264
2265         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2266 errout:
2267         return err;
2268 }
2269
2270 static int rtnl_group_dellink(const struct net *net, int group)
2271 {
2272         struct net_device *dev, *aux;
2273         LIST_HEAD(list_kill);
2274         bool found = false;
2275
2276         if (!group)
2277                 return -EPERM;
2278
2279         for_each_netdev(net, dev) {
2280                 if (dev->group == group) {
2281                         const struct rtnl_link_ops *ops;
2282
2283                         found = true;
2284                         ops = dev->rtnl_link_ops;
2285                         if (!ops || !ops->dellink)
2286                                 return -EOPNOTSUPP;
2287                 }
2288         }
2289
2290         if (!found)
2291                 return -ENODEV;
2292
2293         for_each_netdev_safe(net, dev, aux) {
2294                 if (dev->group == group) {
2295                         const struct rtnl_link_ops *ops;
2296
2297                         ops = dev->rtnl_link_ops;
2298                         ops->dellink(dev, &list_kill);
2299                 }
2300         }
2301         unregister_netdevice_many(&list_kill);
2302
2303         return 0;
2304 }
2305
2306 int rtnl_delete_link(struct net_device *dev)
2307 {
2308         const struct rtnl_link_ops *ops;
2309         LIST_HEAD(list_kill);
2310
2311         ops = dev->rtnl_link_ops;
2312         if (!ops || !ops->dellink)
2313                 return -EOPNOTSUPP;
2314
2315         ops->dellink(dev, &list_kill);
2316         unregister_netdevice_many(&list_kill);
2317
2318         return 0;
2319 }
2320 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2321
2322 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2323                         struct netlink_ext_ack *extack)
2324 {
2325         struct net *net = sock_net(skb->sk);
2326         struct net_device *dev;
2327         struct ifinfomsg *ifm;
2328         char ifname[IFNAMSIZ];
2329         struct nlattr *tb[IFLA_MAX+1];
2330         int err;
2331
2332         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2333         if (err < 0)
2334                 return err;
2335
2336         if (tb[IFLA_IFNAME])
2337                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2338
2339         ifm = nlmsg_data(nlh);
2340         if (ifm->ifi_index > 0)
2341                 dev = __dev_get_by_index(net, ifm->ifi_index);
2342         else if (tb[IFLA_IFNAME])
2343                 dev = __dev_get_by_name(net, ifname);
2344         else if (tb[IFLA_GROUP])
2345                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2346         else
2347                 return -EINVAL;
2348
2349         if (!dev)
2350                 return -ENODEV;
2351
2352         return rtnl_delete_link(dev);
2353 }
2354
2355 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2356 {
2357         unsigned int old_flags;
2358         int err;
2359
2360         old_flags = dev->flags;
2361         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2362                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2363                 if (err < 0)
2364                         return err;
2365         }
2366
2367         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2368
2369         __dev_notify_flags(dev, old_flags, ~0U);
2370         return 0;
2371 }
2372 EXPORT_SYMBOL(rtnl_configure_link);
2373
2374 struct net_device *rtnl_create_link(struct net *net,
2375         const char *ifname, unsigned char name_assign_type,
2376         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2377 {
2378         struct net_device *dev;
2379         unsigned int num_tx_queues = 1;
2380         unsigned int num_rx_queues = 1;
2381
2382         if (tb[IFLA_NUM_TX_QUEUES])
2383                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2384         else if (ops->get_num_tx_queues)
2385                 num_tx_queues = ops->get_num_tx_queues();
2386
2387         if (tb[IFLA_NUM_RX_QUEUES])
2388                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2389         else if (ops->get_num_rx_queues)
2390                 num_rx_queues = ops->get_num_rx_queues();
2391
2392         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2393                                ops->setup, num_tx_queues, num_rx_queues);
2394         if (!dev)
2395                 return ERR_PTR(-ENOMEM);
2396
2397         dev_net_set(dev, net);
2398         dev->rtnl_link_ops = ops;
2399         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2400
2401         if (tb[IFLA_MTU])
2402                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2403         if (tb[IFLA_ADDRESS]) {
2404                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2405                                 nla_len(tb[IFLA_ADDRESS]));
2406                 dev->addr_assign_type = NET_ADDR_SET;
2407         }
2408         if (tb[IFLA_BROADCAST])
2409                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2410                                 nla_len(tb[IFLA_BROADCAST]));
2411         if (tb[IFLA_TXQLEN])
2412                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2413         if (tb[IFLA_OPERSTATE])
2414                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2415         if (tb[IFLA_LINKMODE])
2416                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2417         if (tb[IFLA_GROUP])
2418                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2419
2420         return dev;
2421 }
2422 EXPORT_SYMBOL(rtnl_create_link);
2423
2424 static int rtnl_group_changelink(const struct sk_buff *skb,
2425                 struct net *net, int group,
2426                 struct ifinfomsg *ifm,
2427                 struct netlink_ext_ack *extack,
2428                 struct nlattr **tb)
2429 {
2430         struct net_device *dev, *aux;
2431         int err;
2432
2433         for_each_netdev_safe(net, dev, aux) {
2434                 if (dev->group == group) {
2435                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2436                         if (err < 0)
2437                                 return err;
2438                 }
2439         }
2440
2441         return 0;
2442 }
2443
2444 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2445                         struct netlink_ext_ack *extack)
2446 {
2447         struct net *net = sock_net(skb->sk);
2448         const struct rtnl_link_ops *ops;
2449         const struct rtnl_link_ops *m_ops = NULL;
2450         struct net_device *dev;
2451         struct net_device *master_dev = NULL;
2452         struct ifinfomsg *ifm;
2453         char kind[MODULE_NAME_LEN];
2454         char ifname[IFNAMSIZ];
2455         struct nlattr *tb[IFLA_MAX+1];
2456         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2457         unsigned char name_assign_type = NET_NAME_USER;
2458         int err;
2459
2460 #ifdef CONFIG_MODULES
2461 replay:
2462 #endif
2463         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2464         if (err < 0)
2465                 return err;
2466
2467         if (tb[IFLA_IFNAME])
2468                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2469         else
2470                 ifname[0] = '\0';
2471
2472         ifm = nlmsg_data(nlh);
2473         if (ifm->ifi_index > 0)
2474                 dev = __dev_get_by_index(net, ifm->ifi_index);
2475         else {
2476                 if (ifname[0])
2477                         dev = __dev_get_by_name(net, ifname);
2478                 else
2479                         dev = NULL;
2480         }
2481
2482         if (dev) {
2483                 master_dev = netdev_master_upper_dev_get(dev);
2484                 if (master_dev)
2485                         m_ops = master_dev->rtnl_link_ops;
2486         }
2487
2488         err = validate_linkmsg(dev, tb);
2489         if (err < 0)
2490                 return err;
2491
2492         if (tb[IFLA_LINKINFO]) {
2493                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2494                                        tb[IFLA_LINKINFO], ifla_info_policy,
2495                                        NULL);
2496                 if (err < 0)
2497                         return err;
2498         } else
2499                 memset(linkinfo, 0, sizeof(linkinfo));
2500
2501         if (linkinfo[IFLA_INFO_KIND]) {
2502                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2503                 ops = rtnl_link_ops_get(kind);
2504         } else {
2505                 kind[0] = '\0';
2506                 ops = NULL;
2507         }
2508
2509         if (1) {
2510                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2511                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2512                 struct nlattr **data = NULL;
2513                 struct nlattr **slave_data = NULL;
2514                 struct net *dest_net, *link_net = NULL;
2515
2516                 if (ops) {
2517                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2518                                 err = nla_parse_nested(attr, ops->maxtype,
2519                                                        linkinfo[IFLA_INFO_DATA],
2520                                                        ops->policy, NULL);
2521                                 if (err < 0)
2522                                         return err;
2523                                 data = attr;
2524                         }
2525                         if (ops->validate) {
2526                                 err = ops->validate(tb, data);
2527                                 if (err < 0)
2528                                         return err;
2529                         }
2530                 }
2531
2532                 if (m_ops) {
2533                         if (m_ops->slave_maxtype &&
2534                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2535                                 err = nla_parse_nested(slave_attr,
2536                                                        m_ops->slave_maxtype,
2537                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2538                                                        m_ops->slave_policy,
2539                                                        NULL);
2540                                 if (err < 0)
2541                                         return err;
2542                                 slave_data = slave_attr;
2543                         }
2544                         if (m_ops->slave_validate) {
2545                                 err = m_ops->slave_validate(tb, slave_data);
2546                                 if (err < 0)
2547                                         return err;
2548                         }
2549                 }
2550
2551                 if (dev) {
2552                         int status = 0;
2553
2554                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2555                                 return -EEXIST;
2556                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2557                                 return -EOPNOTSUPP;
2558
2559                         if (linkinfo[IFLA_INFO_DATA]) {
2560                                 if (!ops || ops != dev->rtnl_link_ops ||
2561                                     !ops->changelink)
2562                                         return -EOPNOTSUPP;
2563
2564                                 err = ops->changelink(dev, tb, data);
2565                                 if (err < 0)
2566                                         return err;
2567                                 status |= DO_SETLINK_NOTIFY;
2568                         }
2569
2570                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2571                                 if (!m_ops || !m_ops->slave_changelink)
2572                                         return -EOPNOTSUPP;
2573
2574                                 err = m_ops->slave_changelink(master_dev, dev,
2575                                                               tb, slave_data);
2576                                 if (err < 0)
2577                                         return err;
2578                                 status |= DO_SETLINK_NOTIFY;
2579                         }
2580
2581                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2582                                           status);
2583                 }
2584
2585                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2586                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2587                                 return rtnl_group_changelink(skb, net,
2588                                                 nla_get_u32(tb[IFLA_GROUP]),
2589                                                 ifm, extack, tb);
2590                         return -ENODEV;
2591                 }
2592
2593                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2594                         return -EOPNOTSUPP;
2595
2596                 if (!ops) {
2597 #ifdef CONFIG_MODULES
2598                         if (kind[0]) {
2599                                 __rtnl_unlock();
2600                                 request_module("rtnl-link-%s", kind);
2601                                 rtnl_lock();
2602                                 ops = rtnl_link_ops_get(kind);
2603                                 if (ops)
2604                                         goto replay;
2605                         }
2606 #endif
2607                         return -EOPNOTSUPP;
2608                 }
2609
2610                 if (!ops->setup)
2611                         return -EOPNOTSUPP;
2612
2613                 if (!ifname[0]) {
2614                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2615                         name_assign_type = NET_NAME_ENUM;
2616                 }
2617
2618                 dest_net = rtnl_link_get_net(net, tb);
2619                 if (IS_ERR(dest_net))
2620                         return PTR_ERR(dest_net);
2621
2622                 err = -EPERM;
2623                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2624                         goto out;
2625
2626                 if (tb[IFLA_LINK_NETNSID]) {
2627                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2628
2629                         link_net = get_net_ns_by_id(dest_net, id);
2630                         if (!link_net) {
2631                                 err =  -EINVAL;
2632                                 goto out;
2633                         }
2634                         err = -EPERM;
2635                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2636                                 goto out;
2637                 }
2638
2639                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2640                                        name_assign_type, ops, tb);
2641                 if (IS_ERR(dev)) {
2642                         err = PTR_ERR(dev);
2643                         goto out;
2644                 }
2645
2646                 dev->ifindex = ifm->ifi_index;
2647
2648                 if (ops->newlink) {
2649                         err = ops->newlink(link_net ? : net, dev, tb, data);
2650                         /* Drivers should call free_netdev() in ->destructor
2651                          * and unregister it on failure after registration
2652                          * so that device could be finally freed in rtnl_unlock.
2653                          */
2654                         if (err < 0) {
2655                                 /* If device is not registered at all, free it now */
2656                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2657                                         free_netdev(dev);
2658                                 goto out;
2659                         }
2660                 } else {
2661                         err = register_netdevice(dev);
2662                         if (err < 0) {
2663                                 free_netdev(dev);
2664                                 goto out;
2665                         }
2666                 }
2667                 err = rtnl_configure_link(dev, ifm);
2668                 if (err < 0)
2669                         goto out_unregister;
2670                 if (link_net) {
2671                         err = dev_change_net_namespace(dev, dest_net, ifname);
2672                         if (err < 0)
2673                                 goto out_unregister;
2674                 }
2675                 if (tb[IFLA_MASTER]) {
2676                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2677                         if (err)
2678                                 goto out_unregister;
2679                 }
2680 out:
2681                 if (link_net)
2682                         put_net(link_net);
2683                 put_net(dest_net);
2684                 return err;
2685 out_unregister:
2686                 if (ops->newlink) {
2687                         LIST_HEAD(list_kill);
2688
2689                         ops->dellink(dev, &list_kill);
2690                         unregister_netdevice_many(&list_kill);
2691                 } else {
2692                         unregister_netdevice(dev);
2693                 }
2694                 goto out;
2695         }
2696 }
2697
2698 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2699                         struct netlink_ext_ack *extack)
2700 {
2701         struct net *net = sock_net(skb->sk);
2702         struct ifinfomsg *ifm;
2703         char ifname[IFNAMSIZ];
2704         struct nlattr *tb[IFLA_MAX+1];
2705         struct net_device *dev = NULL;
2706         struct sk_buff *nskb;
2707         int err;
2708         u32 ext_filter_mask = 0;
2709
2710         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2711         if (err < 0)
2712                 return err;
2713
2714         if (tb[IFLA_IFNAME])
2715                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2716
2717         if (tb[IFLA_EXT_MASK])
2718                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2719
2720         ifm = nlmsg_data(nlh);
2721         if (ifm->ifi_index > 0)
2722                 dev = __dev_get_by_index(net, ifm->ifi_index);
2723         else if (tb[IFLA_IFNAME])
2724                 dev = __dev_get_by_name(net, ifname);
2725         else
2726                 return -EINVAL;
2727
2728         if (dev == NULL)
2729                 return -ENODEV;
2730
2731         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2732         if (nskb == NULL)
2733                 return -ENOBUFS;
2734
2735         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2736                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2737         if (err < 0) {
2738                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2739                 WARN_ON(err == -EMSGSIZE);
2740                 kfree_skb(nskb);
2741         } else
2742                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2743
2744         return err;
2745 }
2746
2747 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2748 {
2749         struct net *net = sock_net(skb->sk);
2750         struct net_device *dev;
2751         struct nlattr *tb[IFLA_MAX+1];
2752         u32 ext_filter_mask = 0;
2753         u16 min_ifinfo_dump_size = 0;
2754         int hdrlen;
2755
2756         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2757         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2758                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2759
2760         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2761                 if (tb[IFLA_EXT_MASK])
2762                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2763         }
2764
2765         if (!ext_filter_mask)
2766                 return NLMSG_GOODSIZE;
2767         /*
2768          * traverse the list of net devices and compute the minimum
2769          * buffer size based upon the filter mask.
2770          */
2771         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2772                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2773                                              if_nlmsg_size(dev,
2774                                                            ext_filter_mask));
2775         }
2776
2777         return nlmsg_total_size(min_ifinfo_dump_size);
2778 }
2779
2780 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2781 {
2782         int idx;
2783         int s_idx = cb->family;
2784
2785         if (s_idx == 0)
2786                 s_idx = 1;
2787         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2788                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2789                 if (idx < s_idx || idx == PF_PACKET)
2790                         continue;
2791                 if (rtnl_msg_handlers[idx] == NULL ||
2792                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2793                         continue;
2794                 if (idx > s_idx) {
2795                         memset(&cb->args[0], 0, sizeof(cb->args));
2796                         cb->prev_seq = 0;
2797                         cb->seq = 0;
2798                 }
2799                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2800                         break;
2801         }
2802         cb->family = idx;
2803
2804         return skb->len;
2805 }
2806
2807 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2808                                        unsigned int change, gfp_t flags)
2809 {
2810         struct net *net = dev_net(dev);
2811         struct sk_buff *skb;
2812         int err = -ENOBUFS;
2813         size_t if_info_size;
2814
2815         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2816         if (skb == NULL)
2817                 goto errout;
2818
2819         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2820         if (err < 0) {
2821                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2822                 WARN_ON(err == -EMSGSIZE);
2823                 kfree_skb(skb);
2824                 goto errout;
2825         }
2826         return skb;
2827 errout:
2828         if (err < 0)
2829                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2830         return NULL;
2831 }
2832
2833 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2834 {
2835         struct net *net = dev_net(dev);
2836
2837         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2838 }
2839
2840 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2841                   gfp_t flags)
2842 {
2843         struct sk_buff *skb;
2844
2845         if (dev->reg_state != NETREG_REGISTERED)
2846                 return;
2847
2848         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2849         if (skb)
2850                 rtmsg_ifinfo_send(skb, dev, flags);
2851 }
2852 EXPORT_SYMBOL(rtmsg_ifinfo);
2853
2854 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2855                                    struct net_device *dev,
2856                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2857                                    int type, unsigned int flags,
2858                                    int nlflags, u16 ndm_state)
2859 {
2860         struct nlmsghdr *nlh;
2861         struct ndmsg *ndm;
2862
2863         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2864         if (!nlh)
2865                 return -EMSGSIZE;
2866
2867         ndm = nlmsg_data(nlh);
2868         ndm->ndm_family  = AF_BRIDGE;
2869         ndm->ndm_pad1    = 0;
2870         ndm->ndm_pad2    = 0;
2871         ndm->ndm_flags   = flags;
2872         ndm->ndm_type    = 0;
2873         ndm->ndm_ifindex = dev->ifindex;
2874         ndm->ndm_state   = ndm_state;
2875
2876         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2877                 goto nla_put_failure;
2878         if (vid)
2879                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2880                         goto nla_put_failure;
2881
2882         nlmsg_end(skb, nlh);
2883         return 0;
2884
2885 nla_put_failure:
2886         nlmsg_cancel(skb, nlh);
2887         return -EMSGSIZE;
2888 }
2889
2890 static inline size_t rtnl_fdb_nlmsg_size(void)
2891 {
2892         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2893                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2894                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2895                0;
2896 }
2897
2898 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2899                             u16 ndm_state)
2900 {
2901         struct net *net = dev_net(dev);
2902         struct sk_buff *skb;
2903         int err = -ENOBUFS;
2904
2905         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2906         if (!skb)
2907                 goto errout;
2908
2909         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2910                                       0, 0, type, NTF_SELF, 0, ndm_state);
2911         if (err < 0) {
2912                 kfree_skb(skb);
2913                 goto errout;
2914         }
2915
2916         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2917         return;
2918 errout:
2919         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2920 }
2921
2922 /**
2923  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2924  */
2925 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2926                      struct nlattr *tb[],
2927                      struct net_device *dev,
2928                      const unsigned char *addr, u16 vid,
2929                      u16 flags)
2930 {
2931         int err = -EINVAL;
2932
2933         /* If aging addresses are supported device will need to
2934          * implement its own handler for this.
2935          */
2936         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2937                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2938                 return err;
2939         }
2940
2941         if (vid) {
2942                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2943                 return err;
2944         }
2945
2946         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2947                 err = dev_uc_add_excl(dev, addr);
2948         else if (is_multicast_ether_addr(addr))
2949                 err = dev_mc_add_excl(dev, addr);
2950
2951         /* Only return duplicate errors if NLM_F_EXCL is set */
2952         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2953                 err = 0;
2954
2955         return err;
2956 }
2957 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2958
2959 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2960 {
2961         u16 vid = 0;
2962
2963         if (vlan_attr) {
2964                 if (nla_len(vlan_attr) != sizeof(u16)) {
2965                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2966                         return -EINVAL;
2967                 }
2968
2969                 vid = nla_get_u16(vlan_attr);
2970
2971                 if (!vid || vid >= VLAN_VID_MASK) {
2972                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2973                                 vid);
2974                         return -EINVAL;
2975                 }
2976         }
2977         *p_vid = vid;
2978         return 0;
2979 }
2980
2981 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
2982                         struct netlink_ext_ack *extack)
2983 {
2984         struct net *net = sock_net(skb->sk);
2985         struct ndmsg *ndm;
2986         struct nlattr *tb[NDA_MAX+1];
2987         struct net_device *dev;
2988         u8 *addr;
2989         u16 vid;
2990         int err;
2991
2992         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
2993         if (err < 0)
2994                 return err;
2995
2996         ndm = nlmsg_data(nlh);
2997         if (ndm->ndm_ifindex == 0) {
2998                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2999                 return -EINVAL;
3000         }
3001
3002         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3003         if (dev == NULL) {
3004                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3005                 return -ENODEV;
3006         }
3007
3008         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3009                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3010                 return -EINVAL;
3011         }
3012
3013         addr = nla_data(tb[NDA_LLADDR]);
3014
3015         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3016         if (err)
3017                 return err;
3018
3019         err = -EOPNOTSUPP;
3020
3021         /* Support fdb on master device the net/bridge default case */
3022         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3023             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3024                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3025                 const struct net_device_ops *ops = br_dev->netdev_ops;
3026
3027                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3028                                        nlh->nlmsg_flags);
3029                 if (err)
3030                         goto out;
3031                 else
3032                         ndm->ndm_flags &= ~NTF_MASTER;
3033         }
3034
3035         /* Embedded bridge, macvlan, and any other device support */
3036         if ((ndm->ndm_flags & NTF_SELF)) {
3037                 if (dev->netdev_ops->ndo_fdb_add)
3038                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3039                                                            vid,
3040                                                            nlh->nlmsg_flags);
3041                 else
3042                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3043                                                nlh->nlmsg_flags);
3044
3045                 if (!err) {
3046                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3047                                         ndm->ndm_state);
3048                         ndm->ndm_flags &= ~NTF_SELF;
3049                 }
3050         }
3051 out:
3052         return err;
3053 }
3054
3055 /**
3056  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3057  */
3058 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3059                      struct nlattr *tb[],
3060                      struct net_device *dev,
3061                      const unsigned char *addr, u16 vid)
3062 {
3063         int err = -EINVAL;
3064
3065         /* If aging addresses are supported device will need to
3066          * implement its own handler for this.
3067          */
3068         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3069                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3070                 return err;
3071         }
3072
3073         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3074                 err = dev_uc_del(dev, addr);
3075         else if (is_multicast_ether_addr(addr))
3076                 err = dev_mc_del(dev, addr);
3077
3078         return err;
3079 }
3080 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3081
3082 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3083                         struct netlink_ext_ack *extack)
3084 {
3085         struct net *net = sock_net(skb->sk);
3086         struct ndmsg *ndm;
3087         struct nlattr *tb[NDA_MAX+1];
3088         struct net_device *dev;
3089         int err = -EINVAL;
3090         __u8 *addr;
3091         u16 vid;
3092
3093         if (!netlink_capable(skb, CAP_NET_ADMIN))
3094                 return -EPERM;
3095
3096         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3097         if (err < 0)
3098                 return err;
3099
3100         ndm = nlmsg_data(nlh);
3101         if (ndm->ndm_ifindex == 0) {
3102                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3103                 return -EINVAL;
3104         }
3105
3106         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3107         if (dev == NULL) {
3108                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3109                 return -ENODEV;
3110         }
3111
3112         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3113                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3114                 return -EINVAL;
3115         }
3116
3117         addr = nla_data(tb[NDA_LLADDR]);
3118
3119         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3120         if (err)
3121                 return err;
3122
3123         err = -EOPNOTSUPP;
3124
3125         /* Support fdb on master device the net/bridge default case */
3126         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3127             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3128                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3129                 const struct net_device_ops *ops = br_dev->netdev_ops;
3130
3131                 if (ops->ndo_fdb_del)
3132                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3133
3134                 if (err)
3135                         goto out;
3136                 else
3137                         ndm->ndm_flags &= ~NTF_MASTER;
3138         }
3139
3140         /* Embedded bridge, macvlan, and any other device support */
3141         if (ndm->ndm_flags & NTF_SELF) {
3142                 if (dev->netdev_ops->ndo_fdb_del)
3143                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3144                                                            vid);
3145                 else
3146                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3147
3148                 if (!err) {
3149                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3150                                         ndm->ndm_state);
3151                         ndm->ndm_flags &= ~NTF_SELF;
3152                 }
3153         }
3154 out:
3155         return err;
3156 }
3157
3158 static int nlmsg_populate_fdb(struct sk_buff *skb,
3159                               struct netlink_callback *cb,
3160                               struct net_device *dev,
3161                               int *idx,
3162                               struct netdev_hw_addr_list *list)
3163 {
3164         struct netdev_hw_addr *ha;
3165         int err;
3166         u32 portid, seq;
3167
3168         portid = NETLINK_CB(cb->skb).portid;
3169         seq = cb->nlh->nlmsg_seq;
3170
3171         list_for_each_entry(ha, &list->list, list) {
3172                 if (*idx < cb->args[2])
3173                         goto skip;
3174
3175                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3176                                               portid, seq,
3177                                               RTM_NEWNEIGH, NTF_SELF,
3178                                               NLM_F_MULTI, NUD_PERMANENT);
3179                 if (err < 0)
3180                         return err;
3181 skip:
3182                 *idx += 1;
3183         }
3184         return 0;
3185 }
3186
3187 /**
3188  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3189  * @nlh: netlink message header
3190  * @dev: netdevice
3191  *
3192  * Default netdevice operation to dump the existing unicast address list.
3193  * Returns number of addresses from list put in skb.
3194  */
3195 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3196                       struct netlink_callback *cb,
3197                       struct net_device *dev,
3198                       struct net_device *filter_dev,
3199                       int *idx)
3200 {
3201         int err;
3202
3203         netif_addr_lock_bh(dev);
3204         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3205         if (err)
3206                 goto out;
3207         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3208 out:
3209         netif_addr_unlock_bh(dev);
3210         return err;
3211 }
3212 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3213
3214 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3215 {
3216         struct net_device *dev;
3217         struct nlattr *tb[IFLA_MAX+1];
3218         struct net_device *br_dev = NULL;
3219         const struct net_device_ops *ops = NULL;
3220         const struct net_device_ops *cops = NULL;
3221         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3222         struct net *net = sock_net(skb->sk);
3223         struct hlist_head *head;
3224         int brport_idx = 0;
3225         int br_idx = 0;
3226         int h, s_h;
3227         int idx = 0, s_idx;
3228         int err = 0;
3229         int fidx = 0;
3230
3231         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3232                         IFLA_MAX, ifla_policy, NULL) == 0) {
3233                 if (tb[IFLA_MASTER])
3234                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3235         }
3236
3237         brport_idx = ifm->ifi_index;
3238
3239         if (br_idx) {
3240                 br_dev = __dev_get_by_index(net, br_idx);
3241                 if (!br_dev)
3242                         return -ENODEV;
3243
3244                 ops = br_dev->netdev_ops;
3245         }
3246
3247         s_h = cb->args[0];
3248         s_idx = cb->args[1];
3249
3250         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3251                 idx = 0;
3252                 head = &net->dev_index_head[h];
3253                 hlist_for_each_entry(dev, head, index_hlist) {
3254
3255                         if (brport_idx && (dev->ifindex != brport_idx))
3256                                 continue;
3257
3258                         if (!br_idx) { /* user did not specify a specific bridge */
3259                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3260                                         br_dev = netdev_master_upper_dev_get(dev);
3261                                         cops = br_dev->netdev_ops;
3262                                 }
3263                         } else {
3264                                 if (dev != br_dev &&
3265                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3266                                         continue;
3267
3268                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3269                                     !(dev->priv_flags & IFF_EBRIDGE))
3270                                         continue;
3271                                 cops = ops;
3272                         }
3273
3274                         if (idx < s_idx)
3275                                 goto cont;
3276
3277                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3278                                 if (cops && cops->ndo_fdb_dump) {
3279                                         err = cops->ndo_fdb_dump(skb, cb,
3280                                                                 br_dev, dev,
3281                                                                 &fidx);
3282                                         if (err == -EMSGSIZE)
3283                                                 goto out;
3284                                 }
3285                         }
3286
3287                         if (dev->netdev_ops->ndo_fdb_dump)
3288                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3289                                                                     dev, NULL,
3290                                                                     &fidx);
3291                         else
3292                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3293                                                         &fidx);
3294                         if (err == -EMSGSIZE)
3295                                 goto out;
3296
3297                         cops = NULL;
3298
3299                         /* reset fdb offset to 0 for rest of the interfaces */
3300                         cb->args[2] = 0;
3301                         fidx = 0;
3302 cont:
3303                         idx++;
3304                 }
3305         }
3306
3307 out:
3308         cb->args[0] = h;
3309         cb->args[1] = idx;
3310         cb->args[2] = fidx;
3311
3312         return skb->len;
3313 }
3314
3315 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3316                                unsigned int attrnum, unsigned int flag)
3317 {
3318         if (mask & flag)
3319                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3320         return 0;
3321 }
3322
3323 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3324                             struct net_device *dev, u16 mode,
3325                             u32 flags, u32 mask, int nlflags,
3326                             u32 filter_mask,
3327                             int (*vlan_fill)(struct sk_buff *skb,
3328                                              struct net_device *dev,
3329                                              u32 filter_mask))
3330 {
3331         struct nlmsghdr *nlh;
3332         struct ifinfomsg *ifm;
3333         struct nlattr *br_afspec;
3334         struct nlattr *protinfo;
3335         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3336         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3337         int err = 0;
3338
3339         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3340         if (nlh == NULL)
3341                 return -EMSGSIZE;
3342
3343         ifm = nlmsg_data(nlh);
3344         ifm->ifi_family = AF_BRIDGE;
3345         ifm->__ifi_pad = 0;
3346         ifm->ifi_type = dev->type;
3347         ifm->ifi_index = dev->ifindex;
3348         ifm->ifi_flags = dev_get_flags(dev);
3349         ifm->ifi_change = 0;
3350
3351
3352         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3353             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3354             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3355             (br_dev &&
3356              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3357             (dev->addr_len &&
3358              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3359             (dev->ifindex != dev_get_iflink(dev) &&
3360              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3361                 goto nla_put_failure;
3362
3363         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3364         if (!br_afspec)
3365                 goto nla_put_failure;
3366
3367         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3368                 nla_nest_cancel(skb, br_afspec);
3369                 goto nla_put_failure;
3370         }
3371
3372         if (mode != BRIDGE_MODE_UNDEF) {
3373                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3374                         nla_nest_cancel(skb, br_afspec);
3375                         goto nla_put_failure;
3376                 }
3377         }
3378         if (vlan_fill) {
3379                 err = vlan_fill(skb, dev, filter_mask);
3380                 if (err) {
3381                         nla_nest_cancel(skb, br_afspec);
3382                         goto nla_put_failure;
3383                 }
3384         }
3385         nla_nest_end(skb, br_afspec);
3386
3387         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3388         if (!protinfo)
3389                 goto nla_put_failure;
3390
3391         if (brport_nla_put_flag(skb, flags, mask,
3392                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3393             brport_nla_put_flag(skb, flags, mask,
3394                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3395             brport_nla_put_flag(skb, flags, mask,
3396                                 IFLA_BRPORT_FAST_LEAVE,
3397                                 BR_MULTICAST_FAST_LEAVE) ||
3398             brport_nla_put_flag(skb, flags, mask,
3399                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3400             brport_nla_put_flag(skb, flags, mask,
3401                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3402             brport_nla_put_flag(skb, flags, mask,
3403                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3404             brport_nla_put_flag(skb, flags, mask,
3405                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3406             brport_nla_put_flag(skb, flags, mask,
3407                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3408                 nla_nest_cancel(skb, protinfo);
3409                 goto nla_put_failure;
3410         }
3411
3412         nla_nest_end(skb, protinfo);
3413
3414         nlmsg_end(skb, nlh);
3415         return 0;
3416 nla_put_failure:
3417         nlmsg_cancel(skb, nlh);
3418         return err ? err : -EMSGSIZE;
3419 }
3420 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3421
3422 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3423 {
3424         struct net *net = sock_net(skb->sk);
3425         struct net_device *dev;
3426         int idx = 0;
3427         u32 portid = NETLINK_CB(cb->skb).portid;
3428         u32 seq = cb->nlh->nlmsg_seq;
3429         u32 filter_mask = 0;
3430         int err;
3431
3432         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3433                 struct nlattr *extfilt;
3434
3435                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3436                                           IFLA_EXT_MASK);
3437                 if (extfilt) {
3438                         if (nla_len(extfilt) < sizeof(filter_mask))
3439                                 return -EINVAL;
3440
3441                         filter_mask = nla_get_u32(extfilt);
3442                 }
3443         }
3444
3445         rcu_read_lock();
3446         for_each_netdev_rcu(net, dev) {
3447                 const struct net_device_ops *ops = dev->netdev_ops;
3448                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3449
3450                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3451                         if (idx >= cb->args[0]) {
3452                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3453                                                 skb, portid, seq, dev,
3454                                                 filter_mask, NLM_F_MULTI);
3455                                 if (err < 0 && err != -EOPNOTSUPP)
3456                                         break;
3457                         }
3458                         idx++;
3459                 }
3460
3461                 if (ops->ndo_bridge_getlink) {
3462                         if (idx >= cb->args[0]) {
3463                                 err = ops->ndo_bridge_getlink(skb, portid,
3464                                                               seq, dev,
3465                                                               filter_mask,
3466                                                               NLM_F_MULTI);
3467                                 if (err < 0 && err != -EOPNOTSUPP)
3468                                         break;
3469                         }
3470                         idx++;
3471                 }
3472         }
3473         rcu_read_unlock();
3474         cb->args[0] = idx;
3475
3476         return skb->len;
3477 }
3478
3479 static inline size_t bridge_nlmsg_size(void)
3480 {
3481         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3482                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3483                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3484                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3485                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3486                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3487                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3488                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3489                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3490                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3491                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3492 }
3493
3494 static int rtnl_bridge_notify(struct net_device *dev)
3495 {
3496         struct net *net = dev_net(dev);
3497         struct sk_buff *skb;
3498         int err = -EOPNOTSUPP;
3499
3500         if (!dev->netdev_ops->ndo_bridge_getlink)
3501                 return 0;
3502
3503         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3504         if (!skb) {
3505                 err = -ENOMEM;
3506                 goto errout;
3507         }
3508
3509         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3510         if (err < 0)
3511                 goto errout;
3512
3513         if (!skb->len)
3514                 goto errout;
3515
3516         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3517         return 0;
3518 errout:
3519         WARN_ON(err == -EMSGSIZE);
3520         kfree_skb(skb);
3521         if (err)
3522                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3523         return err;
3524 }
3525
3526 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3527                                struct netlink_ext_ack *extack)
3528 {
3529         struct net *net = sock_net(skb->sk);
3530         struct ifinfomsg *ifm;
3531         struct net_device *dev;
3532         struct nlattr *br_spec, *attr = NULL;
3533         int rem, err = -EOPNOTSUPP;
3534         u16 flags = 0;
3535         bool have_flags = false;
3536
3537         if (nlmsg_len(nlh) < sizeof(*ifm))
3538                 return -EINVAL;
3539
3540         ifm = nlmsg_data(nlh);
3541         if (ifm->ifi_family != AF_BRIDGE)
3542                 return -EPFNOSUPPORT;
3543
3544         dev = __dev_get_by_index(net, ifm->ifi_index);
3545         if (!dev) {
3546                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3547                 return -ENODEV;
3548         }
3549
3550         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3551         if (br_spec) {
3552                 nla_for_each_nested(attr, br_spec, rem) {
3553                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3554                                 if (nla_len(attr) < sizeof(flags))
3555                                         return -EINVAL;
3556
3557                                 have_flags = true;
3558                                 flags = nla_get_u16(attr);
3559                                 break;
3560                         }
3561                 }
3562         }
3563
3564         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3565                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3566
3567                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3568                         err = -EOPNOTSUPP;
3569                         goto out;
3570                 }
3571
3572                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3573                 if (err)
3574                         goto out;
3575
3576                 flags &= ~BRIDGE_FLAGS_MASTER;
3577         }
3578
3579         if ((flags & BRIDGE_FLAGS_SELF)) {
3580                 if (!dev->netdev_ops->ndo_bridge_setlink)
3581                         err = -EOPNOTSUPP;
3582                 else
3583                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3584                                                                   flags);
3585                 if (!err) {
3586                         flags &= ~BRIDGE_FLAGS_SELF;
3587
3588                         /* Generate event to notify upper layer of bridge
3589                          * change
3590                          */
3591                         err = rtnl_bridge_notify(dev);
3592                 }
3593         }
3594
3595         if (have_flags)
3596                 memcpy(nla_data(attr), &flags, sizeof(flags));
3597 out:
3598         return err;
3599 }
3600
3601 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3602                                struct netlink_ext_ack *extack)
3603 {
3604         struct net *net = sock_net(skb->sk);
3605         struct ifinfomsg *ifm;
3606         struct net_device *dev;
3607         struct nlattr *br_spec, *attr = NULL;
3608         int rem, err = -EOPNOTSUPP;
3609         u16 flags = 0;
3610         bool have_flags = false;
3611
3612         if (nlmsg_len(nlh) < sizeof(*ifm))
3613                 return -EINVAL;
3614
3615         ifm = nlmsg_data(nlh);
3616         if (ifm->ifi_family != AF_BRIDGE)
3617                 return -EPFNOSUPPORT;
3618
3619         dev = __dev_get_by_index(net, ifm->ifi_index);
3620         if (!dev) {
3621                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3622                 return -ENODEV;
3623         }
3624
3625         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3626         if (br_spec) {
3627                 nla_for_each_nested(attr, br_spec, rem) {
3628                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3629                                 if (nla_len(attr) < sizeof(flags))
3630                                         return -EINVAL;
3631
3632                                 have_flags = true;
3633                                 flags = nla_get_u16(attr);
3634                                 break;
3635                         }
3636                 }
3637         }
3638
3639         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3640                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3641
3642                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3643                         err = -EOPNOTSUPP;
3644                         goto out;
3645                 }
3646
3647                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3648                 if (err)
3649                         goto out;
3650
3651                 flags &= ~BRIDGE_FLAGS_MASTER;
3652         }
3653
3654         if ((flags & BRIDGE_FLAGS_SELF)) {
3655                 if (!dev->netdev_ops->ndo_bridge_dellink)
3656                         err = -EOPNOTSUPP;
3657                 else
3658                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3659                                                                   flags);
3660
3661                 if (!err) {
3662                         flags &= ~BRIDGE_FLAGS_SELF;
3663
3664                         /* Generate event to notify upper layer of bridge
3665                          * change
3666                          */
3667                         err = rtnl_bridge_notify(dev);
3668                 }
3669         }
3670
3671         if (have_flags)
3672                 memcpy(nla_data(attr), &flags, sizeof(flags));
3673 out:
3674         return err;
3675 }
3676
3677 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3678 {
3679         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3680                (!idxattr || idxattr == attrid);
3681 }
3682
3683 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3684 static int rtnl_get_offload_stats_attr_size(int attr_id)
3685 {
3686         switch (attr_id) {
3687         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3688                 return sizeof(struct rtnl_link_stats64);
3689         }
3690
3691         return 0;
3692 }
3693
3694 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3695                                   int *prividx)
3696 {
3697         struct nlattr *attr = NULL;
3698         int attr_id, size;
3699         void *attr_data;
3700         int err;
3701
3702         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3703               dev->netdev_ops->ndo_get_offload_stats))
3704                 return -ENODATA;
3705
3706         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3707              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3708                 if (attr_id < *prividx)
3709                         continue;
3710
3711                 size = rtnl_get_offload_stats_attr_size(attr_id);
3712                 if (!size)
3713                         continue;
3714
3715                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3716                         continue;
3717
3718                 attr = nla_reserve_64bit(skb, attr_id, size,
3719                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3720                 if (!attr)
3721                         goto nla_put_failure;
3722
3723                 attr_data = nla_data(attr);
3724                 memset(attr_data, 0, size);
3725                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3726                                                              attr_data);
3727                 if (err)
3728                         goto get_offload_stats_failure;
3729         }
3730
3731         if (!attr)
3732                 return -ENODATA;
3733
3734         *prividx = 0;
3735         return 0;
3736
3737 nla_put_failure:
3738         err = -EMSGSIZE;
3739 get_offload_stats_failure:
3740         *prividx = attr_id;
3741         return err;
3742 }
3743
3744 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3745 {
3746         int nla_size = 0;
3747         int attr_id;
3748         int size;
3749
3750         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3751               dev->netdev_ops->ndo_get_offload_stats))
3752                 return 0;
3753
3754         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3755              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3756                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3757                         continue;
3758                 size = rtnl_get_offload_stats_attr_size(attr_id);
3759                 nla_size += nla_total_size_64bit(size);
3760         }
3761
3762         if (nla_size != 0)
3763                 nla_size += nla_total_size(0);
3764
3765         return nla_size;
3766 }
3767
3768 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3769                                int type, u32 pid, u32 seq, u32 change,
3770                                unsigned int flags, unsigned int filter_mask,
3771                                int *idxattr, int *prividx)
3772 {
3773         struct if_stats_msg *ifsm;
3774         struct nlmsghdr *nlh;
3775         struct nlattr *attr;
3776         int s_prividx = *prividx;
3777         int err;
3778
3779         ASSERT_RTNL();
3780
3781         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3782         if (!nlh)
3783                 return -EMSGSIZE;
3784
3785         ifsm = nlmsg_data(nlh);
3786         ifsm->ifindex = dev->ifindex;
3787         ifsm->filter_mask = filter_mask;
3788
3789         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3790                 struct rtnl_link_stats64 *sp;
3791
3792                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3793                                          sizeof(struct rtnl_link_stats64),
3794                                          IFLA_STATS_UNSPEC);
3795                 if (!attr)
3796                         goto nla_put_failure;
3797
3798                 sp = nla_data(attr);
3799                 dev_get_stats(dev, sp);
3800         }
3801
3802         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3803                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3804
3805                 if (ops && ops->fill_linkxstats) {
3806                         *idxattr = IFLA_STATS_LINK_XSTATS;
3807                         attr = nla_nest_start(skb,
3808                                               IFLA_STATS_LINK_XSTATS);
3809                         if (!attr)
3810                                 goto nla_put_failure;
3811
3812                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3813                         nla_nest_end(skb, attr);
3814                         if (err)
3815                                 goto nla_put_failure;
3816                         *idxattr = 0;
3817                 }
3818         }
3819
3820         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3821                              *idxattr)) {
3822                 const struct rtnl_link_ops *ops = NULL;
3823                 const struct net_device *master;
3824
3825                 master = netdev_master_upper_dev_get(dev);
3826                 if (master)
3827                         ops = master->rtnl_link_ops;
3828                 if (ops && ops->fill_linkxstats) {
3829                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3830                         attr = nla_nest_start(skb,
3831                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3832                         if (!attr)
3833                                 goto nla_put_failure;
3834
3835                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3836                         nla_nest_end(skb, attr);
3837                         if (err)
3838                                 goto nla_put_failure;
3839                         *idxattr = 0;
3840                 }
3841         }
3842
3843         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3844                              *idxattr)) {
3845                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3846                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3847                 if (!attr)
3848                         goto nla_put_failure;
3849
3850                 err = rtnl_get_offload_stats(skb, dev, prividx);
3851                 if (err == -ENODATA)
3852                         nla_nest_cancel(skb, attr);
3853                 else
3854                         nla_nest_end(skb, attr);
3855
3856                 if (err && err != -ENODATA)
3857                         goto nla_put_failure;
3858                 *idxattr = 0;
3859         }
3860
3861         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3862                 struct rtnl_af_ops *af_ops;
3863
3864                 *idxattr = IFLA_STATS_AF_SPEC;
3865                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3866                 if (!attr)
3867                         goto nla_put_failure;
3868
3869                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3870                         if (af_ops->fill_stats_af) {
3871                                 struct nlattr *af;
3872                                 int err;
3873
3874                                 af = nla_nest_start(skb, af_ops->family);
3875                                 if (!af)
3876                                         goto nla_put_failure;
3877
3878                                 err = af_ops->fill_stats_af(skb, dev);
3879
3880                                 if (err == -ENODATA)
3881                                         nla_nest_cancel(skb, af);
3882                                 else if (err < 0)
3883                                         goto nla_put_failure;
3884
3885                                 nla_nest_end(skb, af);
3886                         }
3887                 }
3888
3889                 nla_nest_end(skb, attr);
3890
3891                 *idxattr = 0;
3892         }
3893
3894         nlmsg_end(skb, nlh);
3895
3896         return 0;
3897
3898 nla_put_failure:
3899         /* not a multi message or no progress mean a real error */
3900         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3901                 nlmsg_cancel(skb, nlh);
3902         else
3903                 nlmsg_end(skb, nlh);
3904
3905         return -EMSGSIZE;
3906 }
3907
3908 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3909                                   u32 filter_mask)
3910 {
3911         size_t size = 0;
3912
3913         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3914                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3915
3916         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3917                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3918                 int attr = IFLA_STATS_LINK_XSTATS;
3919
3920                 if (ops && ops->get_linkxstats_size) {
3921                         size += nla_total_size(ops->get_linkxstats_size(dev,
3922                                                                         attr));
3923                         /* for IFLA_STATS_LINK_XSTATS */
3924                         size += nla_total_size(0);
3925                 }
3926         }
3927
3928         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3929                 struct net_device *_dev = (struct net_device *)dev;
3930                 const struct rtnl_link_ops *ops = NULL;
3931                 const struct net_device *master;
3932
3933                 /* netdev_master_upper_dev_get can't take const */
3934                 master = netdev_master_upper_dev_get(_dev);
3935                 if (master)
3936                         ops = master->rtnl_link_ops;
3937                 if (ops && ops->get_linkxstats_size) {
3938                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3939
3940                         size += nla_total_size(ops->get_linkxstats_size(dev,
3941                                                                         attr));
3942                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3943                         size += nla_total_size(0);
3944                 }
3945         }
3946
3947         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3948                 size += rtnl_get_offload_stats_size(dev);
3949
3950         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3951                 struct rtnl_af_ops *af_ops;
3952
3953                 /* for IFLA_STATS_AF_SPEC */
3954                 size += nla_total_size(0);
3955
3956                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3957                         if (af_ops->get_stats_af_size) {
3958                                 size += nla_total_size(
3959                                         af_ops->get_stats_af_size(dev));
3960
3961                                 /* for AF_* */
3962                                 size += nla_total_size(0);
3963                         }
3964                 }
3965         }
3966
3967         return size;
3968 }
3969
3970 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
3971                           struct netlink_ext_ack *extack)
3972 {
3973         struct net *net = sock_net(skb->sk);
3974         struct net_device *dev = NULL;
3975         int idxattr = 0, prividx = 0;
3976         struct if_stats_msg *ifsm;
3977         struct sk_buff *nskb;
3978         u32 filter_mask;
3979         int err;
3980
3981         if (nlmsg_len(nlh) < sizeof(*ifsm))
3982                 return -EINVAL;
3983
3984         ifsm = nlmsg_data(nlh);
3985         if (ifsm->ifindex > 0)
3986                 dev = __dev_get_by_index(net, ifsm->ifindex);
3987         else
3988                 return -EINVAL;
3989
3990         if (!dev)
3991                 return -ENODEV;
3992
3993         filter_mask = ifsm->filter_mask;
3994         if (!filter_mask)
3995                 return -EINVAL;
3996
3997         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3998         if (!nskb)
3999                 return -ENOBUFS;
4000
4001         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4002                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4003                                   0, filter_mask, &idxattr, &prividx);
4004         if (err < 0) {
4005                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4006                 WARN_ON(err == -EMSGSIZE);
4007                 kfree_skb(nskb);
4008         } else {
4009                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4010         }
4011
4012         return err;
4013 }
4014
4015 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4016 {
4017         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4018         struct net *net = sock_net(skb->sk);
4019         unsigned int flags = NLM_F_MULTI;
4020         struct if_stats_msg *ifsm;
4021         struct hlist_head *head;
4022         struct net_device *dev;
4023         u32 filter_mask = 0;
4024         int idx = 0;
4025
4026         s_h = cb->args[0];
4027         s_idx = cb->args[1];
4028         s_idxattr = cb->args[2];
4029         s_prividx = cb->args[3];
4030
4031         cb->seq = net->dev_base_seq;
4032
4033         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4034                 return -EINVAL;
4035
4036         ifsm = nlmsg_data(cb->nlh);
4037         filter_mask = ifsm->filter_mask;
4038         if (!filter_mask)
4039                 return -EINVAL;
4040
4041         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4042                 idx = 0;
4043                 head = &net->dev_index_head[h];
4044                 hlist_for_each_entry(dev, head, index_hlist) {
4045                         if (idx < s_idx)
4046                                 goto cont;
4047                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4048                                                   NETLINK_CB(cb->skb).portid,
4049                                                   cb->nlh->nlmsg_seq, 0,
4050                                                   flags, filter_mask,
4051                                                   &s_idxattr, &s_prividx);
4052                         /* If we ran out of room on the first message,
4053                          * we're in trouble
4054                          */
4055                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4056
4057                         if (err < 0)
4058                                 goto out;
4059                         s_prividx = 0;
4060                         s_idxattr = 0;
4061                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4062 cont:
4063                         idx++;
4064                 }
4065         }
4066 out:
4067         cb->args[3] = s_prividx;
4068         cb->args[2] = s_idxattr;
4069         cb->args[1] = idx;
4070         cb->args[0] = h;
4071
4072         return skb->len;
4073 }
4074
4075 /* Process one rtnetlink message. */
4076
4077 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4078                              struct netlink_ext_ack *extack)
4079 {
4080         struct net *net = sock_net(skb->sk);
4081         rtnl_doit_func doit;
4082         int kind;
4083         int family;
4084         int type;
4085         int err;
4086
4087         type = nlh->nlmsg_type;
4088         if (type > RTM_MAX)
4089                 return -EOPNOTSUPP;
4090
4091         type -= RTM_BASE;
4092
4093         /* All the messages must have at least 1 byte length */
4094         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4095                 return 0;
4096
4097         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4098         kind = type&3;
4099
4100         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4101                 return -EPERM;
4102
4103         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4104                 struct sock *rtnl;
4105                 rtnl_dumpit_func dumpit;
4106                 rtnl_calcit_func calcit;
4107                 u16 min_dump_alloc = 0;
4108
4109                 dumpit = rtnl_get_dumpit(family, type);
4110                 if (dumpit == NULL)
4111                         return -EOPNOTSUPP;
4112                 calcit = rtnl_get_calcit(family, type);
4113                 if (calcit)
4114                         min_dump_alloc = calcit(skb, nlh);
4115
4116                 __rtnl_unlock();
4117                 rtnl = net->rtnl;
4118                 {
4119                         struct netlink_dump_control c = {
4120                                 .dump           = dumpit,
4121                                 .min_dump_alloc = min_dump_alloc,
4122                         };
4123                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4124                 }
4125                 rtnl_lock();
4126                 return err;
4127         }
4128
4129         doit = rtnl_get_doit(family, type);
4130         if (doit == NULL)
4131                 return -EOPNOTSUPP;
4132
4133         return doit(skb, nlh, extack);
4134 }
4135
4136 static void rtnetlink_rcv(struct sk_buff *skb)
4137 {
4138         rtnl_lock();
4139         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4140         rtnl_unlock();
4141 }
4142
4143 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4144 {
4145         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4146
4147         switch (event) {
4148         case NETDEV_REBOOT:
4149         case NETDEV_CHANGENAME:
4150         case NETDEV_FEAT_CHANGE:
4151         case NETDEV_BONDING_FAILOVER:
4152         case NETDEV_NOTIFY_PEERS:
4153         case NETDEV_RESEND_IGMP:
4154         case NETDEV_CHANGEINFODATA:
4155                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4156                 break;
4157         default:
4158                 break;
4159         }
4160         return NOTIFY_DONE;
4161 }
4162
4163 static struct notifier_block rtnetlink_dev_notifier = {
4164         .notifier_call  = rtnetlink_event,
4165 };
4166
4167
4168 static int __net_init rtnetlink_net_init(struct net *net)
4169 {
4170         struct sock *sk;
4171         struct netlink_kernel_cfg cfg = {
4172                 .groups         = RTNLGRP_MAX,
4173                 .input          = rtnetlink_rcv,
4174                 .cb_mutex       = &rtnl_mutex,
4175                 .flags          = NL_CFG_F_NONROOT_RECV,
4176         };
4177
4178         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4179         if (!sk)
4180                 return -ENOMEM;
4181         net->rtnl = sk;
4182         return 0;
4183 }
4184
4185 static void __net_exit rtnetlink_net_exit(struct net *net)
4186 {
4187         netlink_kernel_release(net->rtnl);
4188         net->rtnl = NULL;
4189 }
4190
4191 static struct pernet_operations rtnetlink_net_ops = {
4192         .init = rtnetlink_net_init,
4193         .exit = rtnetlink_net_exit,
4194 };
4195
4196 void __init rtnetlink_init(void)
4197 {
4198         if (register_pernet_subsys(&rtnetlink_net_ops))
4199                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4200
4201         register_netdevice_notifier(&rtnetlink_dev_notifier);
4202
4203         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4204                       rtnl_dump_ifinfo, rtnl_calcit);
4205         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4206         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4207         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4208
4209         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4210         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4211         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4212
4213         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4214         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4215         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4216
4217         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4218         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4219         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4220
4221         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4222                       NULL);
4223 }