1/*
   2 * INET		An implementation of the TCP/IP protocol suite for the LINUX
   3 *		operating system.  INET is implemented using the  BSD Socket
   4 *		interface as the means of communication with the user level.
   5 *
   6 *		Routing netlink socket interface: protocol independent part.
   7 *
   8 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9 *
  10 *		This program is free software; you can redistribute it and/or
  11 *		modify it under the terms of the GNU General Public License
  12 *		as published by the Free Software Foundation; either version
  13 *		2 of the License, or (at your option) any later version.
  14 *
  15 *	Fixes:
  16 *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
  17 */
  18
  19#include <linux/errno.h>
  20#include <linux/module.h>
  21#include <linux/types.h>
  22#include <linux/socket.h>
  23#include <linux/kernel.h>
  24#include <linux/timer.h>
  25#include <linux/string.h>
  26#include <linux/sockios.h>
  27#include <linux/net.h>
  28#include <linux/fcntl.h>
  29#include <linux/mm.h>
  30#include <linux/slab.h>
  31#include <linux/interrupt.h>
  32#include <linux/capability.h>
  33#include <linux/skbuff.h>
  34#include <linux/init.h>
  35#include <linux/security.h>
  36#include <linux/mutex.h>
  37#include <linux/if_addr.h>
  38#include <linux/if_bridge.h>
  39#include <linux/pci.h>
  40#include <linux/etherdevice.h>
  41
  42#include <asm/uaccess.h>
 
  43
  44#include <linux/inet.h>
  45#include <linux/netdevice.h>
  46#include <net/ip.h>
  47#include <net/protocol.h>
  48#include <net/arp.h>
  49#include <net/route.h>
  50#include <net/udp.h>
  51#include <net/sock.h>
  52#include <net/pkt_sched.h>
  53#include <net/fib_rules.h>
  54#include <net/rtnetlink.h>
  55#include <net/net_namespace.h>
  56
  57struct rtnl_link {
  58	rtnl_doit_func		doit;
  59	rtnl_dumpit_func	dumpit;
  60	rtnl_calcit_func 	calcit;
  61};
  62
  63static DEFINE_MUTEX(rtnl_mutex);
 
  64
  65void rtnl_lock(void)
  66{
  67	mutex_lock(&rtnl_mutex);
  68}
  69EXPORT_SYMBOL(rtnl_lock);
  70
  71void __rtnl_unlock(void)
  72{
  73	mutex_unlock(&rtnl_mutex);
  74}
  75
  76void rtnl_unlock(void)
  77{
  78	/* This fellow will unlock it for us. */
  79	netdev_run_todo();
  80}
  81EXPORT_SYMBOL(rtnl_unlock);
  82
  83int rtnl_trylock(void)
  84{
  85	return mutex_trylock(&rtnl_mutex);
  86}
  87EXPORT_SYMBOL(rtnl_trylock);
  88
  89int rtnl_is_locked(void)
  90{
  91	return mutex_is_locked(&rtnl_mutex);
  92}
  93EXPORT_SYMBOL(rtnl_is_locked);
  94
  95#ifdef CONFIG_PROVE_LOCKING
  96int lockdep_rtnl_is_held(void)
  97{
  98	return lockdep_is_held(&rtnl_mutex);
  99}
 100EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101#endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105static inline int rtm_msgindex(int msgtype)
 106{
 107	int msgindex = msgtype - RTM_BASE;
 108
 109	/*
 110	 * msgindex < 0 implies someone tried to register a netlink
 111	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112	 * the message type has not been added to linux/rtnetlink.h
 113	 */
 114	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116	return msgindex;
 117}
 118
 119static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120{
 121	struct rtnl_link *tab;
 122
 123	if (protocol <= RTNL_FAMILY_MAX)
 124		tab = rtnl_msg_handlers[protocol];
 125	else
 126		tab = NULL;
 127
 128	if (tab == NULL || tab[msgindex].doit == NULL)
 129		tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131	return tab ? tab[msgindex].doit : NULL;
 132}
 133
 134static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135{
 136	struct rtnl_link *tab;
 137
 138	if (protocol <= RTNL_FAMILY_MAX)
 139		tab = rtnl_msg_handlers[protocol];
 140	else
 141		tab = NULL;
 142
 143	if (tab == NULL || tab[msgindex].dumpit == NULL)
 144		tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146	return tab ? tab[msgindex].dumpit : NULL;
 147}
 148
 149static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150{
 151	struct rtnl_link *tab;
 152
 153	if (protocol <= RTNL_FAMILY_MAX)
 154		tab = rtnl_msg_handlers[protocol];
 155	else
 156		tab = NULL;
 157
 158	if (tab == NULL || tab[msgindex].calcit == NULL)
 159		tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161	return tab ? tab[msgindex].calcit : NULL;
 162}
 163
 164/**
 165 * __rtnl_register - Register a rtnetlink message type
 166 * @protocol: Protocol family or PF_UNSPEC
 167 * @msgtype: rtnetlink message type
 168 * @doit: Function pointer called for each request message
 169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170 * @calcit: Function pointer to calc size of dump message
 171 *
 172 * Registers the specified function pointers (at least one of them has
 173 * to be non-NULL) to be called whenever a request message for the
 174 * specified protocol family and message type is received.
 175 *
 176 * The special protocol family PF_UNSPEC may be used to define fallback
 177 * function pointers for the case when no entry for the specific protocol
 178 * family exists.
 179 *
 180 * Returns 0 on success or a negative error code.
 181 */
 182int __rtnl_register(int protocol, int msgtype,
 183		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184		    rtnl_calcit_func calcit)
 185{
 186	struct rtnl_link *tab;
 187	int msgindex;
 188
 189	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190	msgindex = rtm_msgindex(msgtype);
 191
 192	tab = rtnl_msg_handlers[protocol];
 193	if (tab == NULL) {
 194		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195		if (tab == NULL)
 196			return -ENOBUFS;
 197
 198		rtnl_msg_handlers[protocol] = tab;
 199	}
 200
 201	if (doit)
 202		tab[msgindex].doit = doit;
 203
 204	if (dumpit)
 205		tab[msgindex].dumpit = dumpit;
 206
 207	if (calcit)
 208		tab[msgindex].calcit = calcit;
 209
 210	return 0;
 211}
 212EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214/**
 215 * rtnl_register - Register a rtnetlink message type
 216 *
 217 * Identical to __rtnl_register() but panics on failure. This is useful
 218 * as failure of this function is very unlikely, it can only happen due
 219 * to lack of memory when allocating the chain to store all message
 220 * handlers for a protocol. Meant for use in init functions where lack
 221 * of memory implies no sense in continuing.
 222 */
 223void rtnl_register(int protocol, int msgtype,
 224		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225		   rtnl_calcit_func calcit)
 226{
 227	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228		panic("Unable to register rtnetlink message handler, "
 229		      "protocol = %d, message type = %d\n",
 230		      protocol, msgtype);
 231}
 232EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234/**
 235 * rtnl_unregister - Unregister a rtnetlink message type
 236 * @protocol: Protocol family or PF_UNSPEC
 237 * @msgtype: rtnetlink message type
 238 *
 239 * Returns 0 on success or a negative error code.
 240 */
 241int rtnl_unregister(int protocol, int msgtype)
 242{
 243	int msgindex;
 244
 245	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246	msgindex = rtm_msgindex(msgtype);
 247
 248	if (rtnl_msg_handlers[protocol] == NULL)
 249		return -ENOENT;
 250
 251	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254	return 0;
 255}
 256EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258/**
 259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260 * @protocol : Protocol family or PF_UNSPEC
 261 *
 262 * Identical to calling rtnl_unregster() for all registered message types
 263 * of a certain protocol family.
 264 */
 265void rtnl_unregister_all(int protocol)
 266{
 267	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269	kfree(rtnl_msg_handlers[protocol]);
 270	rtnl_msg_handlers[protocol] = NULL;
 271}
 272EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274static LIST_HEAD(link_ops);
 275
 276static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277{
 278	const struct rtnl_link_ops *ops;
 279
 280	list_for_each_entry(ops, &link_ops, list) {
 281		if (!strcmp(ops->kind, kind))
 282			return ops;
 283	}
 284	return NULL;
 285}
 286
 287/**
 288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289 * @ops: struct rtnl_link_ops * to register
 290 *
 291 * The caller must hold the rtnl_mutex. This function should be used
 292 * by drivers that create devices during module initialization. It
 293 * must be called before registering the devices.
 294 *
 295 * Returns 0 on success or a negative error code.
 296 */
 297int __rtnl_link_register(struct rtnl_link_ops *ops)
 298{
 299	if (rtnl_link_ops_get(ops->kind))
 300		return -EEXIST;
 301
 302	if (!ops->dellink)
 303		ops->dellink = unregister_netdevice_queue;
 304
 305	list_add_tail(&ops->list, &link_ops);
 306	return 0;
 307}
 308EXPORT_SYMBOL_GPL(__rtnl_link_register);
 309
 310/**
 311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 312 * @ops: struct rtnl_link_ops * to register
 313 *
 314 * Returns 0 on success or a negative error code.
 315 */
 316int rtnl_link_register(struct rtnl_link_ops *ops)
 317{
 318	int err;
 319
 320	rtnl_lock();
 321	err = __rtnl_link_register(ops);
 322	rtnl_unlock();
 323	return err;
 324}
 325EXPORT_SYMBOL_GPL(rtnl_link_register);
 326
 327static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 328{
 329	struct net_device *dev;
 330	LIST_HEAD(list_kill);
 331
 332	for_each_netdev(net, dev) {
 333		if (dev->rtnl_link_ops == ops)
 334			ops->dellink(dev, &list_kill);
 335	}
 336	unregister_netdevice_many(&list_kill);
 337}
 338
 339/**
 340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 341 * @ops: struct rtnl_link_ops * to unregister
 342 *
 343 * The caller must hold the rtnl_mutex.
 344 */
 345void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 346{
 347	struct net *net;
 348
 349	for_each_net(net) {
 350		__rtnl_kill_links(net, ops);
 351	}
 352	list_del(&ops->list);
 353}
 354EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 355
 356/**
 357 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 358 * @ops: struct rtnl_link_ops * to unregister
 359 */
 360void rtnl_link_unregister(struct rtnl_link_ops *ops)
 361{
 362	rtnl_lock();
 363	__rtnl_link_unregister(ops);
 364	rtnl_unlock();
 365}
 366EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 367
 
 
 
 
 
 
 
 
 
 
 
 368static size_t rtnl_link_get_size(const struct net_device *dev)
 369{
 370	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 371	size_t size;
 372
 373	if (!ops)
 374		return 0;
 375
 376	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 377	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 378
 379	if (ops->get_size)
 380		/* IFLA_INFO_DATA + nested data */
 381		size += nla_total_size(sizeof(struct nlattr)) +
 382			ops->get_size(dev);
 383
 384	if (ops->get_xstats_size)
 385		/* IFLA_INFO_XSTATS */
 386		size += nla_total_size(ops->get_xstats_size(dev));
 387
 388	return size;
 389}
 390
 391static LIST_HEAD(rtnl_af_ops);
 392
 393static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 394{
 395	const struct rtnl_af_ops *ops;
 396
 397	list_for_each_entry(ops, &rtnl_af_ops, list) {
 398		if (ops->family == family)
 399			return ops;
 400	}
 401
 402	return NULL;
 403}
 404
 405/**
 406 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 407 * @ops: struct rtnl_af_ops * to register
 408 *
 409 * The caller must hold the rtnl_mutex.
 410 *
 411 * Returns 0 on success or a negative error code.
 412 */
 413int __rtnl_af_register(struct rtnl_af_ops *ops)
 414{
 415	list_add_tail(&ops->list, &rtnl_af_ops);
 416	return 0;
 417}
 418EXPORT_SYMBOL_GPL(__rtnl_af_register);
 419
 420/**
 421 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 422 * @ops: struct rtnl_af_ops * to register
 423 *
 424 * Returns 0 on success or a negative error code.
 425 */
 426int rtnl_af_register(struct rtnl_af_ops *ops)
 427{
 428	int err;
 429
 430	rtnl_lock();
 431	err = __rtnl_af_register(ops);
 432	rtnl_unlock();
 433	return err;
 434}
 435EXPORT_SYMBOL_GPL(rtnl_af_register);
 436
 437/**
 438 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 439 * @ops: struct rtnl_af_ops * to unregister
 440 *
 441 * The caller must hold the rtnl_mutex.
 442 */
 443void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 444{
 445	list_del(&ops->list);
 446}
 447EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 448
 449/**
 450 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 451 * @ops: struct rtnl_af_ops * to unregister
 452 */
 453void rtnl_af_unregister(struct rtnl_af_ops *ops)
 454{
 455	rtnl_lock();
 456	__rtnl_af_unregister(ops);
 457	rtnl_unlock();
 458}
 459EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 460
 461static size_t rtnl_link_get_af_size(const struct net_device *dev)
 462{
 463	struct rtnl_af_ops *af_ops;
 464	size_t size;
 465
 466	/* IFLA_AF_SPEC */
 467	size = nla_total_size(sizeof(struct nlattr));
 468
 469	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 470		if (af_ops->get_link_af_size) {
 471			/* AF_* + nested data */
 472			size += nla_total_size(sizeof(struct nlattr)) +
 473				af_ops->get_link_af_size(dev);
 474		}
 475	}
 476
 477	return size;
 478}
 479
 480static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 481{
 482	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 483	struct nlattr *linkinfo, *data;
 484	int err = -EMSGSIZE;
 485
 486	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 487	if (linkinfo == NULL)
 488		goto out;
 489
 490	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 491		goto err_cancel_link;
 492	if (ops->fill_xstats) {
 493		err = ops->fill_xstats(skb, dev);
 494		if (err < 0)
 495			goto err_cancel_link;
 496	}
 497	if (ops->fill_info) {
 498		data = nla_nest_start(skb, IFLA_INFO_DATA);
 499		if (data == NULL)
 500			goto err_cancel_link;
 501		err = ops->fill_info(skb, dev);
 502		if (err < 0)
 503			goto err_cancel_data;
 504		nla_nest_end(skb, data);
 505	}
 506
 507	nla_nest_end(skb, linkinfo);
 508	return 0;
 509
 510err_cancel_data:
 511	nla_nest_cancel(skb, data);
 512err_cancel_link:
 513	nla_nest_cancel(skb, linkinfo);
 514out:
 515	return err;
 516}
 517
 518static const int rtm_min[RTM_NR_FAMILIES] =
 519{
 520	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 521	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 522	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 523	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 524	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 525	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 526	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 527	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 528	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 529	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 530};
 531
 532static const int rta_max[RTM_NR_FAMILIES] =
 533{
 534	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 535	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 536	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 537	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 538	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 539	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 540	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 541	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 542};
 543
 544void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 545{
 546	struct rtattr *rta;
 547	int size = RTA_LENGTH(attrlen);
 548
 549	rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
 550	rta->rta_type = attrtype;
 551	rta->rta_len = size;
 552	memcpy(RTA_DATA(rta), data, attrlen);
 553	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 554}
 555EXPORT_SYMBOL(__rta_fill);
 556
 557int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 558{
 559	struct sock *rtnl = net->rtnl;
 560	int err = 0;
 561
 562	NETLINK_CB(skb).dst_group = group;
 563	if (echo)
 564		atomic_inc(&skb->users);
 565	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 566	if (echo)
 567		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 568	return err;
 569}
 570
 571int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 572{
 573	struct sock *rtnl = net->rtnl;
 574
 575	return nlmsg_unicast(rtnl, skb, pid);
 576}
 577EXPORT_SYMBOL(rtnl_unicast);
 578
 579void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 580		 struct nlmsghdr *nlh, gfp_t flags)
 581{
 582	struct sock *rtnl = net->rtnl;
 583	int report = 0;
 584
 585	if (nlh)
 586		report = nlmsg_report(nlh);
 587
 588	nlmsg_notify(rtnl, skb, pid, group, report, flags);
 589}
 590EXPORT_SYMBOL(rtnl_notify);
 591
 592void rtnl_set_sk_err(struct net *net, u32 group, int error)
 593{
 594	struct sock *rtnl = net->rtnl;
 595
 596	netlink_set_err(rtnl, 0, group, error);
 597}
 598EXPORT_SYMBOL(rtnl_set_sk_err);
 599
 600int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 601{
 602	struct nlattr *mx;
 603	int i, valid = 0;
 604
 605	mx = nla_nest_start(skb, RTA_METRICS);
 606	if (mx == NULL)
 607		return -ENOBUFS;
 608
 609	for (i = 0; i < RTAX_MAX; i++) {
 610		if (metrics[i]) {
 611			valid++;
 612			if (nla_put_u32(skb, i+1, metrics[i]))
 613				goto nla_put_failure;
 614		}
 615	}
 616
 617	if (!valid) {
 618		nla_nest_cancel(skb, mx);
 619		return 0;
 620	}
 621
 622	return nla_nest_end(skb, mx);
 623
 624nla_put_failure:
 625	nla_nest_cancel(skb, mx);
 626	return -EMSGSIZE;
 627}
 628EXPORT_SYMBOL(rtnetlink_put_metrics);
 629
 630int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 631		       u32 ts, u32 tsage, long expires, u32 error)
 632{
 633	struct rta_cacheinfo ci = {
 634		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 635		.rta_used = dst->__use,
 636		.rta_clntref = atomic_read(&(dst->__refcnt)),
 637		.rta_error = error,
 638		.rta_id =  id,
 639		.rta_ts = ts,
 640		.rta_tsage = tsage,
 641	};
 642
 643	if (expires)
 644		ci.rta_expires = jiffies_to_clock_t(expires);
 645
 646	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 647}
 648EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 649
 650static void set_operstate(struct net_device *dev, unsigned char transition)
 651{
 652	unsigned char operstate = dev->operstate;
 653
 654	switch (transition) {
 655	case IF_OPER_UP:
 656		if ((operstate == IF_OPER_DORMANT ||
 657		     operstate == IF_OPER_UNKNOWN) &&
 658		    !netif_dormant(dev))
 659			operstate = IF_OPER_UP;
 660		break;
 661
 662	case IF_OPER_DORMANT:
 663		if (operstate == IF_OPER_UP ||
 664		    operstate == IF_OPER_UNKNOWN)
 665			operstate = IF_OPER_DORMANT;
 666		break;
 667	}
 668
 669	if (dev->operstate != operstate) {
 670		write_lock_bh(&dev_base_lock);
 671		dev->operstate = operstate;
 672		write_unlock_bh(&dev_base_lock);
 673		netdev_state_change(dev);
 674	}
 675}
 676
 677static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 678{
 679	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 680	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 681}
 682
 683static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 684					   const struct ifinfomsg *ifm)
 685{
 686	unsigned int flags = ifm->ifi_flags;
 687
 688	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 689	if (ifm->ifi_change)
 690		flags = (flags & ifm->ifi_change) |
 691			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 692
 693	return flags;
 694}
 695
 696static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 697				 const struct rtnl_link_stats64 *b)
 698{
 699	a->rx_packets = b->rx_packets;
 700	a->tx_packets = b->tx_packets;
 701	a->rx_bytes = b->rx_bytes;
 702	a->tx_bytes = b->tx_bytes;
 703	a->rx_errors = b->rx_errors;
 704	a->tx_errors = b->tx_errors;
 705	a->rx_dropped = b->rx_dropped;
 706	a->tx_dropped = b->tx_dropped;
 707
 708	a->multicast = b->multicast;
 709	a->collisions = b->collisions;
 710
 711	a->rx_length_errors = b->rx_length_errors;
 712	a->rx_over_errors = b->rx_over_errors;
 713	a->rx_crc_errors = b->rx_crc_errors;
 714	a->rx_frame_errors = b->rx_frame_errors;
 715	a->rx_fifo_errors = b->rx_fifo_errors;
 716	a->rx_missed_errors = b->rx_missed_errors;
 717
 718	a->tx_aborted_errors = b->tx_aborted_errors;
 719	a->tx_carrier_errors = b->tx_carrier_errors;
 720	a->tx_fifo_errors = b->tx_fifo_errors;
 721	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 722	a->tx_window_errors = b->tx_window_errors;
 723
 724	a->rx_compressed = b->rx_compressed;
 725	a->tx_compressed = b->tx_compressed;
 726}
 727
 728static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 729{
 730	memcpy(v, b, sizeof(*b));
 731}
 732
 733/* All VF info */
 734static inline int rtnl_vfinfo_size(const struct net_device *dev,
 735				   u32 ext_filter_mask)
 736{
 737	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 738	    (ext_filter_mask & RTEXT_FILTER_VF)) {
 739		int num_vfs = dev_num_vf(dev->dev.parent);
 740		size_t size = nla_total_size(sizeof(struct nlattr));
 741		size += nla_total_size(num_vfs * sizeof(struct nlattr));
 742		size += num_vfs *
 743			(nla_total_size(sizeof(struct ifla_vf_mac)) +
 744			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
 745			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 746			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
 747		return size;
 748	} else
 749		return 0;
 750}
 751
 752static size_t rtnl_port_size(const struct net_device *dev)
 753{
 754	size_t port_size = nla_total_size(4)		/* PORT_VF */
 755		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
 756		+ nla_total_size(sizeof(struct ifla_port_vsi))
 757							/* PORT_VSI_TYPE */
 758		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
 759		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
 760		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
 761		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
 762	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 763	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 764		+ port_size;
 765	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 766		+ port_size;
 767
 768	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 769		return 0;
 770	if (dev_num_vf(dev->dev.parent))
 771		return port_self_size + vf_ports_size +
 772			vf_port_size * dev_num_vf(dev->dev.parent);
 773	else
 774		return port_self_size;
 775}
 776
 777static noinline size_t if_nlmsg_size(const struct net_device *dev,
 778				     u32 ext_filter_mask)
 779{
 780	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 781	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 782	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 783	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 784	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
 785	       + nla_total_size(sizeof(struct rtnl_link_stats))
 786	       + nla_total_size(sizeof(struct rtnl_link_stats64))
 787	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 788	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 789	       + nla_total_size(4) /* IFLA_TXQLEN */
 790	       + nla_total_size(4) /* IFLA_WEIGHT */
 791	       + nla_total_size(4) /* IFLA_MTU */
 792	       + nla_total_size(4) /* IFLA_LINK */
 793	       + nla_total_size(4) /* IFLA_MASTER */
 794	       + nla_total_size(4) /* IFLA_PROMISCUITY */
 795	       + nla_total_size(1) /* IFLA_OPERSTATE */
 796	       + nla_total_size(1) /* IFLA_LINKMODE */
 797	       + nla_total_size(ext_filter_mask
 798			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 799	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 800	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 801	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 802	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
 803}
 804
 805static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 806{
 807	struct nlattr *vf_ports;
 808	struct nlattr *vf_port;
 809	int vf;
 810	int err;
 811
 812	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 813	if (!vf_ports)
 814		return -EMSGSIZE;
 815
 816	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 817		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 818		if (!vf_port)
 819			goto nla_put_failure;
 820		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 821			goto nla_put_failure;
 822		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 823		if (err == -EMSGSIZE)
 824			goto nla_put_failure;
 825		if (err) {
 826			nla_nest_cancel(skb, vf_port);
 827			continue;
 828		}
 829		nla_nest_end(skb, vf_port);
 830	}
 831
 832	nla_nest_end(skb, vf_ports);
 833
 834	return 0;
 835
 836nla_put_failure:
 837	nla_nest_cancel(skb, vf_ports);
 838	return -EMSGSIZE;
 839}
 840
 841static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 842{
 843	struct nlattr *port_self;
 844	int err;
 845
 846	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 847	if (!port_self)
 848		return -EMSGSIZE;
 849
 850	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 851	if (err) {
 852		nla_nest_cancel(skb, port_self);
 853		return (err == -EMSGSIZE) ? err : 0;
 854	}
 855
 856	nla_nest_end(skb, port_self);
 857
 858	return 0;
 859}
 860
 861static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
 862{
 863	int err;
 864
 865	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
 866		return 0;
 867
 868	err = rtnl_port_self_fill(skb, dev);
 869	if (err)
 870		return err;
 871
 872	if (dev_num_vf(dev->dev.parent)) {
 873		err = rtnl_vf_ports_fill(skb, dev);
 874		if (err)
 875			return err;
 876	}
 877
 878	return 0;
 879}
 880
 881static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 882			    int type, u32 pid, u32 seq, u32 change,
 883			    unsigned int flags, u32 ext_filter_mask)
 884{
 885	struct ifinfomsg *ifm;
 886	struct nlmsghdr *nlh;
 887	struct rtnl_link_stats64 temp;
 888	const struct rtnl_link_stats64 *stats;
 889	struct nlattr *attr, *af_spec;
 890	struct rtnl_af_ops *af_ops;
 891
 892	ASSERT_RTNL();
 893	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 894	if (nlh == NULL)
 895		return -EMSGSIZE;
 896
 897	ifm = nlmsg_data(nlh);
 898	ifm->ifi_family = AF_UNSPEC;
 899	ifm->__ifi_pad = 0;
 900	ifm->ifi_type = dev->type;
 901	ifm->ifi_index = dev->ifindex;
 902	ifm->ifi_flags = dev_get_flags(dev);
 903	ifm->ifi_change = change;
 904
 905	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 906	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 907	    nla_put_u8(skb, IFLA_OPERSTATE,
 908		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 909	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
 910	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
 911	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
 912	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
 913	    (dev->ifindex != dev->iflink &&
 914	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
 915	    (dev->master &&
 916	     nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
 917	    (dev->qdisc &&
 918	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
 919	    (dev->ifalias &&
 920	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
 921		goto nla_put_failure;
 
 
 922
 923	if (1) {
 924		struct rtnl_link_ifmap map = {
 925			.mem_start   = dev->mem_start,
 926			.mem_end     = dev->mem_end,
 927			.base_addr   = dev->base_addr,
 928			.irq         = dev->irq,
 929			.dma         = dev->dma,
 930			.port        = dev->if_port,
 931		};
 932		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
 933			goto nla_put_failure;
 934	}
 935
 936	if (dev->addr_len) {
 937		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
 938		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
 939			goto nla_put_failure;
 940	}
 941
 942	attr = nla_reserve(skb, IFLA_STATS,
 943			sizeof(struct rtnl_link_stats));
 944	if (attr == NULL)
 945		goto nla_put_failure;
 946
 947	stats = dev_get_stats(dev, &temp);
 948	copy_rtnl_link_stats(nla_data(attr), stats);
 949
 950	attr = nla_reserve(skb, IFLA_STATS64,
 951			sizeof(struct rtnl_link_stats64));
 952	if (attr == NULL)
 953		goto nla_put_failure;
 954	copy_rtnl_link_stats64(nla_data(attr), stats);
 955
 956	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
 957	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
 958		goto nla_put_failure;
 959
 960	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
 961	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
 962		int i;
 963
 964		struct nlattr *vfinfo, *vf;
 965		int num_vfs = dev_num_vf(dev->dev.parent);
 966
 967		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
 968		if (!vfinfo)
 969			goto nla_put_failure;
 970		for (i = 0; i < num_vfs; i++) {
 971			struct ifla_vf_info ivi;
 972			struct ifla_vf_mac vf_mac;
 973			struct ifla_vf_vlan vf_vlan;
 974			struct ifla_vf_tx_rate vf_tx_rate;
 975			struct ifla_vf_spoofchk vf_spoofchk;
 976
 977			/*
 978			 * Not all SR-IOV capable drivers support the
 979			 * spoofcheck query.  Preset to -1 so the user
 980			 * space tool can detect that the driver didn't
 981			 * report anything.
 982			 */
 983			ivi.spoofchk = -1;
 984			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
 985				break;
 986			vf_mac.vf =
 987				vf_vlan.vf =
 988				vf_tx_rate.vf =
 989				vf_spoofchk.vf = ivi.vf;
 990
 991			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
 992			vf_vlan.vlan = ivi.vlan;
 993			vf_vlan.qos = ivi.qos;
 994			vf_tx_rate.rate = ivi.tx_rate;
 995			vf_spoofchk.setting = ivi.spoofchk;
 996			vf = nla_nest_start(skb, IFLA_VF_INFO);
 997			if (!vf) {
 998				nla_nest_cancel(skb, vfinfo);
 999				goto nla_put_failure;
1000			}
1001			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1002			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1003			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1004				    &vf_tx_rate) ||
1005			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1006				    &vf_spoofchk))
1007				goto nla_put_failure;
1008			nla_nest_end(skb, vf);
1009		}
1010		nla_nest_end(skb, vfinfo);
1011	}
1012
1013	if (rtnl_port_fill(skb, dev))
1014		goto nla_put_failure;
1015
1016	if (dev->rtnl_link_ops) {
1017		if (rtnl_link_fill(skb, dev) < 0)
1018			goto nla_put_failure;
1019	}
1020
1021	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1022		goto nla_put_failure;
1023
1024	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1025		if (af_ops->fill_link_af) {
1026			struct nlattr *af;
1027			int err;
1028
1029			if (!(af = nla_nest_start(skb, af_ops->family)))
1030				goto nla_put_failure;
1031
1032			err = af_ops->fill_link_af(skb, dev);
1033
1034			/*
1035			 * Caller may return ENODATA to indicate that there
1036			 * was no data to be dumped. This is not an error, it
1037			 * means we should trim the attribute header and
1038			 * continue.
1039			 */
1040			if (err == -ENODATA)
1041				nla_nest_cancel(skb, af);
1042			else if (err < 0)
1043				goto nla_put_failure;
1044
1045			nla_nest_end(skb, af);
1046		}
1047	}
1048
1049	nla_nest_end(skb, af_spec);
1050
1051	return nlmsg_end(skb, nlh);
1052
1053nla_put_failure:
1054	nlmsg_cancel(skb, nlh);
1055	return -EMSGSIZE;
1056}
1057
1058static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1059{
1060	struct net *net = sock_net(skb->sk);
1061	int h, s_h;
1062	int idx = 0, s_idx;
1063	struct net_device *dev;
1064	struct hlist_head *head;
1065	struct hlist_node *node;
1066	struct nlattr *tb[IFLA_MAX+1];
1067	u32 ext_filter_mask = 0;
1068
1069	s_h = cb->args[0];
1070	s_idx = cb->args[1];
1071
1072	rcu_read_lock();
1073	cb->seq = net->dev_base_seq;
1074
1075	if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1076			ifla_policy) >= 0) {
1077
1078		if (tb[IFLA_EXT_MASK])
1079			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1080	}
1081
1082	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1083		idx = 0;
1084		head = &net->dev_index_head[h];
1085		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1086			if (idx < s_idx)
1087				goto cont;
1088			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1089					     NETLINK_CB(cb->skb).pid,
1090					     cb->nlh->nlmsg_seq, 0,
1091					     NLM_F_MULTI,
1092					     ext_filter_mask) <= 0)
1093				goto out;
1094
1095			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1096cont:
1097			idx++;
1098		}
1099	}
1100out:
1101	rcu_read_unlock();
1102	cb->args[1] = idx;
1103	cb->args[0] = h;
1104
1105	return skb->len;
1106}
1107
1108const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1109	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1110	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1111	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1112	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1113	[IFLA_MTU]		= { .type = NLA_U32 },
1114	[IFLA_LINK]		= { .type = NLA_U32 },
1115	[IFLA_MASTER]		= { .type = NLA_U32 },
1116	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1117	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1118	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1119	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1120	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1121	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1122	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1123	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1124	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1125	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1126	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1127	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1128	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1129	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1130};
1131EXPORT_SYMBOL(ifla_policy);
1132
1133static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1134	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1135	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1136};
1137
1138static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1139	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1140};
1141
1142static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1143	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1144				    .len = sizeof(struct ifla_vf_mac) },
1145	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1146				    .len = sizeof(struct ifla_vf_vlan) },
1147	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1148				    .len = sizeof(struct ifla_vf_tx_rate) },
1149	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1150				    .len = sizeof(struct ifla_vf_spoofchk) },
1151};
1152
1153static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1154	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1155	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1156				    .len = PORT_PROFILE_MAX },
1157	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1158				    .len = sizeof(struct ifla_port_vsi)},
1159	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1160				      .len = PORT_UUID_MAX },
1161	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1162				    .len = PORT_UUID_MAX },
1163	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1164	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1165};
1166
1167struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1168{
1169	struct net *net;
1170	/* Examine the link attributes and figure out which
1171	 * network namespace we are talking about.
1172	 */
1173	if (tb[IFLA_NET_NS_PID])
1174		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1175	else if (tb[IFLA_NET_NS_FD])
1176		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1177	else
1178		net = get_net(src_net);
1179	return net;
1180}
1181EXPORT_SYMBOL(rtnl_link_get_net);
1182
1183static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1184{
1185	if (dev) {
1186		if (tb[IFLA_ADDRESS] &&
1187		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1188			return -EINVAL;
1189
1190		if (tb[IFLA_BROADCAST] &&
1191		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1192			return -EINVAL;
1193	}
1194
1195	if (tb[IFLA_AF_SPEC]) {
1196		struct nlattr *af;
1197		int rem, err;
1198
1199		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1200			const struct rtnl_af_ops *af_ops;
1201
1202			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1203				return -EAFNOSUPPORT;
1204
1205			if (!af_ops->set_link_af)
1206				return -EOPNOTSUPP;
1207
1208			if (af_ops->validate_link_af) {
1209				err = af_ops->validate_link_af(dev, af);
1210				if (err < 0)
1211					return err;
1212			}
1213		}
1214	}
1215
1216	return 0;
1217}
1218
1219static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1220{
1221	int rem, err = -EINVAL;
1222	struct nlattr *vf;
1223	const struct net_device_ops *ops = dev->netdev_ops;
1224
1225	nla_for_each_nested(vf, attr, rem) {
1226		switch (nla_type(vf)) {
1227		case IFLA_VF_MAC: {
1228			struct ifla_vf_mac *ivm;
1229			ivm = nla_data(vf);
1230			err = -EOPNOTSUPP;
1231			if (ops->ndo_set_vf_mac)
1232				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1233							  ivm->mac);
1234			break;
1235		}
1236		case IFLA_VF_VLAN: {
1237			struct ifla_vf_vlan *ivv;
1238			ivv = nla_data(vf);
1239			err = -EOPNOTSUPP;
1240			if (ops->ndo_set_vf_vlan)
1241				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1242							   ivv->vlan,
1243							   ivv->qos);
1244			break;
1245		}
1246		case IFLA_VF_TX_RATE: {
1247			struct ifla_vf_tx_rate *ivt;
1248			ivt = nla_data(vf);
1249			err = -EOPNOTSUPP;
1250			if (ops->ndo_set_vf_tx_rate)
1251				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1252							      ivt->rate);
1253			break;
1254		}
1255		case IFLA_VF_SPOOFCHK: {
1256			struct ifla_vf_spoofchk *ivs;
1257			ivs = nla_data(vf);
1258			err = -EOPNOTSUPP;
1259			if (ops->ndo_set_vf_spoofchk)
1260				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1261							       ivs->setting);
1262			break;
1263		}
1264		default:
1265			err = -EINVAL;
1266			break;
1267		}
1268		if (err)
1269			break;
1270	}
1271	return err;
1272}
1273
1274static int do_set_master(struct net_device *dev, int ifindex)
1275{
1276	struct net_device *master_dev;
1277	const struct net_device_ops *ops;
1278	int err;
1279
1280	if (dev->master) {
1281		if (dev->master->ifindex == ifindex)
1282			return 0;
1283		ops = dev->master->netdev_ops;
1284		if (ops->ndo_del_slave) {
1285			err = ops->ndo_del_slave(dev->master, dev);
1286			if (err)
1287				return err;
1288		} else {
1289			return -EOPNOTSUPP;
1290		}
1291	}
1292
1293	if (ifindex) {
1294		master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1295		if (!master_dev)
1296			return -EINVAL;
1297		ops = master_dev->netdev_ops;
1298		if (ops->ndo_add_slave) {
1299			err = ops->ndo_add_slave(master_dev, dev);
1300			if (err)
1301				return err;
1302		} else {
1303			return -EOPNOTSUPP;
1304		}
1305	}
1306	return 0;
1307}
1308
1309static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1310		      struct nlattr **tb, char *ifname, int modified)
1311{
1312	const struct net_device_ops *ops = dev->netdev_ops;
1313	int send_addr_notify = 0;
1314	int err;
1315
1316	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1317		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1318		if (IS_ERR(net)) {
1319			err = PTR_ERR(net);
1320			goto errout;
1321		}
1322		err = dev_change_net_namespace(dev, net, ifname);
1323		put_net(net);
1324		if (err)
1325			goto errout;
1326		modified = 1;
1327	}
1328
1329	if (tb[IFLA_MAP]) {
1330		struct rtnl_link_ifmap *u_map;
1331		struct ifmap k_map;
1332
1333		if (!ops->ndo_set_config) {
1334			err = -EOPNOTSUPP;
1335			goto errout;
1336		}
1337
1338		if (!netif_device_present(dev)) {
1339			err = -ENODEV;
1340			goto errout;
1341		}
1342
1343		u_map = nla_data(tb[IFLA_MAP]);
1344		k_map.mem_start = (unsigned long) u_map->mem_start;
1345		k_map.mem_end = (unsigned long) u_map->mem_end;
1346		k_map.base_addr = (unsigned short) u_map->base_addr;
1347		k_map.irq = (unsigned char) u_map->irq;
1348		k_map.dma = (unsigned char) u_map->dma;
1349		k_map.port = (unsigned char) u_map->port;
1350
1351		err = ops->ndo_set_config(dev, &k_map);
1352		if (err < 0)
1353			goto errout;
1354
1355		modified = 1;
1356	}
1357
1358	if (tb[IFLA_ADDRESS]) {
1359		struct sockaddr *sa;
1360		int len;
1361
1362		if (!ops->ndo_set_mac_address) {
1363			err = -EOPNOTSUPP;
1364			goto errout;
1365		}
1366
1367		if (!netif_device_present(dev)) {
1368			err = -ENODEV;
1369			goto errout;
1370		}
1371
1372		len = sizeof(sa_family_t) + dev->addr_len;
1373		sa = kmalloc(len, GFP_KERNEL);
1374		if (!sa) {
1375			err = -ENOMEM;
1376			goto errout;
1377		}
1378		sa->sa_family = dev->type;
1379		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1380		       dev->addr_len);
1381		err = ops->ndo_set_mac_address(dev, sa);
1382		kfree(sa);
1383		if (err)
1384			goto errout;
1385		send_addr_notify = 1;
1386		modified = 1;
1387		add_device_randomness(dev->dev_addr, dev->addr_len);
1388	}
1389
1390	if (tb[IFLA_MTU]) {
1391		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1392		if (err < 0)
1393			goto errout;
1394		modified = 1;
1395	}
1396
1397	if (tb[IFLA_GROUP]) {
1398		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1399		modified = 1;
1400	}
1401
1402	/*
1403	 * Interface selected by interface index but interface
1404	 * name provided implies that a name change has been
1405	 * requested.
1406	 */
1407	if (ifm->ifi_index > 0 && ifname[0]) {
1408		err = dev_change_name(dev, ifname);
1409		if (err < 0)
1410			goto errout;
1411		modified = 1;
1412	}
1413
1414	if (tb[IFLA_IFALIAS]) {
1415		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1416				    nla_len(tb[IFLA_IFALIAS]));
1417		if (err < 0)
1418			goto errout;
1419		modified = 1;
1420	}
1421
1422	if (tb[IFLA_BROADCAST]) {
1423		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1424		send_addr_notify = 1;
1425	}
1426
1427	if (ifm->ifi_flags || ifm->ifi_change) {
1428		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1429		if (err < 0)
1430			goto errout;
1431	}
1432
1433	if (tb[IFLA_MASTER]) {
1434		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1435		if (err)
1436			goto errout;
1437		modified = 1;
1438	}
1439
1440	if (tb[IFLA_TXQLEN])
1441		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1442
1443	if (tb[IFLA_OPERSTATE])
1444		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1445
1446	if (tb[IFLA_LINKMODE]) {
1447		write_lock_bh(&dev_base_lock);
1448		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1449		write_unlock_bh(&dev_base_lock);
1450	}
1451
1452	if (tb[IFLA_VFINFO_LIST]) {
1453		struct nlattr *attr;
1454		int rem;
1455		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1456			if (nla_type(attr) != IFLA_VF_INFO) {
1457				err = -EINVAL;
1458				goto errout;
1459			}
1460			err = do_setvfinfo(dev, attr);
1461			if (err < 0)
1462				goto errout;
1463			modified = 1;
1464		}
1465	}
1466	err = 0;
1467
1468	if (tb[IFLA_VF_PORTS]) {
1469		struct nlattr *port[IFLA_PORT_MAX+1];
1470		struct nlattr *attr;
1471		int vf;
1472		int rem;
1473
1474		err = -EOPNOTSUPP;
1475		if (!ops->ndo_set_vf_port)
1476			goto errout;
1477
1478		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1479			if (nla_type(attr) != IFLA_VF_PORT)
1480				continue;
1481			err = nla_parse_nested(port, IFLA_PORT_MAX,
1482				attr, ifla_port_policy);
1483			if (err < 0)
1484				goto errout;
1485			if (!port[IFLA_PORT_VF]) {
1486				err = -EOPNOTSUPP;
1487				goto errout;
1488			}
1489			vf = nla_get_u32(port[IFLA_PORT_VF]);
1490			err = ops->ndo_set_vf_port(dev, vf, port);
1491			if (err < 0)
1492				goto errout;
1493			modified = 1;
1494		}
1495	}
1496	err = 0;
1497
1498	if (tb[IFLA_PORT_SELF]) {
1499		struct nlattr *port[IFLA_PORT_MAX+1];
1500
1501		err = nla_parse_nested(port, IFLA_PORT_MAX,
1502			tb[IFLA_PORT_SELF], ifla_port_policy);
1503		if (err < 0)
1504			goto errout;
1505
1506		err = -EOPNOTSUPP;
1507		if (ops->ndo_set_vf_port)
1508			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1509		if (err < 0)
1510			goto errout;
1511		modified = 1;
1512	}
1513
1514	if (tb[IFLA_AF_SPEC]) {
1515		struct nlattr *af;
1516		int rem;
1517
1518		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1519			const struct rtnl_af_ops *af_ops;
1520
1521			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1522				BUG();
1523
1524			err = af_ops->set_link_af(dev, af);
1525			if (err < 0)
1526				goto errout;
1527
1528			modified = 1;
1529		}
1530	}
1531	err = 0;
1532
1533errout:
1534	if (err < 0 && modified)
1535		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",
1536				     dev->name);
 
 
1537
1538	if (send_addr_notify)
1539		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1540
1541	return err;
1542}
1543
1544static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1545{
1546	struct net *net = sock_net(skb->sk);
1547	struct ifinfomsg *ifm;
1548	struct net_device *dev;
1549	int err;
1550	struct nlattr *tb[IFLA_MAX+1];
1551	char ifname[IFNAMSIZ];
1552
1553	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1554	if (err < 0)
1555		goto errout;
1556
1557	if (tb[IFLA_IFNAME])
1558		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1559	else
1560		ifname[0] = '\0';
1561
1562	err = -EINVAL;
1563	ifm = nlmsg_data(nlh);
1564	if (ifm->ifi_index > 0)
1565		dev = __dev_get_by_index(net, ifm->ifi_index);
1566	else if (tb[IFLA_IFNAME])
1567		dev = __dev_get_by_name(net, ifname);
1568	else
1569		goto errout;
1570
1571	if (dev == NULL) {
1572		err = -ENODEV;
1573		goto errout;
1574	}
1575
1576	err = validate_linkmsg(dev, tb);
1577	if (err < 0)
1578		goto errout;
1579
1580	err = do_setlink(dev, ifm, tb, ifname, 0);
1581errout:
1582	return err;
1583}
1584
1585static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1586{
1587	struct net *net = sock_net(skb->sk);
1588	const struct rtnl_link_ops *ops;
1589	struct net_device *dev;
1590	struct ifinfomsg *ifm;
1591	char ifname[IFNAMSIZ];
1592	struct nlattr *tb[IFLA_MAX+1];
1593	int err;
1594	LIST_HEAD(list_kill);
1595
1596	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1597	if (err < 0)
1598		return err;
1599
1600	if (tb[IFLA_IFNAME])
1601		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1602
1603	ifm = nlmsg_data(nlh);
1604	if (ifm->ifi_index > 0)
1605		dev = __dev_get_by_index(net, ifm->ifi_index);
1606	else if (tb[IFLA_IFNAME])
1607		dev = __dev_get_by_name(net, ifname);
1608	else
1609		return -EINVAL;
1610
1611	if (!dev)
1612		return -ENODEV;
1613
1614	ops = dev->rtnl_link_ops;
1615	if (!ops)
1616		return -EOPNOTSUPP;
1617
1618	ops->dellink(dev, &list_kill);
1619	unregister_netdevice_many(&list_kill);
1620	list_del(&list_kill);
1621	return 0;
1622}
1623
1624int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1625{
1626	unsigned int old_flags;
1627	int err;
1628
1629	old_flags = dev->flags;
1630	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1631		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1632		if (err < 0)
1633			return err;
1634	}
1635
1636	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1637	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1638
1639	__dev_notify_flags(dev, old_flags);
1640	return 0;
1641}
1642EXPORT_SYMBOL(rtnl_configure_link);
1643
1644struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1645	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1646{
1647	int err;
1648	struct net_device *dev;
1649	unsigned int num_queues = 1;
 
1650
1651	if (ops->get_tx_queues) {
1652		err = ops->get_tx_queues(src_net, tb);
1653		if (err < 0)
 
1654			goto err;
1655		num_queues = err;
1656	}
1657
1658	err = -ENOMEM;
1659	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1660	if (!dev)
1661		goto err;
1662
1663	dev_net_set(dev, net);
1664	dev->rtnl_link_ops = ops;
1665	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
 
1666
1667	if (tb[IFLA_MTU])
1668		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1669	if (tb[IFLA_ADDRESS])
1670		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1671				nla_len(tb[IFLA_ADDRESS]));
1672	if (tb[IFLA_BROADCAST])
1673		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1674				nla_len(tb[IFLA_BROADCAST]));
1675	if (tb[IFLA_TXQLEN])
1676		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1677	if (tb[IFLA_OPERSTATE])
1678		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1679	if (tb[IFLA_LINKMODE])
1680		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1681	if (tb[IFLA_GROUP])
1682		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1683
1684	return dev;
1685
1686err:
1687	return ERR_PTR(err);
1688}
1689EXPORT_SYMBOL(rtnl_create_link);
1690
1691static int rtnl_group_changelink(struct net *net, int group,
1692		struct ifinfomsg *ifm,
1693		struct nlattr **tb)
1694{
1695	struct net_device *dev;
1696	int err;
1697
1698	for_each_netdev(net, dev) {
1699		if (dev->group == group) {
1700			err = do_setlink(dev, ifm, tb, NULL, 0);
1701			if (err < 0)
1702				return err;
1703		}
1704	}
1705
1706	return 0;
1707}
1708
1709static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1710{
1711	struct net *net = sock_net(skb->sk);
1712	const struct rtnl_link_ops *ops;
1713	struct net_device *dev;
1714	struct ifinfomsg *ifm;
1715	char kind[MODULE_NAME_LEN];
1716	char ifname[IFNAMSIZ];
1717	struct nlattr *tb[IFLA_MAX+1];
1718	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1719	int err;
1720
1721#ifdef CONFIG_MODULES
1722replay:
1723#endif
1724	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1725	if (err < 0)
1726		return err;
1727
1728	if (tb[IFLA_IFNAME])
1729		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1730	else
1731		ifname[0] = '\0';
1732
1733	ifm = nlmsg_data(nlh);
1734	if (ifm->ifi_index > 0)
1735		dev = __dev_get_by_index(net, ifm->ifi_index);
1736	else {
1737		if (ifname[0])
1738			dev = __dev_get_by_name(net, ifname);
1739		else
1740			dev = NULL;
1741	}
1742
1743	err = validate_linkmsg(dev, tb);
1744	if (err < 0)
1745		return err;
1746
1747	if (tb[IFLA_LINKINFO]) {
1748		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1749				       tb[IFLA_LINKINFO], ifla_info_policy);
1750		if (err < 0)
1751			return err;
1752	} else
1753		memset(linkinfo, 0, sizeof(linkinfo));
1754
1755	if (linkinfo[IFLA_INFO_KIND]) {
1756		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1757		ops = rtnl_link_ops_get(kind);
1758	} else {
1759		kind[0] = '\0';
1760		ops = NULL;
1761	}
1762
1763	if (1) {
1764		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1765		struct net *dest_net;
1766
1767		if (ops) {
1768			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1769				err = nla_parse_nested(attr, ops->maxtype,
1770						       linkinfo[IFLA_INFO_DATA],
1771						       ops->policy);
1772				if (err < 0)
1773					return err;
1774				data = attr;
1775			}
1776			if (ops->validate) {
1777				err = ops->validate(tb, data);
1778				if (err < 0)
1779					return err;
1780			}
1781		}
1782
1783		if (dev) {
1784			int modified = 0;
1785
1786			if (nlh->nlmsg_flags & NLM_F_EXCL)
1787				return -EEXIST;
1788			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1789				return -EOPNOTSUPP;
1790
1791			if (linkinfo[IFLA_INFO_DATA]) {
1792				if (!ops || ops != dev->rtnl_link_ops ||
1793				    !ops->changelink)
1794					return -EOPNOTSUPP;
1795
1796				err = ops->changelink(dev, tb, data);
1797				if (err < 0)
1798					return err;
1799				modified = 1;
1800			}
1801
1802			return do_setlink(dev, ifm, tb, ifname, modified);
1803		}
1804
1805		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1806			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1807				return rtnl_group_changelink(net,
1808						nla_get_u32(tb[IFLA_GROUP]),
1809						ifm, tb);
1810			return -ENODEV;
1811		}
1812
1813		if (ifm->ifi_index)
1814			return -EOPNOTSUPP;
1815		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1816			return -EOPNOTSUPP;
1817
1818		if (!ops) {
1819#ifdef CONFIG_MODULES
1820			if (kind[0]) {
1821				__rtnl_unlock();
1822				request_module("rtnl-link-%s", kind);
1823				rtnl_lock();
1824				ops = rtnl_link_ops_get(kind);
1825				if (ops)
1826					goto replay;
1827			}
1828#endif
1829			return -EOPNOTSUPP;
1830		}
1831
1832		if (!ifname[0])
1833			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1834
1835		dest_net = rtnl_link_get_net(net, tb);
1836		if (IS_ERR(dest_net))
1837			return PTR_ERR(dest_net);
1838
1839		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1840
1841		if (IS_ERR(dev))
1842			err = PTR_ERR(dev);
1843		else if (ops->newlink)
1844			err = ops->newlink(net, dev, tb, data);
1845		else
1846			err = register_netdevice(dev);
1847
1848		if (err < 0 && !IS_ERR(dev))
1849			free_netdev(dev);
1850		if (err < 0)
1851			goto out;
1852
1853		err = rtnl_configure_link(dev, ifm);
1854		if (err < 0)
1855			unregister_netdevice(dev);
1856out:
1857		put_net(dest_net);
1858		return err;
1859	}
1860}
1861
1862static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1863{
1864	struct net *net = sock_net(skb->sk);
1865	struct ifinfomsg *ifm;
1866	char ifname[IFNAMSIZ];
1867	struct nlattr *tb[IFLA_MAX+1];
1868	struct net_device *dev = NULL;
1869	struct sk_buff *nskb;
1870	int err;
1871	u32 ext_filter_mask = 0;
1872
1873	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1874	if (err < 0)
1875		return err;
1876
1877	if (tb[IFLA_IFNAME])
1878		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1879
1880	if (tb[IFLA_EXT_MASK])
1881		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1882
1883	ifm = nlmsg_data(nlh);
1884	if (ifm->ifi_index > 0)
1885		dev = __dev_get_by_index(net, ifm->ifi_index);
1886	else if (tb[IFLA_IFNAME])
1887		dev = __dev_get_by_name(net, ifname);
1888	else
1889		return -EINVAL;
1890
1891	if (dev == NULL)
1892		return -ENODEV;
1893
1894	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1895	if (nskb == NULL)
1896		return -ENOBUFS;
1897
1898	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1899			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1900	if (err < 0) {
1901		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1902		WARN_ON(err == -EMSGSIZE);
1903		kfree_skb(nskb);
1904	} else
1905		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1906
1907	return err;
1908}
1909
1910static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1911{
1912	struct net *net = sock_net(skb->sk);
1913	struct net_device *dev;
1914	struct nlattr *tb[IFLA_MAX+1];
1915	u32 ext_filter_mask = 0;
1916	u16 min_ifinfo_dump_size = 0;
1917
1918	if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1919			ifla_policy) >= 0) {
1920		if (tb[IFLA_EXT_MASK])
1921			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1922	}
1923
1924	if (!ext_filter_mask)
1925		return NLMSG_GOODSIZE;
1926	/*
1927	 * traverse the list of net devices and compute the minimum
1928	 * buffer size based upon the filter mask.
1929	 */
1930	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1931		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1932					     if_nlmsg_size(dev,
1933						           ext_filter_mask));
1934	}
1935
1936	return min_ifinfo_dump_size;
1937}
1938
1939static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1940{
1941	int idx;
1942	int s_idx = cb->family;
1943
1944	if (s_idx == 0)
1945		s_idx = 1;
1946	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1947		int type = cb->nlh->nlmsg_type-RTM_BASE;
1948		if (idx < s_idx || idx == PF_PACKET)
1949			continue;
1950		if (rtnl_msg_handlers[idx] == NULL ||
1951		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1952			continue;
1953		if (idx > s_idx)
1954			memset(&cb->args[0], 0, sizeof(cb->args));
1955		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1956			break;
1957	}
1958	cb->family = idx;
1959
1960	return skb->len;
1961}
1962
1963void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1964{
1965	struct net *net = dev_net(dev);
1966	struct sk_buff *skb;
1967	int err = -ENOBUFS;
1968	size_t if_info_size;
1969
1970	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1971	if (skb == NULL)
1972		goto errout;
1973
1974	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
 
 
1975	if (err < 0) {
1976		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1977		WARN_ON(err == -EMSGSIZE);
1978		kfree_skb(skb);
1979		goto errout;
1980	}
1981	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1982	return;
1983errout:
1984	if (err < 0)
1985		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1986}
1987
1988static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1989				   struct net_device *dev,
1990				   u8 *addr, u32 pid, u32 seq,
1991				   int type, unsigned int flags)
1992{
1993	struct nlmsghdr *nlh;
1994	struct ndmsg *ndm;
1995
1996	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1997	if (!nlh)
1998		return -EMSGSIZE;
1999
2000	ndm = nlmsg_data(nlh);
2001	ndm->ndm_family  = AF_BRIDGE;
2002	ndm->ndm_pad1	 = 0;
2003	ndm->ndm_pad2    = 0;
2004	ndm->ndm_flags	 = flags;
2005	ndm->ndm_type	 = 0;
2006	ndm->ndm_ifindex = dev->ifindex;
2007	ndm->ndm_state   = NUD_PERMANENT;
2008
2009	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2010		goto nla_put_failure;
2011
2012	return nlmsg_end(skb, nlh);
2013
2014nla_put_failure:
2015	nlmsg_cancel(skb, nlh);
2016	return -EMSGSIZE;
2017}
2018
2019static inline size_t rtnl_fdb_nlmsg_size(void)
2020{
2021	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2022}
2023
2024static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2025{
2026	struct net *net = dev_net(dev);
2027	struct sk_buff *skb;
2028	int err = -ENOBUFS;
2029
2030	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2031	if (!skb)
2032		goto errout;
2033
2034	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2035	if (err < 0) {
2036		kfree_skb(skb);
2037		goto errout;
2038	}
2039
2040	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2041	return;
2042errout:
2043	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2044}
2045
2046static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2047{
2048	struct net *net = sock_net(skb->sk);
2049	struct net_device *master = NULL;
2050	struct ndmsg *ndm;
2051	struct nlattr *tb[NDA_MAX+1];
2052	struct net_device *dev;
2053	u8 *addr;
2054	int err;
2055
2056	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2057	if (err < 0)
2058		return err;
2059
2060	ndm = nlmsg_data(nlh);
2061	if (ndm->ndm_ifindex == 0) {
2062		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2063		return -EINVAL;
2064	}
2065
2066	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2067	if (dev == NULL) {
2068		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2069		return -ENODEV;
2070	}
2071
2072	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2073		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2074		return -EINVAL;
2075	}
2076
2077	addr = nla_data(tb[NDA_LLADDR]);
2078	if (!is_valid_ether_addr(addr)) {
2079		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2080		return -EINVAL;
2081	}
2082
2083	err = -EOPNOTSUPP;
2084
2085	/* Support fdb on master device the net/bridge default case */
2086	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2087	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2088		master = dev->master;
2089		err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2090						      nlh->nlmsg_flags);
2091		if (err)
2092			goto out;
2093		else
2094			ndm->ndm_flags &= ~NTF_MASTER;
2095	}
2096
2097	/* Embedded bridge, macvlan, and any other device support */
2098	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2099		err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2100						   nlh->nlmsg_flags);
2101
2102		if (!err) {
2103			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2104			ndm->ndm_flags &= ~NTF_SELF;
2105		}
2106	}
2107out:
2108	return err;
2109}
2110
2111static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2112{
2113	struct net *net = sock_net(skb->sk);
2114	struct ndmsg *ndm;
2115	struct nlattr *llattr;
2116	struct net_device *dev;
2117	int err = -EINVAL;
2118	__u8 *addr;
2119
2120	if (nlmsg_len(nlh) < sizeof(*ndm))
2121		return -EINVAL;
2122
2123	ndm = nlmsg_data(nlh);
2124	if (ndm->ndm_ifindex == 0) {
2125		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2126		return -EINVAL;
2127	}
2128
2129	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2130	if (dev == NULL) {
2131		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2132		return -ENODEV;
2133	}
2134
2135	llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2136	if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2137		pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2138		return -EINVAL;
2139	}
2140
2141	addr = nla_data(llattr);
2142	err = -EOPNOTSUPP;
2143
2144	/* Support fdb on master device the net/bridge default case */
2145	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2146	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2147		struct net_device *master = dev->master;
2148
2149		if (master->netdev_ops->ndo_fdb_del)
2150			err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2151
2152		if (err)
2153			goto out;
2154		else
2155			ndm->ndm_flags &= ~NTF_MASTER;
2156	}
2157
2158	/* Embedded bridge, macvlan, and any other device support */
2159	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2160		err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2161
2162		if (!err) {
2163			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2164			ndm->ndm_flags &= ~NTF_SELF;
2165		}
2166	}
2167out:
2168	return err;
2169}
2170
2171static int nlmsg_populate_fdb(struct sk_buff *skb,
2172			      struct netlink_callback *cb,
2173			      struct net_device *dev,
2174			      int *idx,
2175			      struct netdev_hw_addr_list *list)
2176{
2177	struct netdev_hw_addr *ha;
2178	int err;
2179	u32 pid, seq;
2180
2181	pid = NETLINK_CB(cb->skb).pid;
2182	seq = cb->nlh->nlmsg_seq;
2183
2184	list_for_each_entry(ha, &list->list, list) {
2185		if (*idx < cb->args[0])
2186			goto skip;
2187
2188		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2189					      pid, seq, 0, NTF_SELF);
2190		if (err < 0)
2191			return err;
2192skip:
2193		*idx += 1;
2194	}
2195	return 0;
2196}
2197
2198/**
2199 * ndo_dflt_fdb_dump: default netdevice operation to dump an FDB table.
2200 * @nlh: netlink message header
2201 * @dev: netdevice
2202 *
2203 * Default netdevice operation to dump the existing unicast address list.
2204 * Returns zero on success.
2205 */
2206int ndo_dflt_fdb_dump(struct sk_buff *skb,
2207		      struct netlink_callback *cb,
2208		      struct net_device *dev,
2209		      int idx)
2210{
2211	int err;
2212
2213	netif_addr_lock_bh(dev);
2214	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2215	if (err)
2216		goto out;
2217	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2218out:
2219	netif_addr_unlock_bh(dev);
2220	return idx;
2221}
2222EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2223
2224static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2225{
2226	int idx = 0;
2227	struct net *net = sock_net(skb->sk);
2228	struct net_device *dev;
2229
2230	rcu_read_lock();
2231	for_each_netdev_rcu(net, dev) {
2232		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2233			struct net_device *master = dev->master;
2234			const struct net_device_ops *ops = master->netdev_ops;
2235
2236			if (ops->ndo_fdb_dump)
2237				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2238		}
2239
2240		if (dev->netdev_ops->ndo_fdb_dump)
2241			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2242	}
2243	rcu_read_unlock();
2244
2245	cb->args[0] = idx;
2246	return skb->len;
2247}
2248
2249/* Protected by RTNL sempahore.  */
2250static struct rtattr **rta_buf;
2251static int rtattr_max;
2252
2253/* Process one rtnetlink message. */
2254
2255static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2256{
2257	struct net *net = sock_net(skb->sk);
2258	rtnl_doit_func doit;
2259	int sz_idx, kind;
2260	int min_len;
2261	int family;
2262	int type;
2263	int err;
2264
2265	type = nlh->nlmsg_type;
2266	if (type > RTM_MAX)
2267		return -EOPNOTSUPP;
2268
2269	type -= RTM_BASE;
2270
2271	/* All the messages must have at least 1 byte length */
2272	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2273		return 0;
2274
2275	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2276	sz_idx = type>>2;
2277	kind = type&3;
2278
2279	if (kind != 2 && !capable(CAP_NET_ADMIN))
2280		return -EPERM;
2281
2282	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2283		struct sock *rtnl;
2284		rtnl_dumpit_func dumpit;
2285		rtnl_calcit_func calcit;
2286		u16 min_dump_alloc = 0;
2287
2288		dumpit = rtnl_get_dumpit(family, type);
2289		if (dumpit == NULL)
2290			return -EOPNOTSUPP;
2291		calcit = rtnl_get_calcit(family, type);
2292		if (calcit)
2293			min_dump_alloc = calcit(skb, nlh);
2294
2295		__rtnl_unlock();
2296		rtnl = net->rtnl;
2297		{
2298			struct netlink_dump_control c = {
2299				.dump		= dumpit,
2300				.min_dump_alloc	= min_dump_alloc,
2301			};
2302			err = netlink_dump_start(rtnl, skb, nlh, &c);
2303		}
2304		rtnl_lock();
2305		return err;
2306	}
2307
2308	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2309
2310	min_len = rtm_min[sz_idx];
2311	if (nlh->nlmsg_len < min_len)
2312		return -EINVAL;
2313
2314	if (nlh->nlmsg_len > min_len) {
2315		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2316		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2317
2318		while (RTA_OK(attr, attrlen)) {
2319			unsigned int flavor = attr->rta_type;
2320			if (flavor) {
2321				if (flavor > rta_max[sz_idx])
2322					return -EINVAL;
2323				rta_buf[flavor-1] = attr;
2324			}
2325			attr = RTA_NEXT(attr, attrlen);
2326		}
2327	}
2328
2329	doit = rtnl_get_doit(family, type);
2330	if (doit == NULL)
2331		return -EOPNOTSUPP;
2332
2333	return doit(skb, nlh, (void *)&rta_buf[0]);
2334}
2335
2336static void rtnetlink_rcv(struct sk_buff *skb)
2337{
2338	rtnl_lock();
2339	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2340	rtnl_unlock();
2341}
2342
2343static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2344{
2345	struct net_device *dev = ptr;
2346
2347	switch (event) {
2348	case NETDEV_UP:
2349	case NETDEV_DOWN:
2350	case NETDEV_PRE_UP:
2351	case NETDEV_POST_INIT:
2352	case NETDEV_REGISTER:
2353	case NETDEV_CHANGE:
2354	case NETDEV_PRE_TYPE_CHANGE:
2355	case NETDEV_GOING_DOWN:
2356	case NETDEV_UNREGISTER:
2357	case NETDEV_UNREGISTER_BATCH:
2358	case NETDEV_RELEASE:
2359	case NETDEV_JOIN:
2360		break;
2361	default:
2362		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2363		break;
2364	}
2365	return NOTIFY_DONE;
2366}
2367
2368static struct notifier_block rtnetlink_dev_notifier = {
2369	.notifier_call	= rtnetlink_event,
2370};
2371
2372
2373static int __net_init rtnetlink_net_init(struct net *net)
2374{
2375	struct sock *sk;
2376	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2377				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2378	if (!sk)
2379		return -ENOMEM;
2380	net->rtnl = sk;
2381	return 0;
2382}
2383
2384static void __net_exit rtnetlink_net_exit(struct net *net)
2385{
2386	netlink_kernel_release(net->rtnl);
2387	net->rtnl = NULL;
2388}
2389
2390static struct pernet_operations rtnetlink_net_ops = {
2391	.init = rtnetlink_net_init,
2392	.exit = rtnetlink_net_exit,
2393};
2394
2395void __init rtnetlink_init(void)
2396{
2397	int i;
2398
2399	rtattr_max = 0;
2400	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2401		if (rta_max[i] > rtattr_max)
2402			rtattr_max = rta_max[i];
2403	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2404	if (!rta_buf)
2405		panic("rtnetlink_init: cannot allocate rta_buf\n");
2406
2407	if (register_pernet_subsys(&rtnetlink_net_ops))
2408		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2409
2410	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2411	register_netdevice_notifier(&rtnetlink_dev_notifier);
2412
2413	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2414		      rtnl_dump_ifinfo, rtnl_calcit);
2415	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2416	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2417	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2418
2419	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2420	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2421
2422	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2423	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2424	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2425}
2426