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                         /* If we ran out of room on the first message,
1631                          * we're in trouble
1632                          */
1633                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1634
1635                         if (err < 0)
1636                                 goto out;
1637
1638                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1639 cont:
1640                         idx++;
1641                 }
1642         }
1643 out:
1644         cb->args[1] = idx;
1645         cb->args[0] = h;
1646
1647         return skb->len;
1648 }
1649
1650 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1651                         struct netlink_ext_ack *exterr)
1652 {
1653         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1654 }
1655 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1656
1657 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1658 {
1659         struct net *net;
1660         /* Examine the link attributes and figure out which
1661          * network namespace we are talking about.
1662          */
1663         if (tb[IFLA_NET_NS_PID])
1664                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1665         else if (tb[IFLA_NET_NS_FD])
1666                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1667         else
1668                 net = get_net(src_net);
1669         return net;
1670 }
1671 EXPORT_SYMBOL(rtnl_link_get_net);
1672
1673 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1674 {
1675         if (dev) {
1676                 if (tb[IFLA_ADDRESS] &&
1677                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1678                         return -EINVAL;
1679
1680                 if (tb[IFLA_BROADCAST] &&
1681                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1682                         return -EINVAL;
1683         }
1684
1685         if (tb[IFLA_AF_SPEC]) {
1686                 struct nlattr *af;
1687                 int rem, err;
1688
1689                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1690                         const struct rtnl_af_ops *af_ops;
1691
1692                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1693                                 return -EAFNOSUPPORT;
1694
1695                         if (!af_ops->set_link_af)
1696                                 return -EOPNOTSUPP;
1697
1698                         if (af_ops->validate_link_af) {
1699                                 err = af_ops->validate_link_af(dev, af);
1700                                 if (err < 0)
1701                                         return err;
1702                         }
1703                 }
1704         }
1705
1706         return 0;
1707 }
1708
1709 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1710                                   int guid_type)
1711 {
1712         const struct net_device_ops *ops = dev->netdev_ops;
1713
1714         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1715 }
1716
1717 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1718 {
1719         if (dev->type != ARPHRD_INFINIBAND)
1720                 return -EOPNOTSUPP;
1721
1722         return handle_infiniband_guid(dev, ivt, guid_type);
1723 }
1724
1725 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1726 {
1727         const struct net_device_ops *ops = dev->netdev_ops;
1728         int err = -EINVAL;
1729
1730         if (tb[IFLA_VF_MAC]) {
1731                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1732
1733                 err = -EOPNOTSUPP;
1734                 if (ops->ndo_set_vf_mac)
1735                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1736                                                   ivm->mac);
1737                 if (err < 0)
1738                         return err;
1739         }
1740
1741         if (tb[IFLA_VF_VLAN]) {
1742                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1743
1744                 err = -EOPNOTSUPP;
1745                 if (ops->ndo_set_vf_vlan)
1746                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1747                                                    ivv->qos,
1748                                                    htons(ETH_P_8021Q));
1749                 if (err < 0)
1750                         return err;
1751         }
1752
1753         if (tb[IFLA_VF_VLAN_LIST]) {
1754                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1755                 struct nlattr *attr;
1756                 int rem, len = 0;
1757
1758                 err = -EOPNOTSUPP;
1759                 if (!ops->ndo_set_vf_vlan)
1760                         return err;
1761
1762                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1763                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1764                             nla_len(attr) < NLA_HDRLEN) {
1765                                 return -EINVAL;
1766                         }
1767                         if (len >= MAX_VLAN_LIST_LEN)
1768                                 return -EOPNOTSUPP;
1769                         ivvl[len] = nla_data(attr);
1770
1771                         len++;
1772                 }
1773                 if (len == 0)
1774                         return -EINVAL;
1775
1776                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1777                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1778                 if (err < 0)
1779                         return err;
1780         }
1781
1782         if (tb[IFLA_VF_TX_RATE]) {
1783                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1784                 struct ifla_vf_info ivf;
1785
1786                 err = -EOPNOTSUPP;
1787                 if (ops->ndo_get_vf_config)
1788                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1789                 if (err < 0)
1790                         return err;
1791
1792                 err = -EOPNOTSUPP;
1793                 if (ops->ndo_set_vf_rate)
1794                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1795                                                    ivf.min_tx_rate,
1796                                                    ivt->rate);
1797                 if (err < 0)
1798                         return err;
1799         }
1800
1801         if (tb[IFLA_VF_RATE]) {
1802                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1803
1804                 err = -EOPNOTSUPP;
1805                 if (ops->ndo_set_vf_rate)
1806                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1807                                                    ivt->min_tx_rate,
1808                                                    ivt->max_tx_rate);
1809                 if (err < 0)
1810                         return err;
1811         }
1812
1813         if (tb[IFLA_VF_SPOOFCHK]) {
1814                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1815
1816                 err = -EOPNOTSUPP;
1817                 if (ops->ndo_set_vf_spoofchk)
1818                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1819                                                        ivs->setting);
1820                 if (err < 0)
1821                         return err;
1822         }
1823
1824         if (tb[IFLA_VF_LINK_STATE]) {
1825                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1826
1827                 err = -EOPNOTSUPP;
1828                 if (ops->ndo_set_vf_link_state)
1829                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1830                                                          ivl->link_state);
1831                 if (err < 0)
1832                         return err;
1833         }
1834
1835         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1836                 struct ifla_vf_rss_query_en *ivrssq_en;
1837
1838                 err = -EOPNOTSUPP;
1839                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1840                 if (ops->ndo_set_vf_rss_query_en)
1841                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1842                                                            ivrssq_en->setting);
1843                 if (err < 0)
1844                         return err;
1845         }
1846
1847         if (tb[IFLA_VF_TRUST]) {
1848                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1849
1850                 err = -EOPNOTSUPP;
1851                 if (ops->ndo_set_vf_trust)
1852                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1853                 if (err < 0)
1854                         return err;
1855         }
1856
1857         if (tb[IFLA_VF_IB_NODE_GUID]) {
1858                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1859
1860                 if (!ops->ndo_set_vf_guid)
1861                         return -EOPNOTSUPP;
1862
1863                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1864         }
1865
1866         if (tb[IFLA_VF_IB_PORT_GUID]) {
1867                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1868
1869                 if (!ops->ndo_set_vf_guid)
1870                         return -EOPNOTSUPP;
1871
1872                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1873         }
1874
1875         return err;
1876 }
1877
1878 static int do_set_master(struct net_device *dev, int ifindex)
1879 {
1880         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1881         const struct net_device_ops *ops;
1882         int err;
1883
1884         if (upper_dev) {
1885                 if (upper_dev->ifindex == ifindex)
1886                         return 0;
1887                 ops = upper_dev->netdev_ops;
1888                 if (ops->ndo_del_slave) {
1889                         err = ops->ndo_del_slave(upper_dev, dev);
1890                         if (err)
1891                                 return err;
1892                 } else {
1893                         return -EOPNOTSUPP;
1894                 }
1895         }
1896
1897         if (ifindex) {
1898                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1899                 if (!upper_dev)
1900                         return -EINVAL;
1901                 ops = upper_dev->netdev_ops;
1902                 if (ops->ndo_add_slave) {
1903                         err = ops->ndo_add_slave(upper_dev, dev);
1904                         if (err)
1905                                 return err;
1906                 } else {
1907                         return -EOPNOTSUPP;
1908                 }
1909         }
1910         return 0;
1911 }
1912
1913 #define DO_SETLINK_MODIFIED     0x01
1914 /* notify flag means notify + modified. */
1915 #define DO_SETLINK_NOTIFY       0x03
1916 static int do_setlink(const struct sk_buff *skb,
1917                       struct net_device *dev, struct ifinfomsg *ifm,
1918                       struct netlink_ext_ack *extack,
1919                       struct nlattr **tb, char *ifname, int status)
1920 {
1921         const struct net_device_ops *ops = dev->netdev_ops;
1922         int err;
1923
1924         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1925                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1926                 if (IS_ERR(net)) {
1927                         err = PTR_ERR(net);
1928                         goto errout;
1929                 }
1930                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1931                         put_net(net);
1932                         err = -EPERM;
1933                         goto errout;
1934                 }
1935                 err = dev_change_net_namespace(dev, net, ifname);
1936                 put_net(net);
1937                 if (err)
1938                         goto errout;
1939                 status |= DO_SETLINK_MODIFIED;
1940         }
1941
1942         if (tb[IFLA_MAP]) {
1943                 struct rtnl_link_ifmap *u_map;
1944                 struct ifmap k_map;
1945
1946                 if (!ops->ndo_set_config) {
1947                         err = -EOPNOTSUPP;
1948                         goto errout;
1949                 }
1950
1951                 if (!netif_device_present(dev)) {
1952                         err = -ENODEV;
1953                         goto errout;
1954                 }
1955
1956                 u_map = nla_data(tb[IFLA_MAP]);
1957                 k_map.mem_start = (unsigned long) u_map->mem_start;
1958                 k_map.mem_end = (unsigned long) u_map->mem_end;
1959                 k_map.base_addr = (unsigned short) u_map->base_addr;
1960                 k_map.irq = (unsigned char) u_map->irq;
1961                 k_map.dma = (unsigned char) u_map->dma;
1962                 k_map.port = (unsigned char) u_map->port;
1963
1964                 err = ops->ndo_set_config(dev, &k_map);
1965                 if (err < 0)
1966                         goto errout;
1967
1968                 status |= DO_SETLINK_NOTIFY;
1969         }
1970
1971         if (tb[IFLA_ADDRESS]) {
1972                 struct sockaddr *sa;
1973                 int len;
1974
1975                 len = sizeof(sa_family_t) + dev->addr_len;
1976                 sa = kmalloc(len, GFP_KERNEL);
1977                 if (!sa) {
1978                         err = -ENOMEM;
1979                         goto errout;
1980                 }
1981                 sa->sa_family = dev->type;
1982                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1983                        dev->addr_len);
1984                 err = dev_set_mac_address(dev, sa);
1985                 kfree(sa);
1986                 if (err)
1987                         goto errout;
1988                 status |= DO_SETLINK_MODIFIED;
1989         }
1990
1991         if (tb[IFLA_MTU]) {
1992                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1993                 if (err < 0)
1994                         goto errout;
1995                 status |= DO_SETLINK_MODIFIED;
1996         }
1997
1998         if (tb[IFLA_GROUP]) {
1999                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2000                 status |= DO_SETLINK_NOTIFY;
2001         }
2002
2003         /*
2004          * Interface selected by interface index but interface
2005          * name provided implies that a name change has been
2006          * requested.
2007          */
2008         if (ifm->ifi_index > 0 && ifname[0]) {
2009                 err = dev_change_name(dev, ifname);
2010                 if (err < 0)
2011                         goto errout;
2012                 status |= DO_SETLINK_MODIFIED;
2013         }
2014
2015         if (tb[IFLA_IFALIAS]) {
2016                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2017                                     nla_len(tb[IFLA_IFALIAS]));
2018                 if (err < 0)
2019                         goto errout;
2020                 status |= DO_SETLINK_NOTIFY;
2021         }
2022
2023         if (tb[IFLA_BROADCAST]) {
2024                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2025                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2026         }
2027
2028         if (ifm->ifi_flags || ifm->ifi_change) {
2029                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2030                 if (err < 0)
2031                         goto errout;
2032         }
2033
2034         if (tb[IFLA_MASTER]) {
2035                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2036                 if (err)
2037                         goto errout;
2038                 status |= DO_SETLINK_MODIFIED;
2039         }
2040
2041         if (tb[IFLA_CARRIER]) {
2042                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2043                 if (err)
2044                         goto errout;
2045                 status |= DO_SETLINK_MODIFIED;
2046         }
2047
2048         if (tb[IFLA_TXQLEN]) {
2049                 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2050                 unsigned long orig_len = dev->tx_queue_len;
2051
2052                 if (dev->tx_queue_len ^ value) {
2053                         dev->tx_queue_len = value;
2054                         err = call_netdevice_notifiers(
2055                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2056                         err = notifier_to_errno(err);
2057                         if (err) {
2058                                 dev->tx_queue_len = orig_len;
2059                                 goto errout;
2060                         }
2061                         status |= DO_SETLINK_NOTIFY;
2062                 }
2063         }
2064
2065         if (tb[IFLA_OPERSTATE])
2066                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2067
2068         if (tb[IFLA_LINKMODE]) {
2069                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2070
2071                 write_lock_bh(&dev_base_lock);
2072                 if (dev->link_mode ^ value)
2073                         status |= DO_SETLINK_NOTIFY;
2074                 dev->link_mode = value;
2075                 write_unlock_bh(&dev_base_lock);
2076         }
2077
2078         if (tb[IFLA_VFINFO_LIST]) {
2079                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2080                 struct nlattr *attr;
2081                 int rem;
2082
2083                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2084                         if (nla_type(attr) != IFLA_VF_INFO ||
2085                             nla_len(attr) < NLA_HDRLEN) {
2086                                 err = -EINVAL;
2087                                 goto errout;
2088                         }
2089                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2090                                                ifla_vf_policy, NULL);
2091                         if (err < 0)
2092                                 goto errout;
2093                         err = do_setvfinfo(dev, vfinfo);
2094                         if (err < 0)
2095                                 goto errout;
2096                         status |= DO_SETLINK_NOTIFY;
2097                 }
2098         }
2099         err = 0;
2100
2101         if (tb[IFLA_VF_PORTS]) {
2102                 struct nlattr *port[IFLA_PORT_MAX+1];
2103                 struct nlattr *attr;
2104                 int vf;
2105                 int rem;
2106
2107                 err = -EOPNOTSUPP;
2108                 if (!ops->ndo_set_vf_port)
2109                         goto errout;
2110
2111                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2112                         if (nla_type(attr) != IFLA_VF_PORT ||
2113                             nla_len(attr) < NLA_HDRLEN) {
2114                                 err = -EINVAL;
2115                                 goto errout;
2116                         }
2117                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2118                                                ifla_port_policy, NULL);
2119                         if (err < 0)
2120                                 goto errout;
2121                         if (!port[IFLA_PORT_VF]) {
2122                                 err = -EOPNOTSUPP;
2123                                 goto errout;
2124                         }
2125                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2126                         err = ops->ndo_set_vf_port(dev, vf, port);
2127                         if (err < 0)
2128                                 goto errout;
2129                         status |= DO_SETLINK_NOTIFY;
2130                 }
2131         }
2132         err = 0;
2133
2134         if (tb[IFLA_PORT_SELF]) {
2135                 struct nlattr *port[IFLA_PORT_MAX+1];
2136
2137                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2138                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2139                                        NULL);
2140                 if (err < 0)
2141                         goto errout;
2142
2143                 err = -EOPNOTSUPP;
2144                 if (ops->ndo_set_vf_port)
2145                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2146                 if (err < 0)
2147                         goto errout;
2148                 status |= DO_SETLINK_NOTIFY;
2149         }
2150
2151         if (tb[IFLA_AF_SPEC]) {
2152                 struct nlattr *af;
2153                 int rem;
2154
2155                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2156                         const struct rtnl_af_ops *af_ops;
2157
2158                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2159                                 BUG();
2160
2161                         err = af_ops->set_link_af(dev, af);
2162                         if (err < 0)
2163                                 goto errout;
2164
2165                         status |= DO_SETLINK_NOTIFY;
2166                 }
2167         }
2168         err = 0;
2169
2170         if (tb[IFLA_PROTO_DOWN]) {
2171                 err = dev_change_proto_down(dev,
2172                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2173                 if (err)
2174                         goto errout;
2175                 status |= DO_SETLINK_NOTIFY;
2176         }
2177
2178         if (tb[IFLA_XDP]) {
2179                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2180                 u32 xdp_flags = 0;
2181
2182                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2183                                        ifla_xdp_policy, NULL);
2184                 if (err < 0)
2185                         goto errout;
2186
2187                 if (xdp[IFLA_XDP_ATTACHED]) {
2188                         err = -EINVAL;
2189                         goto errout;
2190                 }
2191
2192                 if (xdp[IFLA_XDP_FLAGS]) {
2193                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2194                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2195                                 err = -EINVAL;
2196                                 goto errout;
2197                         }
2198                         if ((xdp_flags & XDP_FLAGS_SKB_MODE) &&
2199                             (xdp_flags & XDP_FLAGS_DRV_MODE)) {
2200                                 err = -EINVAL;
2201                                 goto errout;
2202                         }
2203                 }
2204
2205                 if (xdp[IFLA_XDP_FD]) {
2206                         err = dev_change_xdp_fd(dev, extack,
2207                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2208                                                 xdp_flags);
2209                         if (err)
2210                                 goto errout;
2211                         status |= DO_SETLINK_NOTIFY;
2212                 }
2213         }
2214
2215 errout:
2216         if (status & DO_SETLINK_MODIFIED) {
2217                 if (status & DO_SETLINK_NOTIFY)
2218                         netdev_state_change(dev);
2219
2220                 if (err < 0)
2221                         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",
2222                                              dev->name);
2223         }
2224
2225         return err;
2226 }
2227
2228 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2229                         struct netlink_ext_ack *extack)
2230 {
2231         struct net *net = sock_net(skb->sk);
2232         struct ifinfomsg *ifm;
2233         struct net_device *dev;
2234         int err;
2235         struct nlattr *tb[IFLA_MAX+1];
2236         char ifname[IFNAMSIZ];
2237
2238         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2239                           extack);
2240         if (err < 0)
2241                 goto errout;
2242
2243         if (tb[IFLA_IFNAME])
2244                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2245         else
2246                 ifname[0] = '\0';
2247
2248         err = -EINVAL;
2249         ifm = nlmsg_data(nlh);
2250         if (ifm->ifi_index > 0)
2251                 dev = __dev_get_by_index(net, ifm->ifi_index);
2252         else if (tb[IFLA_IFNAME])
2253                 dev = __dev_get_by_name(net, ifname);
2254         else
2255                 goto errout;
2256
2257         if (dev == NULL) {
2258                 err = -ENODEV;
2259                 goto errout;
2260         }
2261
2262         err = validate_linkmsg(dev, tb);
2263         if (err < 0)
2264                 goto errout;
2265
2266         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2267 errout:
2268         return err;
2269 }
2270
2271 static int rtnl_group_dellink(const struct net *net, int group)
2272 {
2273         struct net_device *dev, *aux;
2274         LIST_HEAD(list_kill);
2275         bool found = false;
2276
2277         if (!group)
2278                 return -EPERM;
2279
2280         for_each_netdev(net, dev) {
2281                 if (dev->group == group) {
2282                         const struct rtnl_link_ops *ops;
2283
2284                         found = true;
2285                         ops = dev->rtnl_link_ops;
2286                         if (!ops || !ops->dellink)
2287                                 return -EOPNOTSUPP;
2288                 }
2289         }
2290
2291         if (!found)
2292                 return -ENODEV;
2293
2294         for_each_netdev_safe(net, dev, aux) {
2295                 if (dev->group == group) {
2296                         const struct rtnl_link_ops *ops;
2297
2298                         ops = dev->rtnl_link_ops;
2299                         ops->dellink(dev, &list_kill);
2300                 }
2301         }
2302         unregister_netdevice_many(&list_kill);
2303
2304         return 0;
2305 }
2306
2307 int rtnl_delete_link(struct net_device *dev)
2308 {
2309         const struct rtnl_link_ops *ops;
2310         LIST_HEAD(list_kill);
2311
2312         ops = dev->rtnl_link_ops;
2313         if (!ops || !ops->dellink)
2314                 return -EOPNOTSUPP;
2315
2316         ops->dellink(dev, &list_kill);
2317         unregister_netdevice_many(&list_kill);
2318
2319         return 0;
2320 }
2321 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2322
2323 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2324                         struct netlink_ext_ack *extack)
2325 {
2326         struct net *net = sock_net(skb->sk);
2327         struct net_device *dev;
2328         struct ifinfomsg *ifm;
2329         char ifname[IFNAMSIZ];
2330         struct nlattr *tb[IFLA_MAX+1];
2331         int err;
2332
2333         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2334         if (err < 0)
2335                 return err;
2336
2337         if (tb[IFLA_IFNAME])
2338                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2339
2340         ifm = nlmsg_data(nlh);
2341         if (ifm->ifi_index > 0)
2342                 dev = __dev_get_by_index(net, ifm->ifi_index);
2343         else if (tb[IFLA_IFNAME])
2344                 dev = __dev_get_by_name(net, ifname);
2345         else if (tb[IFLA_GROUP])
2346                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2347         else
2348                 return -EINVAL;
2349
2350         if (!dev)
2351                 return -ENODEV;
2352
2353         return rtnl_delete_link(dev);
2354 }
2355
2356 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2357 {
2358         unsigned int old_flags;
2359         int err;
2360
2361         old_flags = dev->flags;
2362         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2363                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2364                 if (err < 0)
2365                         return err;
2366         }
2367
2368         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2369
2370         __dev_notify_flags(dev, old_flags, ~0U);
2371         return 0;
2372 }
2373 EXPORT_SYMBOL(rtnl_configure_link);
2374
2375 struct net_device *rtnl_create_link(struct net *net,
2376         const char *ifname, unsigned char name_assign_type,
2377         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2378 {
2379         struct net_device *dev;
2380         unsigned int num_tx_queues = 1;
2381         unsigned int num_rx_queues = 1;
2382
2383         if (tb[IFLA_NUM_TX_QUEUES])
2384                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2385         else if (ops->get_num_tx_queues)
2386                 num_tx_queues = ops->get_num_tx_queues();
2387
2388         if (tb[IFLA_NUM_RX_QUEUES])
2389                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2390         else if (ops->get_num_rx_queues)
2391                 num_rx_queues = ops->get_num_rx_queues();
2392
2393         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2394                                ops->setup, num_tx_queues, num_rx_queues);
2395         if (!dev)
2396                 return ERR_PTR(-ENOMEM);
2397
2398         dev_net_set(dev, net);
2399         dev->rtnl_link_ops = ops;
2400         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2401
2402         if (tb[IFLA_MTU])
2403                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2404         if (tb[IFLA_ADDRESS]) {
2405                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2406                                 nla_len(tb[IFLA_ADDRESS]));
2407                 dev->addr_assign_type = NET_ADDR_SET;
2408         }
2409         if (tb[IFLA_BROADCAST])
2410                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2411                                 nla_len(tb[IFLA_BROADCAST]));
2412         if (tb[IFLA_TXQLEN])
2413                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2414         if (tb[IFLA_OPERSTATE])
2415                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2416         if (tb[IFLA_LINKMODE])
2417                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2418         if (tb[IFLA_GROUP])
2419                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2420
2421         return dev;
2422 }
2423 EXPORT_SYMBOL(rtnl_create_link);
2424
2425 static int rtnl_group_changelink(const struct sk_buff *skb,
2426                 struct net *net, int group,
2427                 struct ifinfomsg *ifm,
2428                 struct netlink_ext_ack *extack,
2429                 struct nlattr **tb)
2430 {
2431         struct net_device *dev, *aux;
2432         int err;
2433
2434         for_each_netdev_safe(net, dev, aux) {
2435                 if (dev->group == group) {
2436                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2437                         if (err < 0)
2438                                 return err;
2439                 }
2440         }
2441
2442         return 0;
2443 }
2444
2445 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2446                         struct netlink_ext_ack *extack)
2447 {
2448         struct net *net = sock_net(skb->sk);
2449         const struct rtnl_link_ops *ops;
2450         const struct rtnl_link_ops *m_ops = NULL;
2451         struct net_device *dev;
2452         struct net_device *master_dev = NULL;
2453         struct ifinfomsg *ifm;
2454         char kind[MODULE_NAME_LEN];
2455         char ifname[IFNAMSIZ];
2456         struct nlattr *tb[IFLA_MAX+1];
2457         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2458         unsigned char name_assign_type = NET_NAME_USER;
2459         int err;
2460
2461 #ifdef CONFIG_MODULES
2462 replay:
2463 #endif
2464         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2465         if (err < 0)
2466                 return err;
2467
2468         if (tb[IFLA_IFNAME])
2469                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2470         else
2471                 ifname[0] = '\0';
2472
2473         ifm = nlmsg_data(nlh);
2474         if (ifm->ifi_index > 0)
2475                 dev = __dev_get_by_index(net, ifm->ifi_index);
2476         else {
2477                 if (ifname[0])
2478                         dev = __dev_get_by_name(net, ifname);
2479                 else
2480                         dev = NULL;
2481         }
2482
2483         if (dev) {
2484                 master_dev = netdev_master_upper_dev_get(dev);
2485                 if (master_dev)
2486                         m_ops = master_dev->rtnl_link_ops;
2487         }
2488
2489         err = validate_linkmsg(dev, tb);
2490         if (err < 0)
2491                 return err;
2492
2493         if (tb[IFLA_LINKINFO]) {
2494                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2495                                        tb[IFLA_LINKINFO], ifla_info_policy,
2496                                        NULL);
2497                 if (err < 0)
2498                         return err;
2499         } else
2500                 memset(linkinfo, 0, sizeof(linkinfo));
2501
2502         if (linkinfo[IFLA_INFO_KIND]) {
2503                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2504                 ops = rtnl_link_ops_get(kind);
2505         } else {
2506                 kind[0] = '\0';
2507                 ops = NULL;
2508         }
2509
2510         if (1) {
2511                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2512                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2513                 struct nlattr **data = NULL;
2514                 struct nlattr **slave_data = NULL;
2515                 struct net *dest_net, *link_net = NULL;
2516
2517                 if (ops) {
2518                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2519                                 err = nla_parse_nested(attr, ops->maxtype,
2520                                                        linkinfo[IFLA_INFO_DATA],
2521                                                        ops->policy, NULL);
2522                                 if (err < 0)
2523                                         return err;
2524                                 data = attr;
2525                         }
2526                         if (ops->validate) {
2527                                 err = ops->validate(tb, data);
2528                                 if (err < 0)
2529                                         return err;
2530                         }
2531                 }
2532
2533                 if (m_ops) {
2534                         if (m_ops->slave_maxtype &&
2535                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2536                                 err = nla_parse_nested(slave_attr,
2537                                                        m_ops->slave_maxtype,
2538                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2539                                                        m_ops->slave_policy,
2540                                                        NULL);
2541                                 if (err < 0)
2542                                         return err;
2543                                 slave_data = slave_attr;
2544                         }
2545                         if (m_ops->slave_validate) {
2546                                 err = m_ops->slave_validate(tb, slave_data);
2547                                 if (err < 0)
2548                                         return err;
2549                         }
2550                 }
2551
2552                 if (dev) {
2553                         int status = 0;
2554
2555                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2556                                 return -EEXIST;
2557                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2558                                 return -EOPNOTSUPP;
2559
2560                         if (linkinfo[IFLA_INFO_DATA]) {
2561                                 if (!ops || ops != dev->rtnl_link_ops ||
2562                                     !ops->changelink)
2563                                         return -EOPNOTSUPP;
2564
2565                                 err = ops->changelink(dev, tb, data);
2566                                 if (err < 0)
2567                                         return err;
2568                                 status |= DO_SETLINK_NOTIFY;
2569                         }
2570
2571                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2572                                 if (!m_ops || !m_ops->slave_changelink)
2573                                         return -EOPNOTSUPP;
2574
2575                                 err = m_ops->slave_changelink(master_dev, dev,
2576                                                               tb, slave_data);
2577                                 if (err < 0)
2578                                         return err;
2579                                 status |= DO_SETLINK_NOTIFY;
2580                         }
2581
2582                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2583                                           status);
2584                 }
2585
2586                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2587                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2588                                 return rtnl_group_changelink(skb, net,
2589                                                 nla_get_u32(tb[IFLA_GROUP]),
2590                                                 ifm, extack, tb);
2591                         return -ENODEV;
2592                 }
2593
2594                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2595                         return -EOPNOTSUPP;
2596
2597                 if (!ops) {
2598 #ifdef CONFIG_MODULES
2599                         if (kind[0]) {
2600                                 __rtnl_unlock();
2601                                 request_module("rtnl-link-%s", kind);
2602                                 rtnl_lock();
2603                                 ops = rtnl_link_ops_get(kind);
2604                                 if (ops)
2605                                         goto replay;
2606                         }
2607 #endif
2608                         return -EOPNOTSUPP;
2609                 }
2610
2611                 if (!ops->setup)
2612                         return -EOPNOTSUPP;
2613
2614                 if (!ifname[0]) {
2615                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2616                         name_assign_type = NET_NAME_ENUM;
2617                 }
2618
2619                 dest_net = rtnl_link_get_net(net, tb);
2620                 if (IS_ERR(dest_net))
2621                         return PTR_ERR(dest_net);
2622
2623                 err = -EPERM;
2624                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2625                         goto out;
2626
2627                 if (tb[IFLA_LINK_NETNSID]) {
2628                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2629
2630                         link_net = get_net_ns_by_id(dest_net, id);
2631                         if (!link_net) {
2632                                 err =  -EINVAL;
2633                                 goto out;
2634                         }
2635                         err = -EPERM;
2636                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2637                                 goto out;
2638                 }
2639
2640                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2641                                        name_assign_type, ops, tb);
2642                 if (IS_ERR(dev)) {
2643                         err = PTR_ERR(dev);
2644                         goto out;
2645                 }
2646
2647                 dev->ifindex = ifm->ifi_index;
2648
2649                 if (ops->newlink) {
2650                         err = ops->newlink(link_net ? : net, dev, tb, data);
2651                         /* Drivers should call free_netdev() in ->destructor
2652                          * and unregister it on failure after registration
2653                          * so that device could be finally freed in rtnl_unlock.
2654                          */
2655                         if (err < 0) {
2656                                 /* If device is not registered at all, free it now */
2657                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2658                                         free_netdev(dev);
2659                                 goto out;
2660                         }
2661                 } else {
2662                         err = register_netdevice(dev);
2663                         if (err < 0) {
2664                                 free_netdev(dev);
2665                                 goto out;
2666                         }
2667                 }
2668                 err = rtnl_configure_link(dev, ifm);
2669                 if (err < 0)
2670                         goto out_unregister;
2671                 if (link_net) {
2672                         err = dev_change_net_namespace(dev, dest_net, ifname);
2673                         if (err < 0)
2674                                 goto out_unregister;
2675                 }
2676                 if (tb[IFLA_MASTER]) {
2677                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2678                         if (err)
2679                                 goto out_unregister;
2680                 }
2681 out:
2682                 if (link_net)
2683                         put_net(link_net);
2684                 put_net(dest_net);
2685                 return err;
2686 out_unregister:
2687                 if (ops->newlink) {
2688                         LIST_HEAD(list_kill);
2689
2690                         ops->dellink(dev, &list_kill);
2691                         unregister_netdevice_many(&list_kill);
2692                 } else {
2693                         unregister_netdevice(dev);
2694                 }
2695                 goto out;
2696         }
2697 }
2698
2699 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2700                         struct netlink_ext_ack *extack)
2701 {
2702         struct net *net = sock_net(skb->sk);
2703         struct ifinfomsg *ifm;
2704         char ifname[IFNAMSIZ];
2705         struct nlattr *tb[IFLA_MAX+1];
2706         struct net_device *dev = NULL;
2707         struct sk_buff *nskb;
2708         int err;
2709         u32 ext_filter_mask = 0;
2710
2711         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2712         if (err < 0)
2713                 return err;
2714
2715         if (tb[IFLA_IFNAME])
2716                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2717
2718         if (tb[IFLA_EXT_MASK])
2719                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2720
2721         ifm = nlmsg_data(nlh);
2722         if (ifm->ifi_index > 0)
2723                 dev = __dev_get_by_index(net, ifm->ifi_index);
2724         else if (tb[IFLA_IFNAME])
2725                 dev = __dev_get_by_name(net, ifname);
2726         else
2727                 return -EINVAL;
2728
2729         if (dev == NULL)
2730                 return -ENODEV;
2731
2732         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2733         if (nskb == NULL)
2734                 return -ENOBUFS;
2735
2736         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2737                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2738         if (err < 0) {
2739                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2740                 WARN_ON(err == -EMSGSIZE);
2741                 kfree_skb(nskb);
2742         } else
2743                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2744
2745         return err;
2746 }
2747
2748 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2749 {
2750         struct net *net = sock_net(skb->sk);
2751         struct net_device *dev;
2752         struct nlattr *tb[IFLA_MAX+1];
2753         u32 ext_filter_mask = 0;
2754         u16 min_ifinfo_dump_size = 0;
2755         int hdrlen;
2756
2757         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2758         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2759                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2760
2761         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2762                 if (tb[IFLA_EXT_MASK])
2763                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2764         }
2765
2766         if (!ext_filter_mask)
2767                 return NLMSG_GOODSIZE;
2768         /*
2769          * traverse the list of net devices and compute the minimum
2770          * buffer size based upon the filter mask.
2771          */
2772         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2773                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2774                                              if_nlmsg_size(dev,
2775                                                            ext_filter_mask));
2776         }
2777
2778         return nlmsg_total_size(min_ifinfo_dump_size);
2779 }
2780
2781 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2782 {
2783         int idx;
2784         int s_idx = cb->family;
2785
2786         if (s_idx == 0)
2787                 s_idx = 1;
2788         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2789                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2790                 if (idx < s_idx || idx == PF_PACKET)
2791                         continue;
2792                 if (rtnl_msg_handlers[idx] == NULL ||
2793                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2794                         continue;
2795                 if (idx > s_idx) {
2796                         memset(&cb->args[0], 0, sizeof(cb->args));
2797                         cb->prev_seq = 0;
2798                         cb->seq = 0;
2799                 }
2800                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2801                         break;
2802         }
2803         cb->family = idx;
2804
2805         return skb->len;
2806 }
2807
2808 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2809                                        unsigned int change, gfp_t flags)
2810 {
2811         struct net *net = dev_net(dev);
2812         struct sk_buff *skb;
2813         int err = -ENOBUFS;
2814         size_t if_info_size;
2815
2816         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2817         if (skb == NULL)
2818                 goto errout;
2819
2820         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2821         if (err < 0) {
2822                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2823                 WARN_ON(err == -EMSGSIZE);
2824                 kfree_skb(skb);
2825                 goto errout;
2826         }
2827         return skb;
2828 errout:
2829         if (err < 0)
2830                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2831         return NULL;
2832 }
2833
2834 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2835 {
2836         struct net *net = dev_net(dev);
2837
2838         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2839 }
2840
2841 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2842                   gfp_t flags)
2843 {
2844         struct sk_buff *skb;
2845
2846         if (dev->reg_state != NETREG_REGISTERED)
2847                 return;
2848
2849         skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2850         if (skb)
2851                 rtmsg_ifinfo_send(skb, dev, flags);
2852 }
2853 EXPORT_SYMBOL(rtmsg_ifinfo);
2854
2855 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2856                                    struct net_device *dev,
2857                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2858                                    int type, unsigned int flags,
2859                                    int nlflags, u16 ndm_state)
2860 {
2861         struct nlmsghdr *nlh;
2862         struct ndmsg *ndm;
2863
2864         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2865         if (!nlh)
2866                 return -EMSGSIZE;
2867
2868         ndm = nlmsg_data(nlh);
2869         ndm->ndm_family  = AF_BRIDGE;
2870         ndm->ndm_pad1    = 0;
2871         ndm->ndm_pad2    = 0;
2872         ndm->ndm_flags   = flags;
2873         ndm->ndm_type    = 0;
2874         ndm->ndm_ifindex = dev->ifindex;
2875         ndm->ndm_state   = ndm_state;
2876
2877         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2878                 goto nla_put_failure;
2879         if (vid)
2880                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2881                         goto nla_put_failure;
2882
2883         nlmsg_end(skb, nlh);
2884         return 0;
2885
2886 nla_put_failure:
2887         nlmsg_cancel(skb, nlh);
2888         return -EMSGSIZE;
2889 }
2890
2891 static inline size_t rtnl_fdb_nlmsg_size(void)
2892 {
2893         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2894                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2895                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2896                0;
2897 }
2898
2899 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2900                             u16 ndm_state)
2901 {
2902         struct net *net = dev_net(dev);
2903         struct sk_buff *skb;
2904         int err = -ENOBUFS;
2905
2906         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2907         if (!skb)
2908                 goto errout;
2909
2910         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2911                                       0, 0, type, NTF_SELF, 0, ndm_state);
2912         if (err < 0) {
2913                 kfree_skb(skb);
2914                 goto errout;
2915         }
2916
2917         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2918         return;
2919 errout:
2920         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2921 }
2922
2923 /**
2924  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2925  */
2926 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2927                      struct nlattr *tb[],
2928                      struct net_device *dev,
2929                      const unsigned char *addr, u16 vid,
2930                      u16 flags)
2931 {
2932         int err = -EINVAL;
2933
2934         /* If aging addresses are supported device will need to
2935          * implement its own handler for this.
2936          */
2937         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2938                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2939                 return err;
2940         }
2941
2942         if (vid) {
2943                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2944                 return err;
2945         }
2946
2947         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2948                 err = dev_uc_add_excl(dev, addr);
2949         else if (is_multicast_ether_addr(addr))
2950                 err = dev_mc_add_excl(dev, addr);
2951
2952         /* Only return duplicate errors if NLM_F_EXCL is set */
2953         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2954                 err = 0;
2955
2956         return err;
2957 }
2958 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2959
2960 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2961 {
2962         u16 vid = 0;
2963
2964         if (vlan_attr) {
2965                 if (nla_len(vlan_attr) != sizeof(u16)) {
2966                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2967                         return -EINVAL;
2968                 }
2969
2970                 vid = nla_get_u16(vlan_attr);
2971
2972                 if (!vid || vid >= VLAN_VID_MASK) {
2973                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2974                                 vid);
2975                         return -EINVAL;
2976                 }
2977         }
2978         *p_vid = vid;
2979         return 0;
2980 }
2981
2982 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
2983                         struct netlink_ext_ack *extack)
2984 {
2985         struct net *net = sock_net(skb->sk);
2986         struct ndmsg *ndm;
2987         struct nlattr *tb[NDA_MAX+1];
2988         struct net_device *dev;
2989         u8 *addr;
2990         u16 vid;
2991         int err;
2992
2993         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
2994         if (err < 0)
2995                 return err;
2996
2997         ndm = nlmsg_data(nlh);
2998         if (ndm->ndm_ifindex == 0) {
2999                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3000                 return -EINVAL;
3001         }
3002
3003         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3004         if (dev == NULL) {
3005                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3006                 return -ENODEV;
3007         }
3008
3009         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3010                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3011                 return -EINVAL;
3012         }
3013
3014         addr = nla_data(tb[NDA_LLADDR]);
3015
3016         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3017         if (err)
3018                 return err;
3019
3020         err = -EOPNOTSUPP;
3021
3022         /* Support fdb on master device the net/bridge default case */
3023         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3024             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3025                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3026                 const struct net_device_ops *ops = br_dev->netdev_ops;
3027
3028                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3029                                        nlh->nlmsg_flags);
3030                 if (err)
3031                         goto out;
3032                 else
3033                         ndm->ndm_flags &= ~NTF_MASTER;
3034         }
3035
3036         /* Embedded bridge, macvlan, and any other device support */
3037         if ((ndm->ndm_flags & NTF_SELF)) {
3038                 if (dev->netdev_ops->ndo_fdb_add)
3039                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3040                                                            vid,
3041                                                            nlh->nlmsg_flags);
3042                 else
3043                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3044                                                nlh->nlmsg_flags);
3045
3046                 if (!err) {
3047                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3048                                         ndm->ndm_state);
3049                         ndm->ndm_flags &= ~NTF_SELF;
3050                 }
3051         }
3052 out:
3053         return err;
3054 }
3055
3056 /**
3057  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3058  */
3059 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3060                      struct nlattr *tb[],
3061                      struct net_device *dev,
3062                      const unsigned char *addr, u16 vid)
3063 {
3064         int err = -EINVAL;
3065
3066         /* If aging addresses are supported device will need to
3067          * implement its own handler for this.
3068          */
3069         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3070                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3071                 return err;
3072         }
3073
3074         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3075                 err = dev_uc_del(dev, addr);
3076         else if (is_multicast_ether_addr(addr))
3077                 err = dev_mc_del(dev, addr);
3078
3079         return err;
3080 }
3081 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3082
3083 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3084                         struct netlink_ext_ack *extack)
3085 {
3086         struct net *net = sock_net(skb->sk);
3087         struct ndmsg *ndm;
3088         struct nlattr *tb[NDA_MAX+1];
3089         struct net_device *dev;
3090         int err = -EINVAL;
3091         __u8 *addr;
3092         u16 vid;
3093
3094         if (!netlink_capable(skb, CAP_NET_ADMIN))
3095                 return -EPERM;
3096
3097         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3098         if (err < 0)
3099                 return err;
3100
3101         ndm = nlmsg_data(nlh);
3102         if (ndm->ndm_ifindex == 0) {
3103                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3104                 return -EINVAL;
3105         }
3106
3107         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3108         if (dev == NULL) {
3109                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3110                 return -ENODEV;
3111         }
3112
3113         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3114                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3115                 return -EINVAL;
3116         }
3117
3118         addr = nla_data(tb[NDA_LLADDR]);
3119
3120         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3121         if (err)
3122                 return err;
3123
3124         err = -EOPNOTSUPP;
3125
3126         /* Support fdb on master device the net/bridge default case */
3127         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3128             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3129                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3130                 const struct net_device_ops *ops = br_dev->netdev_ops;
3131
3132                 if (ops->ndo_fdb_del)
3133                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3134
3135                 if (err)
3136                         goto out;
3137                 else
3138                         ndm->ndm_flags &= ~NTF_MASTER;
3139         }
3140
3141         /* Embedded bridge, macvlan, and any other device support */
3142         if (ndm->ndm_flags & NTF_SELF) {
3143                 if (dev->netdev_ops->ndo_fdb_del)
3144                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3145                                                            vid);
3146                 else
3147                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3148
3149                 if (!err) {
3150                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3151                                         ndm->ndm_state);
3152                         ndm->ndm_flags &= ~NTF_SELF;
3153                 }
3154         }
3155 out:
3156         return err;
3157 }
3158
3159 static int nlmsg_populate_fdb(struct sk_buff *skb,
3160                               struct netlink_callback *cb,
3161                               struct net_device *dev,
3162                               int *idx,
3163                               struct netdev_hw_addr_list *list)
3164 {
3165         struct netdev_hw_addr *ha;
3166         int err;
3167         u32 portid, seq;
3168
3169         portid = NETLINK_CB(cb->skb).portid;
3170         seq = cb->nlh->nlmsg_seq;
3171
3172         list_for_each_entry(ha, &list->list, list) {
3173                 if (*idx < cb->args[2])
3174                         goto skip;
3175
3176                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3177                                               portid, seq,
3178                                               RTM_NEWNEIGH, NTF_SELF,
3179                                               NLM_F_MULTI, NUD_PERMANENT);
3180                 if (err < 0)
3181                         return err;
3182 skip:
3183                 *idx += 1;
3184         }
3185         return 0;
3186 }
3187
3188 /**
3189  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3190  * @nlh: netlink message header
3191  * @dev: netdevice
3192  *
3193  * Default netdevice operation to dump the existing unicast address list.
3194  * Returns number of addresses from list put in skb.
3195  */
3196 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3197                       struct netlink_callback *cb,
3198                       struct net_device *dev,
3199                       struct net_device *filter_dev,
3200                       int *idx)
3201 {
3202         int err;
3203
3204         netif_addr_lock_bh(dev);
3205         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3206         if (err)
3207                 goto out;
3208         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3209 out:
3210         netif_addr_unlock_bh(dev);
3211         return err;
3212 }
3213 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3214
3215 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3216 {
3217         struct net_device *dev;
3218         struct nlattr *tb[IFLA_MAX+1];
3219         struct net_device *br_dev = NULL;
3220         const struct net_device_ops *ops = NULL;
3221         const struct net_device_ops *cops = NULL;
3222         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3223         struct net *net = sock_net(skb->sk);
3224         struct hlist_head *head;
3225         int brport_idx = 0;
3226         int br_idx = 0;
3227         int h, s_h;
3228         int idx = 0, s_idx;
3229         int err = 0;
3230         int fidx = 0;
3231
3232         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3233                         IFLA_MAX, ifla_policy, NULL) == 0) {
3234                 if (tb[IFLA_MASTER])
3235                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3236         }
3237
3238         brport_idx = ifm->ifi_index;
3239
3240         if (br_idx) {
3241                 br_dev = __dev_get_by_index(net, br_idx);
3242                 if (!br_dev)
3243                         return -ENODEV;
3244
3245                 ops = br_dev->netdev_ops;
3246         }
3247
3248         s_h = cb->args[0];
3249         s_idx = cb->args[1];
3250
3251         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3252                 idx = 0;
3253                 head = &net->dev_index_head[h];
3254                 hlist_for_each_entry(dev, head, index_hlist) {
3255
3256                         if (brport_idx && (dev->ifindex != brport_idx))
3257                                 continue;
3258
3259                         if (!br_idx) { /* user did not specify a specific bridge */
3260                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3261                                         br_dev = netdev_master_upper_dev_get(dev);
3262                                         cops = br_dev->netdev_ops;
3263                                 }
3264                         } else {
3265                                 if (dev != br_dev &&
3266                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3267                                         continue;
3268
3269                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3270                                     !(dev->priv_flags & IFF_EBRIDGE))
3271                                         continue;
3272                                 cops = ops;
3273                         }
3274
3275                         if (idx < s_idx)
3276                                 goto cont;
3277
3278                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3279                                 if (cops && cops->ndo_fdb_dump) {
3280                                         err = cops->ndo_fdb_dump(skb, cb,
3281                                                                 br_dev, dev,
3282                                                                 &fidx);
3283                                         if (err == -EMSGSIZE)
3284                                                 goto out;
3285                                 }
3286                         }
3287
3288                         if (dev->netdev_ops->ndo_fdb_dump)
3289                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3290                                                                     dev, NULL,
3291                                                                     &fidx);
3292                         else
3293                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3294                                                         &fidx);
3295                         if (err == -EMSGSIZE)
3296                                 goto out;
3297
3298                         cops = NULL;
3299
3300                         /* reset fdb offset to 0 for rest of the interfaces */
3301                         cb->args[2] = 0;
3302                         fidx = 0;
3303 cont:
3304                         idx++;
3305                 }
3306         }
3307
3308 out:
3309         cb->args[0] = h;
3310         cb->args[1] = idx;
3311         cb->args[2] = fidx;
3312
3313         return skb->len;
3314 }
3315
3316 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3317                                unsigned int attrnum, unsigned int flag)
3318 {
3319         if (mask & flag)
3320                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3321         return 0;
3322 }
3323
3324 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3325                             struct net_device *dev, u16 mode,
3326                             u32 flags, u32 mask, int nlflags,
3327                             u32 filter_mask,
3328                             int (*vlan_fill)(struct sk_buff *skb,
3329                                              struct net_device *dev,
3330                                              u32 filter_mask))
3331 {
3332         struct nlmsghdr *nlh;
3333         struct ifinfomsg *ifm;
3334         struct nlattr *br_afspec;
3335         struct nlattr *protinfo;
3336         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3337         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3338         int err = 0;
3339
3340         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3341         if (nlh == NULL)
3342                 return -EMSGSIZE;
3343
3344         ifm = nlmsg_data(nlh);
3345         ifm->ifi_family = AF_BRIDGE;
3346         ifm->__ifi_pad = 0;
3347         ifm->ifi_type = dev->type;
3348         ifm->ifi_index = dev->ifindex;
3349         ifm->ifi_flags = dev_get_flags(dev);
3350         ifm->ifi_change = 0;
3351
3352
3353         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3354             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3355             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3356             (br_dev &&
3357              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3358             (dev->addr_len &&
3359              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3360             (dev->ifindex != dev_get_iflink(dev) &&
3361              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3362                 goto nla_put_failure;
3363
3364         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3365         if (!br_afspec)
3366                 goto nla_put_failure;
3367
3368         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3369                 nla_nest_cancel(skb, br_afspec);
3370                 goto nla_put_failure;
3371         }
3372
3373         if (mode != BRIDGE_MODE_UNDEF) {
3374                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3375                         nla_nest_cancel(skb, br_afspec);
3376                         goto nla_put_failure;
3377                 }
3378         }
3379         if (vlan_fill) {
3380                 err = vlan_fill(skb, dev, filter_mask);
3381                 if (err) {
3382                         nla_nest_cancel(skb, br_afspec);
3383                         goto nla_put_failure;
3384                 }
3385         }
3386         nla_nest_end(skb, br_afspec);
3387
3388         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3389         if (!protinfo)
3390                 goto nla_put_failure;
3391
3392         if (brport_nla_put_flag(skb, flags, mask,
3393                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3394             brport_nla_put_flag(skb, flags, mask,
3395                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3396             brport_nla_put_flag(skb, flags, mask,
3397                                 IFLA_BRPORT_FAST_LEAVE,
3398                                 BR_MULTICAST_FAST_LEAVE) ||
3399             brport_nla_put_flag(skb, flags, mask,
3400                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3401             brport_nla_put_flag(skb, flags, mask,
3402                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3403             brport_nla_put_flag(skb, flags, mask,
3404                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3405             brport_nla_put_flag(skb, flags, mask,
3406                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3407             brport_nla_put_flag(skb, flags, mask,
3408                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3409                 nla_nest_cancel(skb, protinfo);
3410                 goto nla_put_failure;
3411         }
3412
3413         nla_nest_end(skb, protinfo);
3414
3415         nlmsg_end(skb, nlh);
3416         return 0;
3417 nla_put_failure:
3418         nlmsg_cancel(skb, nlh);
3419         return err ? err : -EMSGSIZE;
3420 }
3421 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3422
3423 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3424 {
3425         struct net *net = sock_net(skb->sk);
3426         struct net_device *dev;
3427         int idx = 0;
3428         u32 portid = NETLINK_CB(cb->skb).portid;
3429         u32 seq = cb->nlh->nlmsg_seq;
3430         u32 filter_mask = 0;
3431         int err;
3432
3433         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3434                 struct nlattr *extfilt;
3435
3436                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3437                                           IFLA_EXT_MASK);
3438                 if (extfilt) {
3439                         if (nla_len(extfilt) < sizeof(filter_mask))
3440                                 return -EINVAL;
3441
3442                         filter_mask = nla_get_u32(extfilt);
3443                 }
3444         }
3445
3446         rcu_read_lock();
3447         for_each_netdev_rcu(net, dev) {
3448                 const struct net_device_ops *ops = dev->netdev_ops;
3449                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3450
3451                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3452                         if (idx >= cb->args[0]) {
3453                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3454                                                 skb, portid, seq, dev,
3455                                                 filter_mask, NLM_F_MULTI);
3456                                 if (err < 0 && err != -EOPNOTSUPP)
3457                                         break;
3458                         }
3459                         idx++;
3460                 }
3461
3462                 if (ops->ndo_bridge_getlink) {
3463                         if (idx >= cb->args[0]) {
3464                                 err = ops->ndo_bridge_getlink(skb, portid,
3465                                                               seq, dev,
3466                                                               filter_mask,
3467                                                               NLM_F_MULTI);
3468                                 if (err < 0 && err != -EOPNOTSUPP)
3469                                         break;
3470                         }
3471                         idx++;
3472                 }
3473         }
3474         rcu_read_unlock();
3475         cb->args[0] = idx;
3476
3477         return skb->len;
3478 }
3479
3480 static inline size_t bridge_nlmsg_size(void)
3481 {
3482         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3483                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3484                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3485                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3486                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3487                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3488                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3489                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3490                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3491                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3492                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3493 }
3494
3495 static int rtnl_bridge_notify(struct net_device *dev)
3496 {
3497         struct net *net = dev_net(dev);
3498         struct sk_buff *skb;
3499         int err = -EOPNOTSUPP;
3500
3501         if (!dev->netdev_ops->ndo_bridge_getlink)
3502                 return 0;
3503
3504         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3505         if (!skb) {
3506                 err = -ENOMEM;
3507                 goto errout;
3508         }
3509
3510         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3511         if (err < 0)
3512                 goto errout;
3513
3514         if (!skb->len)
3515                 goto errout;
3516
3517         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3518         return 0;
3519 errout:
3520         WARN_ON(err == -EMSGSIZE);
3521         kfree_skb(skb);
3522         if (err)
3523                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3524         return err;
3525 }
3526
3527 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3528                                struct netlink_ext_ack *extack)
3529 {
3530         struct net *net = sock_net(skb->sk);
3531         struct ifinfomsg *ifm;
3532         struct net_device *dev;
3533         struct nlattr *br_spec, *attr = NULL;
3534         int rem, err = -EOPNOTSUPP;
3535         u16 flags = 0;
3536         bool have_flags = false;
3537
3538         if (nlmsg_len(nlh) < sizeof(*ifm))
3539                 return -EINVAL;
3540
3541         ifm = nlmsg_data(nlh);
3542         if (ifm->ifi_family != AF_BRIDGE)
3543                 return -EPFNOSUPPORT;
3544
3545         dev = __dev_get_by_index(net, ifm->ifi_index);
3546         if (!dev) {
3547                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3548                 return -ENODEV;
3549         }
3550
3551         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3552         if (br_spec) {
3553                 nla_for_each_nested(attr, br_spec, rem) {
3554                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3555                                 if (nla_len(attr) < sizeof(flags))
3556                                         return -EINVAL;
3557
3558                                 have_flags = true;
3559                                 flags = nla_get_u16(attr);
3560                                 break;
3561                         }
3562                 }
3563         }
3564
3565         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3566                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3567
3568                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3569                         err = -EOPNOTSUPP;
3570                         goto out;
3571                 }
3572
3573                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3574                 if (err)
3575                         goto out;
3576
3577                 flags &= ~BRIDGE_FLAGS_MASTER;
3578         }
3579
3580         if ((flags & BRIDGE_FLAGS_SELF)) {
3581                 if (!dev->netdev_ops->ndo_bridge_setlink)
3582                         err = -EOPNOTSUPP;
3583                 else
3584                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3585                                                                   flags);
3586                 if (!err) {
3587                         flags &= ~BRIDGE_FLAGS_SELF;
3588
3589                         /* Generate event to notify upper layer of bridge
3590                          * change
3591                          */
3592                         err = rtnl_bridge_notify(dev);
3593                 }
3594         }
3595
3596         if (have_flags)
3597                 memcpy(nla_data(attr), &flags, sizeof(flags));
3598 out:
3599         return err;
3600 }
3601
3602 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3603                                struct netlink_ext_ack *extack)
3604 {
3605         struct net *net = sock_net(skb->sk);
3606         struct ifinfomsg *ifm;
3607         struct net_device *dev;
3608         struct nlattr *br_spec, *attr = NULL;
3609         int rem, err = -EOPNOTSUPP;
3610         u16 flags = 0;
3611         bool have_flags = false;
3612
3613         if (nlmsg_len(nlh) < sizeof(*ifm))
3614                 return -EINVAL;
3615
3616         ifm = nlmsg_data(nlh);
3617         if (ifm->ifi_family != AF_BRIDGE)
3618                 return -EPFNOSUPPORT;
3619
3620         dev = __dev_get_by_index(net, ifm->ifi_index);
3621         if (!dev) {
3622                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3623                 return -ENODEV;
3624         }
3625
3626         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3627         if (br_spec) {
3628                 nla_for_each_nested(attr, br_spec, rem) {
3629                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3630                                 if (nla_len(attr) < sizeof(flags))
3631                                         return -EINVAL;
3632
3633                                 have_flags = true;
3634                                 flags = nla_get_u16(attr);
3635                                 break;
3636                         }
3637                 }
3638         }
3639
3640         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3641                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3642
3643                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3644                         err = -EOPNOTSUPP;
3645                         goto out;
3646                 }
3647
3648                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3649                 if (err)
3650                         goto out;
3651
3652                 flags &= ~BRIDGE_FLAGS_MASTER;
3653         }
3654
3655         if ((flags & BRIDGE_FLAGS_SELF)) {
3656                 if (!dev->netdev_ops->ndo_bridge_dellink)
3657                         err = -EOPNOTSUPP;
3658                 else
3659                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3660                                                                   flags);
3661
3662                 if (!err) {
3663                         flags &= ~BRIDGE_FLAGS_SELF;
3664
3665                         /* Generate event to notify upper layer of bridge
3666                          * change
3667                          */
3668                         err = rtnl_bridge_notify(dev);
3669                 }
3670         }
3671
3672         if (have_flags)
3673                 memcpy(nla_data(attr), &flags, sizeof(flags));
3674 out:
3675         return err;
3676 }
3677
3678 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3679 {
3680         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3681                (!idxattr || idxattr == attrid);
3682 }
3683
3684 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3685 static int rtnl_get_offload_stats_attr_size(int attr_id)
3686 {
3687         switch (attr_id) {
3688         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3689                 return sizeof(struct rtnl_link_stats64);
3690         }
3691
3692         return 0;
3693 }
3694
3695 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3696                                   int *prividx)
3697 {
3698         struct nlattr *attr = NULL;
3699         int attr_id, size;
3700         void *attr_data;
3701         int err;
3702
3703         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3704               dev->netdev_ops->ndo_get_offload_stats))
3705                 return -ENODATA;
3706
3707         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3708              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3709                 if (attr_id < *prividx)
3710                         continue;
3711
3712                 size = rtnl_get_offload_stats_attr_size(attr_id);
3713                 if (!size)
3714                         continue;
3715
3716                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3717                         continue;
3718
3719                 attr = nla_reserve_64bit(skb, attr_id, size,
3720                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3721                 if (!attr)
3722                         goto nla_put_failure;
3723
3724                 attr_data = nla_data(attr);
3725                 memset(attr_data, 0, size);
3726                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3727                                                              attr_data);
3728                 if (err)
3729                         goto get_offload_stats_failure;
3730         }
3731
3732         if (!attr)
3733                 return -ENODATA;
3734
3735         *prividx = 0;
3736         return 0;
3737
3738 nla_put_failure:
3739         err = -EMSGSIZE;
3740 get_offload_stats_failure:
3741         *prividx = attr_id;
3742         return err;
3743 }
3744
3745 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3746 {
3747         int nla_size = 0;
3748         int attr_id;
3749         int size;
3750
3751         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3752               dev->netdev_ops->ndo_get_offload_stats))
3753                 return 0;
3754
3755         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3756              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3757                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3758                         continue;
3759                 size = rtnl_get_offload_stats_attr_size(attr_id);
3760                 nla_size += nla_total_size_64bit(size);
3761         }
3762
3763         if (nla_size != 0)
3764                 nla_size += nla_total_size(0);
3765
3766         return nla_size;
3767 }
3768
3769 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3770                                int type, u32 pid, u32 seq, u32 change,
3771                                unsigned int flags, unsigned int filter_mask,
3772                                int *idxattr, int *prividx)
3773 {
3774         struct if_stats_msg *ifsm;
3775         struct nlmsghdr *nlh;
3776         struct nlattr *attr;
3777         int s_prividx = *prividx;
3778         int err;
3779
3780         ASSERT_RTNL();
3781
3782         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3783         if (!nlh)
3784                 return -EMSGSIZE;
3785
3786         ifsm = nlmsg_data(nlh);
3787         ifsm->ifindex = dev->ifindex;
3788         ifsm->filter_mask = filter_mask;
3789
3790         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3791                 struct rtnl_link_stats64 *sp;
3792
3793                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3794                                          sizeof(struct rtnl_link_stats64),
3795                                          IFLA_STATS_UNSPEC);
3796                 if (!attr)
3797                         goto nla_put_failure;
3798
3799                 sp = nla_data(attr);
3800                 dev_get_stats(dev, sp);
3801         }
3802
3803         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3804                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3805
3806                 if (ops && ops->fill_linkxstats) {
3807                         *idxattr = IFLA_STATS_LINK_XSTATS;
3808                         attr = nla_nest_start(skb,
3809                                               IFLA_STATS_LINK_XSTATS);
3810                         if (!attr)
3811                                 goto nla_put_failure;
3812
3813                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3814                         nla_nest_end(skb, attr);
3815                         if (err)
3816                                 goto nla_put_failure;
3817                         *idxattr = 0;
3818                 }
3819         }
3820
3821         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3822                              *idxattr)) {
3823                 const struct rtnl_link_ops *ops = NULL;
3824                 const struct net_device *master;
3825
3826                 master = netdev_master_upper_dev_get(dev);
3827                 if (master)
3828                         ops = master->rtnl_link_ops;
3829                 if (ops && ops->fill_linkxstats) {
3830                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3831                         attr = nla_nest_start(skb,
3832                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3833                         if (!attr)
3834                                 goto nla_put_failure;
3835
3836                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3837                         nla_nest_end(skb, attr);
3838                         if (err)
3839                                 goto nla_put_failure;
3840                         *idxattr = 0;
3841                 }
3842         }
3843
3844         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3845                              *idxattr)) {
3846                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3847                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3848                 if (!attr)
3849                         goto nla_put_failure;
3850
3851                 err = rtnl_get_offload_stats(skb, dev, prividx);
3852                 if (err == -ENODATA)
3853                         nla_nest_cancel(skb, attr);
3854                 else
3855                         nla_nest_end(skb, attr);
3856
3857                 if (err && err != -ENODATA)
3858                         goto nla_put_failure;
3859                 *idxattr = 0;
3860         }
3861
3862         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3863                 struct rtnl_af_ops *af_ops;
3864
3865                 *idxattr = IFLA_STATS_AF_SPEC;
3866                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3867                 if (!attr)
3868                         goto nla_put_failure;
3869
3870                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3871                         if (af_ops->fill_stats_af) {
3872                                 struct nlattr *af;
3873                                 int err;
3874
3875                                 af = nla_nest_start(skb, af_ops->family);
3876                                 if (!af)
3877                                         goto nla_put_failure;
3878
3879                                 err = af_ops->fill_stats_af(skb, dev);
3880
3881                                 if (err == -ENODATA)
3882                                         nla_nest_cancel(skb, af);
3883                                 else if (err < 0)
3884                                         goto nla_put_failure;
3885
3886                                 nla_nest_end(skb, af);
3887                         }
3888                 }
3889
3890                 nla_nest_end(skb, attr);
3891
3892                 *idxattr = 0;
3893         }
3894
3895         nlmsg_end(skb, nlh);
3896
3897         return 0;
3898
3899 nla_put_failure:
3900         /* not a multi message or no progress mean a real error */
3901         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3902                 nlmsg_cancel(skb, nlh);
3903         else
3904                 nlmsg_end(skb, nlh);
3905
3906         return -EMSGSIZE;
3907 }
3908
3909 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3910                                   u32 filter_mask)
3911 {
3912         size_t size = 0;
3913
3914         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3915                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3916
3917         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3918                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3919                 int attr = IFLA_STATS_LINK_XSTATS;
3920
3921                 if (ops && ops->get_linkxstats_size) {
3922                         size += nla_total_size(ops->get_linkxstats_size(dev,
3923                                                                         attr));
3924                         /* for IFLA_STATS_LINK_XSTATS */
3925                         size += nla_total_size(0);
3926                 }
3927         }
3928
3929         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3930                 struct net_device *_dev = (struct net_device *)dev;
3931                 const struct rtnl_link_ops *ops = NULL;
3932                 const struct net_device *master;
3933
3934                 /* netdev_master_upper_dev_get can't take const */
3935                 master = netdev_master_upper_dev_get(_dev);
3936                 if (master)
3937                         ops = master->rtnl_link_ops;
3938                 if (ops && ops->get_linkxstats_size) {
3939                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3940
3941                         size += nla_total_size(ops->get_linkxstats_size(dev,
3942                                                                         attr));
3943                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3944                         size += nla_total_size(0);
3945                 }
3946         }
3947
3948         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3949                 size += rtnl_get_offload_stats_size(dev);
3950
3951         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3952                 struct rtnl_af_ops *af_ops;
3953
3954                 /* for IFLA_STATS_AF_SPEC */
3955                 size += nla_total_size(0);
3956
3957                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3958                         if (af_ops->get_stats_af_size) {
3959                                 size += nla_total_size(
3960                                         af_ops->get_stats_af_size(dev));
3961
3962                                 /* for AF_* */
3963                                 size += nla_total_size(0);
3964                         }
3965                 }
3966         }
3967
3968         return size;
3969 }
3970
3971 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
3972                           struct netlink_ext_ack *extack)
3973 {
3974         struct net *net = sock_net(skb->sk);
3975         struct net_device *dev = NULL;
3976         int idxattr = 0, prividx = 0;
3977         struct if_stats_msg *ifsm;
3978         struct sk_buff *nskb;
3979         u32 filter_mask;
3980         int err;
3981
3982         if (nlmsg_len(nlh) < sizeof(*ifsm))
3983                 return -EINVAL;
3984
3985         ifsm = nlmsg_data(nlh);
3986         if (ifsm->ifindex > 0)
3987                 dev = __dev_get_by_index(net, ifsm->ifindex);
3988         else
3989                 return -EINVAL;
3990
3991         if (!dev)
3992                 return -ENODEV;
3993
3994         filter_mask = ifsm->filter_mask;
3995         if (!filter_mask)
3996                 return -EINVAL;
3997
3998         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3999         if (!nskb)
4000                 return -ENOBUFS;
4001
4002         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4003                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4004                                   0, filter_mask, &idxattr, &prividx);
4005         if (err < 0) {
4006                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4007                 WARN_ON(err == -EMSGSIZE);
4008                 kfree_skb(nskb);
4009         } else {
4010                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4011         }
4012
4013         return err;
4014 }
4015
4016 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4017 {
4018         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4019         struct net *net = sock_net(skb->sk);
4020         unsigned int flags = NLM_F_MULTI;
4021         struct if_stats_msg *ifsm;
4022         struct hlist_head *head;
4023         struct net_device *dev;
4024         u32 filter_mask = 0;
4025         int idx = 0;
4026
4027         s_h = cb->args[0];
4028         s_idx = cb->args[1];
4029         s_idxattr = cb->args[2];
4030         s_prividx = cb->args[3];
4031
4032         cb->seq = net->dev_base_seq;
4033
4034         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4035                 return -EINVAL;
4036
4037         ifsm = nlmsg_data(cb->nlh);
4038         filter_mask = ifsm->filter_mask;
4039         if (!filter_mask)
4040                 return -EINVAL;
4041
4042         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4043                 idx = 0;
4044                 head = &net->dev_index_head[h];
4045                 hlist_for_each_entry(dev, head, index_hlist) {
4046                         if (idx < s_idx)
4047                                 goto cont;
4048                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4049                                                   NETLINK_CB(cb->skb).portid,
4050                                                   cb->nlh->nlmsg_seq, 0,
4051                                                   flags, filter_mask,
4052                                                   &s_idxattr, &s_prividx);
4053                         /* If we ran out of room on the first message,
4054                          * we're in trouble
4055                          */
4056                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4057
4058                         if (err < 0)
4059                                 goto out;
4060                         s_prividx = 0;
4061                         s_idxattr = 0;
4062                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4063 cont:
4064                         idx++;
4065                 }
4066         }
4067 out:
4068         cb->args[3] = s_prividx;
4069         cb->args[2] = s_idxattr;
4070         cb->args[1] = idx;
4071         cb->args[0] = h;
4072
4073         return skb->len;
4074 }
4075
4076 /* Process one rtnetlink message. */
4077
4078 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4079                              struct netlink_ext_ack *extack)
4080 {
4081         struct net *net = sock_net(skb->sk);
4082         rtnl_doit_func doit;
4083         int kind;
4084         int family;
4085         int type;
4086         int err;
4087
4088         type = nlh->nlmsg_type;
4089         if (type > RTM_MAX)
4090                 return -EOPNOTSUPP;
4091
4092         type -= RTM_BASE;
4093
4094         /* All the messages must have at least 1 byte length */
4095         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4096                 return 0;
4097
4098         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4099         kind = type&3;
4100
4101         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4102                 return -EPERM;
4103
4104         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4105                 struct sock *rtnl;
4106                 rtnl_dumpit_func dumpit;
4107                 rtnl_calcit_func calcit;
4108                 u16 min_dump_alloc = 0;
4109
4110                 dumpit = rtnl_get_dumpit(family, type);
4111                 if (dumpit == NULL)
4112                         return -EOPNOTSUPP;
4113                 calcit = rtnl_get_calcit(family, type);
4114                 if (calcit)
4115                         min_dump_alloc = calcit(skb, nlh);
4116
4117                 __rtnl_unlock();
4118                 rtnl = net->rtnl;
4119                 {
4120                         struct netlink_dump_control c = {
4121                                 .dump           = dumpit,
4122                                 .min_dump_alloc = min_dump_alloc,
4123                         };
4124                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4125                 }
4126                 rtnl_lock();
4127                 return err;
4128         }
4129
4130         doit = rtnl_get_doit(family, type);
4131         if (doit == NULL)
4132                 return -EOPNOTSUPP;
4133
4134         return doit(skb, nlh, extack);
4135 }
4136
4137 static void rtnetlink_rcv(struct sk_buff *skb)
4138 {
4139         rtnl_lock();
4140         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4141         rtnl_unlock();
4142 }
4143
4144 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4145 {
4146         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4147
4148         switch (event) {
4149         case NETDEV_REBOOT:
4150         case NETDEV_CHANGENAME:
4151         case NETDEV_FEAT_CHANGE:
4152         case NETDEV_BONDING_FAILOVER:
4153         case NETDEV_NOTIFY_PEERS:
4154         case NETDEV_RESEND_IGMP:
4155         case NETDEV_CHANGEINFODATA:
4156                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4157                 break;
4158         default:
4159                 break;
4160         }
4161         return NOTIFY_DONE;
4162 }
4163
4164 static struct notifier_block rtnetlink_dev_notifier = {
4165         .notifier_call  = rtnetlink_event,
4166 };
4167
4168
4169 static int __net_init rtnetlink_net_init(struct net *net)
4170 {
4171         struct sock *sk;
4172         struct netlink_kernel_cfg cfg = {
4173                 .groups         = RTNLGRP_MAX,
4174                 .input          = rtnetlink_rcv,
4175                 .cb_mutex       = &rtnl_mutex,
4176                 .flags          = NL_CFG_F_NONROOT_RECV,
4177         };
4178
4179         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4180         if (!sk)
4181                 return -ENOMEM;
4182         net->rtnl = sk;
4183         return 0;
4184 }
4185
4186 static void __net_exit rtnetlink_net_exit(struct net *net)
4187 {
4188         netlink_kernel_release(net->rtnl);
4189         net->rtnl = NULL;
4190 }
4191
4192 static struct pernet_operations rtnetlink_net_ops = {
4193         .init = rtnetlink_net_init,
4194         .exit = rtnetlink_net_exit,
4195 };
4196
4197 void __init rtnetlink_init(void)
4198 {
4199         if (register_pernet_subsys(&rtnetlink_net_ops))
4200                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4201
4202         register_netdevice_notifier(&rtnetlink_dev_notifier);
4203
4204         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4205                       rtnl_dump_ifinfo, rtnl_calcit);
4206         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4207         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4208         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4209
4210         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4211         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4212         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4213
4214         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4215         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4216         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4217
4218         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4219         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4220         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4221
4222         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4223                       NULL);
4224 }