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/bitops.h>
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/kernel.h>
  25 #include <linux/timer.h>
  26 #include <linux/string.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/fcntl.h>
  30 #include <linux/mm.h>
  31 #include <linux/slab.h>
  32 #include <linux/interrupt.h>
  33 #include <linux/capability.h>
  34 #include <linux/skbuff.h>
  35 #include <linux/init.h>
  36 #include <linux/security.h>
  37 #include <linux/mutex.h>
  38 #include <linux/if_addr.h>
  39 #include <linux/if_bridge.h>
  40 #include <linux/if_vlan.h>
  41 #include <linux/pci.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/bpf.h>
  44
  45 #include <linux/uaccess.h>
  46
  47 #include <linux/inet.h>
  48 #include <linux/netdevice.h>
  49 #include <net/switchdev.h>
  50 #include <net/ip.h>
  51 #include <net/protocol.h>
  52 #include <net/arp.h>
  53 #include <net/route.h>
  54 #include <net/udp.h>
  55 #include <net/tcp.h>
  56 #include <net/sock.h>
  57 #include <net/pkt_sched.h>
  58 #include <net/fib_rules.h>
  59 #include <net/rtnetlink.h>
  60 #include <net/net_namespace.h>
  61
  62 struct rtnl_link {
  63         rtnl_doit_func          doit;
  64         rtnl_dumpit_func        dumpit;
  65         unsigned int            flags;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 static struct sk_buff *defer_kfree_skb_list;
  77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  78 {
  79         if (head && tail) {
  80                 tail->next = defer_kfree_skb_list;
  81                 defer_kfree_skb_list = head;
  82         }
  83 }
  84 EXPORT_SYMBOL(rtnl_kfree_skbs);
  85
  86 void __rtnl_unlock(void)
  87 {
  88         struct sk_buff *head = defer_kfree_skb_list;
  89
  90         defer_kfree_skb_list = NULL;
  91
  92         mutex_unlock(&rtnl_mutex);
  93
  94         while (head) {
  95                 struct sk_buff *next = head->next;
  96
  97                 kfree_skb(head);
  98                 cond_resched();
  99                 head = next;
 100         }
 101 }
 102
 103 void rtnl_unlock(void)
 104 {
 105         /* This fellow will unlock it for us. */
 106         netdev_run_todo();
 107 }
 108 EXPORT_SYMBOL(rtnl_unlock);
 109
 110 int rtnl_trylock(void)
 111 {
 112         return mutex_trylock(&rtnl_mutex);
 113 }
 114 EXPORT_SYMBOL(rtnl_trylock);
 115
 116 int rtnl_is_locked(void)
 117 {
 118         return mutex_is_locked(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_is_locked);
 121
 122 #ifdef CONFIG_PROVE_LOCKING
 123 bool lockdep_rtnl_is_held(void)
 124 {
 125         return lockdep_is_held(&rtnl_mutex);
 126 }
 127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 129
 130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
 132
 133 static inline int rtm_msgindex(int msgtype)
 134 {
 135         int msgindex = msgtype - RTM_BASE;
 136
 137         /*
 138          * msgindex < 0 implies someone tried to register a netlink
 139          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 140          * the message type has not been added to linux/rtnetlink.h
 141          */
 142         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 143
 144         return msgindex;
 145 }
 146
 147 /**
 148  * __rtnl_register - Register a rtnetlink message type
 149  * @protocol: Protocol family or PF_UNSPEC
 150  * @msgtype: rtnetlink message type
 151  * @doit: Function pointer called for each request message
 152  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 153  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 154  *
 155  * Registers the specified function pointers (at least one of them has
 156  * to be non-NULL) to be called whenever a request message for the
 157  * specified protocol family and message type is received.
 158  *
 159  * The special protocol family PF_UNSPEC may be used to define fallback
 160  * function pointers for the case when no entry for the specific protocol
 161  * family exists.
 162  *
 163  * Returns 0 on success or a negative error code.
 164  */
 165 int __rtnl_register(int protocol, int msgtype,
 166                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 167                     unsigned int flags)
 168 {
 169         struct rtnl_link *tab;
 170         int msgindex;
 171
 172         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 173         msgindex = rtm_msgindex(msgtype);
 174
 175         tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
 176         if (tab == NULL) {
 177                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 178                 if (tab == NULL)
 179                         return -ENOBUFS;
 180
 181                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 182         }
 183
 184         if (doit)
 185                 tab[msgindex].doit = doit;
 186         if (dumpit)
 187                 tab[msgindex].dumpit = dumpit;
 188         tab[msgindex].flags |= flags;
 189
 190         return 0;
 191 }
 192 EXPORT_SYMBOL_GPL(__rtnl_register);
 193
 194 /**
 195  * rtnl_register - Register a rtnetlink message type
 196  *
 197  * Identical to __rtnl_register() but panics on failure. This is useful
 198  * as failure of this function is very unlikely, it can only happen due
 199  * to lack of memory when allocating the chain to store all message
 200  * handlers for a protocol. Meant for use in init functions where lack
 201  * of memory implies no sense in continuing.
 202  */
 203 void rtnl_register(int protocol, int msgtype,
 204                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 205                    unsigned int flags)
 206 {
 207         if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
 208                 panic("Unable to register rtnetlink message handler, "
 209                       "protocol = %d, message type = %d\n",
 210                       protocol, msgtype);
 211 }
 212 EXPORT_SYMBOL_GPL(rtnl_register);
 213
 214 /**
 215  * rtnl_unregister - Unregister a rtnetlink message type
 216  * @protocol: Protocol family or PF_UNSPEC
 217  * @msgtype: rtnetlink message type
 218  *
 219  * Returns 0 on success or a negative error code.
 220  */
 221 int rtnl_unregister(int protocol, int msgtype)
 222 {
 223         struct rtnl_link *handlers;
 224         int msgindex;
 225
 226         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 227         msgindex = rtm_msgindex(msgtype);
 228
 229         rtnl_lock();
 230         handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
 231         if (!handlers) {
 232                 rtnl_unlock();
 233                 return -ENOENT;
 234         }
 235
 236         handlers[msgindex].doit = NULL;
 237         handlers[msgindex].dumpit = NULL;
 238         handlers[msgindex].flags = 0;
 239         rtnl_unlock();
 240
 241         return 0;
 242 }
 243 EXPORT_SYMBOL_GPL(rtnl_unregister);
 244
 245 /**
 246  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 247  * @protocol : Protocol family or PF_UNSPEC
 248  *
 249  * Identical to calling rtnl_unregster() for all registered message types
 250  * of a certain protocol family.
 251  */
 252 void rtnl_unregister_all(int protocol)
 253 {
 254         struct rtnl_link *handlers;
 255
 256         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 257
 258         rtnl_lock();
 259         handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
 260         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 261         rtnl_unlock();
 262
 263         synchronize_net();
 264
 265         while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
 266                 schedule();
 267         kfree(handlers);
 268 }
 269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 270
 271 static LIST_HEAD(link_ops);
 272
 273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 274 {
 275         const struct rtnl_link_ops *ops;
 276
 277         list_for_each_entry(ops, &link_ops, list) {
 278                 if (!strcmp(ops->kind, kind))
 279                         return ops;
 280         }
 281         return NULL;
 282 }
 283
 284 /**
 285  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 286  * @ops: struct rtnl_link_ops * to register
 287  *
 288  * The caller must hold the rtnl_mutex. This function should be used
 289  * by drivers that create devices during module initialization. It
 290  * must be called before registering the devices.
 291  *
 292  * Returns 0 on success or a negative error code.
 293  */
 294 int __rtnl_link_register(struct rtnl_link_ops *ops)
 295 {
 296         if (rtnl_link_ops_get(ops->kind))
 297                 return -EEXIST;
 298
 299         /* The check for setup is here because if ops
 300          * does not have that filled up, it is not possible
 301          * to use the ops for creating device. So do not
 302          * fill up dellink as well. That disables rtnl_dellink.
 303          */
 304         if (ops->setup && !ops->dellink)
 305                 ops->dellink = unregister_netdevice_queue;
 306
 307         list_add_tail(&ops->list, &link_ops);
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 311
 312 /**
 313  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 314  * @ops: struct rtnl_link_ops * to register
 315  *
 316  * Returns 0 on success or a negative error code.
 317  */
 318 int rtnl_link_register(struct rtnl_link_ops *ops)
 319 {
 320         int err;
 321
 322         rtnl_lock();
 323         err = __rtnl_link_register(ops);
 324         rtnl_unlock();
 325         return err;
 326 }
 327 EXPORT_SYMBOL_GPL(rtnl_link_register);
 328
 329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 330 {
 331         struct net_device *dev;
 332         LIST_HEAD(list_kill);
 333
 334         for_each_netdev(net, dev) {
 335                 if (dev->rtnl_link_ops == ops)
 336                         ops->dellink(dev, &list_kill);
 337         }
 338         unregister_netdevice_many(&list_kill);
 339 }
 340
 341 /**
 342  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 343  * @ops: struct rtnl_link_ops * to unregister
 344  *
 345  * The caller must hold the rtnl_mutex.
 346  */
 347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 348 {
 349         struct net *net;
 350
 351         for_each_net(net) {
 352                 __rtnl_kill_links(net, ops);
 353         }
 354         list_del(&ops->list);
 355 }
 356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 357
 358 /* Return with the rtnl_lock held when there are no network
 359  * devices unregistering in any network namespace.
 360  */
 361 static void rtnl_lock_unregistering_all(void)
 362 {
 363         struct net *net;
 364         bool unregistering;
 365         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 366
 367         add_wait_queue(&netdev_unregistering_wq, &wait);
 368         for (;;) {
 369                 unregistering = false;
 370                 rtnl_lock();
 371                 for_each_net(net) {
 372                         if (net->dev_unreg_count > 0) {
 373                                 unregistering = true;
 374                                 break;
 375                         }
 376                 }
 377                 if (!unregistering)
 378                         break;
 379                 __rtnl_unlock();
 380
 381                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 382         }
 383         remove_wait_queue(&netdev_unregistering_wq, &wait);
 384 }
 385
 386 /**
 387  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 388  * @ops: struct rtnl_link_ops * to unregister
 389  */
 390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 391 {
 392         /* Close the race with cleanup_net() */
 393         mutex_lock(&net_mutex);
 394         rtnl_lock_unregistering_all();
 395         __rtnl_link_unregister(ops);
 396         rtnl_unlock();
 397         mutex_unlock(&net_mutex);
 398 }
 399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 400
 401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 402 {
 403         struct net_device *master_dev;
 404         const struct rtnl_link_ops *ops;
 405         size_t size = 0;
 406
 407         rcu_read_lock();
 408
 409         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 410         if (!master_dev)
 411                 goto out;
 412
 413         ops = master_dev->rtnl_link_ops;
 414         if (!ops || !ops->get_slave_size)
 415                 goto out;
 416         /* IFLA_INFO_SLAVE_DATA + nested data */
 417         size = nla_total_size(sizeof(struct nlattr)) +
 418                ops->get_slave_size(master_dev, dev);
 419
 420 out:
 421         rcu_read_unlock();
 422         return size;
 423 }
 424
 425 static size_t rtnl_link_get_size(const struct net_device *dev)
 426 {
 427         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 428         size_t size;
 429
 430         if (!ops)
 431                 return 0;
 432
 433         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 434                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 435
 436         if (ops->get_size)
 437                 /* IFLA_INFO_DATA + nested data */
 438                 size += nla_total_size(sizeof(struct nlattr)) +
 439                         ops->get_size(dev);
 440
 441         if (ops->get_xstats_size)
 442                 /* IFLA_INFO_XSTATS */
 443                 size += nla_total_size(ops->get_xstats_size(dev));
 444
 445         size += rtnl_link_get_slave_info_data_size(dev);
 446
 447         return size;
 448 }
 449
 450 static LIST_HEAD(rtnl_af_ops);
 451
 452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 453 {
 454         const struct rtnl_af_ops *ops;
 455
 456         list_for_each_entry(ops, &rtnl_af_ops, list) {
 457                 if (ops->family == family)
 458                         return ops;
 459         }
 460
 461         return NULL;
 462 }
 463
 464 /**
 465  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 466  * @ops: struct rtnl_af_ops * to register
 467  *
 468  * Returns 0 on success or a negative error code.
 469  */
 470 void rtnl_af_register(struct rtnl_af_ops *ops)
 471 {
 472         rtnl_lock();
 473         list_add_tail(&ops->list, &rtnl_af_ops);
 474         rtnl_unlock();
 475 }
 476 EXPORT_SYMBOL_GPL(rtnl_af_register);
 477
 478 /**
 479  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 480  * @ops: struct rtnl_af_ops * to unregister
 481  *
 482  * The caller must hold the rtnl_mutex.
 483  */
 484 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 485 {
 486         list_del(&ops->list);
 487 }
 488 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 489
 490 /**
 491  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 492  * @ops: struct rtnl_af_ops * to unregister
 493  */
 494 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 495 {
 496         rtnl_lock();
 497         __rtnl_af_unregister(ops);
 498         rtnl_unlock();
 499 }
 500 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 501
 502 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 503                                     u32 ext_filter_mask)
 504 {
 505         struct rtnl_af_ops *af_ops;
 506         size_t size;
 507
 508         /* IFLA_AF_SPEC */
 509         size = nla_total_size(sizeof(struct nlattr));
 510
 511         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 512                 if (af_ops->get_link_af_size) {
 513                         /* AF_* + nested data */
 514                         size += nla_total_size(sizeof(struct nlattr)) +
 515                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 516                 }
 517         }
 518
 519         return size;
 520 }
 521
 522 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 523 {
 524         struct net_device *master_dev;
 525
 526         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 527         if (master_dev && master_dev->rtnl_link_ops)
 528                 return true;
 529         return false;
 530 }
 531
 532 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 533                                      const struct net_device *dev)
 534 {
 535         struct net_device *master_dev;
 536         const struct rtnl_link_ops *ops;
 537         struct nlattr *slave_data;
 538         int err;
 539
 540         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 541         if (!master_dev)
 542                 return 0;
 543         ops = master_dev->rtnl_link_ops;
 544         if (!ops)
 545                 return 0;
 546         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 547                 return -EMSGSIZE;
 548         if (ops->fill_slave_info) {
 549                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 550                 if (!slave_data)
 551                         return -EMSGSIZE;
 552                 err = ops->fill_slave_info(skb, master_dev, dev);
 553                 if (err < 0)
 554                         goto err_cancel_slave_data;
 555                 nla_nest_end(skb, slave_data);
 556         }
 557         return 0;
 558
 559 err_cancel_slave_data:
 560         nla_nest_cancel(skb, slave_data);
 561         return err;
 562 }
 563
 564 static int rtnl_link_info_fill(struct sk_buff *skb,
 565                                const struct net_device *dev)
 566 {
 567         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 568         struct nlattr *data;
 569         int err;
 570
 571         if (!ops)
 572                 return 0;
 573         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 574                 return -EMSGSIZE;
 575         if (ops->fill_xstats) {
 576                 err = ops->fill_xstats(skb, dev);
 577                 if (err < 0)
 578                         return err;
 579         }
 580         if (ops->fill_info) {
 581                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 582                 if (data == NULL)
 583                         return -EMSGSIZE;
 584                 err = ops->fill_info(skb, dev);
 585                 if (err < 0)
 586                         goto err_cancel_data;
 587                 nla_nest_end(skb, data);
 588         }
 589         return 0;
 590
 591 err_cancel_data:
 592         nla_nest_cancel(skb, data);
 593         return err;
 594 }
 595
 596 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 597 {
 598         struct nlattr *linkinfo;
 599         int err = -EMSGSIZE;
 600
 601         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 602         if (linkinfo == NULL)
 603                 goto out;
 604
 605         err = rtnl_link_info_fill(skb, dev);
 606         if (err < 0)
 607                 goto err_cancel_link;
 608
 609         err = rtnl_link_slave_info_fill(skb, dev);
 610         if (err < 0)
 611                 goto err_cancel_link;
 612
 613         nla_nest_end(skb, linkinfo);
 614         return 0;
 615
 616 err_cancel_link:
 617         nla_nest_cancel(skb, linkinfo);
 618 out:
 619         return err;
 620 }
 621
 622 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 623 {
 624         struct sock *rtnl = net->rtnl;
 625         int err = 0;
 626
 627         NETLINK_CB(skb).dst_group = group;
 628         if (echo)
 629                 refcount_inc(&skb->users);
 630         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 631         if (echo)
 632                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 633         return err;
 634 }
 635
 636 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 637 {
 638         struct sock *rtnl = net->rtnl;
 639
 640         return nlmsg_unicast(rtnl, skb, pid);
 641 }
 642 EXPORT_SYMBOL(rtnl_unicast);
 643
 644 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 645                  struct nlmsghdr *nlh, gfp_t flags)
 646 {
 647         struct sock *rtnl = net->rtnl;
 648         int report = 0;
 649
 650         if (nlh)
 651                 report = nlmsg_report(nlh);
 652
 653         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 654 }
 655 EXPORT_SYMBOL(rtnl_notify);
 656
 657 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 658 {
 659         struct sock *rtnl = net->rtnl;
 660
 661         netlink_set_err(rtnl, 0, group, error);
 662 }
 663 EXPORT_SYMBOL(rtnl_set_sk_err);
 664
 665 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 666 {
 667         struct nlattr *mx;
 668         int i, valid = 0;
 669
 670         mx = nla_nest_start(skb, RTA_METRICS);
 671         if (mx == NULL)
 672                 return -ENOBUFS;
 673
 674         for (i = 0; i < RTAX_MAX; i++) {
 675                 if (metrics[i]) {
 676                         if (i == RTAX_CC_ALGO - 1) {
 677                                 char tmp[TCP_CA_NAME_MAX], *name;
 678
 679                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 680                                 if (!name)
 681                                         continue;
 682                                 if (nla_put_string(skb, i + 1, name))
 683                                         goto nla_put_failure;
 684                         } else if (i == RTAX_FEATURES - 1) {
 685                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 686
 687                                 if (!user_features)
 688                                         continue;
 689                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 690                                 if (nla_put_u32(skb, i + 1, user_features))
 691                                         goto nla_put_failure;
 692                         } else {
 693                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 694                                         goto nla_put_failure;
 695                         }
 696                         valid++;
 697                 }
 698         }
 699
 700         if (!valid) {
 701                 nla_nest_cancel(skb, mx);
 702                 return 0;
 703         }
 704
 705         return nla_nest_end(skb, mx);
 706
 707 nla_put_failure:
 708         nla_nest_cancel(skb, mx);
 709         return -EMSGSIZE;
 710 }
 711 EXPORT_SYMBOL(rtnetlink_put_metrics);
 712
 713 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 714                        long expires, u32 error)
 715 {
 716         struct rta_cacheinfo ci = {
 717                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 718                 .rta_used = dst->__use,
 719                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 720                 .rta_error = error,
 721                 .rta_id =  id,
 722         };
 723
 724         if (expires) {
 725                 unsigned long clock;
 726
 727                 clock = jiffies_to_clock_t(abs(expires));
 728                 clock = min_t(unsigned long, clock, INT_MAX);
 729                 ci.rta_expires = (expires > 0) ? clock : -clock;
 730         }
 731         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 732 }
 733 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 734
 735 static void set_operstate(struct net_device *dev, unsigned char transition)
 736 {
 737         unsigned char operstate = dev->operstate;
 738
 739         switch (transition) {
 740         case IF_OPER_UP:
 741                 if ((operstate == IF_OPER_DORMANT ||
 742                      operstate == IF_OPER_UNKNOWN) &&
 743                     !netif_dormant(dev))
 744                         operstate = IF_OPER_UP;
 745                 break;
 746
 747         case IF_OPER_DORMANT:
 748                 if (operstate == IF_OPER_UP ||
 749                     operstate == IF_OPER_UNKNOWN)
 750                         operstate = IF_OPER_DORMANT;
 751                 break;
 752         }
 753
 754         if (dev->operstate != operstate) {
 755                 write_lock_bh(&dev_base_lock);
 756                 dev->operstate = operstate;
 757                 write_unlock_bh(&dev_base_lock);
 758                 netdev_state_change(dev);
 759         }
 760 }
 761
 762 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 763 {
 764         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 765                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 766 }
 767
 768 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 769                                            const struct ifinfomsg *ifm)
 770 {
 771         unsigned int flags = ifm->ifi_flags;
 772
 773         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 774         if (ifm->ifi_change)
 775                 flags = (flags & ifm->ifi_change) |
 776                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 777
 778         return flags;
 779 }
 780
 781 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 782                                  const struct rtnl_link_stats64 *b)
 783 {
 784         a->rx_packets = b->rx_packets;
 785         a->tx_packets = b->tx_packets;
 786         a->rx_bytes = b->rx_bytes;
 787         a->tx_bytes = b->tx_bytes;
 788         a->rx_errors = b->rx_errors;
 789         a->tx_errors = b->tx_errors;
 790         a->rx_dropped = b->rx_dropped;
 791         a->tx_dropped = b->tx_dropped;
 792
 793         a->multicast = b->multicast;
 794         a->collisions = b->collisions;
 795
 796         a->rx_length_errors = b->rx_length_errors;
 797         a->rx_over_errors = b->rx_over_errors;
 798         a->rx_crc_errors = b->rx_crc_errors;
 799         a->rx_frame_errors = b->rx_frame_errors;
 800         a->rx_fifo_errors = b->rx_fifo_errors;
 801         a->rx_missed_errors = b->rx_missed_errors;
 802
 803         a->tx_aborted_errors = b->tx_aborted_errors;
 804         a->tx_carrier_errors = b->tx_carrier_errors;
 805         a->tx_fifo_errors = b->tx_fifo_errors;
 806         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 807         a->tx_window_errors = b->tx_window_errors;
 808
 809         a->rx_compressed = b->rx_compressed;
 810         a->tx_compressed = b->tx_compressed;
 811
 812         a->rx_nohandler = b->rx_nohandler;
 813 }
 814
 815 /* All VF info */
 816 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 817                                    u32 ext_filter_mask)
 818 {
 819         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 820                 int num_vfs = dev_num_vf(dev->dev.parent);
 821                 size_t size = nla_total_size(0);
 822                 size += num_vfs *
 823                         (nla_total_size(0) +
 824                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 825                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 826                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 827                          nla_total_size(MAX_VLAN_LIST_LEN *
 828                                         sizeof(struct ifla_vf_vlan_info)) +
 829                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 830                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 831                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 832                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 833                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 834                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 835                          /* IFLA_VF_STATS_RX_PACKETS */
 836                          nla_total_size_64bit(sizeof(__u64)) +
 837                          /* IFLA_VF_STATS_TX_PACKETS */
 838                          nla_total_size_64bit(sizeof(__u64)) +
 839                          /* IFLA_VF_STATS_RX_BYTES */
 840                          nla_total_size_64bit(sizeof(__u64)) +
 841                          /* IFLA_VF_STATS_TX_BYTES */
 842                          nla_total_size_64bit(sizeof(__u64)) +
 843                          /* IFLA_VF_STATS_BROADCAST */
 844                          nla_total_size_64bit(sizeof(__u64)) +
 845                          /* IFLA_VF_STATS_MULTICAST */
 846                          nla_total_size_64bit(sizeof(__u64)) +
 847                          nla_total_size(sizeof(struct ifla_vf_trust)));
 848                 return size;
 849         } else
 850                 return 0;
 851 }
 852
 853 static size_t rtnl_port_size(const struct net_device *dev,
 854                              u32 ext_filter_mask)
 855 {
 856         size_t port_size = nla_total_size(4)            /* PORT_VF */
 857                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 858                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 859                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 860                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 861                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 862         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 863         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 864                 + port_size;
 865         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 866                 + port_size;
 867
 868         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 869             !(ext_filter_mask & RTEXT_FILTER_VF))
 870                 return 0;
 871         if (dev_num_vf(dev->dev.parent))
 872                 return port_self_size + vf_ports_size +
 873                         vf_port_size * dev_num_vf(dev->dev.parent);
 874         else
 875                 return port_self_size;
 876 }
 877
 878 static size_t rtnl_xdp_size(void)
 879 {
 880         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 881                           nla_total_size(1) +   /* XDP_ATTACHED */
 882                           nla_total_size(4);    /* XDP_PROG_ID */
 883
 884         return xdp_size;
 885 }
 886
 887 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 888                                      u32 ext_filter_mask)
 889 {
 890         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 891                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 892                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 893                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 894                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 895                + nla_total_size(sizeof(struct rtnl_link_stats))
 896                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 897                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 898                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 899                + nla_total_size(4) /* IFLA_TXQLEN */
 900                + nla_total_size(4) /* IFLA_WEIGHT */
 901                + nla_total_size(4) /* IFLA_MTU */
 902                + nla_total_size(4) /* IFLA_LINK */
 903                + nla_total_size(4) /* IFLA_MASTER */
 904                + nla_total_size(1) /* IFLA_CARRIER */
 905                + nla_total_size(4) /* IFLA_PROMISCUITY */
 906                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 907                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 908                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 909                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 910                + nla_total_size(1) /* IFLA_OPERSTATE */
 911                + nla_total_size(1) /* IFLA_LINKMODE */
 912                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 913                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 914                + nla_total_size(4) /* IFLA_GROUP */
 915                + nla_total_size(ext_filter_mask
 916                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 917                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 918                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 919                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 920                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 921                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 922                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 923                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 924                + rtnl_xdp_size() /* IFLA_XDP */
 925                + nla_total_size(4)  /* IFLA_EVENT */
 926                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 927
 928 }
 929
 930 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 931 {
 932         struct nlattr *vf_ports;
 933         struct nlattr *vf_port;
 934         int vf;
 935         int err;
 936
 937         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 938         if (!vf_ports)
 939                 return -EMSGSIZE;
 940
 941         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 942                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 943                 if (!vf_port)
 944                         goto nla_put_failure;
 945                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 946                         goto nla_put_failure;
 947                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 948                 if (err == -EMSGSIZE)
 949                         goto nla_put_failure;
 950                 if (err) {
 951                         nla_nest_cancel(skb, vf_port);
 952                         continue;
 953                 }
 954                 nla_nest_end(skb, vf_port);
 955         }
 956
 957         nla_nest_end(skb, vf_ports);
 958
 959         return 0;
 960
 961 nla_put_failure:
 962         nla_nest_cancel(skb, vf_ports);
 963         return -EMSGSIZE;
 964 }
 965
 966 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 967 {
 968         struct nlattr *port_self;
 969         int err;
 970
 971         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 972         if (!port_self)
 973                 return -EMSGSIZE;
 974
 975         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 976         if (err) {
 977                 nla_nest_cancel(skb, port_self);
 978                 return (err == -EMSGSIZE) ? err : 0;
 979         }
 980
 981         nla_nest_end(skb, port_self);
 982
 983         return 0;
 984 }
 985
 986 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 987                           u32 ext_filter_mask)
 988 {
 989         int err;
 990
 991         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 992             !(ext_filter_mask & RTEXT_FILTER_VF))
 993                 return 0;
 994
 995         err = rtnl_port_self_fill(skb, dev);
 996         if (err)
 997                 return err;
 998
 999         if (dev_num_vf(dev->dev.parent)) {
1000                 err = rtnl_vf_ports_fill(skb, dev);
1001                 if (err)
1002                         return err;
1003         }
1004
1005         return 0;
1006 }
1007
1008 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1009 {
1010         int err;
1011         struct netdev_phys_item_id ppid;
1012
1013         err = dev_get_phys_port_id(dev, &ppid);
1014         if (err) {
1015                 if (err == -EOPNOTSUPP)
1016                         return 0;
1017                 return err;
1018         }
1019
1020         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1021                 return -EMSGSIZE;
1022
1023         return 0;
1024 }
1025
1026 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028         char name[IFNAMSIZ];
1029         int err;
1030
1031         err = dev_get_phys_port_name(dev, name, sizeof(name));
1032         if (err) {
1033                 if (err == -EOPNOTSUPP)
1034                         return 0;
1035                 return err;
1036         }
1037
1038         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1039                 return -EMSGSIZE;
1040
1041         return 0;
1042 }
1043
1044 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1045 {
1046         int err;
1047         struct switchdev_attr attr = {
1048                 .orig_dev = dev,
1049                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1050                 .flags = SWITCHDEV_F_NO_RECURSE,
1051         };
1052
1053         err = switchdev_port_attr_get(dev, &attr);
1054         if (err) {
1055                 if (err == -EOPNOTSUPP)
1056                         return 0;
1057                 return err;
1058         }
1059
1060         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1061                     attr.u.ppid.id))
1062                 return -EMSGSIZE;
1063
1064         return 0;
1065 }
1066
1067 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1068                                               struct net_device *dev)
1069 {
1070         struct rtnl_link_stats64 *sp;
1071         struct nlattr *attr;
1072
1073         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1074                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1075         if (!attr)
1076                 return -EMSGSIZE;
1077
1078         sp = nla_data(attr);
1079         dev_get_stats(dev, sp);
1080
1081         attr = nla_reserve(skb, IFLA_STATS,
1082                            sizeof(struct rtnl_link_stats));
1083         if (!attr)
1084                 return -EMSGSIZE;
1085
1086         copy_rtnl_link_stats(nla_data(attr), sp);
1087
1088         return 0;
1089 }
1090
1091 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1092                                                struct net_device *dev,
1093                                                int vfs_num,
1094                                                struct nlattr *vfinfo)
1095 {
1096         struct ifla_vf_rss_query_en vf_rss_query_en;
1097         struct nlattr *vf, *vfstats, *vfvlanlist;
1098         struct ifla_vf_link_state vf_linkstate;
1099         struct ifla_vf_vlan_info vf_vlan_info;
1100         struct ifla_vf_spoofchk vf_spoofchk;
1101         struct ifla_vf_tx_rate vf_tx_rate;
1102         struct ifla_vf_stats vf_stats;
1103         struct ifla_vf_trust vf_trust;
1104         struct ifla_vf_vlan vf_vlan;
1105         struct ifla_vf_rate vf_rate;
1106         struct ifla_vf_mac vf_mac;
1107         struct ifla_vf_info ivi;
1108
1109         memset(&ivi, 0, sizeof(ivi));
1110
1111         /* Not all SR-IOV capable drivers support the
1112          * spoofcheck and "RSS query enable" query.  Preset to
1113          * -1 so the user space tool can detect that the driver
1114          * didn't report anything.
1115          */
1116         ivi.spoofchk = -1;
1117         ivi.rss_query_en = -1;
1118         ivi.trusted = -1;
1119         /* The default value for VF link state is "auto"
1120          * IFLA_VF_LINK_STATE_AUTO which equals zero
1121          */
1122         ivi.linkstate = 0;
1123         /* VLAN Protocol by default is 802.1Q */
1124         ivi.vlan_proto = htons(ETH_P_8021Q);
1125         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1126                 return 0;
1127
1128         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1129
1130         vf_mac.vf =
1131                 vf_vlan.vf =
1132                 vf_vlan_info.vf =
1133                 vf_rate.vf =
1134                 vf_tx_rate.vf =
1135                 vf_spoofchk.vf =
1136                 vf_linkstate.vf =
1137                 vf_rss_query_en.vf =
1138                 vf_trust.vf = ivi.vf;
1139
1140         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1141         vf_vlan.vlan = ivi.vlan;
1142         vf_vlan.qos = ivi.qos;
1143         vf_vlan_info.vlan = ivi.vlan;
1144         vf_vlan_info.qos = ivi.qos;
1145         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1146         vf_tx_rate.rate = ivi.max_tx_rate;
1147         vf_rate.min_tx_rate = ivi.min_tx_rate;
1148         vf_rate.max_tx_rate = ivi.max_tx_rate;
1149         vf_spoofchk.setting = ivi.spoofchk;
1150         vf_linkstate.link_state = ivi.linkstate;
1151         vf_rss_query_en.setting = ivi.rss_query_en;
1152         vf_trust.setting = ivi.trusted;
1153         vf = nla_nest_start(skb, IFLA_VF_INFO);
1154         if (!vf)
1155                 goto nla_put_vfinfo_failure;
1156         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1157             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1158             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1159                     &vf_rate) ||
1160             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1161                     &vf_tx_rate) ||
1162             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1163                     &vf_spoofchk) ||
1164             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1165                     &vf_linkstate) ||
1166             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1167                     sizeof(vf_rss_query_en),
1168                     &vf_rss_query_en) ||
1169             nla_put(skb, IFLA_VF_TRUST,
1170                     sizeof(vf_trust), &vf_trust))
1171                 goto nla_put_vf_failure;
1172         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1173         if (!vfvlanlist)
1174                 goto nla_put_vf_failure;
1175         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1176                     &vf_vlan_info)) {
1177                 nla_nest_cancel(skb, vfvlanlist);
1178                 goto nla_put_vf_failure;
1179         }
1180         nla_nest_end(skb, vfvlanlist);
1181         memset(&vf_stats, 0, sizeof(vf_stats));
1182         if (dev->netdev_ops->ndo_get_vf_stats)
1183                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1184                                                 &vf_stats);
1185         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1186         if (!vfstats)
1187                 goto nla_put_vf_failure;
1188         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1189                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1190             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1191                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1192             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1193                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1194             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1195                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1196             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1197                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1198             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1199                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1200                 nla_nest_cancel(skb, vfstats);
1201                 goto nla_put_vf_failure;
1202         }
1203         nla_nest_end(skb, vfstats);
1204         nla_nest_end(skb, vf);
1205         return 0;
1206
1207 nla_put_vf_failure:
1208         nla_nest_cancel(skb, vf);
1209 nla_put_vfinfo_failure:
1210         nla_nest_cancel(skb, vfinfo);
1211         return -EMSGSIZE;
1212 }
1213
1214 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1215 {
1216         struct rtnl_link_ifmap map;
1217
1218         memset(&map, 0, sizeof(map));
1219         map.mem_start   = dev->mem_start;
1220         map.mem_end     = dev->mem_end;
1221         map.base_addr   = dev->base_addr;
1222         map.irq         = dev->irq;
1223         map.dma         = dev->dma;
1224         map.port        = dev->if_port;
1225
1226         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1227                 return -EMSGSIZE;
1228
1229         return 0;
1230 }
1231
1232 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1233 {
1234         const struct net_device_ops *ops = dev->netdev_ops;
1235         const struct bpf_prog *generic_xdp_prog;
1236
1237         ASSERT_RTNL();
1238
1239         *prog_id = 0;
1240         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1241         if (generic_xdp_prog) {
1242                 *prog_id = generic_xdp_prog->aux->id;
1243                 return XDP_ATTACHED_SKB;
1244         }
1245         if (!ops->ndo_xdp)
1246                 return XDP_ATTACHED_NONE;
1247
1248         return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1249 }
1250
1251 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1252 {
1253         struct nlattr *xdp;
1254         u32 prog_id;
1255         int err;
1256
1257         xdp = nla_nest_start(skb, IFLA_XDP);
1258         if (!xdp)
1259                 return -EMSGSIZE;
1260
1261         err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1262                          rtnl_xdp_attached_mode(dev, &prog_id));
1263         if (err)
1264                 goto err_cancel;
1265
1266         if (prog_id) {
1267                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1268                 if (err)
1269                         goto err_cancel;
1270         }
1271
1272         nla_nest_end(skb, xdp);
1273         return 0;
1274
1275 err_cancel:
1276         nla_nest_cancel(skb, xdp);
1277         return err;
1278 }
1279
1280 static u32 rtnl_get_event(unsigned long event)
1281 {
1282         u32 rtnl_event_type = IFLA_EVENT_NONE;
1283
1284         switch (event) {
1285         case NETDEV_REBOOT:
1286                 rtnl_event_type = IFLA_EVENT_REBOOT;
1287                 break;
1288         case NETDEV_FEAT_CHANGE:
1289                 rtnl_event_type = IFLA_EVENT_FEATURES;
1290                 break;
1291         case NETDEV_BONDING_FAILOVER:
1292                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1293                 break;
1294         case NETDEV_NOTIFY_PEERS:
1295                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1296                 break;
1297         case NETDEV_RESEND_IGMP:
1298                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1299                 break;
1300         case NETDEV_CHANGEINFODATA:
1301                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1302                 break;
1303         default:
1304                 break;
1305         }
1306
1307         return rtnl_event_type;
1308 }
1309
1310 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1311                             int type, u32 pid, u32 seq, u32 change,
1312                             unsigned int flags, u32 ext_filter_mask,
1313                             u32 event)
1314 {
1315         struct ifinfomsg *ifm;
1316         struct nlmsghdr *nlh;
1317         struct nlattr *af_spec;
1318         struct rtnl_af_ops *af_ops;
1319         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1320
1321         ASSERT_RTNL();
1322         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1323         if (nlh == NULL)
1324                 return -EMSGSIZE;
1325
1326         ifm = nlmsg_data(nlh);
1327         ifm->ifi_family = AF_UNSPEC;
1328         ifm->__ifi_pad = 0;
1329         ifm->ifi_type = dev->type;
1330         ifm->ifi_index = dev->ifindex;
1331         ifm->ifi_flags = dev_get_flags(dev);
1332         ifm->ifi_change = change;
1333
1334         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1335             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1336             nla_put_u8(skb, IFLA_OPERSTATE,
1337                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1338             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1339             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1340             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1341             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1342             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1343             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1344             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1345 #ifdef CONFIG_RPS
1346             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1347 #endif
1348             (dev->ifindex != dev_get_iflink(dev) &&
1349              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1350             (upper_dev &&
1351              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1352             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1353             (dev->qdisc &&
1354              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1355             (dev->ifalias &&
1356              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1357             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1358                         atomic_read(&dev->carrier_changes)) ||
1359             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1360                 goto nla_put_failure;
1361
1362         if (event != IFLA_EVENT_NONE) {
1363                 if (nla_put_u32(skb, IFLA_EVENT, event))
1364                         goto nla_put_failure;
1365         }
1366
1367         if (rtnl_fill_link_ifmap(skb, dev))
1368                 goto nla_put_failure;
1369
1370         if (dev->addr_len) {
1371                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1372                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1373                         goto nla_put_failure;
1374         }
1375
1376         if (rtnl_phys_port_id_fill(skb, dev))
1377                 goto nla_put_failure;
1378
1379         if (rtnl_phys_port_name_fill(skb, dev))
1380                 goto nla_put_failure;
1381
1382         if (rtnl_phys_switch_id_fill(skb, dev))
1383                 goto nla_put_failure;
1384
1385         if (rtnl_fill_stats(skb, dev))
1386                 goto nla_put_failure;
1387
1388         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1389             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1390                 goto nla_put_failure;
1391
1392         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1393             ext_filter_mask & RTEXT_FILTER_VF) {
1394                 int i;
1395                 struct nlattr *vfinfo;
1396                 int num_vfs = dev_num_vf(dev->dev.parent);
1397
1398                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1399                 if (!vfinfo)
1400                         goto nla_put_failure;
1401                 for (i = 0; i < num_vfs; i++) {
1402                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1403                                 goto nla_put_failure;
1404                 }
1405
1406                 nla_nest_end(skb, vfinfo);
1407         }
1408
1409         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1410                 goto nla_put_failure;
1411
1412         if (rtnl_xdp_fill(skb, dev))
1413                 goto nla_put_failure;
1414
1415         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1416                 if (rtnl_link_fill(skb, dev) < 0)
1417                         goto nla_put_failure;
1418         }
1419
1420         if (dev->rtnl_link_ops &&
1421             dev->rtnl_link_ops->get_link_net) {
1422                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1423
1424                 if (!net_eq(dev_net(dev), link_net)) {
1425                         int id = peernet2id_alloc(dev_net(dev), link_net);
1426
1427                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1428                                 goto nla_put_failure;
1429                 }
1430         }
1431
1432         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1433                 goto nla_put_failure;
1434
1435         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1436                 if (af_ops->fill_link_af) {
1437                         struct nlattr *af;
1438                         int err;
1439
1440                         if (!(af = nla_nest_start(skb, af_ops->family)))
1441                                 goto nla_put_failure;
1442
1443                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1444
1445                         /*
1446                          * Caller may return ENODATA to indicate that there
1447                          * was no data to be dumped. This is not an error, it
1448                          * means we should trim the attribute header and
1449                          * continue.
1450                          */
1451                         if (err == -ENODATA)
1452                                 nla_nest_cancel(skb, af);
1453                         else if (err < 0)
1454                                 goto nla_put_failure;
1455
1456                         nla_nest_end(skb, af);
1457                 }
1458         }
1459
1460         nla_nest_end(skb, af_spec);
1461
1462         nlmsg_end(skb, nlh);
1463         return 0;
1464
1465 nla_put_failure:
1466         nlmsg_cancel(skb, nlh);
1467         return -EMSGSIZE;
1468 }
1469
1470 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1471         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1472         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1473         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1474         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1475         [IFLA_MTU]              = { .type = NLA_U32 },
1476         [IFLA_LINK]             = { .type = NLA_U32 },
1477         [IFLA_MASTER]           = { .type = NLA_U32 },
1478         [IFLA_CARRIER]          = { .type = NLA_U8 },
1479         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1480         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1481         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1482         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1483         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1484         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1485         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1486         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1487          * allow 0-length string (needed to remove an alias).
1488          */
1489         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1490         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1491         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1492         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1493         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1494         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1495         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1496         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1497         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1498         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1499         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1500         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1501         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1502         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1503         [IFLA_XDP]              = { .type = NLA_NESTED },
1504         [IFLA_EVENT]            = { .type = NLA_U32 },
1505         [IFLA_GROUP]            = { .type = NLA_U32 },
1506 };
1507
1508 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1509         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1510         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1511         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1512         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1513 };
1514
1515 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1516         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1517         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1518         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1519         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1520         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1521         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1522         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1523         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1524         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1525         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1526         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1527         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1528 };
1529
1530 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1531         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1532         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1533                                     .len = PORT_PROFILE_MAX },
1534         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1535                                       .len = PORT_UUID_MAX },
1536         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1537                                     .len = PORT_UUID_MAX },
1538         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1539         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1540
1541         /* Unused, but we need to keep it here since user space could
1542          * fill it. It's also broken with regard to NLA_BINARY use in
1543          * combination with structs.
1544          */
1545         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1546                                     .len = sizeof(struct ifla_port_vsi) },
1547 };
1548
1549 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1550         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1551         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1552         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1553         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1554 };
1555
1556 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1557 {
1558         const struct rtnl_link_ops *ops = NULL;
1559         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1560
1561         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1562                              ifla_info_policy, NULL) < 0)
1563                 return NULL;
1564
1565         if (linfo[IFLA_INFO_KIND]) {
1566                 char kind[MODULE_NAME_LEN];
1567
1568                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1569                 ops = rtnl_link_ops_get(kind);
1570         }
1571
1572         return ops;
1573 }
1574
1575 static bool link_master_filtered(struct net_device *dev, int master_idx)
1576 {
1577         struct net_device *master;
1578
1579         if (!master_idx)
1580                 return false;
1581
1582         master = netdev_master_upper_dev_get(dev);
1583         if (!master || master->ifindex != master_idx)
1584                 return true;
1585
1586         return false;
1587 }
1588
1589 static bool link_kind_filtered(const struct net_device *dev,
1590                                const struct rtnl_link_ops *kind_ops)
1591 {
1592         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1593                 return true;
1594
1595         return false;
1596 }
1597
1598 static bool link_dump_filtered(struct net_device *dev,
1599                                int master_idx,
1600                                const struct rtnl_link_ops *kind_ops)
1601 {
1602         if (link_master_filtered(dev, master_idx) ||
1603             link_kind_filtered(dev, kind_ops))
1604                 return true;
1605
1606         return false;
1607 }
1608
1609 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1610 {
1611         struct net *net = sock_net(skb->sk);
1612         int h, s_h;
1613         int idx = 0, s_idx;
1614         struct net_device *dev;
1615         struct hlist_head *head;
1616         struct nlattr *tb[IFLA_MAX+1];
1617         u32 ext_filter_mask = 0;
1618         const struct rtnl_link_ops *kind_ops = NULL;
1619         unsigned int flags = NLM_F_MULTI;
1620         int master_idx = 0;
1621         int err;
1622         int hdrlen;
1623
1624         s_h = cb->args[0];
1625         s_idx = cb->args[1];
1626
1627         /* A hack to preserve kernel<->userspace interface.
1628          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1629          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1630          * what iproute2 < v3.9.0 used.
1631          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1632          * attribute, its netlink message is shorter than struct ifinfomsg.
1633          */
1634         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1635                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1636
1637         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1638                         ifla_policy, NULL) >= 0) {
1639                 if (tb[IFLA_EXT_MASK])
1640                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1641
1642                 if (tb[IFLA_MASTER])
1643                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1644
1645                 if (tb[IFLA_LINKINFO])
1646                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1647
1648                 if (master_idx || kind_ops)
1649                         flags |= NLM_F_DUMP_FILTERED;
1650         }
1651
1652         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1653                 idx = 0;
1654                 head = &net->dev_index_head[h];
1655                 hlist_for_each_entry(dev, head, index_hlist) {
1656                         if (link_dump_filtered(dev, master_idx, kind_ops))
1657                                 goto cont;
1658                         if (idx < s_idx)
1659                                 goto cont;
1660                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1661                                                NETLINK_CB(cb->skb).portid,
1662                                                cb->nlh->nlmsg_seq, 0,
1663                                                flags,
1664                                                ext_filter_mask, 0);
1665
1666                         if (err < 0) {
1667                                 if (likely(skb->len))
1668                                         goto out;
1669
1670                                 goto out_err;
1671                         }
1672 cont:
1673                         idx++;
1674                 }
1675         }
1676 out:
1677         err = skb->len;
1678 out_err:
1679         cb->args[1] = idx;
1680         cb->args[0] = h;
1681         cb->seq = net->dev_base_seq;
1682         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1683
1684         return err;
1685 }
1686
1687 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1688                         struct netlink_ext_ack *exterr)
1689 {
1690         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1691 }
1692 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1693
1694 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1695 {
1696         struct net *net;
1697         /* Examine the link attributes and figure out which
1698          * network namespace we are talking about.
1699          */
1700         if (tb[IFLA_NET_NS_PID])
1701                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1702         else if (tb[IFLA_NET_NS_FD])
1703                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1704         else
1705                 net = get_net(src_net);
1706         return net;
1707 }
1708 EXPORT_SYMBOL(rtnl_link_get_net);
1709
1710 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1711 {
1712         if (dev) {
1713                 if (tb[IFLA_ADDRESS] &&
1714                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1715                         return -EINVAL;
1716
1717                 if (tb[IFLA_BROADCAST] &&
1718                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1719                         return -EINVAL;
1720         }
1721
1722         if (tb[IFLA_AF_SPEC]) {
1723                 struct nlattr *af;
1724                 int rem, err;
1725
1726                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1727                         const struct rtnl_af_ops *af_ops;
1728
1729                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1730                                 return -EAFNOSUPPORT;
1731
1732                         if (!af_ops->set_link_af)
1733                                 return -EOPNOTSUPP;
1734
1735                         if (af_ops->validate_link_af) {
1736                                 err = af_ops->validate_link_af(dev, af);
1737                                 if (err < 0)
1738                                         return err;
1739                         }
1740                 }
1741         }
1742
1743         return 0;
1744 }
1745
1746 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1747                                   int guid_type)
1748 {
1749         const struct net_device_ops *ops = dev->netdev_ops;
1750
1751         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1752 }
1753
1754 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1755 {
1756         if (dev->type != ARPHRD_INFINIBAND)
1757                 return -EOPNOTSUPP;
1758
1759         return handle_infiniband_guid(dev, ivt, guid_type);
1760 }
1761
1762 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1763 {
1764         const struct net_device_ops *ops = dev->netdev_ops;
1765         int err = -EINVAL;
1766
1767         if (tb[IFLA_VF_MAC]) {
1768                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1769
1770                 if (ivm->vf >= INT_MAX)
1771                         return -EINVAL;
1772                 err = -EOPNOTSUPP;
1773                 if (ops->ndo_set_vf_mac)
1774                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1775                                                   ivm->mac);
1776                 if (err < 0)
1777                         return err;
1778         }
1779
1780         if (tb[IFLA_VF_VLAN]) {
1781                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1782
1783                 if (ivv->vf >= INT_MAX)
1784                         return -EINVAL;
1785                 err = -EOPNOTSUPP;
1786                 if (ops->ndo_set_vf_vlan)
1787                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1788                                                    ivv->qos,
1789                                                    htons(ETH_P_8021Q));
1790                 if (err < 0)
1791                         return err;
1792         }
1793
1794         if (tb[IFLA_VF_VLAN_LIST]) {
1795                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1796                 struct nlattr *attr;
1797                 int rem, len = 0;
1798
1799                 err = -EOPNOTSUPP;
1800                 if (!ops->ndo_set_vf_vlan)
1801                         return err;
1802
1803                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1804                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1805                             nla_len(attr) < NLA_HDRLEN) {
1806                                 return -EINVAL;
1807                         }
1808                         if (len >= MAX_VLAN_LIST_LEN)
1809                                 return -EOPNOTSUPP;
1810                         ivvl[len] = nla_data(attr);
1811
1812                         len++;
1813                 }
1814                 if (len == 0)
1815                         return -EINVAL;
1816
1817                 if (ivvl[0]->vf >= INT_MAX)
1818                         return -EINVAL;
1819                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1820                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1821                 if (err < 0)
1822                         return err;
1823         }
1824
1825         if (tb[IFLA_VF_TX_RATE]) {
1826                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1827                 struct ifla_vf_info ivf;
1828
1829                 if (ivt->vf >= INT_MAX)
1830                         return -EINVAL;
1831                 err = -EOPNOTSUPP;
1832                 if (ops->ndo_get_vf_config)
1833                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1834                 if (err < 0)
1835                         return err;
1836
1837                 err = -EOPNOTSUPP;
1838                 if (ops->ndo_set_vf_rate)
1839                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1840                                                    ivf.min_tx_rate,
1841                                                    ivt->rate);
1842                 if (err < 0)
1843                         return err;
1844         }
1845
1846         if (tb[IFLA_VF_RATE]) {
1847                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1848
1849                 if (ivt->vf >= INT_MAX)
1850                         return -EINVAL;
1851                 err = -EOPNOTSUPP;
1852                 if (ops->ndo_set_vf_rate)
1853                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1854                                                    ivt->min_tx_rate,
1855                                                    ivt->max_tx_rate);
1856                 if (err < 0)
1857                         return err;
1858         }
1859
1860         if (tb[IFLA_VF_SPOOFCHK]) {
1861                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1862
1863                 if (ivs->vf >= INT_MAX)
1864                         return -EINVAL;
1865                 err = -EOPNOTSUPP;
1866                 if (ops->ndo_set_vf_spoofchk)
1867                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1868                                                        ivs->setting);
1869                 if (err < 0)
1870                         return err;
1871         }
1872
1873         if (tb[IFLA_VF_LINK_STATE]) {
1874                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1875
1876                 if (ivl->vf >= INT_MAX)
1877                         return -EINVAL;
1878                 err = -EOPNOTSUPP;
1879                 if (ops->ndo_set_vf_link_state)
1880                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1881                                                          ivl->link_state);
1882                 if (err < 0)
1883                         return err;
1884         }
1885
1886         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1887                 struct ifla_vf_rss_query_en *ivrssq_en;
1888
1889                 err = -EOPNOTSUPP;
1890                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1891                 if (ivrssq_en->vf >= INT_MAX)
1892                         return -EINVAL;
1893                 if (ops->ndo_set_vf_rss_query_en)
1894                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1895                                                            ivrssq_en->setting);
1896                 if (err < 0)
1897                         return err;
1898         }
1899
1900         if (tb[IFLA_VF_TRUST]) {
1901                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1902
1903                 if (ivt->vf >= INT_MAX)
1904                         return -EINVAL;
1905                 err = -EOPNOTSUPP;
1906                 if (ops->ndo_set_vf_trust)
1907                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1908                 if (err < 0)
1909                         return err;
1910         }
1911
1912         if (tb[IFLA_VF_IB_NODE_GUID]) {
1913                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1914
1915                 if (ivt->vf >= INT_MAX)
1916                         return -EINVAL;
1917                 if (!ops->ndo_set_vf_guid)
1918                         return -EOPNOTSUPP;
1919                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1920         }
1921
1922         if (tb[IFLA_VF_IB_PORT_GUID]) {
1923                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1924
1925                 if (ivt->vf >= INT_MAX)
1926                         return -EINVAL;
1927                 if (!ops->ndo_set_vf_guid)
1928                         return -EOPNOTSUPP;
1929
1930                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1931         }
1932
1933         return err;
1934 }
1935
1936 static int do_set_master(struct net_device *dev, int ifindex)
1937 {
1938         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1939         const struct net_device_ops *ops;
1940         int err;
1941
1942         if (upper_dev) {
1943                 if (upper_dev->ifindex == ifindex)
1944                         return 0;
1945                 ops = upper_dev->netdev_ops;
1946                 if (ops->ndo_del_slave) {
1947                         err = ops->ndo_del_slave(upper_dev, dev);
1948                         if (err)
1949                                 return err;
1950                 } else {
1951                         return -EOPNOTSUPP;
1952                 }
1953         }
1954
1955         if (ifindex) {
1956                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1957                 if (!upper_dev)
1958                         return -EINVAL;
1959                 ops = upper_dev->netdev_ops;
1960                 if (ops->ndo_add_slave) {
1961                         err = ops->ndo_add_slave(upper_dev, dev);
1962                         if (err)
1963                                 return err;
1964                 } else {
1965                         return -EOPNOTSUPP;
1966                 }
1967         }
1968         return 0;
1969 }
1970
1971 #define DO_SETLINK_MODIFIED     0x01
1972 /* notify flag means notify + modified. */
1973 #define DO_SETLINK_NOTIFY       0x03
1974 static int do_setlink(const struct sk_buff *skb,
1975                       struct net_device *dev, struct ifinfomsg *ifm,
1976                       struct netlink_ext_ack *extack,
1977                       struct nlattr **tb, char *ifname, int status)
1978 {
1979         const struct net_device_ops *ops = dev->netdev_ops;
1980         int err;
1981
1982         err = validate_linkmsg(dev, tb);
1983         if (err < 0)
1984                 return err;
1985
1986         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1987                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1988                 if (IS_ERR(net)) {
1989                         err = PTR_ERR(net);
1990                         goto errout;
1991                 }
1992                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1993                         put_net(net);
1994                         err = -EPERM;
1995                         goto errout;
1996                 }
1997                 err = dev_change_net_namespace(dev, net, ifname);
1998                 put_net(net);
1999                 if (err)
2000                         goto errout;
2001                 status |= DO_SETLINK_MODIFIED;
2002         }
2003
2004         if (tb[IFLA_MAP]) {
2005                 struct rtnl_link_ifmap *u_map;
2006                 struct ifmap k_map;
2007
2008                 if (!ops->ndo_set_config) {
2009                         err = -EOPNOTSUPP;
2010                         goto errout;
2011                 }
2012
2013                 if (!netif_device_present(dev)) {
2014                         err = -ENODEV;
2015                         goto errout;
2016                 }
2017
2018                 u_map = nla_data(tb[IFLA_MAP]);
2019                 k_map.mem_start = (unsigned long) u_map->mem_start;
2020                 k_map.mem_end = (unsigned long) u_map->mem_end;
2021                 k_map.base_addr = (unsigned short) u_map->base_addr;
2022                 k_map.irq = (unsigned char) u_map->irq;
2023                 k_map.dma = (unsigned char) u_map->dma;
2024                 k_map.port = (unsigned char) u_map->port;
2025
2026                 err = ops->ndo_set_config(dev, &k_map);
2027                 if (err < 0)
2028                         goto errout;
2029
2030                 status |= DO_SETLINK_NOTIFY;
2031         }
2032
2033         if (tb[IFLA_ADDRESS]) {
2034                 struct sockaddr *sa;
2035                 int len;
2036
2037                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2038                                                   sizeof(*sa));
2039                 sa = kmalloc(len, GFP_KERNEL);
2040                 if (!sa) {
2041                         err = -ENOMEM;
2042                         goto errout;
2043                 }
2044                 sa->sa_family = dev->type;
2045                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2046                        dev->addr_len);
2047                 err = dev_set_mac_address(dev, sa);
2048                 kfree(sa);
2049                 if (err)
2050                         goto errout;
2051                 status |= DO_SETLINK_MODIFIED;
2052         }
2053
2054         if (tb[IFLA_MTU]) {
2055                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2056                 if (err < 0)
2057                         goto errout;
2058                 status |= DO_SETLINK_MODIFIED;
2059         }
2060
2061         if (tb[IFLA_GROUP]) {
2062                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2063                 status |= DO_SETLINK_NOTIFY;
2064         }
2065
2066         /*
2067          * Interface selected by interface index but interface
2068          * name provided implies that a name change has been
2069          * requested.
2070          */
2071         if (ifm->ifi_index > 0 && ifname[0]) {
2072                 err = dev_change_name(dev, ifname);
2073                 if (err < 0)
2074                         goto errout;
2075                 status |= DO_SETLINK_MODIFIED;
2076         }
2077
2078         if (tb[IFLA_IFALIAS]) {
2079                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2080                                     nla_len(tb[IFLA_IFALIAS]));
2081                 if (err < 0)
2082                         goto errout;
2083                 status |= DO_SETLINK_NOTIFY;
2084         }
2085
2086         if (tb[IFLA_BROADCAST]) {
2087                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2088                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2089         }
2090
2091         if (ifm->ifi_flags || ifm->ifi_change) {
2092                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2093                 if (err < 0)
2094                         goto errout;
2095         }
2096
2097         if (tb[IFLA_MASTER]) {
2098                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2099                 if (err)
2100                         goto errout;
2101                 status |= DO_SETLINK_MODIFIED;
2102         }
2103
2104         if (tb[IFLA_CARRIER]) {
2105                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2106                 if (err)
2107                         goto errout;
2108                 status |= DO_SETLINK_MODIFIED;
2109         }
2110
2111         if (tb[IFLA_TXQLEN]) {
2112                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2113                 unsigned int orig_len = dev->tx_queue_len;
2114
2115                 if (dev->tx_queue_len ^ value) {
2116                         dev->tx_queue_len = value;
2117                         err = call_netdevice_notifiers(
2118                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2119                         err = notifier_to_errno(err);
2120                         if (err) {
2121                                 dev->tx_queue_len = orig_len;
2122                                 goto errout;
2123                         }
2124                         status |= DO_SETLINK_MODIFIED;
2125                 }
2126         }
2127
2128         if (tb[IFLA_OPERSTATE])
2129                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2130
2131         if (tb[IFLA_LINKMODE]) {
2132                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2133
2134                 write_lock_bh(&dev_base_lock);
2135                 if (dev->link_mode ^ value)
2136                         status |= DO_SETLINK_NOTIFY;
2137                 dev->link_mode = value;
2138                 write_unlock_bh(&dev_base_lock);
2139         }
2140
2141         if (tb[IFLA_VFINFO_LIST]) {
2142                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2143                 struct nlattr *attr;
2144                 int rem;
2145
2146                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2147                         if (nla_type(attr) != IFLA_VF_INFO ||
2148                             nla_len(attr) < NLA_HDRLEN) {
2149                                 err = -EINVAL;
2150                                 goto errout;
2151                         }
2152                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2153                                                ifla_vf_policy, NULL);
2154                         if (err < 0)
2155                                 goto errout;
2156                         err = do_setvfinfo(dev, vfinfo);
2157                         if (err < 0)
2158                                 goto errout;
2159                         status |= DO_SETLINK_NOTIFY;
2160                 }
2161         }
2162         err = 0;
2163
2164         if (tb[IFLA_VF_PORTS]) {
2165                 struct nlattr *port[IFLA_PORT_MAX+1];
2166                 struct nlattr *attr;
2167                 int vf;
2168                 int rem;
2169
2170                 err = -EOPNOTSUPP;
2171                 if (!ops->ndo_set_vf_port)
2172                         goto errout;
2173
2174                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2175                         if (nla_type(attr) != IFLA_VF_PORT ||
2176                             nla_len(attr) < NLA_HDRLEN) {
2177                                 err = -EINVAL;
2178                                 goto errout;
2179                         }
2180                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2181                                                ifla_port_policy, NULL);
2182                         if (err < 0)
2183                                 goto errout;
2184                         if (!port[IFLA_PORT_VF]) {
2185                                 err = -EOPNOTSUPP;
2186                                 goto errout;
2187                         }
2188                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2189                         err = ops->ndo_set_vf_port(dev, vf, port);
2190                         if (err < 0)
2191                                 goto errout;
2192                         status |= DO_SETLINK_NOTIFY;
2193                 }
2194         }
2195         err = 0;
2196
2197         if (tb[IFLA_PORT_SELF]) {
2198                 struct nlattr *port[IFLA_PORT_MAX+1];
2199
2200                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2201                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2202                                        NULL);
2203                 if (err < 0)
2204                         goto errout;
2205
2206                 err = -EOPNOTSUPP;
2207                 if (ops->ndo_set_vf_port)
2208                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2209                 if (err < 0)
2210                         goto errout;
2211                 status |= DO_SETLINK_NOTIFY;
2212         }
2213
2214         if (tb[IFLA_AF_SPEC]) {
2215                 struct nlattr *af;
2216                 int rem;
2217
2218                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2219                         const struct rtnl_af_ops *af_ops;
2220
2221                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2222                                 BUG();
2223
2224                         err = af_ops->set_link_af(dev, af);
2225                         if (err < 0)
2226                                 goto errout;
2227
2228                         status |= DO_SETLINK_NOTIFY;
2229                 }
2230         }
2231         err = 0;
2232
2233         if (tb[IFLA_PROTO_DOWN]) {
2234                 err = dev_change_proto_down(dev,
2235                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2236                 if (err)
2237                         goto errout;
2238                 status |= DO_SETLINK_NOTIFY;
2239         }
2240
2241         if (tb[IFLA_XDP]) {
2242                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2243                 u32 xdp_flags = 0;
2244
2245                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2246                                        ifla_xdp_policy, NULL);
2247                 if (err < 0)
2248                         goto errout;
2249
2250                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2251                         err = -EINVAL;
2252                         goto errout;
2253                 }
2254
2255                 if (xdp[IFLA_XDP_FLAGS]) {
2256                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2257                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2258                                 err = -EINVAL;
2259                                 goto errout;
2260                         }
2261                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2262                                 err = -EINVAL;
2263                                 goto errout;
2264                         }
2265                 }
2266
2267                 if (xdp[IFLA_XDP_FD]) {
2268                         err = dev_change_xdp_fd(dev, extack,
2269                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2270                                                 xdp_flags);
2271                         if (err)
2272                                 goto errout;
2273                         status |= DO_SETLINK_NOTIFY;
2274                 }
2275         }
2276
2277 errout:
2278         if (status & DO_SETLINK_MODIFIED) {
2279                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2280                         netdev_state_change(dev);
2281
2282                 if (err < 0)
2283                         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",
2284                                              dev->name);
2285         }
2286
2287         return err;
2288 }
2289
2290 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2291                         struct netlink_ext_ack *extack)
2292 {
2293         struct net *net = sock_net(skb->sk);
2294         struct ifinfomsg *ifm;
2295         struct net_device *dev;
2296         int err;
2297         struct nlattr *tb[IFLA_MAX+1];
2298         char ifname[IFNAMSIZ];
2299
2300         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2301                           extack);
2302         if (err < 0)
2303                 goto errout;
2304
2305         if (tb[IFLA_IFNAME])
2306                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2307         else
2308                 ifname[0] = '\0';
2309
2310         err = -EINVAL;
2311         ifm = nlmsg_data(nlh);
2312         if (ifm->ifi_index > 0)
2313                 dev = __dev_get_by_index(net, ifm->ifi_index);
2314         else if (tb[IFLA_IFNAME])
2315                 dev = __dev_get_by_name(net, ifname);
2316         else
2317                 goto errout;
2318
2319         if (dev == NULL) {
2320                 err = -ENODEV;
2321                 goto errout;
2322         }
2323
2324         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2325 errout:
2326         return err;
2327 }
2328
2329 static int rtnl_group_dellink(const struct net *net, int group)
2330 {
2331         struct net_device *dev, *aux;
2332         LIST_HEAD(list_kill);
2333         bool found = false;
2334
2335         if (!group)
2336                 return -EPERM;
2337
2338         for_each_netdev(net, dev) {
2339                 if (dev->group == group) {
2340                         const struct rtnl_link_ops *ops;
2341
2342                         found = true;
2343                         ops = dev->rtnl_link_ops;
2344                         if (!ops || !ops->dellink)
2345                                 return -EOPNOTSUPP;
2346                 }
2347         }
2348
2349         if (!found)
2350                 return -ENODEV;
2351
2352         for_each_netdev_safe(net, dev, aux) {
2353                 if (dev->group == group) {
2354                         const struct rtnl_link_ops *ops;
2355
2356                         ops = dev->rtnl_link_ops;
2357                         ops->dellink(dev, &list_kill);
2358                 }
2359         }
2360         unregister_netdevice_many(&list_kill);
2361
2362         return 0;
2363 }
2364
2365 int rtnl_delete_link(struct net_device *dev)
2366 {
2367         const struct rtnl_link_ops *ops;
2368         LIST_HEAD(list_kill);
2369
2370         ops = dev->rtnl_link_ops;
2371         if (!ops || !ops->dellink)
2372                 return -EOPNOTSUPP;
2373
2374         ops->dellink(dev, &list_kill);
2375         unregister_netdevice_many(&list_kill);
2376
2377         return 0;
2378 }
2379 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2380
2381 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2382                         struct netlink_ext_ack *extack)
2383 {
2384         struct net *net = sock_net(skb->sk);
2385         struct net_device *dev;
2386         struct ifinfomsg *ifm;
2387         char ifname[IFNAMSIZ];
2388         struct nlattr *tb[IFLA_MAX+1];
2389         int err;
2390
2391         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2392         if (err < 0)
2393                 return err;
2394
2395         if (tb[IFLA_IFNAME])
2396                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2397
2398         ifm = nlmsg_data(nlh);
2399         if (ifm->ifi_index > 0)
2400                 dev = __dev_get_by_index(net, ifm->ifi_index);
2401         else if (tb[IFLA_IFNAME])
2402                 dev = __dev_get_by_name(net, ifname);
2403         else if (tb[IFLA_GROUP])
2404                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2405         else
2406                 return -EINVAL;
2407
2408         if (!dev)
2409                 return -ENODEV;
2410
2411         return rtnl_delete_link(dev);
2412 }
2413
2414 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2415 {
2416         unsigned int old_flags;
2417         int err;
2418
2419         old_flags = dev->flags;
2420         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2421                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2422                 if (err < 0)
2423                         return err;
2424         }
2425
2426         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2427                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2428         } else {
2429                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2430                 __dev_notify_flags(dev, old_flags, ~0U);
2431         }
2432         return 0;
2433 }
2434 EXPORT_SYMBOL(rtnl_configure_link);
2435
2436 struct net_device *rtnl_create_link(struct net *net,
2437         const char *ifname, unsigned char name_assign_type,
2438         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2439 {
2440         struct net_device *dev;
2441         unsigned int num_tx_queues = 1;
2442         unsigned int num_rx_queues = 1;
2443
2444         if (tb[IFLA_NUM_TX_QUEUES])
2445                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2446         else if (ops->get_num_tx_queues)
2447                 num_tx_queues = ops->get_num_tx_queues();
2448
2449         if (tb[IFLA_NUM_RX_QUEUES])
2450                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2451         else if (ops->get_num_rx_queues)
2452                 num_rx_queues = ops->get_num_rx_queues();
2453
2454         if (num_tx_queues < 1 || num_tx_queues > 4096)
2455                 return ERR_PTR(-EINVAL);
2456
2457         if (num_rx_queues < 1 || num_rx_queues > 4096)
2458                 return ERR_PTR(-EINVAL);
2459
2460         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2461                                ops->setup, num_tx_queues, num_rx_queues);
2462         if (!dev)
2463                 return ERR_PTR(-ENOMEM);
2464
2465         dev_net_set(dev, net);
2466         dev->rtnl_link_ops = ops;
2467         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2468
2469         if (tb[IFLA_MTU]) {
2470                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2471                 int err;
2472
2473                 err = dev_validate_mtu(dev, mtu);
2474                 if (err) {
2475                         free_netdev(dev);
2476                         return ERR_PTR(err);
2477                 }
2478                 dev->mtu = mtu;
2479         }
2480         if (tb[IFLA_ADDRESS]) {
2481                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2482                                 nla_len(tb[IFLA_ADDRESS]));
2483                 dev->addr_assign_type = NET_ADDR_SET;
2484         }
2485         if (tb[IFLA_BROADCAST])
2486                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2487                                 nla_len(tb[IFLA_BROADCAST]));
2488         if (tb[IFLA_TXQLEN])
2489                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2490         if (tb[IFLA_OPERSTATE])
2491                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2492         if (tb[IFLA_LINKMODE])
2493                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2494         if (tb[IFLA_GROUP])
2495                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2496
2497         return dev;
2498 }
2499 EXPORT_SYMBOL(rtnl_create_link);
2500
2501 static int rtnl_group_changelink(const struct sk_buff *skb,
2502                 struct net *net, int group,
2503                 struct ifinfomsg *ifm,
2504                 struct netlink_ext_ack *extack,
2505                 struct nlattr **tb)
2506 {
2507         struct net_device *dev, *aux;
2508         int err;
2509
2510         for_each_netdev_safe(net, dev, aux) {
2511                 if (dev->group == group) {
2512                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2513                         if (err < 0)
2514                                 return err;
2515                 }
2516         }
2517
2518         return 0;
2519 }
2520
2521 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2522                         struct netlink_ext_ack *extack)
2523 {
2524         struct net *net = sock_net(skb->sk);
2525         const struct rtnl_link_ops *ops;
2526         const struct rtnl_link_ops *m_ops;
2527         struct net_device *dev;
2528         struct net_device *master_dev;
2529         struct ifinfomsg *ifm;
2530         char kind[MODULE_NAME_LEN];
2531         char ifname[IFNAMSIZ];
2532         struct nlattr *tb[IFLA_MAX+1];
2533         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2534         unsigned char name_assign_type = NET_NAME_USER;
2535         int err;
2536
2537 #ifdef CONFIG_MODULES
2538 replay:
2539 #endif
2540         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2541         if (err < 0)
2542                 return err;
2543
2544         if (tb[IFLA_IFNAME])
2545                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2546         else
2547                 ifname[0] = '\0';
2548
2549         ifm = nlmsg_data(nlh);
2550         if (ifm->ifi_index > 0)
2551                 dev = __dev_get_by_index(net, ifm->ifi_index);
2552         else {
2553                 if (ifname[0])
2554                         dev = __dev_get_by_name(net, ifname);
2555                 else
2556                         dev = NULL;
2557         }
2558
2559         master_dev = NULL;
2560         m_ops = NULL;
2561         if (dev) {
2562                 master_dev = netdev_master_upper_dev_get(dev);
2563                 if (master_dev)
2564                         m_ops = master_dev->rtnl_link_ops;
2565         }
2566
2567         err = validate_linkmsg(dev, tb);
2568         if (err < 0)
2569                 return err;
2570
2571         if (tb[IFLA_LINKINFO]) {
2572                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2573                                        tb[IFLA_LINKINFO], ifla_info_policy,
2574                                        NULL);
2575                 if (err < 0)
2576                         return err;
2577         } else
2578                 memset(linkinfo, 0, sizeof(linkinfo));
2579
2580         if (linkinfo[IFLA_INFO_KIND]) {
2581                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2582                 ops = rtnl_link_ops_get(kind);
2583         } else {
2584                 kind[0] = '\0';
2585                 ops = NULL;
2586         }
2587
2588         if (1) {
2589                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2590                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2591                 struct nlattr **data = NULL;
2592                 struct nlattr **slave_data = NULL;
2593                 struct net *dest_net, *link_net = NULL;
2594
2595                 if (ops) {
2596                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2597                                 err = nla_parse_nested(attr, ops->maxtype,
2598                                                        linkinfo[IFLA_INFO_DATA],
2599                                                        ops->policy, NULL);
2600                                 if (err < 0)
2601                                         return err;
2602                                 data = attr;
2603                         }
2604                         if (ops->validate) {
2605                                 err = ops->validate(tb, data, extack);
2606                                 if (err < 0)
2607                                         return err;
2608                         }
2609                 }
2610
2611                 if (m_ops) {
2612                         if (m_ops->slave_maxtype &&
2613                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2614                                 err = nla_parse_nested(slave_attr,
2615                                                        m_ops->slave_maxtype,
2616                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2617                                                        m_ops->slave_policy,
2618                                                        NULL);
2619                                 if (err < 0)
2620                                         return err;
2621                                 slave_data = slave_attr;
2622                         }
2623                         if (m_ops->slave_validate) {
2624                                 err = m_ops->slave_validate(tb, slave_data,
2625                                                             extack);
2626                                 if (err < 0)
2627                                         return err;
2628                         }
2629                 }
2630
2631                 if (dev) {
2632                         int status = 0;
2633
2634                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2635                                 return -EEXIST;
2636                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2637                                 return -EOPNOTSUPP;
2638
2639                         if (linkinfo[IFLA_INFO_DATA]) {
2640                                 if (!ops || ops != dev->rtnl_link_ops ||
2641                                     !ops->changelink)
2642                                         return -EOPNOTSUPP;
2643
2644                                 err = ops->changelink(dev, tb, data, extack);
2645                                 if (err < 0)
2646                                         return err;
2647                                 status |= DO_SETLINK_NOTIFY;
2648                         }
2649
2650                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2651                                 if (!m_ops || !m_ops->slave_changelink)
2652                                         return -EOPNOTSUPP;
2653
2654                                 err = m_ops->slave_changelink(master_dev, dev,
2655                                                               tb, slave_data,
2656                                                               extack);
2657                                 if (err < 0)
2658                                         return err;
2659                                 status |= DO_SETLINK_NOTIFY;
2660                         }
2661
2662                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2663                                           status);
2664                 }
2665
2666                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2667                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2668                                 return rtnl_group_changelink(skb, net,
2669                                                 nla_get_u32(tb[IFLA_GROUP]),
2670                                                 ifm, extack, tb);
2671                         return -ENODEV;
2672                 }
2673
2674                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2675                         return -EOPNOTSUPP;
2676
2677                 if (!ops) {
2678 #ifdef CONFIG_MODULES
2679                         if (kind[0]) {
2680                                 __rtnl_unlock();
2681                                 request_module("rtnl-link-%s", kind);
2682                                 rtnl_lock();
2683                                 ops = rtnl_link_ops_get(kind);
2684                                 if (ops)
2685                                         goto replay;
2686                         }
2687 #endif
2688                         return -EOPNOTSUPP;
2689                 }
2690
2691                 if (!ops->setup)
2692                         return -EOPNOTSUPP;
2693
2694                 if (!ifname[0]) {
2695                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2696                         name_assign_type = NET_NAME_ENUM;
2697                 }
2698
2699                 dest_net = rtnl_link_get_net(net, tb);
2700                 if (IS_ERR(dest_net))
2701                         return PTR_ERR(dest_net);
2702
2703                 err = -EPERM;
2704                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2705                         goto out;
2706
2707                 if (tb[IFLA_LINK_NETNSID]) {
2708                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2709
2710                         link_net = get_net_ns_by_id(dest_net, id);
2711                         if (!link_net) {
2712                                 err =  -EINVAL;
2713                                 goto out;
2714                         }
2715                         err = -EPERM;
2716                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2717                                 goto out;
2718                 }
2719
2720                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2721                                        name_assign_type, ops, tb);
2722                 if (IS_ERR(dev)) {
2723                         err = PTR_ERR(dev);
2724                         goto out;
2725                 }
2726
2727                 dev->ifindex = ifm->ifi_index;
2728
2729                 if (ops->newlink) {
2730                         err = ops->newlink(link_net ? : net, dev, tb, data,
2731                                            extack);
2732                         /* Drivers should call free_netdev() in ->destructor
2733                          * and unregister it on failure after registration
2734                          * so that device could be finally freed in rtnl_unlock.
2735                          */
2736                         if (err < 0) {
2737                                 /* If device is not registered at all, free it now */
2738                                 if (dev->reg_state == NETREG_UNINITIALIZED ||
2739                                     dev->reg_state == NETREG_UNREGISTERED)
2740                                         free_netdev(dev);
2741                                 goto out;
2742                         }
2743                 } else {
2744                         err = register_netdevice(dev);
2745                         if (err < 0) {
2746                                 free_netdev(dev);
2747                                 goto out;
2748                         }
2749                 }
2750                 err = rtnl_configure_link(dev, ifm);
2751                 if (err < 0)
2752                         goto out_unregister;
2753                 if (link_net) {
2754                         err = dev_change_net_namespace(dev, dest_net, ifname);
2755                         if (err < 0)
2756                                 goto out_unregister;
2757                 }
2758                 if (tb[IFLA_MASTER]) {
2759                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2760                         if (err)
2761                                 goto out_unregister;
2762                 }
2763 out:
2764                 if (link_net)
2765                         put_net(link_net);
2766                 put_net(dest_net);
2767                 return err;
2768 out_unregister:
2769                 if (ops->newlink) {
2770                         LIST_HEAD(list_kill);
2771
2772                         ops->dellink(dev, &list_kill);
2773                         unregister_netdevice_many(&list_kill);
2774                 } else {
2775                         unregister_netdevice(dev);
2776                 }
2777                 goto out;
2778         }
2779 }
2780
2781 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2782                         struct netlink_ext_ack *extack)
2783 {
2784         struct net *net = sock_net(skb->sk);
2785         struct ifinfomsg *ifm;
2786         char ifname[IFNAMSIZ];
2787         struct nlattr *tb[IFLA_MAX+1];
2788         struct net_device *dev = NULL;
2789         struct sk_buff *nskb;
2790         int err;
2791         u32 ext_filter_mask = 0;
2792
2793         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2794         if (err < 0)
2795                 return err;
2796
2797         if (tb[IFLA_IFNAME])
2798                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2799
2800         if (tb[IFLA_EXT_MASK])
2801                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2802
2803         ifm = nlmsg_data(nlh);
2804         if (ifm->ifi_index > 0)
2805                 dev = __dev_get_by_index(net, ifm->ifi_index);
2806         else if (tb[IFLA_IFNAME])
2807                 dev = __dev_get_by_name(net, ifname);
2808         else
2809                 return -EINVAL;
2810
2811         if (dev == NULL)
2812                 return -ENODEV;
2813
2814         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2815         if (nskb == NULL)
2816                 return -ENOBUFS;
2817
2818         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2819                                nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2820         if (err < 0) {
2821                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2822                 WARN_ON(err == -EMSGSIZE);
2823                 kfree_skb(nskb);
2824         } else
2825                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2826
2827         return err;
2828 }
2829
2830 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2831 {
2832         struct net *net = sock_net(skb->sk);
2833         struct net_device *dev;
2834         struct nlattr *tb[IFLA_MAX+1];
2835         u32 ext_filter_mask = 0;
2836         u16 min_ifinfo_dump_size = 0;
2837         int hdrlen;
2838
2839         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2840         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2841                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2842
2843         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2844                 if (tb[IFLA_EXT_MASK])
2845                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2846         }
2847
2848         if (!ext_filter_mask)
2849                 return NLMSG_GOODSIZE;
2850         /*
2851          * traverse the list of net devices and compute the minimum
2852          * buffer size based upon the filter mask.
2853          */
2854         rcu_read_lock();
2855         for_each_netdev_rcu(net, dev) {
2856                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2857                                              if_nlmsg_size(dev,
2858                                                            ext_filter_mask));
2859         }
2860         rcu_read_unlock();
2861
2862         return nlmsg_total_size(min_ifinfo_dump_size);
2863 }
2864
2865 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2866 {
2867         int idx;
2868         int s_idx = cb->family;
2869
2870         if (s_idx == 0)
2871                 s_idx = 1;
2872
2873         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2874                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2875                 struct rtnl_link *handlers;
2876                 rtnl_dumpit_func dumpit;
2877
2878                 if (idx < s_idx || idx == PF_PACKET)
2879                         continue;
2880
2881                 handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2882                 if (!handlers)
2883                         continue;
2884
2885                 dumpit = READ_ONCE(handlers[type].dumpit);
2886                 if (!dumpit)
2887                         continue;
2888
2889                 if (idx > s_idx) {
2890                         memset(&cb->args[0], 0, sizeof(cb->args));
2891                         cb->prev_seq = 0;
2892                         cb->seq = 0;
2893                 }
2894                 if (dumpit(skb, cb))
2895                         break;
2896         }
2897         cb->family = idx;
2898
2899         return skb->len;
2900 }
2901
2902 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2903                                        unsigned int change,
2904                                        u32 event, gfp_t flags)
2905 {
2906         struct net *net = dev_net(dev);
2907         struct sk_buff *skb;
2908         int err = -ENOBUFS;
2909         size_t if_info_size;
2910
2911         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2912         if (skb == NULL)
2913                 goto errout;
2914
2915         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2916         if (err < 0) {
2917                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2918                 WARN_ON(err == -EMSGSIZE);
2919                 kfree_skb(skb);
2920                 goto errout;
2921         }
2922         return skb;
2923 errout:
2924         if (err < 0)
2925                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2926         return NULL;
2927 }
2928
2929 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2930 {
2931         struct net *net = dev_net(dev);
2932
2933         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2934 }
2935
2936 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2937                                unsigned int change, u32 event,
2938                                gfp_t flags)
2939 {
2940         struct sk_buff *skb;
2941
2942         if (dev->reg_state != NETREG_REGISTERED)
2943                 return;
2944
2945         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2946         if (skb)
2947                 rtmsg_ifinfo_send(skb, dev, flags);
2948 }
2949
2950 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2951                   gfp_t flags)
2952 {
2953         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2954 }
2955 EXPORT_SYMBOL(rtmsg_ifinfo);
2956
2957 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2958                                    struct net_device *dev,
2959                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2960                                    int type, unsigned int flags,
2961                                    int nlflags, u16 ndm_state)
2962 {
2963         struct nlmsghdr *nlh;
2964         struct ndmsg *ndm;
2965
2966         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2967         if (!nlh)
2968                 return -EMSGSIZE;
2969
2970         ndm = nlmsg_data(nlh);
2971         ndm->ndm_family  = AF_BRIDGE;
2972         ndm->ndm_pad1    = 0;
2973         ndm->ndm_pad2    = 0;
2974         ndm->ndm_flags   = flags;
2975         ndm->ndm_type    = 0;
2976         ndm->ndm_ifindex = dev->ifindex;
2977         ndm->ndm_state   = ndm_state;
2978
2979         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2980                 goto nla_put_failure;
2981         if (vid)
2982                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2983                         goto nla_put_failure;
2984
2985         nlmsg_end(skb, nlh);
2986         return 0;
2987
2988 nla_put_failure:
2989         nlmsg_cancel(skb, nlh);
2990         return -EMSGSIZE;
2991 }
2992
2993 static inline size_t rtnl_fdb_nlmsg_size(void)
2994 {
2995         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2996                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2997                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2998                0;
2999 }
3000
3001 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3002                             u16 ndm_state)
3003 {
3004         struct net *net = dev_net(dev);
3005         struct sk_buff *skb;
3006         int err = -ENOBUFS;
3007
3008         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3009         if (!skb)
3010                 goto errout;
3011
3012         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3013                                       0, 0, type, NTF_SELF, 0, ndm_state);
3014         if (err < 0) {
3015                 kfree_skb(skb);
3016                 goto errout;
3017         }
3018
3019         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3020         return;
3021 errout:
3022         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3023 }
3024
3025 /**
3026  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3027  */
3028 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3029                      struct nlattr *tb[],
3030                      struct net_device *dev,
3031                      const unsigned char *addr, u16 vid,
3032                      u16 flags)
3033 {
3034         int err = -EINVAL;
3035
3036         /* If aging addresses are supported device will need to
3037          * implement its own handler for this.
3038          */
3039         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3040                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3041                 return err;
3042         }
3043
3044         if (vid) {
3045                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3046                 return err;
3047         }
3048
3049         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3050                 err = dev_uc_add_excl(dev, addr);
3051         else if (is_multicast_ether_addr(addr))
3052                 err = dev_mc_add_excl(dev, addr);
3053
3054         /* Only return duplicate errors if NLM_F_EXCL is set */
3055         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3056                 err = 0;
3057
3058         return err;
3059 }
3060 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3061
3062 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3063 {
3064         u16 vid = 0;
3065
3066         if (vlan_attr) {
3067                 if (nla_len(vlan_attr) != sizeof(u16)) {
3068                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3069                         return -EINVAL;
3070                 }
3071
3072                 vid = nla_get_u16(vlan_attr);
3073
3074                 if (!vid || vid >= VLAN_VID_MASK) {
3075                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3076                                 vid);
3077                         return -EINVAL;
3078                 }
3079         }
3080         *p_vid = vid;
3081         return 0;
3082 }
3083
3084 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3085                         struct netlink_ext_ack *extack)
3086 {
3087         struct net *net = sock_net(skb->sk);
3088         struct ndmsg *ndm;
3089         struct nlattr *tb[NDA_MAX+1];
3090         struct net_device *dev;
3091         u8 *addr;
3092         u16 vid;
3093         int err;
3094
3095         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3096         if (err < 0)
3097                 return err;
3098
3099         ndm = nlmsg_data(nlh);
3100         if (ndm->ndm_ifindex == 0) {
3101                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3102                 return -EINVAL;
3103         }
3104
3105         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3106         if (dev == NULL) {
3107                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3108                 return -ENODEV;
3109         }
3110
3111         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3112                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3113                 return -EINVAL;
3114         }
3115
3116         if (dev->type != ARPHRD_ETHER) {
3117                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3118                 return -EINVAL;
3119         }
3120
3121         addr = nla_data(tb[NDA_LLADDR]);
3122
3123         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3124         if (err)
3125                 return err;
3126
3127         err = -EOPNOTSUPP;
3128
3129         /* Support fdb on master device the net/bridge default case */
3130         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3131             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3132                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3133                 const struct net_device_ops *ops = br_dev->netdev_ops;
3134
3135                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3136                                        nlh->nlmsg_flags);
3137                 if (err)
3138                         goto out;
3139                 else
3140                         ndm->ndm_flags &= ~NTF_MASTER;
3141         }
3142
3143         /* Embedded bridge, macvlan, and any other device support */
3144         if ((ndm->ndm_flags & NTF_SELF)) {
3145                 if (dev->netdev_ops->ndo_fdb_add)
3146                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3147                                                            vid,
3148                                                            nlh->nlmsg_flags);
3149                 else
3150                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3151                                                nlh->nlmsg_flags);
3152
3153                 if (!err) {
3154                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3155                                         ndm->ndm_state);
3156                         ndm->ndm_flags &= ~NTF_SELF;
3157                 }
3158         }
3159 out:
3160         return err;
3161 }
3162
3163 /**
3164  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3165  */
3166 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3167                      struct nlattr *tb[],
3168                      struct net_device *dev,
3169                      const unsigned char *addr, u16 vid)
3170 {
3171         int err = -EINVAL;
3172
3173         /* If aging addresses are supported device will need to
3174          * implement its own handler for this.
3175          */
3176         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3177                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3178                 return err;
3179         }
3180
3181         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3182                 err = dev_uc_del(dev, addr);
3183         else if (is_multicast_ether_addr(addr))
3184                 err = dev_mc_del(dev, addr);
3185
3186         return err;
3187 }
3188 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3189
3190 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3191                         struct netlink_ext_ack *extack)
3192 {
3193         struct net *net = sock_net(skb->sk);
3194         struct ndmsg *ndm;
3195         struct nlattr *tb[NDA_MAX+1];
3196         struct net_device *dev;
3197         int err = -EINVAL;
3198         __u8 *addr;
3199         u16 vid;
3200
3201         if (!netlink_capable(skb, CAP_NET_ADMIN))
3202                 return -EPERM;
3203
3204         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3205         if (err < 0)
3206                 return err;
3207
3208         ndm = nlmsg_data(nlh);
3209         if (ndm->ndm_ifindex == 0) {
3210                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3211                 return -EINVAL;
3212         }
3213
3214         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3215         if (dev == NULL) {
3216                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3217                 return -ENODEV;
3218         }
3219
3220         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3221                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3222                 return -EINVAL;
3223         }
3224
3225         if (dev->type != ARPHRD_ETHER) {
3226                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3227                 return -EINVAL;
3228         }
3229
3230         addr = nla_data(tb[NDA_LLADDR]);
3231
3232         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3233         if (err)
3234                 return err;
3235
3236         err = -EOPNOTSUPP;
3237
3238         /* Support fdb on master device the net/bridge default case */
3239         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3240             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3241                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3242                 const struct net_device_ops *ops = br_dev->netdev_ops;
3243
3244                 if (ops->ndo_fdb_del)
3245                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3246
3247                 if (err)
3248                         goto out;
3249                 else
3250                         ndm->ndm_flags &= ~NTF_MASTER;
3251         }
3252
3253         /* Embedded bridge, macvlan, and any other device support */
3254         if (ndm->ndm_flags & NTF_SELF) {
3255                 if (dev->netdev_ops->ndo_fdb_del)
3256                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3257                                                            vid);
3258                 else
3259                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3260
3261                 if (!err) {
3262                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3263                                         ndm->ndm_state);
3264                         ndm->ndm_flags &= ~NTF_SELF;
3265                 }
3266         }
3267 out:
3268         return err;
3269 }
3270
3271 static int nlmsg_populate_fdb(struct sk_buff *skb,
3272                               struct netlink_callback *cb,
3273                               struct net_device *dev,
3274                               int *idx,
3275                               struct netdev_hw_addr_list *list)
3276 {
3277         struct netdev_hw_addr *ha;
3278         int err;
3279         u32 portid, seq;
3280
3281         portid = NETLINK_CB(cb->skb).portid;
3282         seq = cb->nlh->nlmsg_seq;
3283
3284         list_for_each_entry(ha, &list->list, list) {
3285                 if (*idx < cb->args[2])
3286                         goto skip;
3287
3288                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3289                                               portid, seq,
3290                                               RTM_NEWNEIGH, NTF_SELF,
3291                                               NLM_F_MULTI, NUD_PERMANENT);
3292                 if (err < 0)
3293                         return err;
3294 skip:
3295                 *idx += 1;
3296         }
3297         return 0;
3298 }
3299
3300 /**
3301  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3302  * @nlh: netlink message header
3303  * @dev: netdevice
3304  *
3305  * Default netdevice operation to dump the existing unicast address list.
3306  * Returns number of addresses from list put in skb.
3307  */
3308 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3309                       struct netlink_callback *cb,
3310                       struct net_device *dev,
3311                       struct net_device *filter_dev,
3312                       int *idx)
3313 {
3314         int err;
3315
3316         if (dev->type != ARPHRD_ETHER)
3317                 return -EINVAL;
3318
3319         netif_addr_lock_bh(dev);
3320         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3321         if (err)
3322                 goto out;
3323         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3324 out:
3325         netif_addr_unlock_bh(dev);
3326         return err;
3327 }
3328 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3329
3330 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3331 {
3332         struct net_device *dev;
3333         struct nlattr *tb[IFLA_MAX+1];
3334         struct net_device *br_dev = NULL;
3335         const struct net_device_ops *ops = NULL;
3336         const struct net_device_ops *cops = NULL;
3337         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3338         struct net *net = sock_net(skb->sk);
3339         struct hlist_head *head;
3340         int brport_idx = 0;
3341         int br_idx = 0;
3342         int h, s_h;
3343         int idx = 0, s_idx;
3344         int err = 0;
3345         int fidx = 0;
3346
3347         /* A hack to preserve kernel<->userspace interface.
3348          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3349          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3350          * So, check for ndmsg with an optional u32 attribute (not used here).
3351          * Fortunately these sizes don't conflict with the size of ifinfomsg
3352          * with an optional attribute.
3353          */
3354         if (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) &&
3355             (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) +
3356              nla_attr_size(sizeof(u32)))) {
3357                 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3358                                   IFLA_MAX, ifla_policy, NULL);
3359                 if (err < 0) {
3360                         return -EINVAL;
3361                 } else if (err == 0) {
3362                         if (tb[IFLA_MASTER])
3363                                 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3364                 }
3365
3366                 brport_idx = ifm->ifi_index;
3367         }
3368
3369         if (br_idx) {
3370                 br_dev = __dev_get_by_index(net, br_idx);
3371                 if (!br_dev)
3372                         return -ENODEV;
3373
3374                 ops = br_dev->netdev_ops;
3375         }
3376
3377         s_h = cb->args[0];
3378         s_idx = cb->args[1];
3379
3380         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3381                 idx = 0;
3382                 head = &net->dev_index_head[h];
3383                 hlist_for_each_entry(dev, head, index_hlist) {
3384
3385                         if (brport_idx && (dev->ifindex != brport_idx))
3386                                 continue;
3387
3388                         if (!br_idx) { /* user did not specify a specific bridge */
3389                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3390                                         br_dev = netdev_master_upper_dev_get(dev);
3391                                         cops = br_dev->netdev_ops;
3392                                 }
3393                         } else {
3394                                 if (dev != br_dev &&
3395                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3396                                         continue;
3397
3398                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3399                                     !(dev->priv_flags & IFF_EBRIDGE))
3400                                         continue;
3401                                 cops = ops;
3402                         }
3403
3404                         if (idx < s_idx)
3405                                 goto cont;
3406
3407                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3408                                 if (cops && cops->ndo_fdb_dump) {
3409                                         err = cops->ndo_fdb_dump(skb, cb,
3410                                                                 br_dev, dev,
3411                                                                 &fidx);
3412                                         if (err == -EMSGSIZE)
3413                                                 goto out;
3414                                 }
3415                         }
3416
3417                         if (dev->netdev_ops->ndo_fdb_dump)
3418                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3419                                                                     dev, NULL,
3420                                                                     &fidx);
3421                         else
3422                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3423                                                         &fidx);
3424                         if (err == -EMSGSIZE)
3425                                 goto out;
3426
3427                         cops = NULL;
3428
3429                         /* reset fdb offset to 0 for rest of the interfaces */
3430                         cb->args[2] = 0;
3431                         fidx = 0;
3432 cont:
3433                         idx++;
3434                 }
3435         }
3436
3437 out:
3438         cb->args[0] = h;
3439         cb->args[1] = idx;
3440         cb->args[2] = fidx;
3441
3442         return skb->len;
3443 }
3444
3445 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3446                                unsigned int attrnum, unsigned int flag)
3447 {
3448         if (mask & flag)
3449                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3450         return 0;
3451 }
3452
3453 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3454                             struct net_device *dev, u16 mode,
3455                             u32 flags, u32 mask, int nlflags,
3456                             u32 filter_mask,
3457                             int (*vlan_fill)(struct sk_buff *skb,
3458                                              struct net_device *dev,
3459                                              u32 filter_mask))
3460 {
3461         struct nlmsghdr *nlh;
3462         struct ifinfomsg *ifm;
3463         struct nlattr *br_afspec;
3464         struct nlattr *protinfo;
3465         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3466         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3467         int err = 0;
3468
3469         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3470         if (nlh == NULL)
3471                 return -EMSGSIZE;
3472
3473         ifm = nlmsg_data(nlh);
3474         ifm->ifi_family = AF_BRIDGE;
3475         ifm->__ifi_pad = 0;
3476         ifm->ifi_type = dev->type;
3477         ifm->ifi_index = dev->ifindex;
3478         ifm->ifi_flags = dev_get_flags(dev);
3479         ifm->ifi_change = 0;
3480
3481
3482         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3483             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3484             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3485             (br_dev &&
3486              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3487             (dev->addr_len &&
3488              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3489             (dev->ifindex != dev_get_iflink(dev) &&
3490              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3491                 goto nla_put_failure;
3492
3493         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3494         if (!br_afspec)
3495                 goto nla_put_failure;
3496
3497         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3498                 nla_nest_cancel(skb, br_afspec);
3499                 goto nla_put_failure;
3500         }
3501
3502         if (mode != BRIDGE_MODE_UNDEF) {
3503                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3504                         nla_nest_cancel(skb, br_afspec);
3505                         goto nla_put_failure;
3506                 }
3507         }
3508         if (vlan_fill) {
3509                 err = vlan_fill(skb, dev, filter_mask);
3510                 if (err) {
3511                         nla_nest_cancel(skb, br_afspec);
3512                         goto nla_put_failure;
3513                 }
3514         }
3515         nla_nest_end(skb, br_afspec);
3516
3517         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3518         if (!protinfo)
3519                 goto nla_put_failure;
3520
3521         if (brport_nla_put_flag(skb, flags, mask,
3522                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3523             brport_nla_put_flag(skb, flags, mask,
3524                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3525             brport_nla_put_flag(skb, flags, mask,
3526                                 IFLA_BRPORT_FAST_LEAVE,
3527                                 BR_MULTICAST_FAST_LEAVE) ||
3528             brport_nla_put_flag(skb, flags, mask,
3529                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3530             brport_nla_put_flag(skb, flags, mask,
3531                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3532             brport_nla_put_flag(skb, flags, mask,
3533                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3534             brport_nla_put_flag(skb, flags, mask,
3535                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3536             brport_nla_put_flag(skb, flags, mask,
3537                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3538                 nla_nest_cancel(skb, protinfo);
3539                 goto nla_put_failure;
3540         }
3541
3542         nla_nest_end(skb, protinfo);
3543
3544         nlmsg_end(skb, nlh);
3545         return 0;
3546 nla_put_failure:
3547         nlmsg_cancel(skb, nlh);
3548         return err ? err : -EMSGSIZE;
3549 }
3550 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3551
3552 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3553 {
3554         struct net *net = sock_net(skb->sk);
3555         struct net_device *dev;
3556         int idx = 0;
3557         u32 portid = NETLINK_CB(cb->skb).portid;
3558         u32 seq = cb->nlh->nlmsg_seq;
3559         u32 filter_mask = 0;
3560         int err;
3561
3562         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3563                 struct nlattr *extfilt;
3564
3565                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3566                                           IFLA_EXT_MASK);
3567                 if (extfilt) {
3568                         if (nla_len(extfilt) < sizeof(filter_mask))
3569                                 return -EINVAL;
3570
3571                         filter_mask = nla_get_u32(extfilt);
3572                 }
3573         }
3574
3575         rcu_read_lock();
3576         for_each_netdev_rcu(net, dev) {
3577                 const struct net_device_ops *ops = dev->netdev_ops;
3578                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3579
3580                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3581                         if (idx >= cb->args[0]) {
3582                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3583                                                 skb, portid, seq, dev,
3584                                                 filter_mask, NLM_F_MULTI);
3585                                 if (err < 0 && err != -EOPNOTSUPP) {
3586                                         if (likely(skb->len))
3587                                                 break;
3588
3589                                         goto out_err;
3590                                 }
3591                         }
3592                         idx++;
3593                 }
3594
3595                 if (ops->ndo_bridge_getlink) {
3596                         if (idx >= cb->args[0]) {
3597                                 err = ops->ndo_bridge_getlink(skb, portid,
3598                                                               seq, dev,
3599                                                               filter_mask,
3600                                                               NLM_F_MULTI);
3601                                 if (err < 0 && err != -EOPNOTSUPP) {
3602                                         if (likely(skb->len))
3603                                                 break;
3604
3605                                         goto out_err;
3606                                 }
3607                         }
3608                         idx++;
3609                 }
3610         }
3611         err = skb->len;
3612 out_err:
3613         rcu_read_unlock();
3614         cb->args[0] = idx;
3615
3616         return err;
3617 }
3618
3619 static inline size_t bridge_nlmsg_size(void)
3620 {
3621         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3622                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3623                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3624                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3625                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3626                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3627                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3628                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3629                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3630                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3631                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3632 }
3633
3634 static int rtnl_bridge_notify(struct net_device *dev)
3635 {
3636         struct net *net = dev_net(dev);
3637         struct sk_buff *skb;
3638         int err = -EOPNOTSUPP;
3639
3640         if (!dev->netdev_ops->ndo_bridge_getlink)
3641                 return 0;
3642
3643         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3644         if (!skb) {
3645                 err = -ENOMEM;
3646                 goto errout;
3647         }
3648
3649         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3650         if (err < 0)
3651                 goto errout;
3652
3653         /* Notification info is only filled for bridge ports, not the bridge
3654          * device itself. Therefore, a zero notification length is valid and
3655          * should not result in an error.
3656          */
3657         if (!skb->len)
3658                 goto errout;
3659
3660         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3661         return 0;
3662 errout:
3663         WARN_ON(err == -EMSGSIZE);
3664         kfree_skb(skb);
3665         if (err)
3666                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3667         return err;
3668 }
3669
3670 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3671                                struct netlink_ext_ack *extack)
3672 {
3673         struct net *net = sock_net(skb->sk);
3674         struct ifinfomsg *ifm;
3675         struct net_device *dev;
3676         struct nlattr *br_spec, *attr = NULL;
3677         int rem, err = -EOPNOTSUPP;
3678         u16 flags = 0;
3679         bool have_flags = false;
3680
3681         if (nlmsg_len(nlh) < sizeof(*ifm))
3682                 return -EINVAL;
3683
3684         ifm = nlmsg_data(nlh);
3685         if (ifm->ifi_family != AF_BRIDGE)
3686                 return -EPFNOSUPPORT;
3687
3688         dev = __dev_get_by_index(net, ifm->ifi_index);
3689         if (!dev) {
3690                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3691                 return -ENODEV;
3692         }
3693
3694         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3695         if (br_spec) {
3696                 nla_for_each_nested(attr, br_spec, rem) {
3697                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3698                                 if (nla_len(attr) < sizeof(flags))
3699                                         return -EINVAL;
3700
3701                                 have_flags = true;
3702                                 flags = nla_get_u16(attr);
3703                                 break;
3704                         }
3705                 }
3706         }
3707
3708         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3709                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3710
3711                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3712                         err = -EOPNOTSUPP;
3713                         goto out;
3714                 }
3715
3716                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3717                 if (err)
3718                         goto out;
3719
3720                 flags &= ~BRIDGE_FLAGS_MASTER;
3721         }
3722
3723         if ((flags & BRIDGE_FLAGS_SELF)) {
3724                 if (!dev->netdev_ops->ndo_bridge_setlink)
3725                         err = -EOPNOTSUPP;
3726                 else
3727                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3728                                                                   flags);
3729                 if (!err) {
3730                         flags &= ~BRIDGE_FLAGS_SELF;
3731
3732                         /* Generate event to notify upper layer of bridge
3733                          * change
3734                          */
3735                         err = rtnl_bridge_notify(dev);
3736                 }
3737         }
3738
3739         if (have_flags)
3740                 memcpy(nla_data(attr), &flags, sizeof(flags));
3741 out:
3742         return err;
3743 }
3744
3745 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3746                                struct netlink_ext_ack *extack)
3747 {
3748         struct net *net = sock_net(skb->sk);
3749         struct ifinfomsg *ifm;
3750         struct net_device *dev;
3751         struct nlattr *br_spec, *attr = NULL;
3752         int rem, err = -EOPNOTSUPP;
3753         u16 flags = 0;
3754         bool have_flags = false;
3755
3756         if (nlmsg_len(nlh) < sizeof(*ifm))
3757                 return -EINVAL;
3758
3759         ifm = nlmsg_data(nlh);
3760         if (ifm->ifi_family != AF_BRIDGE)
3761                 return -EPFNOSUPPORT;
3762
3763         dev = __dev_get_by_index(net, ifm->ifi_index);
3764         if (!dev) {
3765                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3766                 return -ENODEV;
3767         }
3768
3769         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3770         if (br_spec) {
3771                 nla_for_each_nested(attr, br_spec, rem) {
3772                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3773                                 if (nla_len(attr) < sizeof(flags))
3774                                         return -EINVAL;
3775
3776                                 have_flags = true;
3777                                 flags = nla_get_u16(attr);
3778                                 break;
3779                         }
3780                 }
3781         }
3782
3783         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3784                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3785
3786                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3787                         err = -EOPNOTSUPP;
3788                         goto out;
3789                 }
3790
3791                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3792                 if (err)
3793                         goto out;
3794
3795                 flags &= ~BRIDGE_FLAGS_MASTER;
3796         }
3797
3798         if ((flags & BRIDGE_FLAGS_SELF)) {
3799                 if (!dev->netdev_ops->ndo_bridge_dellink)
3800                         err = -EOPNOTSUPP;
3801                 else
3802                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3803                                                                   flags);
3804
3805                 if (!err) {
3806                         flags &= ~BRIDGE_FLAGS_SELF;
3807
3808                         /* Generate event to notify upper layer of bridge
3809                          * change
3810                          */
3811                         err = rtnl_bridge_notify(dev);
3812                 }
3813         }
3814
3815         if (have_flags)
3816                 memcpy(nla_data(attr), &flags, sizeof(flags));
3817 out:
3818         return err;
3819 }
3820
3821 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3822 {
3823         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3824                (!idxattr || idxattr == attrid);
3825 }
3826
3827 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3828 static int rtnl_get_offload_stats_attr_size(int attr_id)
3829 {
3830         switch (attr_id) {
3831         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3832                 return sizeof(struct rtnl_link_stats64);
3833         }
3834
3835         return 0;
3836 }
3837
3838 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3839                                   int *prividx)
3840 {
3841         struct nlattr *attr = NULL;
3842         int attr_id, size;
3843         void *attr_data;
3844         int err;
3845
3846         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3847               dev->netdev_ops->ndo_get_offload_stats))
3848                 return -ENODATA;
3849
3850         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3851              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3852                 if (attr_id < *prividx)
3853                         continue;
3854
3855                 size = rtnl_get_offload_stats_attr_size(attr_id);
3856                 if (!size)
3857                         continue;
3858
3859                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3860                         continue;
3861
3862                 attr = nla_reserve_64bit(skb, attr_id, size,
3863                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3864                 if (!attr)
3865                         goto nla_put_failure;
3866
3867                 attr_data = nla_data(attr);
3868                 memset(attr_data, 0, size);
3869                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3870                                                              attr_data);
3871                 if (err)
3872                         goto get_offload_stats_failure;
3873         }
3874
3875         if (!attr)
3876                 return -ENODATA;
3877
3878         *prividx = 0;
3879         return 0;
3880
3881 nla_put_failure:
3882         err = -EMSGSIZE;
3883 get_offload_stats_failure:
3884         *prividx = attr_id;
3885         return err;
3886 }
3887
3888 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3889 {
3890         int nla_size = 0;
3891         int attr_id;
3892         int size;
3893
3894         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3895               dev->netdev_ops->ndo_get_offload_stats))
3896                 return 0;
3897
3898         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3899              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3900                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3901                         continue;
3902                 size = rtnl_get_offload_stats_attr_size(attr_id);
3903                 nla_size += nla_total_size_64bit(size);
3904         }
3905
3906         if (nla_size != 0)
3907                 nla_size += nla_total_size(0);
3908
3909         return nla_size;
3910 }
3911
3912 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3913                                int type, u32 pid, u32 seq, u32 change,
3914                                unsigned int flags, unsigned int filter_mask,
3915                                int *idxattr, int *prividx)
3916 {
3917         struct if_stats_msg *ifsm;
3918         struct nlmsghdr *nlh;
3919         struct nlattr *attr;
3920         int s_prividx = *prividx;
3921         int err;
3922
3923         ASSERT_RTNL();
3924
3925         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3926         if (!nlh)
3927                 return -EMSGSIZE;
3928
3929         ifsm = nlmsg_data(nlh);
3930         ifsm->family = PF_UNSPEC;
3931         ifsm->pad1 = 0;
3932         ifsm->pad2 = 0;
3933         ifsm->ifindex = dev->ifindex;
3934         ifsm->filter_mask = filter_mask;
3935
3936         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3937                 struct rtnl_link_stats64 *sp;
3938
3939                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3940                                          sizeof(struct rtnl_link_stats64),
3941                                          IFLA_STATS_UNSPEC);
3942                 if (!attr)
3943                         goto nla_put_failure;
3944
3945                 sp = nla_data(attr);
3946                 dev_get_stats(dev, sp);
3947         }
3948
3949         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3950                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3951
3952                 if (ops && ops->fill_linkxstats) {
3953                         *idxattr = IFLA_STATS_LINK_XSTATS;
3954                         attr = nla_nest_start(skb,
3955                                               IFLA_STATS_LINK_XSTATS);
3956                         if (!attr)
3957                                 goto nla_put_failure;
3958
3959                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3960                         nla_nest_end(skb, attr);
3961                         if (err)
3962                                 goto nla_put_failure;
3963                         *idxattr = 0;
3964                 }
3965         }
3966
3967         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3968                              *idxattr)) {
3969                 const struct rtnl_link_ops *ops = NULL;
3970                 const struct net_device *master;
3971
3972                 master = netdev_master_upper_dev_get(dev);
3973                 if (master)
3974                         ops = master->rtnl_link_ops;
3975                 if (ops && ops->fill_linkxstats) {
3976                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3977                         attr = nla_nest_start(skb,
3978                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3979                         if (!attr)
3980                                 goto nla_put_failure;
3981
3982                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3983                         nla_nest_end(skb, attr);
3984                         if (err)
3985                                 goto nla_put_failure;
3986                         *idxattr = 0;
3987                 }
3988         }
3989
3990         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3991                              *idxattr)) {
3992                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3993                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3994                 if (!attr)
3995                         goto nla_put_failure;
3996
3997                 err = rtnl_get_offload_stats(skb, dev, prividx);
3998                 if (err == -ENODATA)
3999                         nla_nest_cancel(skb, attr);
4000                 else
4001                         nla_nest_end(skb, attr);
4002
4003                 if (err && err != -ENODATA)
4004                         goto nla_put_failure;
4005                 *idxattr = 0;
4006         }
4007
4008         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4009                 struct rtnl_af_ops *af_ops;
4010
4011                 *idxattr = IFLA_STATS_AF_SPEC;
4012                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4013                 if (!attr)
4014                         goto nla_put_failure;
4015
4016                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4017                         if (af_ops->fill_stats_af) {
4018                                 struct nlattr *af;
4019                                 int err;
4020
4021                                 af = nla_nest_start(skb, af_ops->family);
4022                                 if (!af)
4023                                         goto nla_put_failure;
4024
4025                                 err = af_ops->fill_stats_af(skb, dev);
4026
4027                                 if (err == -ENODATA)
4028                                         nla_nest_cancel(skb, af);
4029                                 else if (err < 0)
4030                                         goto nla_put_failure;
4031
4032                                 nla_nest_end(skb, af);
4033                         }
4034                 }
4035
4036                 nla_nest_end(skb, attr);
4037
4038                 *idxattr = 0;
4039         }
4040
4041         nlmsg_end(skb, nlh);
4042
4043         return 0;
4044
4045 nla_put_failure:
4046         /* not a multi message or no progress mean a real error */
4047         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4048                 nlmsg_cancel(skb, nlh);
4049         else
4050                 nlmsg_end(skb, nlh);
4051
4052         return -EMSGSIZE;
4053 }
4054
4055 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4056                                   u32 filter_mask)
4057 {
4058         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
4059
4060         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4061                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4062
4063         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4064                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4065                 int attr = IFLA_STATS_LINK_XSTATS;
4066
4067                 if (ops && ops->get_linkxstats_size) {
4068                         size += nla_total_size(ops->get_linkxstats_size(dev,
4069                                                                         attr));
4070                         /* for IFLA_STATS_LINK_XSTATS */
4071                         size += nla_total_size(0);
4072                 }
4073         }
4074
4075         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4076                 struct net_device *_dev = (struct net_device *)dev;
4077                 const struct rtnl_link_ops *ops = NULL;
4078                 const struct net_device *master;
4079
4080                 /* netdev_master_upper_dev_get can't take const */
4081                 master = netdev_master_upper_dev_get(_dev);
4082                 if (master)
4083                         ops = master->rtnl_link_ops;
4084                 if (ops && ops->get_linkxstats_size) {
4085                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4086
4087                         size += nla_total_size(ops->get_linkxstats_size(dev,
4088                                                                         attr));
4089                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4090                         size += nla_total_size(0);
4091                 }
4092         }
4093
4094         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4095                 size += rtnl_get_offload_stats_size(dev);
4096
4097         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4098                 struct rtnl_af_ops *af_ops;
4099
4100                 /* for IFLA_STATS_AF_SPEC */
4101                 size += nla_total_size(0);
4102
4103                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4104                         if (af_ops->get_stats_af_size) {
4105                                 size += nla_total_size(
4106                                         af_ops->get_stats_af_size(dev));
4107
4108                                 /* for AF_* */
4109                                 size += nla_total_size(0);
4110                         }
4111                 }
4112         }
4113
4114         return size;
4115 }
4116
4117 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4118                           struct netlink_ext_ack *extack)
4119 {
4120         struct net *net = sock_net(skb->sk);
4121         struct net_device *dev = NULL;
4122         int idxattr = 0, prividx = 0;
4123         struct if_stats_msg *ifsm;
4124         struct sk_buff *nskb;
4125         u32 filter_mask;
4126         int err;
4127
4128         if (nlmsg_len(nlh) < sizeof(*ifsm))
4129                 return -EINVAL;
4130
4131         ifsm = nlmsg_data(nlh);
4132         if (ifsm->ifindex > 0)
4133                 dev = __dev_get_by_index(net, ifsm->ifindex);
4134         else
4135                 return -EINVAL;
4136
4137         if (!dev)
4138                 return -ENODEV;
4139
4140         filter_mask = ifsm->filter_mask;
4141         if (!filter_mask)
4142                 return -EINVAL;
4143
4144         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4145         if (!nskb)
4146                 return -ENOBUFS;
4147
4148         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4149                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4150                                   0, filter_mask, &idxattr, &prividx);
4151         if (err < 0) {
4152                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4153                 WARN_ON(err == -EMSGSIZE);
4154                 kfree_skb(nskb);
4155         } else {
4156                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4157         }
4158
4159         return err;
4160 }
4161
4162 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4163 {
4164         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4165         struct net *net = sock_net(skb->sk);
4166         unsigned int flags = NLM_F_MULTI;
4167         struct if_stats_msg *ifsm;
4168         struct hlist_head *head;
4169         struct net_device *dev;
4170         u32 filter_mask = 0;
4171         int idx = 0;
4172
4173         s_h = cb->args[0];
4174         s_idx = cb->args[1];
4175         s_idxattr = cb->args[2];
4176         s_prividx = cb->args[3];
4177
4178         cb->seq = net->dev_base_seq;
4179
4180         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4181                 return -EINVAL;
4182
4183         ifsm = nlmsg_data(cb->nlh);
4184         filter_mask = ifsm->filter_mask;
4185         if (!filter_mask)
4186                 return -EINVAL;
4187
4188         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4189                 idx = 0;
4190                 head = &net->dev_index_head[h];
4191                 hlist_for_each_entry(dev, head, index_hlist) {
4192                         if (idx < s_idx)
4193                                 goto cont;
4194                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4195                                                   NETLINK_CB(cb->skb).portid,
4196                                                   cb->nlh->nlmsg_seq, 0,
4197                                                   flags, filter_mask,
4198                                                   &s_idxattr, &s_prividx);
4199                         /* If we ran out of room on the first message,
4200                          * we're in trouble
4201                          */
4202                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4203
4204                         if (err < 0)
4205                                 goto out;
4206                         s_prividx = 0;
4207                         s_idxattr = 0;
4208                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4209 cont:
4210                         idx++;
4211                 }
4212         }
4213 out:
4214         cb->args[3] = s_prividx;
4215         cb->args[2] = s_idxattr;
4216         cb->args[1] = idx;
4217         cb->args[0] = h;
4218
4219         return skb->len;
4220 }
4221
4222 /* Process one rtnetlink message. */
4223
4224 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4225                              struct netlink_ext_ack *extack)
4226 {
4227         struct net *net = sock_net(skb->sk);
4228         struct rtnl_link *handlers;
4229         int err = -EOPNOTSUPP;
4230         rtnl_doit_func doit;
4231         unsigned int flags;
4232         int kind;
4233         int family;
4234         int type;
4235
4236         type = nlh->nlmsg_type;
4237         if (type > RTM_MAX)
4238                 return -EOPNOTSUPP;
4239
4240         type -= RTM_BASE;
4241
4242         /* All the messages must have at least 1 byte length */
4243         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4244                 return 0;
4245
4246         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4247         kind = type&3;
4248
4249         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4250                 return -EPERM;
4251
4252         if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4253                 family = PF_UNSPEC;
4254
4255         rcu_read_lock();
4256         handlers = rcu_dereference(rtnl_msg_handlers[family]);
4257         if (!handlers) {
4258                 family = PF_UNSPEC;
4259                 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4260         }
4261
4262         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4263                 struct sock *rtnl;
4264                 rtnl_dumpit_func dumpit;
4265                 u16 min_dump_alloc = 0;
4266
4267                 dumpit = READ_ONCE(handlers[type].dumpit);
4268                 if (!dumpit) {
4269                         family = PF_UNSPEC;
4270                         handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4271                         if (!handlers)
4272                                 goto err_unlock;
4273
4274                         dumpit = READ_ONCE(handlers[type].dumpit);
4275                         if (!dumpit)
4276                                 goto err_unlock;
4277                 }
4278
4279                 refcount_inc(&rtnl_msg_handlers_ref[family]);
4280
4281                 if (type == RTM_GETLINK - RTM_BASE)
4282                         min_dump_alloc = rtnl_calcit(skb, nlh);
4283
4284                 rcu_read_unlock();
4285
4286                 rtnl = net->rtnl;
4287                 {
4288                         struct netlink_dump_control c = {
4289                                 .dump           = dumpit,
4290                                 .min_dump_alloc = min_dump_alloc,
4291                         };
4292                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4293                 }
4294                 refcount_dec(&rtnl_msg_handlers_ref[family]);
4295                 return err;
4296         }
4297
4298         doit = READ_ONCE(handlers[type].doit);
4299         if (!doit) {
4300                 family = PF_UNSPEC;
4301                 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4302         }
4303
4304         flags = READ_ONCE(handlers[type].flags);
4305         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4306                 refcount_inc(&rtnl_msg_handlers_ref[family]);
4307                 doit = READ_ONCE(handlers[type].doit);
4308                 rcu_read_unlock();
4309                 if (doit)
4310                         err = doit(skb, nlh, extack);
4311                 refcount_dec(&rtnl_msg_handlers_ref[family]);
4312                 return err;
4313         }
4314
4315         rcu_read_unlock();
4316
4317         rtnl_lock();
4318         handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4319         if (handlers) {
4320                 doit = READ_ONCE(handlers[type].doit);
4321                 if (doit)
4322                         err = doit(skb, nlh, extack);
4323         }
4324         rtnl_unlock();
4325         return err;
4326
4327 err_unlock:
4328         rcu_read_unlock();
4329         return -EOPNOTSUPP;
4330 }
4331
4332 static void rtnetlink_rcv(struct sk_buff *skb)
4333 {
4334         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4335 }
4336
4337 static int rtnetlink_bind(struct net *net, int group)
4338 {
4339         switch (group) {
4340         case RTNLGRP_IPV4_MROUTE_R:
4341         case RTNLGRP_IPV6_MROUTE_R:
4342                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4343                         return -EPERM;
4344                 break;
4345         }
4346         return 0;
4347 }
4348
4349 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4350 {
4351         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4352
4353         switch (event) {
4354         case NETDEV_REBOOT:
4355         case NETDEV_CHANGEMTU:
4356         case NETDEV_CHANGEADDR:
4357         case NETDEV_CHANGENAME:
4358         case NETDEV_FEAT_CHANGE:
4359         case NETDEV_BONDING_FAILOVER:
4360         case NETDEV_POST_TYPE_CHANGE:
4361         case NETDEV_NOTIFY_PEERS:
4362         case NETDEV_CHANGEUPPER:
4363         case NETDEV_RESEND_IGMP:
4364         case NETDEV_CHANGEINFODATA:
4365         case NETDEV_CHANGE_TX_QUEUE_LEN:
4366                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4367                                    GFP_KERNEL);
4368                 break;
4369         default:
4370                 break;
4371         }
4372         return NOTIFY_DONE;
4373 }
4374
4375 static struct notifier_block rtnetlink_dev_notifier = {
4376         .notifier_call  = rtnetlink_event,
4377 };
4378
4379
4380 static int __net_init rtnetlink_net_init(struct net *net)
4381 {
4382         struct sock *sk;
4383         struct netlink_kernel_cfg cfg = {
4384                 .groups         = RTNLGRP_MAX,
4385                 .input          = rtnetlink_rcv,
4386                 .cb_mutex       = &rtnl_mutex,
4387                 .flags          = NL_CFG_F_NONROOT_RECV,
4388                 .bind           = rtnetlink_bind,
4389         };
4390
4391         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4392         if (!sk)
4393                 return -ENOMEM;
4394         net->rtnl = sk;
4395         return 0;
4396 }
4397
4398 static void __net_exit rtnetlink_net_exit(struct net *net)
4399 {
4400         netlink_kernel_release(net->rtnl);
4401         net->rtnl = NULL;
4402 }
4403
4404 static struct pernet_operations rtnetlink_net_ops = {
4405         .init = rtnetlink_net_init,
4406         .exit = rtnetlink_net_exit,
4407 };
4408
4409 void __init rtnetlink_init(void)
4410 {
4411         int i;
4412
4413         for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4414                 refcount_set(&rtnl_msg_handlers_ref[i], 1);
4415
4416         if (register_pernet_subsys(&rtnetlink_net_ops))
4417                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4418
4419         register_netdevice_notifier(&rtnetlink_dev_notifier);
4420
4421         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4422                       rtnl_dump_ifinfo, 0);
4423         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4424         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4425         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4426
4427         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4428         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4429         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4430
4431         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4432         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4433         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4434
4435         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4436         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4437         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4438
4439         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4440                       0);
4441 }