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
 904         return xdp_size;
 905 }
 906
 907 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 908                                      u32 ext_filter_mask)
 909 {
 910         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 911                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 912                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 913                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 914                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 915                + nla_total_size(sizeof(struct rtnl_link_stats))
 916                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 917                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 918                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 919                + nla_total_size(4) /* IFLA_TXQLEN */
 920                + nla_total_size(4) /* IFLA_WEIGHT */
 921                + nla_total_size(4) /* IFLA_MTU */
 922                + nla_total_size(4) /* IFLA_LINK */
 923                + nla_total_size(4) /* IFLA_MASTER */
 924                + nla_total_size(1) /* IFLA_CARRIER */
 925                + nla_total_size(4) /* IFLA_PROMISCUITY */
 926                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 927                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 928                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 929                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 930                + nla_total_size(1) /* IFLA_OPERSTATE */
 931                + nla_total_size(1) /* IFLA_LINKMODE */
 932                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 933                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 934                + nla_total_size(ext_filter_mask
 935                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 936                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 937                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 938                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 939                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 940                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 941                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 942                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 943                + rtnl_xdp_size() /* IFLA_XDP */
 944                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 945
 946 }
 947
 948 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 949 {
 950         struct nlattr *vf_ports;
 951         struct nlattr *vf_port;
 952         int vf;
 953         int err;
 954
 955         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 956         if (!vf_ports)
 957                 return -EMSGSIZE;
 958
 959         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 960                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 961                 if (!vf_port)
 962                         goto nla_put_failure;
 963                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 964                         goto nla_put_failure;
 965                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 966                 if (err == -EMSGSIZE)
 967                         goto nla_put_failure;
 968                 if (err) {
 969                         nla_nest_cancel(skb, vf_port);
 970                         continue;
 971                 }
 972                 nla_nest_end(skb, vf_port);
 973         }
 974
 975         nla_nest_end(skb, vf_ports);
 976
 977         return 0;
 978
 979 nla_put_failure:
 980         nla_nest_cancel(skb, vf_ports);
 981         return -EMSGSIZE;
 982 }
 983
 984 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 985 {
 986         struct nlattr *port_self;
 987         int err;
 988
 989         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 990         if (!port_self)
 991                 return -EMSGSIZE;
 992
 993         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 994         if (err) {
 995                 nla_nest_cancel(skb, port_self);
 996                 return (err == -EMSGSIZE) ? err : 0;
 997         }
 998
 999         nla_nest_end(skb, port_self);
1000
1001         return 0;
1002 }
1003
1004 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1005                           u32 ext_filter_mask)
1006 {
1007         int err;
1008
1009         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1010             !(ext_filter_mask & RTEXT_FILTER_VF))
1011                 return 0;
1012
1013         err = rtnl_port_self_fill(skb, dev);
1014         if (err)
1015                 return err;
1016
1017         if (dev_num_vf(dev->dev.parent)) {
1018                 err = rtnl_vf_ports_fill(skb, dev);
1019                 if (err)
1020                         return err;
1021         }
1022
1023         return 0;
1024 }
1025
1026 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028         int err;
1029         struct netdev_phys_item_id ppid;
1030
1031         err = dev_get_phys_port_id(dev, &ppid);
1032         if (err) {
1033                 if (err == -EOPNOTSUPP)
1034                         return 0;
1035                 return err;
1036         }
1037
1038         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1039                 return -EMSGSIZE;
1040
1041         return 0;
1042 }
1043
1044 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1045 {
1046         char name[IFNAMSIZ];
1047         int err;
1048
1049         err = dev_get_phys_port_name(dev, name, sizeof(name));
1050         if (err) {
1051                 if (err == -EOPNOTSUPP)
1052                         return 0;
1053                 return err;
1054         }
1055
1056         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1057                 return -EMSGSIZE;
1058
1059         return 0;
1060 }
1061
1062 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1063 {
1064         int err;
1065         struct switchdev_attr attr = {
1066                 .orig_dev = dev,
1067                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1068                 .flags = SWITCHDEV_F_NO_RECURSE,
1069         };
1070
1071         err = switchdev_port_attr_get(dev, &attr);
1072         if (err) {
1073                 if (err == -EOPNOTSUPP)
1074                         return 0;
1075                 return err;
1076         }
1077
1078         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1079                     attr.u.ppid.id))
1080                 return -EMSGSIZE;
1081
1082         return 0;
1083 }
1084
1085 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1086                                               struct net_device *dev)
1087 {
1088         struct rtnl_link_stats64 *sp;
1089         struct nlattr *attr;
1090
1091         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1092                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1093         if (!attr)
1094                 return -EMSGSIZE;
1095
1096         sp = nla_data(attr);
1097         dev_get_stats(dev, sp);
1098
1099         attr = nla_reserve(skb, IFLA_STATS,
1100                            sizeof(struct rtnl_link_stats));
1101         if (!attr)
1102                 return -EMSGSIZE;
1103
1104         copy_rtnl_link_stats(nla_data(attr), sp);
1105
1106         return 0;
1107 }
1108
1109 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1110                                                struct net_device *dev,
1111                                                int vfs_num,
1112                                                struct nlattr *vfinfo)
1113 {
1114         struct ifla_vf_rss_query_en vf_rss_query_en;
1115         struct nlattr *vf, *vfstats, *vfvlanlist;
1116         struct ifla_vf_link_state vf_linkstate;
1117         struct ifla_vf_vlan_info vf_vlan_info;
1118         struct ifla_vf_spoofchk vf_spoofchk;
1119         struct ifla_vf_tx_rate vf_tx_rate;
1120         struct ifla_vf_stats vf_stats;
1121         struct ifla_vf_trust vf_trust;
1122         struct ifla_vf_vlan vf_vlan;
1123         struct ifla_vf_rate vf_rate;
1124         struct ifla_vf_mac vf_mac;
1125         struct ifla_vf_info ivi;
1126
1127         /* Not all SR-IOV capable drivers support the
1128          * spoofcheck and "RSS query enable" query.  Preset to
1129          * -1 so the user space tool can detect that the driver
1130          * didn't report anything.
1131          */
1132         ivi.spoofchk = -1;
1133         ivi.rss_query_en = -1;
1134         ivi.trusted = -1;
1135         memset(ivi.mac, 0, sizeof(ivi.mac));
1136         /* The default value for VF link state is "auto"
1137          * IFLA_VF_LINK_STATE_AUTO which equals zero
1138          */
1139         ivi.linkstate = 0;
1140         /* VLAN Protocol by default is 802.1Q */
1141         ivi.vlan_proto = htons(ETH_P_8021Q);
1142         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1143                 return 0;
1144
1145         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1146
1147         vf_mac.vf =
1148                 vf_vlan.vf =
1149                 vf_vlan_info.vf =
1150                 vf_rate.vf =
1151                 vf_tx_rate.vf =
1152                 vf_spoofchk.vf =
1153                 vf_linkstate.vf =
1154                 vf_rss_query_en.vf =
1155                 vf_trust.vf = ivi.vf;
1156
1157         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1158         vf_vlan.vlan = ivi.vlan;
1159         vf_vlan.qos = ivi.qos;
1160         vf_vlan_info.vlan = ivi.vlan;
1161         vf_vlan_info.qos = ivi.qos;
1162         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1163         vf_tx_rate.rate = ivi.max_tx_rate;
1164         vf_rate.min_tx_rate = ivi.min_tx_rate;
1165         vf_rate.max_tx_rate = ivi.max_tx_rate;
1166         vf_spoofchk.setting = ivi.spoofchk;
1167         vf_linkstate.link_state = ivi.linkstate;
1168         vf_rss_query_en.setting = ivi.rss_query_en;
1169         vf_trust.setting = ivi.trusted;
1170         vf = nla_nest_start(skb, IFLA_VF_INFO);
1171         if (!vf)
1172                 goto nla_put_vfinfo_failure;
1173         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1174             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1175             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1176                     &vf_rate) ||
1177             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1178                     &vf_tx_rate) ||
1179             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1180                     &vf_spoofchk) ||
1181             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1182                     &vf_linkstate) ||
1183             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1184                     sizeof(vf_rss_query_en),
1185                     &vf_rss_query_en) ||
1186             nla_put(skb, IFLA_VF_TRUST,
1187                     sizeof(vf_trust), &vf_trust))
1188                 goto nla_put_vf_failure;
1189         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1190         if (!vfvlanlist)
1191                 goto nla_put_vf_failure;
1192         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1193                     &vf_vlan_info)) {
1194                 nla_nest_cancel(skb, vfvlanlist);
1195                 goto nla_put_vf_failure;
1196         }
1197         nla_nest_end(skb, vfvlanlist);
1198         memset(&vf_stats, 0, sizeof(vf_stats));
1199         if (dev->netdev_ops->ndo_get_vf_stats)
1200                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1201                                                 &vf_stats);
1202         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1203         if (!vfstats)
1204                 goto nla_put_vf_failure;
1205         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1206                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1207             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1208                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1209             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1210                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1211             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1212                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1214                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1216                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1217                 nla_nest_cancel(skb, vfstats);
1218                 goto nla_put_vf_failure;
1219         }
1220         nla_nest_end(skb, vfstats);
1221         nla_nest_end(skb, vf);
1222         return 0;
1223
1224 nla_put_vf_failure:
1225         nla_nest_cancel(skb, vf);
1226 nla_put_vfinfo_failure:
1227         nla_nest_cancel(skb, vfinfo);
1228         return -EMSGSIZE;
1229 }
1230
1231 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1232 {
1233         struct rtnl_link_ifmap map;
1234
1235         memset(&map, 0, sizeof(map));
1236         map.mem_start   = dev->mem_start;
1237         map.mem_end     = dev->mem_end;
1238         map.base_addr   = dev->base_addr;
1239         map.irq         = dev->irq;
1240         map.dma         = dev->dma;
1241         map.port        = dev->if_port;
1242
1243         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1244                 return -EMSGSIZE;
1245
1246         return 0;
1247 }
1248
1249 static u8 rtnl_xdp_attached_mode(struct net_device *dev)
1250 {
1251         const struct net_device_ops *ops = dev->netdev_ops;
1252
1253         ASSERT_RTNL();
1254
1255         if (rcu_access_pointer(dev->xdp_prog))
1256                 return XDP_ATTACHED_SKB;
1257         if (ops->ndo_xdp && __dev_xdp_attached(dev, ops->ndo_xdp))
1258                 return XDP_ATTACHED_DRV;
1259
1260         return XDP_ATTACHED_NONE;
1261 }
1262
1263 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1264 {
1265         struct nlattr *xdp;
1266         int err;
1267
1268         xdp = nla_nest_start(skb, IFLA_XDP);
1269         if (!xdp)
1270                 return -EMSGSIZE;
1271
1272         err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1273                          rtnl_xdp_attached_mode(dev));
1274         if (err)
1275                 goto err_cancel;
1276
1277         nla_nest_end(skb, xdp);
1278         return 0;
1279
1280 err_cancel:
1281         nla_nest_cancel(skb, xdp);
1282         return err;
1283 }
1284
1285 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1286                             int type, u32 pid, u32 seq, u32 change,
1287                             unsigned int flags, u32 ext_filter_mask)
1288 {
1289         struct ifinfomsg *ifm;
1290         struct nlmsghdr *nlh;
1291         struct nlattr *af_spec;
1292         struct rtnl_af_ops *af_ops;
1293         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1294
1295         ASSERT_RTNL();
1296         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1297         if (nlh == NULL)
1298                 return -EMSGSIZE;
1299
1300         ifm = nlmsg_data(nlh);
1301         ifm->ifi_family = AF_UNSPEC;
1302         ifm->__ifi_pad = 0;
1303         ifm->ifi_type = dev->type;
1304         ifm->ifi_index = dev->ifindex;
1305         ifm->ifi_flags = dev_get_flags(dev);
1306         ifm->ifi_change = change;
1307
1308         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1309             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1310             nla_put_u8(skb, IFLA_OPERSTATE,
1311                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1312             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1313             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1314             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1315             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1316             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1317             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1318             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1319 #ifdef CONFIG_RPS
1320             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1321 #endif
1322             (dev->ifindex != dev_get_iflink(dev) &&
1323              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1324             (upper_dev &&
1325              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1326             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1327             (dev->qdisc &&
1328              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1329             (dev->ifalias &&
1330              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1331             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1332                         atomic_read(&dev->carrier_changes)) ||
1333             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1334                 goto nla_put_failure;
1335
1336         if (rtnl_fill_link_ifmap(skb, dev))
1337                 goto nla_put_failure;
1338
1339         if (dev->addr_len) {
1340                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1341                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1342                         goto nla_put_failure;
1343         }
1344
1345         if (rtnl_phys_port_id_fill(skb, dev))
1346                 goto nla_put_failure;
1347
1348         if (rtnl_phys_port_name_fill(skb, dev))
1349                 goto nla_put_failure;
1350
1351         if (rtnl_phys_switch_id_fill(skb, dev))
1352                 goto nla_put_failure;
1353
1354         if (rtnl_fill_stats(skb, dev))
1355                 goto nla_put_failure;
1356
1357         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1358             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1359                 goto nla_put_failure;
1360
1361         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1362             ext_filter_mask & RTEXT_FILTER_VF) {
1363                 int i;
1364                 struct nlattr *vfinfo;
1365                 int num_vfs = dev_num_vf(dev->dev.parent);
1366
1367                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1368                 if (!vfinfo)
1369                         goto nla_put_failure;
1370                 for (i = 0; i < num_vfs; i++) {
1371                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1372                                 goto nla_put_failure;
1373                 }
1374
1375                 nla_nest_end(skb, vfinfo);
1376         }
1377
1378         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1379                 goto nla_put_failure;
1380
1381         if (rtnl_xdp_fill(skb, dev))
1382                 goto nla_put_failure;
1383
1384         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1385                 if (rtnl_link_fill(skb, dev) < 0)
1386                         goto nla_put_failure;
1387         }
1388
1389         if (dev->rtnl_link_ops &&
1390             dev->rtnl_link_ops->get_link_net) {
1391                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1392
1393                 if (!net_eq(dev_net(dev), link_net)) {
1394                         int id = peernet2id_alloc(dev_net(dev), link_net);
1395
1396                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1397                                 goto nla_put_failure;
1398                 }
1399         }
1400
1401         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1402                 goto nla_put_failure;
1403
1404         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1405                 if (af_ops->fill_link_af) {
1406                         struct nlattr *af;
1407                         int err;
1408
1409                         if (!(af = nla_nest_start(skb, af_ops->family)))
1410                                 goto nla_put_failure;
1411
1412                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1413
1414                         /*
1415                          * Caller may return ENODATA to indicate that there
1416                          * was no data to be dumped. This is not an error, it
1417                          * means we should trim the attribute header and
1418                          * continue.
1419                          */
1420                         if (err == -ENODATA)
1421                                 nla_nest_cancel(skb, af);
1422                         else if (err < 0)
1423                                 goto nla_put_failure;
1424
1425                         nla_nest_end(skb, af);
1426                 }
1427         }
1428
1429         nla_nest_end(skb, af_spec);
1430
1431         nlmsg_end(skb, nlh);
1432         return 0;
1433
1434 nla_put_failure:
1435         nlmsg_cancel(skb, nlh);
1436         return -EMSGSIZE;
1437 }
1438
1439 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1440         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1441         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1442         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1443         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1444         [IFLA_MTU]              = { .type = NLA_U32 },
1445         [IFLA_LINK]             = { .type = NLA_U32 },
1446         [IFLA_MASTER]           = { .type = NLA_U32 },
1447         [IFLA_CARRIER]          = { .type = NLA_U8 },
1448         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1449         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1450         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1451         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1452         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1453         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1454         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1455         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1456         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1457         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1458         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1459         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1460         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1461         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1462         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1463         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1464         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1465         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1466         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1467         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1468         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1469         [IFLA_XDP]              = { .type = NLA_NESTED },
1470 };
1471
1472 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1473         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1474         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1475         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1476         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1477 };
1478
1479 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1480         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1481         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1482         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1483         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1484         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1485         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1486         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1487         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1488         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1489         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1490         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1491         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1492 };
1493
1494 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1495         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1496         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1497                                     .len = PORT_PROFILE_MAX },
1498         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1499                                       .len = PORT_UUID_MAX },
1500         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1501                                     .len = PORT_UUID_MAX },
1502         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1503         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1504
1505         /* Unused, but we need to keep it here since user space could
1506          * fill it. It's also broken with regard to NLA_BINARY use in
1507          * combination with structs.
1508          */
1509         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1510                                     .len = sizeof(struct ifla_port_vsi) },
1511 };
1512
1513 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1514         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1515         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1516         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1517 };
1518
1519 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1520 {
1521         const struct rtnl_link_ops *ops = NULL;
1522         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1523
1524         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1525                              ifla_info_policy, NULL) < 0)
1526                 return NULL;
1527
1528         if (linfo[IFLA_INFO_KIND]) {
1529                 char kind[MODULE_NAME_LEN];
1530
1531                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1532                 ops = rtnl_link_ops_get(kind);
1533         }
1534
1535         return ops;
1536 }
1537
1538 static bool link_master_filtered(struct net_device *dev, int master_idx)
1539 {
1540         struct net_device *master;
1541
1542         if (!master_idx)
1543                 return false;
1544
1545         master = netdev_master_upper_dev_get(dev);
1546         if (!master || master->ifindex != master_idx)
1547                 return true;
1548
1549         return false;
1550 }
1551
1552 static bool link_kind_filtered(const struct net_device *dev,
1553                                const struct rtnl_link_ops *kind_ops)
1554 {
1555         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1556                 return true;
1557
1558         return false;
1559 }
1560
1561 static bool link_dump_filtered(struct net_device *dev,
1562                                int master_idx,
1563                                const struct rtnl_link_ops *kind_ops)
1564 {
1565         if (link_master_filtered(dev, master_idx) ||
1566             link_kind_filtered(dev, kind_ops))
1567                 return true;
1568
1569         return false;
1570 }
1571
1572 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1573 {
1574         struct net *net = sock_net(skb->sk);
1575         int h, s_h;
1576         int idx = 0, s_idx;
1577         struct net_device *dev;
1578         struct hlist_head *head;
1579         struct nlattr *tb[IFLA_MAX+1];
1580         u32 ext_filter_mask = 0;
1581         const struct rtnl_link_ops *kind_ops = NULL;
1582         unsigned int flags = NLM_F_MULTI;
1583         int master_idx = 0;
1584         int err;
1585         int hdrlen;
1586
1587         s_h = cb->args[0];
1588         s_idx = cb->args[1];
1589
1590         cb->seq = net->dev_base_seq;
1591
1592         /* A hack to preserve kernel<->userspace interface.
1593          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1594          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1595          * what iproute2 < v3.9.0 used.
1596          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1597          * attribute, its netlink message is shorter than struct ifinfomsg.
1598          */
1599         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1600                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1601
1602         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1603                         ifla_policy, NULL) >= 0) {
1604                 if (tb[IFLA_EXT_MASK])
1605                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1606
1607                 if (tb[IFLA_MASTER])
1608                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1609
1610                 if (tb[IFLA_LINKINFO])
1611                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1612
1613                 if (master_idx || kind_ops)
1614                         flags |= NLM_F_DUMP_FILTERED;
1615         }
1616
1617         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1618                 idx = 0;
1619                 head = &net->dev_index_head[h];
1620                 hlist_for_each_entry(dev, head, index_hlist) {
1621                         if (link_dump_filtered(dev, master_idx, kind_ops))
1622                                 goto cont;
1623                         if (idx < s_idx)
1624                                 goto cont;
1625                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1626                                                NETLINK_CB(cb->skb).portid,
1627                                                cb->nlh->nlmsg_seq, 0,
1628                                                flags,
1629                                                ext_filter_mask);
1630
1631                         if (err < 0) {
1632                                 if (likely(skb->len))
1633                                         goto out;
1634
1635                                 goto out_err;
1636                         }
1637
1638                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1639 cont:
1640                         idx++;
1641                 }
1642         }
1643 out:
1644         err = skb->len;
1645 out_err:
1646         cb->args[1] = idx;
1647         cb->args[0] = h;
1648
1649         return err;
1650 }
1651
1652 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1653                         struct netlink_ext_ack *exterr)
1654 {
1655         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1656 }
1657 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1658
1659 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1660 {
1661         struct net *net;
1662         /* Examine the link attributes and figure out which
1663          * network namespace we are talking about.
1664          */
1665         if (tb[IFLA_NET_NS_PID])
1666                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1667         else if (tb[IFLA_NET_NS_FD])
1668                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1669         else
1670                 net = get_net(src_net);
1671         return net;
1672 }
1673 EXPORT_SYMBOL(rtnl_link_get_net);
1674
1675 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1676 {
1677         if (dev) {
1678                 if (tb[IFLA_ADDRESS] &&
1679                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1680                         return -EINVAL;
1681
1682                 if (tb[IFLA_BROADCAST] &&
1683                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1684                         return -EINVAL;
1685         }
1686
1687         if (tb[IFLA_AF_SPEC]) {
1688                 struct nlattr *af;
1689                 int rem, err;
1690
1691                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1692                         const struct rtnl_af_ops *af_ops;
1693
1694                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1695                                 return -EAFNOSUPPORT;
1696
1697                         if (!af_ops->set_link_af)
1698                                 return -EOPNOTSUPP;
1699
1700                         if (af_ops->validate_link_af) {
1701                                 err = af_ops->validate_link_af(dev, af);
1702                                 if (err < 0)
1703                                         return err;
1704                         }
1705                 }
1706         }
1707
1708         return 0;
1709 }
1710
1711 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1712                                   int guid_type)
1713 {
1714         const struct net_device_ops *ops = dev->netdev_ops;
1715
1716         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1717 }
1718
1719 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1720 {
1721         if (dev->type != ARPHRD_INFINIBAND)
1722                 return -EOPNOTSUPP;
1723
1724         return handle_infiniband_guid(dev, ivt, guid_type);
1725 }
1726
1727 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1728 {
1729         const struct net_device_ops *ops = dev->netdev_ops;
1730         int err = -EINVAL;
1731
1732         if (tb[IFLA_VF_MAC]) {
1733                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1734
1735                 err = -EOPNOTSUPP;
1736                 if (ops->ndo_set_vf_mac)
1737                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1738                                                   ivm->mac);
1739                 if (err < 0)
1740                         return err;
1741         }
1742
1743         if (tb[IFLA_VF_VLAN]) {
1744                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1745
1746                 err = -EOPNOTSUPP;
1747                 if (ops->ndo_set_vf_vlan)
1748                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1749                                                    ivv->qos,
1750                                                    htons(ETH_P_8021Q));
1751                 if (err < 0)
1752                         return err;
1753         }
1754
1755         if (tb[IFLA_VF_VLAN_LIST]) {
1756                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1757                 struct nlattr *attr;
1758                 int rem, len = 0;
1759
1760                 err = -EOPNOTSUPP;
1761                 if (!ops->ndo_set_vf_vlan)
1762                         return err;
1763
1764                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1765                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1766                             nla_len(attr) < NLA_HDRLEN) {
1767                                 return -EINVAL;
1768                         }
1769                         if (len >= MAX_VLAN_LIST_LEN)
1770                                 return -EOPNOTSUPP;
1771                         ivvl[len] = nla_data(attr);
1772
1773                         len++;
1774                 }
1775                 if (len == 0)
1776                         return -EINVAL;
1777
1778                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1779                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1780                 if (err < 0)
1781                         return err;
1782         }
1783
1784         if (tb[IFLA_VF_TX_RATE]) {
1785                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1786                 struct ifla_vf_info ivf;
1787
1788                 err = -EOPNOTSUPP;
1789                 if (ops->ndo_get_vf_config)
1790                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1791                 if (err < 0)
1792                         return err;
1793
1794                 err = -EOPNOTSUPP;
1795                 if (ops->ndo_set_vf_rate)
1796                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1797                                                    ivf.min_tx_rate,
1798                                                    ivt->rate);
1799                 if (err < 0)
1800                         return err;
1801         }
1802
1803         if (tb[IFLA_VF_RATE]) {
1804                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1805
1806                 err = -EOPNOTSUPP;
1807                 if (ops->ndo_set_vf_rate)
1808                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1809                                                    ivt->min_tx_rate,
1810                                                    ivt->max_tx_rate);
1811                 if (err < 0)
1812                         return err;
1813         }
1814
1815         if (tb[IFLA_VF_SPOOFCHK]) {
1816                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1817
1818                 err = -EOPNOTSUPP;
1819                 if (ops->ndo_set_vf_spoofchk)
1820                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1821                                                        ivs->setting);
1822                 if (err < 0)
1823                         return err;
1824         }
1825
1826         if (tb[IFLA_VF_LINK_STATE]) {
1827                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1828
1829                 err = -EOPNOTSUPP;
1830                 if (ops->ndo_set_vf_link_state)
1831                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1832                                                          ivl->link_state);
1833                 if (err < 0)
1834                         return err;
1835         }
1836
1837         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1838                 struct ifla_vf_rss_query_en *ivrssq_en;
1839
1840                 err = -EOPNOTSUPP;
1841                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1842                 if (ops->ndo_set_vf_rss_query_en)
1843                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1844                                                            ivrssq_en->setting);
1845                 if (err < 0)
1846                         return err;
1847         }
1848
1849         if (tb[IFLA_VF_TRUST]) {
1850                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1851
1852                 err = -EOPNOTSUPP;
1853                 if (ops->ndo_set_vf_trust)
1854                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1855                 if (err < 0)
1856                         return err;
1857         }
1858
1859         if (tb[IFLA_VF_IB_NODE_GUID]) {
1860                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1861
1862                 if (!ops->ndo_set_vf_guid)
1863                         return -EOPNOTSUPP;
1864
1865                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1866         }
1867
1868         if (tb[IFLA_VF_IB_PORT_GUID]) {
1869                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1870
1871                 if (!ops->ndo_set_vf_guid)
1872                         return -EOPNOTSUPP;
1873
1874                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1875         }
1876
1877         return err;
1878 }
1879
1880 static int do_set_master(struct net_device *dev, int ifindex)
1881 {
1882         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1883         const struct net_device_ops *ops;
1884         int err;
1885
1886         if (upper_dev) {
1887                 if (upper_dev->ifindex == ifindex)
1888                         return 0;
1889                 ops = upper_dev->netdev_ops;
1890                 if (ops->ndo_del_slave) {
1891                         err = ops->ndo_del_slave(upper_dev, dev);
1892                         if (err)
1893                                 return err;
1894                 } else {
1895                         return -EOPNOTSUPP;
1896                 }
1897         }
1898
1899         if (ifindex) {
1900                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1901                 if (!upper_dev)
1902                         return -EINVAL;
1903                 ops = upper_dev->netdev_ops;
1904                 if (ops->ndo_add_slave) {
1905                         err = ops->ndo_add_slave(upper_dev, dev);
1906                         if (err)
1907                                 return err;
1908                 } else {
1909                         return -EOPNOTSUPP;
1910                 }
1911         }
1912         return 0;
1913 }
1914
1915 #define DO_SETLINK_MODIFIED     0x01
1916 /* notify flag means notify + modified. */
1917 #define DO_SETLINK_NOTIFY       0x03
1918 static int do_setlink(const struct sk_buff *skb,
1919                       struct net_device *dev, struct ifinfomsg *ifm,
1920                       struct netlink_ext_ack *extack,
1921                       struct nlattr **tb, char *ifname, int status)
1922 {
1923         const struct net_device_ops *ops = dev->netdev_ops;
1924         int err;
1925
1926         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1927                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1928                 if (IS_ERR(net)) {
1929                         err = PTR_ERR(net);
1930                         goto errout;
1931                 }
1932                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1933                         put_net(net);
1934                         err = -EPERM;
1935                         goto errout;
1936                 }
1937                 err = dev_change_net_namespace(dev, net, ifname);
1938                 put_net(net);
1939                 if (err)
1940                         goto errout;
1941                 status |= DO_SETLINK_MODIFIED;
1942         }
1943
1944         if (tb[IFLA_MAP]) {
1945                 struct rtnl_link_ifmap *u_map;
1946                 struct ifmap k_map;
1947
1948                 if (!ops->ndo_set_config) {
1949                         err = -EOPNOTSUPP;
1950                         goto errout;
1951                 }
1952
1953                 if (!netif_device_present(dev)) {
1954                         err = -ENODEV;
1955                         goto errout;
1956                 }
1957
1958                 u_map = nla_data(tb[IFLA_MAP]);
1959                 k_map.mem_start = (unsigned long) u_map->mem_start;
1960                 k_map.mem_end = (unsigned long) u_map->mem_end;
1961                 k_map.base_addr = (unsigned short) u_map->base_addr;
1962                 k_map.irq = (unsigned char) u_map->irq;
1963                 k_map.dma = (unsigned char) u_map->dma;
1964                 k_map.port = (unsigned char) u_map->port;
1965
1966                 err = ops->ndo_set_config(dev, &k_map);
1967                 if (err < 0)
1968                         goto errout;
1969
1970                 status |= DO_SETLINK_NOTIFY;
1971         }
1972
1973         if (tb[IFLA_ADDRESS]) {
1974                 struct sockaddr *sa;
1975                 int len;
1976
1977                 len = sizeof(sa_family_t) + dev->addr_len;
1978                 sa = kmalloc(len, GFP_KERNEL);
1979                 if (!sa) {
1980                         err = -ENOMEM;
1981                         goto errout;
1982                 }
1983                 sa->sa_family = dev->type;
1984                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1985                        dev->addr_len);
1986                 err = dev_set_mac_address(dev, sa);
1987                 kfree(sa);
1988                 if (err)
1989                         goto errout;
1990                 status |= DO_SETLINK_MODIFIED;
1991         }
1992
1993         if (tb[IFLA_MTU]) {
1994                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1995                 if (err < 0)
1996                         goto errout;
1997                 status |= DO_SETLINK_MODIFIED;
1998         }
1999
2000         if (tb[IFLA_GROUP]) {
2001                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2002                 status |= DO_SETLINK_NOTIFY;
2003         }
2004
2005         /*
2006          * Interface selected by interface index but interface
2007          * name provided implies that a name change has been
2008          * requested.
2009          */
2010         if (ifm->ifi_index > 0 && ifname[0]) {
2011                 err = dev_change_name(dev, ifname);
2012                 if (err < 0)
2013                         goto errout;
2014                 status |= DO_SETLINK_MODIFIED;
2015         }
2016
2017         if (tb[IFLA_IFALIAS]) {
2018                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2019                                     nla_len(tb[IFLA_IFALIAS]));
2020                 if (err < 0)
2021                         goto errout;
2022                 status |= DO_SETLINK_NOTIFY;
2023         }
2024
2025         if (tb[IFLA_BROADCAST]) {
2026                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2027                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2028         }
2029
2030         if (ifm->ifi_flags || ifm->ifi_change) {
2031                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2032                 if (err < 0)
2033                         goto errout;
2034         }
2035
2036         if (tb[IFLA_MASTER]) {
2037                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2038                 if (err)
2039                         goto errout;
2040                 status |= DO_SETLINK_MODIFIED;
2041         }
2042
2043         if (tb[IFLA_CARRIER]) {
2044                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2045                 if (err)
2046                         goto errout;
2047                 status |= DO_SETLINK_MODIFIED;
2048         }
2049
2050         if (tb[IFLA_TXQLEN]) {
2051                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2052                 unsigned long orig_len = dev->tx_queue_len;
2053
2054                 if (dev->tx_queue_len ^ value) {
2055                         dev->tx_queue_len = value;
2056                         err = call_netdevice_notifiers(
2057                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2058                         err = notifier_to_errno(err);
2059                         if (err) {
2060                                 dev->tx_queue_len = orig_len;
2061                                 goto errout;
2062                         }
2063                         status |= DO_SETLINK_NOTIFY;
2064                 }
2065         }
2066
2067         if (tb[IFLA_OPERSTATE])
2068                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2069
2070         if (tb[IFLA_LINKMODE]) {
2071                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2072
2073                 write_lock_bh(&dev_base_lock);
2074                 if (dev->link_mode ^ value)
2075                         status |= DO_SETLINK_NOTIFY;
2076                 dev->link_mode = value;
2077                 write_unlock_bh(&dev_base_lock);
2078         }
2079
2080         if (tb[IFLA_VFINFO_LIST]) {
2081                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2082                 struct nlattr *attr;
2083                 int rem;
2084
2085                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2086                         if (nla_type(attr) != IFLA_VF_INFO ||
2087                             nla_len(attr) < NLA_HDRLEN) {
2088                                 err = -EINVAL;
2089                                 goto errout;
2090                         }
2091                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2092                                                ifla_vf_policy, NULL);
2093                         if (err < 0)
2094                                 goto errout;
2095                         err = do_setvfinfo(dev, vfinfo);
2096                         if (err < 0)
2097                                 goto errout;
2098                         status |= DO_SETLINK_NOTIFY;
2099                 }
2100         }
2101         err = 0;
2102
2103         if (tb[IFLA_VF_PORTS]) {
2104                 struct nlattr *port[IFLA_PORT_MAX+1];
2105                 struct nlattr *attr;
2106                 int vf;
2107                 int rem;
2108
2109                 err = -EOPNOTSUPP;
2110                 if (!ops->ndo_set_vf_port)
2111                         goto errout;
2112
2113                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2114                         if (nla_type(attr) != IFLA_VF_PORT ||
2115                             nla_len(attr) < NLA_HDRLEN) {
2116                                 err = -EINVAL;
2117                                 goto errout;
2118                         }
2119                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2120                                                ifla_port_policy, NULL);
2121                         if (err < 0)
2122                                 goto errout;
2123                         if (!port[IFLA_PORT_VF]) {
2124                                 err = -EOPNOTSUPP;
2125                                 goto errout;
2126                         }
2127                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2128                         err = ops->ndo_set_vf_port(dev, vf, port);
2129                         if (err < 0)
2130                                 goto errout;
2131                         status |= DO_SETLINK_NOTIFY;
2132                 }
2133         }
2134         err = 0;
2135
2136         if (tb[IFLA_PORT_SELF]) {
2137                 struct nlattr *port[IFLA_PORT_MAX+1];
2138
2139                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2140                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2141                                        NULL);
2142                 if (err < 0)
2143                         goto errout;
2144
2145                 err = -EOPNOTSUPP;
2146                 if (ops->ndo_set_vf_port)
2147                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2148                 if (err < 0)
2149                         goto errout;
2150                 status |= DO_SETLINK_NOTIFY;
2151         }
2152
2153         if (tb[IFLA_AF_SPEC]) {
2154                 struct nlattr *af;
2155                 int rem;
2156
2157                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2158                         const struct rtnl_af_ops *af_ops;
2159
2160                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2161                                 BUG();
2162
2163                         err = af_ops->set_link_af(dev, af);
2164                         if (err < 0)
2165                                 goto errout;
2166
2167                         status |= DO_SETLINK_NOTIFY;
2168                 }
2169         }
2170         err = 0;
2171
2172         if (tb[IFLA_PROTO_DOWN]) {
2173                 err = dev_change_proto_down(dev,
2174                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2175                 if (err)
2176                         goto errout;
2177                 status |= DO_SETLINK_NOTIFY;
2178         }
2179
2180         if (tb[IFLA_XDP]) {
2181                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2182                 u32 xdp_flags = 0;
2183
2184                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2185                                        ifla_xdp_policy, NULL);
2186                 if (err < 0)
2187                         goto errout;
2188
2189                 if (xdp[IFLA_XDP_ATTACHED]) {
2190                         err = -EINVAL;
2191                         goto errout;
2192                 }
2193
2194                 if (xdp[IFLA_XDP_FLAGS]) {
2195                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2196                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2197                                 err = -EINVAL;
2198                                 goto errout;
2199                         }
2200                         if ((xdp_flags & XDP_FLAGS_SKB_MODE) &&
2201                             (xdp_flags & XDP_FLAGS_DRV_MODE)) {
2202                                 err = -EINVAL;
2203                                 goto errout;
2204                         }
2205                 }
2206
2207                 if (xdp[IFLA_XDP_FD]) {
2208                         err = dev_change_xdp_fd(dev, extack,
2209                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2210                                                 xdp_flags);
2211                         if (err)
2212                                 goto errout;
2213                         status |= DO_SETLINK_NOTIFY;
2214                 }
2215         }
2216
2217 errout:
2218         if (status & DO_SETLINK_MODIFIED) {
2219                 if (status & DO_SETLINK_NOTIFY)
2220                         netdev_state_change(dev);
2221
2222                 if (err < 0)
2223                         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",
2224                                              dev->name);
2225         }
2226
2227         return err;
2228 }
2229
2230 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2231                         struct netlink_ext_ack *extack)
2232 {
2233         struct net *net = sock_net(skb->sk);
2234         struct ifinfomsg *ifm;
2235         struct net_device *dev;
2236         int err;
2237         struct nlattr *tb[IFLA_MAX+1];
2238         char ifname[IFNAMSIZ];
2239
2240         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2241                           extack);
2242         if (err < 0)
2243                 goto errout;
2244
2245         if (tb[IFLA_IFNAME])
2246                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2247         else
2248                 ifname[0] = '\0';
2249
2250         err = -EINVAL;
2251         ifm = nlmsg_data(nlh);
2252         if (ifm->ifi_index > 0)
2253                 dev = __dev_get_by_index(net, ifm->ifi_index);
2254         else if (tb[IFLA_IFNAME])
2255                 dev = __dev_get_by_name(net, ifname);
2256         else
2257                 goto errout;
2258
2259         if (dev == NULL) {
2260                 err = -ENODEV;
2261                 goto errout;
2262         }
2263
2264         err = validate_linkmsg(dev, tb);
2265         if (err < 0)
2266                 goto errout;
2267
2268         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2269 errout:
2270         return err;
2271 }
2272
2273 static int rtnl_group_dellink(const struct net *net, int group)
2274 {
2275         struct net_device *dev, *aux;
2276         LIST_HEAD(list_kill);
2277         bool found = false;
2278
2279         if (!group)
2280                 return -EPERM;
2281
2282         for_each_netdev(net, dev) {
2283                 if (dev->group == group) {
2284                         const struct rtnl_link_ops *ops;
2285
2286                         found = true;
2287                         ops = dev->rtnl_link_ops;
2288                         if (!ops || !ops->dellink)
2289                                 return -EOPNOTSUPP;
2290                 }
2291         }
2292
2293         if (!found)
2294                 return -ENODEV;
2295
2296         for_each_netdev_safe(net, dev, aux) {
2297                 if (dev->group == group) {
2298                         const struct rtnl_link_ops *ops;
2299
2300                         ops = dev->rtnl_link_ops;
2301                         ops->dellink(dev, &list_kill);
2302                 }
2303         }
2304         unregister_netdevice_many(&list_kill);
2305
2306         return 0;
2307 }
2308
2309 int rtnl_delete_link(struct net_device *dev)
2310 {
2311         const struct rtnl_link_ops *ops;
2312         LIST_HEAD(list_kill);
2313
2314         ops = dev->rtnl_link_ops;
2315         if (!ops || !ops->dellink)
2316                 return -EOPNOTSUPP;
2317
2318         ops->dellink(dev, &list_kill);
2319         unregister_netdevice_many(&list_kill);
2320
2321         return 0;
2322 }
2323 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2324
2325 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2326                         struct netlink_ext_ack *extack)
2327 {
2328         struct net *net = sock_net(skb->sk);
2329         struct net_device *dev;
2330         struct ifinfomsg *ifm;
2331         char ifname[IFNAMSIZ];
2332         struct nlattr *tb[IFLA_MAX+1];
2333         int err;
2334
2335         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2336         if (err < 0)
2337                 return err;
2338
2339         if (tb[IFLA_IFNAME])
2340                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2341
2342         ifm = nlmsg_data(nlh);
2343         if (ifm->ifi_index > 0)
2344                 dev = __dev_get_by_index(net, ifm->ifi_index);
2345         else if (tb[IFLA_IFNAME])
2346                 dev = __dev_get_by_name(net, ifname);
2347         else if (tb[IFLA_GROUP])
2348                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2349         else
2350                 return -EINVAL;
2351
2352         if (!dev)
2353                 return -ENODEV;
2354
2355         return rtnl_delete_link(dev);
2356 }
2357
2358 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2359 {
2360         unsigned int old_flags;
2361         int err;
2362
2363         old_flags = dev->flags;
2364         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2365                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2366                 if (err < 0)
2367                         return err;
2368         }
2369
2370         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2371
2372         __dev_notify_flags(dev, old_flags, ~0U);
2373         return 0;
2374 }
2375 EXPORT_SYMBOL(rtnl_configure_link);
2376
2377 struct net_device *rtnl_create_link(struct net *net,
2378         const char *ifname, unsigned char name_assign_type,
2379         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2380 {
2381         struct net_device *dev;
2382         unsigned int num_tx_queues = 1;
2383         unsigned int num_rx_queues = 1;
2384
2385         if (tb[IFLA_NUM_TX_QUEUES])
2386                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2387         else if (ops->get_num_tx_queues)
2388                 num_tx_queues = ops->get_num_tx_queues();
2389
2390         if (tb[IFLA_NUM_RX_QUEUES])
2391                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2392         else if (ops->get_num_rx_queues)
2393                 num_rx_queues = ops->get_num_rx_queues();
2394
2395         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2396                                ops->setup, num_tx_queues, num_rx_queues);
2397         if (!dev)
2398                 return ERR_PTR(-ENOMEM);
2399
2400         dev_net_set(dev, net);
2401         dev->rtnl_link_ops = ops;
2402         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2403
2404         if (tb[IFLA_MTU])
2405                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2406         if (tb[IFLA_ADDRESS]) {
2407                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2408                                 nla_len(tb[IFLA_ADDRESS]));
2409                 dev->addr_assign_type = NET_ADDR_SET;
2410         }
2411         if (tb[IFLA_BROADCAST])
2412                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2413                                 nla_len(tb[IFLA_BROADCAST]));
2414         if (tb[IFLA_TXQLEN])
2415                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2416         if (tb[IFLA_OPERSTATE])
2417                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2418         if (tb[IFLA_LINKMODE])
2419                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2420         if (tb[IFLA_GROUP])
2421                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2422
2423         return dev;
2424 }
2425 EXPORT_SYMBOL(rtnl_create_link);
2426
2427 static int rtnl_group_changelink(const struct sk_buff *skb,
2428                 struct net *net, int group,
2429                 struct ifinfomsg *ifm,
2430                 struct netlink_ext_ack *extack,
2431                 struct nlattr **tb)
2432 {
2433         struct net_device *dev, *aux;
2434         int err;
2435
2436         for_each_netdev_safe(net, dev, aux) {
2437                 if (dev->group == group) {
2438                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2439                         if (err < 0)
2440                                 return err;
2441                 }
2442         }
2443
2444         return 0;
2445 }
2446
2447 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2448                         struct netlink_ext_ack *extack)
2449 {
2450         struct net *net = sock_net(skb->sk);
2451         const struct rtnl_link_ops *ops;
2452         const struct rtnl_link_ops *m_ops = NULL;
2453         struct net_device *dev;
2454         struct net_device *master_dev = NULL;
2455         struct ifinfomsg *ifm;
2456         char kind[MODULE_NAME_LEN];
2457         char ifname[IFNAMSIZ];
2458         struct nlattr *tb[IFLA_MAX+1];
2459         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2460         unsigned char name_assign_type = NET_NAME_USER;
2461         int err;
2462
2463 #ifdef CONFIG_MODULES
2464 replay:
2465 #endif
2466         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2467         if (err < 0)
2468                 return err;
2469
2470         if (tb[IFLA_IFNAME])
2471                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2472         else
2473                 ifname[0] = '\0';
2474
2475         ifm = nlmsg_data(nlh);
2476         if (ifm->ifi_index > 0)
2477                 dev = __dev_get_by_index(net, ifm->ifi_index);
2478         else {
2479                 if (ifname[0])
2480                         dev = __dev_get_by_name(net, ifname);
2481                 else
2482                         dev = NULL;
2483         }
2484
2485         if (dev) {
2486                 master_dev = netdev_master_upper_dev_get(dev);
2487                 if (master_dev)
2488                         m_ops = master_dev->rtnl_link_ops;
2489         }
2490
2491         err = validate_linkmsg(dev, tb);
2492         if (err < 0)
2493                 return err;
2494
2495         if (tb[IFLA_LINKINFO]) {
2496                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2497                                        tb[IFLA_LINKINFO], ifla_info_policy,
2498                                        NULL);
2499                 if (err < 0)
2500                         return err;
2501         } else
2502                 memset(linkinfo, 0, sizeof(linkinfo));
2503
2504         if (linkinfo[IFLA_INFO_KIND]) {
2505                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2506                 ops = rtnl_link_ops_get(kind);
2507         } else {
2508                 kind[0] = '\0';
2509                 ops = NULL;
2510         }
2511
2512         if (1) {
2513                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2514                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2515                 struct nlattr **data = NULL;
2516                 struct nlattr **slave_data = NULL;
2517                 struct net *dest_net, *link_net = NULL;
2518
2519                 if (ops) {
2520                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2521                                 err = nla_parse_nested(attr, ops->maxtype,
2522                                                        linkinfo[IFLA_INFO_DATA],
2523                                                        ops->policy, NULL);
2524                                 if (err < 0)
2525                                         return err;
2526                                 data = attr;
2527                         }
2528                         if (ops->validate) {
2529                                 err = ops->validate(tb, data);
2530                                 if (err < 0)
2531                                         return err;
2532                         }
2533                 }
2534
2535                 if (m_ops) {
2536                         if (m_ops->slave_maxtype &&
2537                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2538                                 err = nla_parse_nested(slave_attr,
2539                                                        m_ops->slave_maxtype,
2540                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2541                                                        m_ops->slave_policy,
2542                                                        NULL);
2543                                 if (err < 0)
2544                                         return err;
2545                                 slave_data = slave_attr;
2546                         }
2547                         if (m_ops->slave_validate) {
2548                                 err = m_ops->slave_validate(tb, slave_data);
2549                                 if (err < 0)
2550                                         return err;
2551                         }
2552                 }
2553
2554                 if (dev) {
2555                         int status = 0;
2556
2557                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2558                                 return -EEXIST;
2559                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2560                                 return -EOPNOTSUPP;
2561
2562                         if (linkinfo[IFLA_INFO_DATA]) {
2563                                 if (!ops || ops != dev->rtnl_link_ops ||
2564                                     !ops->changelink)
2565                                         return -EOPNOTSUPP;
2566
2567                                 err = ops->changelink(dev, tb, data);
2568                                 if (err < 0)
2569                                         return err;
2570                                 status |= DO_SETLINK_NOTIFY;
2571                         }
2572
2573                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2574                                 if (!m_ops || !m_ops->slave_changelink)
2575                                         return -EOPNOTSUPP;
2576
2577                                 err = m_ops->slave_changelink(master_dev, dev,
2578                                                               tb, slave_data);
2579                                 if (err < 0)
2580                                         return err;
2581                                 status |= DO_SETLINK_NOTIFY;
2582                         }
2583
2584                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2585                                           status);
2586                 }
2587
2588                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2589                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2590                                 return rtnl_group_changelink(skb, net,
2591                                                 nla_get_u32(tb[IFLA_GROUP]),
2592                                                 ifm, extack, tb);
2593                         return -ENODEV;
2594                 }
2595
2596                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2597                         return -EOPNOTSUPP;
2598
2599                 if (!ops) {
2600 #ifdef CONFIG_MODULES
2601                         if (kind[0]) {
2602                                 __rtnl_unlock();
2603                                 request_module("rtnl-link-%s", kind);
2604                                 rtnl_lock();
2605                                 ops = rtnl_link_ops_get(kind);
2606                                 if (ops)
2607                                         goto replay;
2608                         }
2609 #endif
2610                         return -EOPNOTSUPP;
2611                 }
2612
2613                 if (!ops->setup)
2614                         return -EOPNOTSUPP;
2615
2616                 if (!ifname[0]) {
2617                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2618                         name_assign_type = NET_NAME_ENUM;
2619                 }
2620
2621                 dest_net = rtnl_link_get_net(net, tb);
2622                 if (IS_ERR(dest_net))
2623                         return PTR_ERR(dest_net);
2624
2625                 err = -EPERM;
2626                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2627                         goto out;
2628
2629                 if (tb[IFLA_LINK_NETNSID]) {
2630                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2631
2632                         link_net = get_net_ns_by_id(dest_net, id);
2633                         if (!link_net) {
2634                                 err =  -EINVAL;
2635                                 goto out;
2636                         }
2637                         err = -EPERM;
2638                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2639                                 goto out;
2640                 }
2641
2642                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2643                                        name_assign_type, ops, tb);
2644                 if (IS_ERR(dev)) {
2645                         err = PTR_ERR(dev);
2646                         goto out;
2647                 }
2648
2649                 dev->ifindex = ifm->ifi_index;
2650
2651                 if (ops->newlink) {
2652                         err = ops->newlink(link_net ? : net, dev, tb, data);
2653                         /* Drivers should call free_netdev() in ->destructor
2654                          * and unregister it on failure after registration
2655                          * so that device could be finally freed in rtnl_unlock.
2656                          */
2657                         if (err < 0) {
2658                                 /* If device is not registered at all, free it now */
2659                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2660                                         free_netdev(dev);
2661                                 goto out;
2662                         }
2663                 } else {
2664                         err = register_netdevice(dev);
2665                         if (err < 0) {
2666                                 free_netdev(dev);
2667                                 goto out;
2668                         }
2669                 }
2670                 err = rtnl_configure_link(dev, ifm);
2671                 if (err < 0)
2672                         goto out_unregister;
2673                 if (link_net) {
2674                         err = dev_change_net_namespace(dev, dest_net, ifname);
2675                         if (err < 0)
2676                                 goto out_unregister;
2677                 }
2678                 if (tb[IFLA_MASTER]) {
2679                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2680                         if (err)
2681                                 goto out_unregister;
2682                 }
2683 out:
2684                 if (link_net)
2685                         put_net(link_net);
2686                 put_net(dest_net);
2687                 return err;
2688 out_unregister:
2689                 if (ops->newlink) {
2690                         LIST_HEAD(list_kill);
2691
2692                         ops->dellink(dev, &list_kill);
2693                         unregister_netdevice_many(&list_kill);
2694                 } else {
2695                         unregister_netdevice(dev);
2696                 }
2697                 goto out;
2698         }
2699 }
2700
2701 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2702                         struct netlink_ext_ack *extack)
2703 {
2704         struct net *net = sock_net(skb->sk);
2705         struct ifinfomsg *ifm;
2706         char ifname[IFNAMSIZ];
2707         struct nlattr *tb[IFLA_MAX+1];
2708         struct net_device *dev = NULL;
2709         struct sk_buff *nskb;
2710         int err;
2711         u32 ext_filter_mask = 0;
2712
2713         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2714         if (err < 0)
2715                 return err;
2716
2717         if (tb[IFLA_IFNAME])
2718                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2719
2720         if (tb[IFLA_EXT_MASK])
2721                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2722
2723         ifm = nlmsg_data(nlh);
2724         if (ifm->ifi_index > 0)
2725                 dev = __dev_get_by_index(net, ifm->ifi_index);
2726         else if (tb[IFLA_IFNAME])
2727                 dev = __dev_get_by_name(net, ifname);
2728         else
2729                 return -EINVAL;
2730
2731         if (dev == NULL)
2732                 return -ENODEV;
2733
2734         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2735         if (nskb == NULL)
2736                 return -ENOBUFS;
2737
2738         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2739                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2740         if (err < 0) {
2741                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2742                 WARN_ON(err == -EMSGSIZE);
2743                 kfree_skb(nskb);
2744         } else
2745                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2746
2747         return err;
2748 }
2749
2750 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2751 {
2752         struct net *net = sock_net(skb->sk);
2753         struct net_device *dev;
2754         struct nlattr *tb[IFLA_MAX+1];
2755         u32 ext_filter_mask = 0;
2756         u16 min_ifinfo_dump_size = 0;
2757         int hdrlen;
2758
2759         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2760         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2761                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2762
2763         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2764                 if (tb[IFLA_EXT_MASK])
2765                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2766         }
2767
2768         if (!ext_filter_mask)
2769                 return NLMSG_GOODSIZE;
2770         /*
2771          * traverse the list of net devices and compute the minimum
2772          * buffer size based upon the filter mask.
2773          */
2774         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2775                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2776                                              if_nlmsg_size(dev,
2777                                                            ext_filter_mask));
2778         }
2779
2780         return nlmsg_total_size(min_ifinfo_dump_size);
2781 }
2782
2783 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2784 {
2785         int idx;
2786         int s_idx = cb->family;
2787
2788         if (s_idx == 0)
2789                 s_idx = 1;
2790         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2791                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2792                 if (idx < s_idx || idx == PF_PACKET)
2793                         continue;
2794                 if (rtnl_msg_handlers[idx] == NULL ||
2795                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2796                         continue;
2797                 if (idx > s_idx) {
2798                         memset(&cb->args[0], 0, sizeof(cb->args));
2799                         cb->prev_seq = 0;
2800                         cb->seq = 0;
2801                 }
2802                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2803                         break;
2804         }
2805         cb->family = idx;
2806
2807         return skb->len;
2808 }
2809
2810 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2811                                        unsigned int change, gfp_t flags)
2812 {
2813         struct net *net = dev_net(dev);
2814         struct sk_buff *skb;
2815         int err = -ENOBUFS;
2816         size_t if_info_size;
2817
2818         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2819         if (skb == NULL)
2820                 goto errout;
2821
2822         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2823         if (err < 0) {
2824                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2825                 WARN_ON(err == -EMSGSIZE);
2826                 kfree_skb(skb);
2827                 goto errout;
2828         }
2829         return skb;
2830 errout:
2831         if (err < 0)
2832                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2833         return NULL;
2834 }
2835
2836 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2837 {
2838         struct net *net = dev_net(dev);
2839
2840         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2841 }
2842
2843 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2844                   gfp_t flags)
2845 {
2846         struct sk_buff *skb;
2847
2848         if (dev->reg_state != NETREG_REGISTERED)
2849                 return;
2850
2851         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2852         if (skb)
2853                 rtmsg_ifinfo_send(skb, dev, flags);
2854 }
2855 EXPORT_SYMBOL(rtmsg_ifinfo);
2856
2857 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2858                                    struct net_device *dev,
2859                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2860                                    int type, unsigned int flags,
2861                                    int nlflags, u16 ndm_state)
2862 {
2863         struct nlmsghdr *nlh;
2864         struct ndmsg *ndm;
2865
2866         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2867         if (!nlh)
2868                 return -EMSGSIZE;
2869
2870         ndm = nlmsg_data(nlh);
2871         ndm->ndm_family  = AF_BRIDGE;
2872         ndm->ndm_pad1    = 0;
2873         ndm->ndm_pad2    = 0;
2874         ndm->ndm_flags   = flags;
2875         ndm->ndm_type    = 0;
2876         ndm->ndm_ifindex = dev->ifindex;
2877         ndm->ndm_state   = ndm_state;
2878
2879         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2880                 goto nla_put_failure;
2881         if (vid)
2882                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2883                         goto nla_put_failure;
2884
2885         nlmsg_end(skb, nlh);
2886         return 0;
2887
2888 nla_put_failure:
2889         nlmsg_cancel(skb, nlh);
2890         return -EMSGSIZE;
2891 }
2892
2893 static inline size_t rtnl_fdb_nlmsg_size(void)
2894 {
2895         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2896                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2897                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2898                0;
2899 }
2900
2901 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2902                             u16 ndm_state)
2903 {
2904         struct net *net = dev_net(dev);
2905         struct sk_buff *skb;
2906         int err = -ENOBUFS;
2907
2908         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2909         if (!skb)
2910                 goto errout;
2911
2912         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2913                                       0, 0, type, NTF_SELF, 0, ndm_state);
2914         if (err < 0) {
2915                 kfree_skb(skb);
2916                 goto errout;
2917         }
2918
2919         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2920         return;
2921 errout:
2922         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2923 }
2924
2925 /**
2926  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2927  */
2928 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2929                      struct nlattr *tb[],
2930                      struct net_device *dev,
2931                      const unsigned char *addr, u16 vid,
2932                      u16 flags)
2933 {
2934         int err = -EINVAL;
2935
2936         /* If aging addresses are supported device will need to
2937          * implement its own handler for this.
2938          */
2939         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2940                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2941                 return err;
2942         }
2943
2944         if (vid) {
2945                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2946                 return err;
2947         }
2948
2949         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2950                 err = dev_uc_add_excl(dev, addr);
2951         else if (is_multicast_ether_addr(addr))
2952                 err = dev_mc_add_excl(dev, addr);
2953
2954         /* Only return duplicate errors if NLM_F_EXCL is set */
2955         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2956                 err = 0;
2957
2958         return err;
2959 }
2960 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2961
2962 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2963 {
2964         u16 vid = 0;
2965
2966         if (vlan_attr) {
2967                 if (nla_len(vlan_attr) != sizeof(u16)) {
2968                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2969                         return -EINVAL;
2970                 }
2971
2972                 vid = nla_get_u16(vlan_attr);
2973
2974                 if (!vid || vid >= VLAN_VID_MASK) {
2975                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2976                                 vid);
2977                         return -EINVAL;
2978                 }
2979         }
2980         *p_vid = vid;
2981         return 0;
2982 }
2983
2984 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
2985                         struct netlink_ext_ack *extack)
2986 {
2987         struct net *net = sock_net(skb->sk);
2988         struct ndmsg *ndm;
2989         struct nlattr *tb[NDA_MAX+1];
2990         struct net_device *dev;
2991         u8 *addr;
2992         u16 vid;
2993         int err;
2994
2995         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
2996         if (err < 0)
2997                 return err;
2998
2999         ndm = nlmsg_data(nlh);
3000         if (ndm->ndm_ifindex == 0) {
3001                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3002                 return -EINVAL;
3003         }
3004
3005         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3006         if (dev == NULL) {
3007                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3008                 return -ENODEV;
3009         }
3010
3011         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3012                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3013                 return -EINVAL;
3014         }
3015
3016         addr = nla_data(tb[NDA_LLADDR]);
3017
3018         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3019         if (err)
3020                 return err;
3021
3022         err = -EOPNOTSUPP;
3023
3024         /* Support fdb on master device the net/bridge default case */
3025         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3026             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3027                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3028                 const struct net_device_ops *ops = br_dev->netdev_ops;
3029
3030                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3031                                        nlh->nlmsg_flags);
3032                 if (err)
3033                         goto out;
3034                 else
3035                         ndm->ndm_flags &= ~NTF_MASTER;
3036         }
3037
3038         /* Embedded bridge, macvlan, and any other device support */
3039         if ((ndm->ndm_flags & NTF_SELF)) {
3040                 if (dev->netdev_ops->ndo_fdb_add)
3041                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3042                                                            vid,
3043                                                            nlh->nlmsg_flags);
3044                 else
3045                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3046                                                nlh->nlmsg_flags);
3047
3048                 if (!err) {
3049                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3050                                         ndm->ndm_state);
3051                         ndm->ndm_flags &= ~NTF_SELF;
3052                 }
3053         }
3054 out:
3055         return err;
3056 }
3057
3058 /**
3059  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3060  */
3061 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3062                      struct nlattr *tb[],
3063                      struct net_device *dev,
3064                      const unsigned char *addr, u16 vid)
3065 {
3066         int err = -EINVAL;
3067
3068         /* If aging addresses are supported device will need to
3069          * implement its own handler for this.
3070          */
3071         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3072                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3073                 return err;
3074         }
3075
3076         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3077                 err = dev_uc_del(dev, addr);
3078         else if (is_multicast_ether_addr(addr))
3079                 err = dev_mc_del(dev, addr);
3080
3081         return err;
3082 }
3083 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3084
3085 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3086                         struct netlink_ext_ack *extack)
3087 {
3088         struct net *net = sock_net(skb->sk);
3089         struct ndmsg *ndm;
3090         struct nlattr *tb[NDA_MAX+1];
3091         struct net_device *dev;
3092         int err = -EINVAL;
3093         __u8 *addr;
3094         u16 vid;
3095
3096         if (!netlink_capable(skb, CAP_NET_ADMIN))
3097                 return -EPERM;
3098
3099         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3100         if (err < 0)
3101                 return err;
3102
3103         ndm = nlmsg_data(nlh);
3104         if (ndm->ndm_ifindex == 0) {
3105                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3106                 return -EINVAL;
3107         }
3108
3109         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3110         if (dev == NULL) {
3111                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3112                 return -ENODEV;
3113         }
3114
3115         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3116                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3117                 return -EINVAL;
3118         }
3119
3120         addr = nla_data(tb[NDA_LLADDR]);
3121
3122         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3123         if (err)
3124                 return err;
3125
3126         err = -EOPNOTSUPP;
3127
3128         /* Support fdb on master device the net/bridge default case */
3129         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3130             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3131                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3132                 const struct net_device_ops *ops = br_dev->netdev_ops;
3133
3134                 if (ops->ndo_fdb_del)
3135                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3136
3137                 if (err)
3138                         goto out;
3139                 else
3140                         ndm->ndm_flags &= ~NTF_MASTER;
3141         }
3142
3143         /* Embedded bridge, macvlan, and any other device support */
3144         if (ndm->ndm_flags & NTF_SELF) {
3145                 if (dev->netdev_ops->ndo_fdb_del)
3146                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3147                                                            vid);
3148                 else
3149                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3150
3151                 if (!err) {
3152                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3153                                         ndm->ndm_state);
3154                         ndm->ndm_flags &= ~NTF_SELF;
3155                 }
3156         }
3157 out:
3158         return err;
3159 }
3160
3161 static int nlmsg_populate_fdb(struct sk_buff *skb,
3162                               struct netlink_callback *cb,
3163                               struct net_device *dev,
3164                               int *idx,
3165                               struct netdev_hw_addr_list *list)
3166 {
3167         struct netdev_hw_addr *ha;
3168         int err;
3169         u32 portid, seq;
3170
3171         portid = NETLINK_CB(cb->skb).portid;
3172         seq = cb->nlh->nlmsg_seq;
3173
3174         list_for_each_entry(ha, &list->list, list) {
3175                 if (*idx < cb->args[2])
3176                         goto skip;
3177
3178                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3179                                               portid, seq,
3180                                               RTM_NEWNEIGH, NTF_SELF,
3181                                               NLM_F_MULTI, NUD_PERMANENT);
3182                 if (err < 0)
3183                         return err;
3184 skip:
3185                 *idx += 1;
3186         }
3187         return 0;
3188 }
3189
3190 /**
3191  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3192  * @nlh: netlink message header
3193  * @dev: netdevice
3194  *
3195  * Default netdevice operation to dump the existing unicast address list.
3196  * Returns number of addresses from list put in skb.
3197  */
3198 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3199                       struct netlink_callback *cb,
3200                       struct net_device *dev,
3201                       struct net_device *filter_dev,
3202                       int *idx)
3203 {
3204         int err;
3205
3206         netif_addr_lock_bh(dev);
3207         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3208         if (err)
3209                 goto out;
3210         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3211 out:
3212         netif_addr_unlock_bh(dev);
3213         return err;
3214 }
3215 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3216
3217 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3218 {
3219         struct net_device *dev;
3220         struct nlattr *tb[IFLA_MAX+1];
3221         struct net_device *br_dev = NULL;
3222         const struct net_device_ops *ops = NULL;
3223         const struct net_device_ops *cops = NULL;
3224         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3225         struct net *net = sock_net(skb->sk);
3226         struct hlist_head *head;
3227         int brport_idx = 0;
3228         int br_idx = 0;
3229         int h, s_h;
3230         int idx = 0, s_idx;
3231         int err = 0;
3232         int fidx = 0;
3233
3234         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3235                           IFLA_MAX, ifla_policy, NULL);
3236         if (err < 0) {
3237                 return -EINVAL;
3238         } else if (err == 0) {
3239                 if (tb[IFLA_MASTER])
3240                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3241         }
3242
3243         brport_idx = ifm->ifi_index;
3244
3245         if (br_idx) {
3246                 br_dev = __dev_get_by_index(net, br_idx);
3247                 if (!br_dev)
3248                         return -ENODEV;
3249
3250                 ops = br_dev->netdev_ops;
3251         }
3252
3253         s_h = cb->args[0];
3254         s_idx = cb->args[1];
3255
3256         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3257                 idx = 0;
3258                 head = &net->dev_index_head[h];
3259                 hlist_for_each_entry(dev, head, index_hlist) {
3260
3261                         if (brport_idx && (dev->ifindex != brport_idx))
3262                                 continue;
3263
3264                         if (!br_idx) { /* user did not specify a specific bridge */
3265                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3266                                         br_dev = netdev_master_upper_dev_get(dev);
3267                                         cops = br_dev->netdev_ops;
3268                                 }
3269                         } else {
3270                                 if (dev != br_dev &&
3271                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3272                                         continue;
3273
3274                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3275                                     !(dev->priv_flags & IFF_EBRIDGE))
3276                                         continue;
3277                                 cops = ops;
3278                         }
3279
3280                         if (idx < s_idx)
3281                                 goto cont;
3282
3283                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3284                                 if (cops && cops->ndo_fdb_dump) {
3285                                         err = cops->ndo_fdb_dump(skb, cb,
3286                                                                 br_dev, dev,
3287                                                                 &fidx);
3288                                         if (err == -EMSGSIZE)
3289                                                 goto out;
3290                                 }
3291                         }
3292
3293                         if (dev->netdev_ops->ndo_fdb_dump)
3294                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3295                                                                     dev, NULL,
3296                                                                     &fidx);
3297                         else
3298                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3299                                                         &fidx);
3300                         if (err == -EMSGSIZE)
3301                                 goto out;
3302
3303                         cops = NULL;
3304
3305                         /* reset fdb offset to 0 for rest of the interfaces */
3306                         cb->args[2] = 0;
3307                         fidx = 0;
3308 cont:
3309                         idx++;
3310                 }
3311         }
3312
3313 out:
3314         cb->args[0] = h;
3315         cb->args[1] = idx;
3316         cb->args[2] = fidx;
3317
3318         return skb->len;
3319 }
3320
3321 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3322                                unsigned int attrnum, unsigned int flag)
3323 {
3324         if (mask & flag)
3325                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3326         return 0;
3327 }
3328
3329 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3330                             struct net_device *dev, u16 mode,
3331                             u32 flags, u32 mask, int nlflags,
3332                             u32 filter_mask,
3333                             int (*vlan_fill)(struct sk_buff *skb,
3334                                              struct net_device *dev,
3335                                              u32 filter_mask))
3336 {
3337         struct nlmsghdr *nlh;
3338         struct ifinfomsg *ifm;
3339         struct nlattr *br_afspec;
3340         struct nlattr *protinfo;
3341         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3342         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3343         int err = 0;
3344
3345         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3346         if (nlh == NULL)
3347                 return -EMSGSIZE;
3348
3349         ifm = nlmsg_data(nlh);
3350         ifm->ifi_family = AF_BRIDGE;
3351         ifm->__ifi_pad = 0;
3352         ifm->ifi_type = dev->type;
3353         ifm->ifi_index = dev->ifindex;
3354         ifm->ifi_flags = dev_get_flags(dev);
3355         ifm->ifi_change = 0;
3356
3357
3358         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3359             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3360             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3361             (br_dev &&
3362              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3363             (dev->addr_len &&
3364              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3365             (dev->ifindex != dev_get_iflink(dev) &&
3366              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3367                 goto nla_put_failure;
3368
3369         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3370         if (!br_afspec)
3371                 goto nla_put_failure;
3372
3373         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3374                 nla_nest_cancel(skb, br_afspec);
3375                 goto nla_put_failure;
3376         }
3377
3378         if (mode != BRIDGE_MODE_UNDEF) {
3379                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3380                         nla_nest_cancel(skb, br_afspec);
3381                         goto nla_put_failure;
3382                 }
3383         }
3384         if (vlan_fill) {
3385                 err = vlan_fill(skb, dev, filter_mask);
3386                 if (err) {
3387                         nla_nest_cancel(skb, br_afspec);
3388                         goto nla_put_failure;
3389                 }
3390         }
3391         nla_nest_end(skb, br_afspec);
3392
3393         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3394         if (!protinfo)
3395                 goto nla_put_failure;
3396
3397         if (brport_nla_put_flag(skb, flags, mask,
3398                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3399             brport_nla_put_flag(skb, flags, mask,
3400                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3401             brport_nla_put_flag(skb, flags, mask,
3402                                 IFLA_BRPORT_FAST_LEAVE,
3403                                 BR_MULTICAST_FAST_LEAVE) ||
3404             brport_nla_put_flag(skb, flags, mask,
3405                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3406             brport_nla_put_flag(skb, flags, mask,
3407                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3408             brport_nla_put_flag(skb, flags, mask,
3409                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3410             brport_nla_put_flag(skb, flags, mask,
3411                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3412             brport_nla_put_flag(skb, flags, mask,
3413                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3414                 nla_nest_cancel(skb, protinfo);
3415                 goto nla_put_failure;
3416         }
3417
3418         nla_nest_end(skb, protinfo);
3419
3420         nlmsg_end(skb, nlh);
3421         return 0;
3422 nla_put_failure:
3423         nlmsg_cancel(skb, nlh);
3424         return err ? err : -EMSGSIZE;
3425 }
3426 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3427
3428 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3429 {
3430         struct net *net = sock_net(skb->sk);
3431         struct net_device *dev;
3432         int idx = 0;
3433         u32 portid = NETLINK_CB(cb->skb).portid;
3434         u32 seq = cb->nlh->nlmsg_seq;
3435         u32 filter_mask = 0;
3436         int err;
3437
3438         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3439                 struct nlattr *extfilt;
3440
3441                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3442                                           IFLA_EXT_MASK);
3443                 if (extfilt) {
3444                         if (nla_len(extfilt) < sizeof(filter_mask))
3445                                 return -EINVAL;
3446
3447                         filter_mask = nla_get_u32(extfilt);
3448                 }
3449         }
3450
3451         rcu_read_lock();
3452         for_each_netdev_rcu(net, dev) {
3453                 const struct net_device_ops *ops = dev->netdev_ops;
3454                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3455
3456                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3457                         if (idx >= cb->args[0]) {
3458                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3459                                                 skb, portid, seq, dev,
3460                                                 filter_mask, NLM_F_MULTI);
3461                                 if (err < 0 && err != -EOPNOTSUPP) {
3462                                         if (likely(skb->len))
3463                                                 break;
3464
3465                                         goto out_err;
3466                                 }
3467                         }
3468                         idx++;
3469                 }
3470
3471                 if (ops->ndo_bridge_getlink) {
3472                         if (idx >= cb->args[0]) {
3473                                 err = ops->ndo_bridge_getlink(skb, portid,
3474                                                               seq, dev,
3475                                                               filter_mask,
3476                                                               NLM_F_MULTI);
3477                                 if (err < 0 && err != -EOPNOTSUPP) {
3478                                         if (likely(skb->len))
3479                                                 break;
3480
3481                                         goto out_err;
3482                                 }
3483                         }
3484                         idx++;
3485                 }
3486         }
3487         err = skb->len;
3488 out_err:
3489         rcu_read_unlock();
3490         cb->args[0] = idx;
3491
3492         return err;
3493 }
3494
3495 static inline size_t bridge_nlmsg_size(void)
3496 {
3497         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3498                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3499                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3500                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3501                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3502                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3503                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3504                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3505                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3506                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3507                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3508 }
3509
3510 static int rtnl_bridge_notify(struct net_device *dev)
3511 {
3512         struct net *net = dev_net(dev);
3513         struct sk_buff *skb;
3514         int err = -EOPNOTSUPP;
3515
3516         if (!dev->netdev_ops->ndo_bridge_getlink)
3517                 return 0;
3518
3519         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3520         if (!skb) {
3521                 err = -ENOMEM;
3522                 goto errout;
3523         }
3524
3525         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3526         if (err < 0)
3527                 goto errout;
3528
3529         if (!skb->len)
3530                 goto errout;
3531
3532         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3533         return 0;
3534 errout:
3535         WARN_ON(err == -EMSGSIZE);
3536         kfree_skb(skb);
3537         if (err)
3538                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3539         return err;
3540 }
3541
3542 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3543                                struct netlink_ext_ack *extack)
3544 {
3545         struct net *net = sock_net(skb->sk);
3546         struct ifinfomsg *ifm;
3547         struct net_device *dev;
3548         struct nlattr *br_spec, *attr = NULL;
3549         int rem, err = -EOPNOTSUPP;
3550         u16 flags = 0;
3551         bool have_flags = false;
3552
3553         if (nlmsg_len(nlh) < sizeof(*ifm))
3554                 return -EINVAL;
3555
3556         ifm = nlmsg_data(nlh);
3557         if (ifm->ifi_family != AF_BRIDGE)
3558                 return -EPFNOSUPPORT;
3559
3560         dev = __dev_get_by_index(net, ifm->ifi_index);
3561         if (!dev) {
3562                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3563                 return -ENODEV;
3564         }
3565
3566         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3567         if (br_spec) {
3568                 nla_for_each_nested(attr, br_spec, rem) {
3569                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3570                                 if (nla_len(attr) < sizeof(flags))
3571                                         return -EINVAL;
3572
3573                                 have_flags = true;
3574                                 flags = nla_get_u16(attr);
3575                                 break;
3576                         }
3577                 }
3578         }
3579
3580         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3581                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3582
3583                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3584                         err = -EOPNOTSUPP;
3585                         goto out;
3586                 }
3587
3588                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3589                 if (err)
3590                         goto out;
3591
3592                 flags &= ~BRIDGE_FLAGS_MASTER;
3593         }
3594
3595         if ((flags & BRIDGE_FLAGS_SELF)) {
3596                 if (!dev->netdev_ops->ndo_bridge_setlink)
3597                         err = -EOPNOTSUPP;
3598                 else
3599                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3600                                                                   flags);
3601                 if (!err) {
3602                         flags &= ~BRIDGE_FLAGS_SELF;
3603
3604                         /* Generate event to notify upper layer of bridge
3605                          * change
3606                          */
3607                         err = rtnl_bridge_notify(dev);
3608                 }
3609         }
3610
3611         if (have_flags)
3612                 memcpy(nla_data(attr), &flags, sizeof(flags));
3613 out:
3614         return err;
3615 }
3616
3617 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3618                                struct netlink_ext_ack *extack)
3619 {
3620         struct net *net = sock_net(skb->sk);
3621         struct ifinfomsg *ifm;
3622         struct net_device *dev;
3623         struct nlattr *br_spec, *attr = NULL;
3624         int rem, err = -EOPNOTSUPP;
3625         u16 flags = 0;
3626         bool have_flags = false;
3627
3628         if (nlmsg_len(nlh) < sizeof(*ifm))
3629                 return -EINVAL;
3630
3631         ifm = nlmsg_data(nlh);
3632         if (ifm->ifi_family != AF_BRIDGE)
3633                 return -EPFNOSUPPORT;
3634
3635         dev = __dev_get_by_index(net, ifm->ifi_index);
3636         if (!dev) {
3637                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3638                 return -ENODEV;
3639         }
3640
3641         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3642         if (br_spec) {
3643                 nla_for_each_nested(attr, br_spec, rem) {
3644                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3645                                 if (nla_len(attr) < sizeof(flags))
3646                                         return -EINVAL;
3647
3648                                 have_flags = true;
3649                                 flags = nla_get_u16(attr);
3650                                 break;
3651                         }
3652                 }
3653         }
3654
3655         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3656                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3657
3658                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3659                         err = -EOPNOTSUPP;
3660                         goto out;
3661                 }
3662
3663                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3664                 if (err)
3665                         goto out;
3666
3667                 flags &= ~BRIDGE_FLAGS_MASTER;
3668         }
3669
3670         if ((flags & BRIDGE_FLAGS_SELF)) {
3671                 if (!dev->netdev_ops->ndo_bridge_dellink)
3672                         err = -EOPNOTSUPP;
3673                 else
3674                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3675                                                                   flags);
3676
3677                 if (!err) {
3678                         flags &= ~BRIDGE_FLAGS_SELF;
3679
3680                         /* Generate event to notify upper layer of bridge
3681                          * change
3682                          */
3683                         err = rtnl_bridge_notify(dev);
3684                 }
3685         }
3686
3687         if (have_flags)
3688                 memcpy(nla_data(attr), &flags, sizeof(flags));
3689 out:
3690         return err;
3691 }
3692
3693 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3694 {
3695         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3696                (!idxattr || idxattr == attrid);
3697 }
3698
3699 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3700 static int rtnl_get_offload_stats_attr_size(int attr_id)
3701 {
3702         switch (attr_id) {
3703         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3704                 return sizeof(struct rtnl_link_stats64);
3705         }
3706
3707         return 0;
3708 }
3709
3710 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3711                                   int *prividx)
3712 {
3713         struct nlattr *attr = NULL;
3714         int attr_id, size;
3715         void *attr_data;
3716         int err;
3717
3718         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3719               dev->netdev_ops->ndo_get_offload_stats))
3720                 return -ENODATA;
3721
3722         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3723              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3724                 if (attr_id < *prividx)
3725                         continue;
3726
3727                 size = rtnl_get_offload_stats_attr_size(attr_id);
3728                 if (!size)
3729                         continue;
3730
3731                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3732                         continue;
3733
3734                 attr = nla_reserve_64bit(skb, attr_id, size,
3735                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3736                 if (!attr)
3737                         goto nla_put_failure;
3738
3739                 attr_data = nla_data(attr);
3740                 memset(attr_data, 0, size);
3741                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3742                                                              attr_data);
3743                 if (err)
3744                         goto get_offload_stats_failure;
3745         }
3746
3747         if (!attr)
3748                 return -ENODATA;
3749
3750         *prividx = 0;
3751         return 0;
3752
3753 nla_put_failure:
3754         err = -EMSGSIZE;
3755 get_offload_stats_failure:
3756         *prividx = attr_id;
3757         return err;
3758 }
3759
3760 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3761 {
3762         int nla_size = 0;
3763         int attr_id;
3764         int size;
3765
3766         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3767               dev->netdev_ops->ndo_get_offload_stats))
3768                 return 0;
3769
3770         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3771              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3772                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3773                         continue;
3774                 size = rtnl_get_offload_stats_attr_size(attr_id);
3775                 nla_size += nla_total_size_64bit(size);
3776         }
3777
3778         if (nla_size != 0)
3779                 nla_size += nla_total_size(0);
3780
3781         return nla_size;
3782 }
3783
3784 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3785                                int type, u32 pid, u32 seq, u32 change,
3786                                unsigned int flags, unsigned int filter_mask,
3787                                int *idxattr, int *prividx)
3788 {
3789         struct if_stats_msg *ifsm;
3790         struct nlmsghdr *nlh;
3791         struct nlattr *attr;
3792         int s_prividx = *prividx;
3793         int err;
3794
3795         ASSERT_RTNL();
3796
3797         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3798         if (!nlh)
3799                 return -EMSGSIZE;
3800
3801         ifsm = nlmsg_data(nlh);
3802         ifsm->ifindex = dev->ifindex;
3803         ifsm->filter_mask = filter_mask;
3804
3805         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3806                 struct rtnl_link_stats64 *sp;
3807
3808                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3809                                          sizeof(struct rtnl_link_stats64),
3810                                          IFLA_STATS_UNSPEC);
3811                 if (!attr)
3812                         goto nla_put_failure;
3813
3814                 sp = nla_data(attr);
3815                 dev_get_stats(dev, sp);
3816         }
3817
3818         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3819                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3820
3821                 if (ops && ops->fill_linkxstats) {
3822                         *idxattr = IFLA_STATS_LINK_XSTATS;
3823                         attr = nla_nest_start(skb,
3824                                               IFLA_STATS_LINK_XSTATS);
3825                         if (!attr)
3826                                 goto nla_put_failure;
3827
3828                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3829                         nla_nest_end(skb, attr);
3830                         if (err)
3831                                 goto nla_put_failure;
3832                         *idxattr = 0;
3833                 }
3834         }
3835
3836         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3837                              *idxattr)) {
3838                 const struct rtnl_link_ops *ops = NULL;
3839                 const struct net_device *master;
3840
3841                 master = netdev_master_upper_dev_get(dev);
3842                 if (master)
3843                         ops = master->rtnl_link_ops;
3844                 if (ops && ops->fill_linkxstats) {
3845                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3846                         attr = nla_nest_start(skb,
3847                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3848                         if (!attr)
3849                                 goto nla_put_failure;
3850
3851                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3852                         nla_nest_end(skb, attr);
3853                         if (err)
3854                                 goto nla_put_failure;
3855                         *idxattr = 0;
3856                 }
3857         }
3858
3859         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3860                              *idxattr)) {
3861                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3862                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3863                 if (!attr)
3864                         goto nla_put_failure;
3865
3866                 err = rtnl_get_offload_stats(skb, dev, prividx);
3867                 if (err == -ENODATA)
3868                         nla_nest_cancel(skb, attr);
3869                 else
3870                         nla_nest_end(skb, attr);
3871
3872                 if (err && err != -ENODATA)
3873                         goto nla_put_failure;
3874                 *idxattr = 0;
3875         }
3876
3877         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3878                 struct rtnl_af_ops *af_ops;
3879
3880                 *idxattr = IFLA_STATS_AF_SPEC;
3881                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3882                 if (!attr)
3883                         goto nla_put_failure;
3884
3885                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3886                         if (af_ops->fill_stats_af) {
3887                                 struct nlattr *af;
3888                                 int err;
3889
3890                                 af = nla_nest_start(skb, af_ops->family);
3891                                 if (!af)
3892                                         goto nla_put_failure;
3893
3894                                 err = af_ops->fill_stats_af(skb, dev);
3895
3896                                 if (err == -ENODATA)
3897                                         nla_nest_cancel(skb, af);
3898                                 else if (err < 0)
3899                                         goto nla_put_failure;
3900
3901                                 nla_nest_end(skb, af);
3902                         }
3903                 }
3904
3905                 nla_nest_end(skb, attr);
3906
3907                 *idxattr = 0;
3908         }
3909
3910         nlmsg_end(skb, nlh);
3911
3912         return 0;
3913
3914 nla_put_failure:
3915         /* not a multi message or no progress mean a real error */
3916         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3917                 nlmsg_cancel(skb, nlh);
3918         else
3919                 nlmsg_end(skb, nlh);
3920
3921         return -EMSGSIZE;
3922 }
3923
3924 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3925                                   u32 filter_mask)
3926 {
3927         size_t size = 0;
3928
3929         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3930                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3931
3932         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3933                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3934                 int attr = IFLA_STATS_LINK_XSTATS;
3935
3936                 if (ops && ops->get_linkxstats_size) {
3937                         size += nla_total_size(ops->get_linkxstats_size(dev,
3938                                                                         attr));
3939                         /* for IFLA_STATS_LINK_XSTATS */
3940                         size += nla_total_size(0);
3941                 }
3942         }
3943
3944         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3945                 struct net_device *_dev = (struct net_device *)dev;
3946                 const struct rtnl_link_ops *ops = NULL;
3947                 const struct net_device *master;
3948
3949                 /* netdev_master_upper_dev_get can't take const */
3950                 master = netdev_master_upper_dev_get(_dev);
3951                 if (master)
3952                         ops = master->rtnl_link_ops;
3953                 if (ops && ops->get_linkxstats_size) {
3954                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3955
3956                         size += nla_total_size(ops->get_linkxstats_size(dev,
3957                                                                         attr));
3958                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3959                         size += nla_total_size(0);
3960                 }
3961         }
3962
3963         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3964                 size += rtnl_get_offload_stats_size(dev);
3965
3966         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3967                 struct rtnl_af_ops *af_ops;
3968
3969                 /* for IFLA_STATS_AF_SPEC */
3970                 size += nla_total_size(0);
3971
3972                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3973                         if (af_ops->get_stats_af_size) {
3974                                 size += nla_total_size(
3975                                         af_ops->get_stats_af_size(dev));
3976
3977                                 /* for AF_* */
3978                                 size += nla_total_size(0);
3979                         }
3980                 }
3981         }
3982
3983         return size;
3984 }
3985
3986 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
3987                           struct netlink_ext_ack *extack)
3988 {
3989         struct net *net = sock_net(skb->sk);
3990         struct net_device *dev = NULL;
3991         int idxattr = 0, prividx = 0;
3992         struct if_stats_msg *ifsm;
3993         struct sk_buff *nskb;
3994         u32 filter_mask;
3995         int err;
3996
3997         if (nlmsg_len(nlh) < sizeof(*ifsm))
3998                 return -EINVAL;
3999
4000         ifsm = nlmsg_data(nlh);
4001         if (ifsm->ifindex > 0)
4002                 dev = __dev_get_by_index(net, ifsm->ifindex);
4003         else
4004                 return -EINVAL;
4005
4006         if (!dev)
4007                 return -ENODEV;
4008
4009         filter_mask = ifsm->filter_mask;
4010         if (!filter_mask)
4011                 return -EINVAL;
4012
4013         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4014         if (!nskb)
4015                 return -ENOBUFS;
4016
4017         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4018                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4019                                   0, filter_mask, &idxattr, &prividx);
4020         if (err < 0) {
4021                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4022                 WARN_ON(err == -EMSGSIZE);
4023                 kfree_skb(nskb);
4024         } else {
4025                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4026         }
4027
4028         return err;
4029 }
4030
4031 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4032 {
4033         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4034         struct net *net = sock_net(skb->sk);
4035         unsigned int flags = NLM_F_MULTI;
4036         struct if_stats_msg *ifsm;
4037         struct hlist_head *head;
4038         struct net_device *dev;
4039         u32 filter_mask = 0;
4040         int idx = 0;
4041
4042         s_h = cb->args[0];
4043         s_idx = cb->args[1];
4044         s_idxattr = cb->args[2];
4045         s_prividx = cb->args[3];
4046
4047         cb->seq = net->dev_base_seq;
4048
4049         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4050                 return -EINVAL;
4051
4052         ifsm = nlmsg_data(cb->nlh);
4053         filter_mask = ifsm->filter_mask;
4054         if (!filter_mask)
4055                 return -EINVAL;
4056
4057         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4058                 idx = 0;
4059                 head = &net->dev_index_head[h];
4060                 hlist_for_each_entry(dev, head, index_hlist) {
4061                         if (idx < s_idx)
4062                                 goto cont;
4063                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4064                                                   NETLINK_CB(cb->skb).portid,
4065                                                   cb->nlh->nlmsg_seq, 0,
4066                                                   flags, filter_mask,
4067                                                   &s_idxattr, &s_prividx);
4068                         /* If we ran out of room on the first message,
4069                          * we're in trouble
4070                          */
4071                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4072
4073                         if (err < 0)
4074                                 goto out;
4075                         s_prividx = 0;
4076                         s_idxattr = 0;
4077                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4078 cont:
4079                         idx++;
4080                 }
4081         }
4082 out:
4083         cb->args[3] = s_prividx;
4084         cb->args[2] = s_idxattr;
4085         cb->args[1] = idx;
4086         cb->args[0] = h;
4087
4088         return skb->len;
4089 }
4090
4091 /* Process one rtnetlink message. */
4092
4093 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4094                              struct netlink_ext_ack *extack)
4095 {
4096         struct net *net = sock_net(skb->sk);
4097         rtnl_doit_func doit;
4098         int kind;
4099         int family;
4100         int type;
4101         int err;
4102
4103         type = nlh->nlmsg_type;
4104         if (type > RTM_MAX)
4105                 return -EOPNOTSUPP;
4106
4107         type -= RTM_BASE;
4108
4109         /* All the messages must have at least 1 byte length */
4110         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4111                 return 0;
4112
4113         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4114         kind = type&3;
4115
4116         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4117                 return -EPERM;
4118
4119         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4120                 struct sock *rtnl;
4121                 rtnl_dumpit_func dumpit;
4122                 rtnl_calcit_func calcit;
4123                 u16 min_dump_alloc = 0;
4124
4125                 dumpit = rtnl_get_dumpit(family, type);
4126                 if (dumpit == NULL)
4127                         return -EOPNOTSUPP;
4128                 calcit = rtnl_get_calcit(family, type);
4129                 if (calcit)
4130                         min_dump_alloc = calcit(skb, nlh);
4131
4132                 __rtnl_unlock();
4133                 rtnl = net->rtnl;
4134                 {
4135                         struct netlink_dump_control c = {
4136                                 .dump           = dumpit,
4137                                 .min_dump_alloc = min_dump_alloc,
4138                         };
4139                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4140                 }
4141                 rtnl_lock();
4142                 return err;
4143         }
4144
4145         doit = rtnl_get_doit(family, type);
4146         if (doit == NULL)
4147                 return -EOPNOTSUPP;
4148
4149         return doit(skb, nlh, extack);
4150 }
4151
4152 static void rtnetlink_rcv(struct sk_buff *skb)
4153 {
4154         rtnl_lock();
4155         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4156         rtnl_unlock();
4157 }
4158
4159 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4160 {
4161         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4162
4163         switch (event) {
4164         case NETDEV_REBOOT:
4165         case NETDEV_CHANGENAME:
4166         case NETDEV_FEAT_CHANGE:
4167         case NETDEV_BONDING_FAILOVER:
4168         case NETDEV_NOTIFY_PEERS:
4169         case NETDEV_RESEND_IGMP:
4170         case NETDEV_CHANGEINFODATA:
4171                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4172                 break;
4173         default:
4174                 break;
4175         }
4176         return NOTIFY_DONE;
4177 }
4178
4179 static struct notifier_block rtnetlink_dev_notifier = {
4180         .notifier_call  = rtnetlink_event,
4181 };
4182
4183
4184 static int __net_init rtnetlink_net_init(struct net *net)
4185 {
4186         struct sock *sk;
4187         struct netlink_kernel_cfg cfg = {
4188                 .groups         = RTNLGRP_MAX,
4189                 .input          = rtnetlink_rcv,
4190                 .cb_mutex       = &rtnl_mutex,
4191                 .flags          = NL_CFG_F_NONROOT_RECV,
4192         };
4193
4194         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4195         if (!sk)
4196                 return -ENOMEM;
4197         net->rtnl = sk;
4198         return 0;
4199 }
4200
4201 static void __net_exit rtnetlink_net_exit(struct net *net)
4202 {
4203         netlink_kernel_release(net->rtnl);
4204         net->rtnl = NULL;
4205 }
4206
4207 static struct pernet_operations rtnetlink_net_ops = {
4208         .init = rtnetlink_net_init,
4209         .exit = rtnetlink_net_exit,
4210 };
4211
4212 void __init rtnetlink_init(void)
4213 {
4214         if (register_pernet_subsys(&rtnetlink_net_ops))
4215                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4216
4217         register_netdevice_notifier(&rtnetlink_dev_notifier);
4218
4219         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4220                       rtnl_dump_ifinfo, rtnl_calcit);
4221         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4222         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4223         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4224
4225         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4226         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4227         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4228
4229         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4230         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4231         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4232
4233         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4234         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4235         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4236
4237         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4238                       NULL);
4239 }