Linux Audio

Check our new training course

Loading...
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	NET3	IP device support routines.
   4 *
 
 
 
 
 
   5 *	Derived from the IP parts of dev.c 1.0.19
   6 * 		Authors:	Ross Biro
   7 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   8 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
   9 *
  10 *	Additional Authors:
  11 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
  12 *		Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  13 *
  14 *	Changes:
  15 *		Alexey Kuznetsov:	pa_* fields are replaced with ifaddr
  16 *					lists.
  17 *		Cyrus Durgin:		updated for kmod
  18 *		Matthias Andree:	in devinet_ioctl, compare label and
  19 *					address (4.4BSD alias style support),
  20 *					fall back to comparing just the label
  21 *					if no match found.
  22 */
  23
  24
  25#include <linux/uaccess.h>
  26#include <linux/bitops.h>
  27#include <linux/capability.h>
  28#include <linux/module.h>
  29#include <linux/types.h>
  30#include <linux/kernel.h>
  31#include <linux/sched/signal.h>
  32#include <linux/string.h>
  33#include <linux/mm.h>
  34#include <linux/socket.h>
  35#include <linux/sockios.h>
  36#include <linux/in.h>
  37#include <linux/errno.h>
  38#include <linux/interrupt.h>
  39#include <linux/if_addr.h>
  40#include <linux/if_ether.h>
  41#include <linux/inet.h>
  42#include <linux/netdevice.h>
  43#include <linux/etherdevice.h>
  44#include <linux/skbuff.h>
  45#include <linux/init.h>
  46#include <linux/notifier.h>
  47#include <linux/inetdevice.h>
  48#include <linux/igmp.h>
  49#include <linux/slab.h>
  50#include <linux/hash.h>
  51#ifdef CONFIG_SYSCTL
  52#include <linux/sysctl.h>
  53#endif
  54#include <linux/kmod.h>
  55#include <linux/netconf.h>
  56
  57#include <net/arp.h>
  58#include <net/ip.h>
  59#include <net/route.h>
  60#include <net/ip_fib.h>
  61#include <net/rtnetlink.h>
  62#include <net/net_namespace.h>
  63#include <net/addrconf.h>
  64
  65#define IPV6ONLY_FLAGS	\
  66		(IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \
  67		 IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \
  68		 IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY)
  69
  70static struct ipv4_devconf ipv4_devconf = {
  71	.data = {
  72		[IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  73		[IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  74		[IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  75		[IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
  76		[IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
  77		[IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] =  1000 /*ms*/,
  78		[IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1,
  79	},
  80};
  81
  82static struct ipv4_devconf ipv4_devconf_dflt = {
  83	.data = {
  84		[IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  85		[IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  86		[IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  87		[IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
  88		[IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
  89		[IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
  90		[IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] =  1000 /*ms*/,
  91		[IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1,
  92	},
  93};
  94
  95#define IPV4_DEVCONF_DFLT(net, attr) \
  96	IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
  97
  98static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
  99	[IFA_LOCAL]     	= { .type = NLA_U32 },
 100	[IFA_ADDRESS]   	= { .type = NLA_U32 },
 101	[IFA_BROADCAST] 	= { .type = NLA_U32 },
 102	[IFA_LABEL]     	= { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
 103	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
 104	[IFA_FLAGS]		= { .type = NLA_U32 },
 105	[IFA_RT_PRIORITY]	= { .type = NLA_U32 },
 106	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
 107	[IFA_PROTO]		= { .type = NLA_U8 },
 108};
 109
 110struct inet_fill_args {
 111	u32 portid;
 112	u32 seq;
 113	int event;
 114	unsigned int flags;
 115	int netnsid;
 116	int ifindex;
 117};
 118
 119#define IN4_ADDR_HSIZE_SHIFT	8
 120#define IN4_ADDR_HSIZE		(1U << IN4_ADDR_HSIZE_SHIFT)
 
 
 
 
 
 121
 122static u32 inet_addr_hash(const struct net *net, __be32 addr)
 123{
 124	u32 val = __ipv4_addr_hash(addr, net_hash_mix(net));
 125
 126	return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
 
 127}
 128
 129static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
 130{
 131	u32 hash = inet_addr_hash(net, ifa->ifa_local);
 132
 133	ASSERT_RTNL();
 134	hlist_add_head_rcu(&ifa->addr_lst, &net->ipv4.inet_addr_lst[hash]);
 
 135}
 136
 137static void inet_hash_remove(struct in_ifaddr *ifa)
 138{
 139	ASSERT_RTNL();
 140	hlist_del_init_rcu(&ifa->addr_lst);
 
 141}
 142
 143/**
 144 * __ip_dev_find - find the first device with a given source address.
 145 * @net: the net namespace
 146 * @addr: the source address
 147 * @devref: if true, take a reference on the found device
 148 *
 149 * If a caller uses devref=false, it should be protected by RCU, or RTNL
 150 */
 151struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
 152{
 
 153	struct net_device *result = NULL;
 154	struct in_ifaddr *ifa;
 
 155
 156	rcu_read_lock();
 157	ifa = inet_lookup_ifaddr_rcu(net, addr);
 158	if (!ifa) {
 
 
 
 
 
 
 
 
 
 159		struct flowi4 fl4 = { .daddr = addr };
 160		struct fib_result res = { 0 };
 161		struct fib_table *local;
 162
 163		/* Fallback to FIB local table so that communication
 164		 * over loopback subnets work.
 165		 */
 166		local = fib_get_table(net, RT_TABLE_LOCAL);
 167		if (local &&
 168		    !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
 169		    res.type == RTN_LOCAL)
 170			result = FIB_RES_DEV(res);
 171	} else {
 172		result = ifa->ifa_dev->dev;
 173	}
 174	if (result && devref)
 175		dev_hold(result);
 176	rcu_read_unlock();
 177	return result;
 178}
 179EXPORT_SYMBOL(__ip_dev_find);
 180
 181/* called under RCU lock */
 182struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
 183{
 184	u32 hash = inet_addr_hash(net, addr);
 185	struct in_ifaddr *ifa;
 186
 187	hlist_for_each_entry_rcu(ifa, &net->ipv4.inet_addr_lst[hash], addr_lst)
 188		if (ifa->ifa_local == addr)
 189			return ifa;
 190
 191	return NULL;
 192}
 193
 194static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
 195
 196static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
 197static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
 198static void inet_del_ifa(struct in_device *in_dev,
 199			 struct in_ifaddr __rcu **ifap,
 200			 int destroy);
 201#ifdef CONFIG_SYSCTL
 202static int devinet_sysctl_register(struct in_device *idev);
 203static void devinet_sysctl_unregister(struct in_device *idev);
 204#else
 205static int devinet_sysctl_register(struct in_device *idev)
 206{
 207	return 0;
 208}
 209static void devinet_sysctl_unregister(struct in_device *idev)
 210{
 211}
 212#endif
 213
 214/* Locks all the inet devices. */
 215
 216static struct in_ifaddr *inet_alloc_ifa(struct in_device *in_dev)
 217{
 218	struct in_ifaddr *ifa;
 219
 220	ifa = kzalloc(sizeof(*ifa), GFP_KERNEL_ACCOUNT);
 221	if (!ifa)
 222		return NULL;
 223
 224	in_dev_hold(in_dev);
 225	ifa->ifa_dev = in_dev;
 226
 227	INIT_HLIST_NODE(&ifa->addr_lst);
 228
 229	return ifa;
 230}
 231
 232static void inet_rcu_free_ifa(struct rcu_head *head)
 233{
 234	struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 235
 236	in_dev_put(ifa->ifa_dev);
 237	kfree(ifa);
 238}
 239
 240static void inet_free_ifa(struct in_ifaddr *ifa)
 241{
 242	/* Our reference to ifa->ifa_dev must be freed ASAP
 243	 * to release the reference to the netdev the same way.
 244	 * in_dev_put() -> in_dev_finish_destroy() -> netdev_put()
 245	 */
 246	call_rcu_hurry(&ifa->rcu_head, inet_rcu_free_ifa);
 247}
 248
 249static void in_dev_free_rcu(struct rcu_head *head)
 250{
 251	struct in_device *idev = container_of(head, struct in_device, rcu_head);
 252
 253	kfree(rcu_dereference_protected(idev->mc_hash, 1));
 254	kfree(idev);
 255}
 256
 257void in_dev_finish_destroy(struct in_device *idev)
 258{
 259	struct net_device *dev = idev->dev;
 260
 261	WARN_ON(idev->ifa_list);
 262	WARN_ON(idev->mc_list);
 263#ifdef NET_REFCNT_DEBUG
 264	pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
 265#endif
 266	netdev_put(dev, &idev->dev_tracker);
 267	if (!idev->dead)
 268		pr_err("Freeing alive in_device %p\n", idev);
 269	else
 270		call_rcu(&idev->rcu_head, in_dev_free_rcu);
 271}
 272EXPORT_SYMBOL(in_dev_finish_destroy);
 273
 274static struct in_device *inetdev_init(struct net_device *dev)
 275{
 276	struct in_device *in_dev;
 277	int err = -ENOMEM;
 278
 279	ASSERT_RTNL();
 280
 281	in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
 282	if (!in_dev)
 283		goto out;
 284	memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
 285			sizeof(in_dev->cnf));
 286	in_dev->cnf.sysctl = NULL;
 287	in_dev->dev = dev;
 288	in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
 289	if (!in_dev->arp_parms)
 290		goto out_kfree;
 291	if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
 292		dev_disable_lro(dev);
 293	/* Reference in_dev->dev */
 294	netdev_hold(dev, &in_dev->dev_tracker, GFP_KERNEL);
 295	/* Account for reference dev->ip_ptr (below) */
 296	refcount_set(&in_dev->refcnt, 1);
 297
 298	if (dev != blackhole_netdev) {
 299		err = devinet_sysctl_register(in_dev);
 300		if (err) {
 301			in_dev->dead = 1;
 302			neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 303			in_dev_put(in_dev);
 304			in_dev = NULL;
 305			goto out;
 306		}
 307		ip_mc_init_dev(in_dev);
 308		if (dev->flags & IFF_UP)
 309			ip_mc_up(in_dev);
 310	}
 311
 312	/* we can receive as soon as ip_ptr is set -- do this last */
 313	rcu_assign_pointer(dev->ip_ptr, in_dev);
 314out:
 315	return in_dev ?: ERR_PTR(err);
 316out_kfree:
 317	kfree(in_dev);
 318	in_dev = NULL;
 319	goto out;
 320}
 321
 
 
 
 
 
 
 322static void inetdev_destroy(struct in_device *in_dev)
 323{
 324	struct net_device *dev;
 325	struct in_ifaddr *ifa;
 
 326
 327	ASSERT_RTNL();
 328
 329	dev = in_dev->dev;
 330
 331	in_dev->dead = 1;
 332
 333	ip_mc_destroy_dev(in_dev);
 334
 335	while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
 336		inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 337		inet_free_ifa(ifa);
 338	}
 339
 340	RCU_INIT_POINTER(dev->ip_ptr, NULL);
 341
 342	devinet_sysctl_unregister(in_dev);
 343	neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 344	arp_ifdown(dev);
 345
 346	in_dev_put(in_dev);
 347}
 348
 349static int __init inet_blackhole_dev_init(void)
 350{
 351	int err = 0;
 352
 353	rtnl_lock();
 354	if (!inetdev_init(blackhole_netdev))
 355		err = -ENOMEM;
 356	rtnl_unlock();
 357
 358	return err;
 359}
 360late_initcall(inet_blackhole_dev_init);
 361
 362int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 363{
 364	const struct in_ifaddr *ifa;
 365
 366	rcu_read_lock();
 367	in_dev_for_each_ifa_rcu(ifa, in_dev) {
 368		if (inet_ifa_match(a, ifa)) {
 369			if (!b || inet_ifa_match(b, ifa)) {
 370				rcu_read_unlock();
 371				return 1;
 372			}
 373		}
 374	}
 375	rcu_read_unlock();
 376	return 0;
 377}
 378
 379static void __inet_del_ifa(struct in_device *in_dev,
 380			   struct in_ifaddr __rcu **ifap,
 381			   int destroy, struct nlmsghdr *nlh, u32 portid)
 382{
 383	struct in_ifaddr *promote = NULL;
 384	struct in_ifaddr *ifa, *ifa1;
 385	struct in_ifaddr __rcu **last_prim;
 386	struct in_ifaddr *prev_prom = NULL;
 387	int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 388
 389	ASSERT_RTNL();
 390
 391	ifa1 = rtnl_dereference(*ifap);
 392	last_prim = ifap;
 393	if (in_dev->dead)
 394		goto no_promotions;
 395
 396	/* 1. Deleting primary ifaddr forces deletion all secondaries
 397	 * unless alias promotion is set
 398	 **/
 399
 400	if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 401		struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
 402
 403		while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
 404			if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 405			    ifa1->ifa_scope <= ifa->ifa_scope)
 406				last_prim = &ifa->ifa_next;
 407
 408			if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 409			    ifa1->ifa_mask != ifa->ifa_mask ||
 410			    !inet_ifa_match(ifa1->ifa_address, ifa)) {
 411				ifap1 = &ifa->ifa_next;
 412				prev_prom = ifa;
 413				continue;
 414			}
 415
 416			if (!do_promote) {
 417				inet_hash_remove(ifa);
 418				*ifap1 = ifa->ifa_next;
 419
 420				rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
 421				blocking_notifier_call_chain(&inetaddr_chain,
 422						NETDEV_DOWN, ifa);
 423				inet_free_ifa(ifa);
 424			} else {
 425				promote = ifa;
 426				break;
 427			}
 428		}
 429	}
 430
 431	/* On promotion all secondaries from subnet are changing
 432	 * the primary IP, we must remove all their routes silently
 433	 * and later to add them back with new prefsrc. Do this
 434	 * while all addresses are on the device list.
 435	 */
 436	for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
 437		if (ifa1->ifa_mask == ifa->ifa_mask &&
 438		    inet_ifa_match(ifa1->ifa_address, ifa))
 439			fib_del_ifaddr(ifa, ifa1);
 440	}
 441
 442no_promotions:
 443	/* 2. Unlink it */
 444
 445	*ifap = ifa1->ifa_next;
 446	inet_hash_remove(ifa1);
 447
 448	/* 3. Announce address deletion */
 449
 450	/* Send message first, then call notifier.
 451	   At first sight, FIB update triggered by notifier
 452	   will refer to already deleted ifaddr, that could confuse
 453	   netlink listeners. It is not true: look, gated sees
 454	   that route deleted and if it still thinks that ifaddr
 455	   is valid, it will try to restore deleted routes... Grr.
 456	   So that, this order is correct.
 457	 */
 458	rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
 459	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 460
 461	if (promote) {
 462		struct in_ifaddr *next_sec;
 463
 464		next_sec = rtnl_dereference(promote->ifa_next);
 465		if (prev_prom) {
 466			struct in_ifaddr *last_sec;
 467
 468			rcu_assign_pointer(prev_prom->ifa_next, next_sec);
 469
 470			last_sec = rtnl_dereference(*last_prim);
 471			rcu_assign_pointer(promote->ifa_next, last_sec);
 472			rcu_assign_pointer(*last_prim, promote);
 473		}
 474
 475		promote->ifa_flags &= ~IFA_F_SECONDARY;
 476		rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
 477		blocking_notifier_call_chain(&inetaddr_chain,
 478				NETDEV_UP, promote);
 479		for (ifa = next_sec; ifa;
 480		     ifa = rtnl_dereference(ifa->ifa_next)) {
 481			if (ifa1->ifa_mask != ifa->ifa_mask ||
 482			    !inet_ifa_match(ifa1->ifa_address, ifa))
 483					continue;
 484			fib_add_ifaddr(ifa);
 485		}
 486
 487	}
 488	if (destroy)
 489		inet_free_ifa(ifa1);
 490}
 491
 492static void inet_del_ifa(struct in_device *in_dev,
 493			 struct in_ifaddr __rcu **ifap,
 494			 int destroy)
 495{
 496	__inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 497}
 498
 499static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 500			     u32 portid, struct netlink_ext_ack *extack)
 501{
 502	struct in_ifaddr __rcu **last_primary, **ifap;
 503	struct in_device *in_dev = ifa->ifa_dev;
 504	struct net *net = dev_net(in_dev->dev);
 505	struct in_validator_info ivi;
 506	struct in_ifaddr *ifa1;
 507	int ret;
 508
 509	ASSERT_RTNL();
 510
 
 
 
 
 
 511	ifa->ifa_flags &= ~IFA_F_SECONDARY;
 512	last_primary = &in_dev->ifa_list;
 513
 514	/* Don't set IPv6 only flags to IPv4 addresses */
 515	ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
 516
 517	ifap = &in_dev->ifa_list;
 518	ifa1 = rtnl_dereference(*ifap);
 519
 520	while (ifa1) {
 521		if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 522		    ifa->ifa_scope <= ifa1->ifa_scope)
 523			last_primary = &ifa1->ifa_next;
 524		if (ifa1->ifa_mask == ifa->ifa_mask &&
 525		    inet_ifa_match(ifa1->ifa_address, ifa)) {
 526			if (ifa1->ifa_local == ifa->ifa_local) {
 527				inet_free_ifa(ifa);
 528				return -EEXIST;
 529			}
 530			if (ifa1->ifa_scope != ifa->ifa_scope) {
 531				NL_SET_ERR_MSG(extack, "ipv4: Invalid scope value");
 532				inet_free_ifa(ifa);
 533				return -EINVAL;
 534			}
 535			ifa->ifa_flags |= IFA_F_SECONDARY;
 536		}
 537
 538		ifap = &ifa1->ifa_next;
 539		ifa1 = rtnl_dereference(*ifap);
 540	}
 541
 542	/* Allow any devices that wish to register ifaddr validtors to weigh
 543	 * in now, before changes are committed.  The rntl lock is serializing
 544	 * access here, so the state should not change between a validator call
 545	 * and a final notify on commit.  This isn't invoked on promotion under
 546	 * the assumption that validators are checking the address itself, and
 547	 * not the flags.
 548	 */
 549	ivi.ivi_addr = ifa->ifa_address;
 550	ivi.ivi_dev = ifa->ifa_dev;
 551	ivi.extack = extack;
 552	ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
 553					   NETDEV_UP, &ivi);
 554	ret = notifier_to_errno(ret);
 555	if (ret) {
 556		inet_free_ifa(ifa);
 557		return ret;
 558	}
 559
 560	if (!(ifa->ifa_flags & IFA_F_SECONDARY))
 561		ifap = last_primary;
 
 562
 563	rcu_assign_pointer(ifa->ifa_next, *ifap);
 564	rcu_assign_pointer(*ifap, ifa);
 565
 566	inet_hash_insert(dev_net(in_dev->dev), ifa);
 567
 568	cancel_delayed_work(&net->ipv4.addr_chk_work);
 569	queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work, 0);
 570
 571	/* Send message first, then call notifier.
 572	   Notifier will trigger FIB update, so that
 573	   listeners of netlink will know about new ifaddr */
 574	rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
 575	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 576
 577	return 0;
 578}
 579
 580static int inet_insert_ifa(struct in_ifaddr *ifa)
 581{
 582	if (!ifa->ifa_local) {
 583		inet_free_ifa(ifa);
 584		return 0;
 585	}
 586
 587	return __inet_insert_ifa(ifa, NULL, 0, NULL);
 588}
 589
 590static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 591{
 592	struct in_device *in_dev = __in_dev_get_rtnl_net(dev);
 593
 594	ipv4_devconf_setall(in_dev);
 595	neigh_parms_data_state_setall(in_dev->arp_parms);
 596
 
 
 
 
 
 
 
 
 
 
 597	if (ipv4_is_loopback(ifa->ifa_local))
 598		ifa->ifa_scope = RT_SCOPE_HOST;
 599	return inet_insert_ifa(ifa);
 600}
 601
 602/* Caller must hold RCU or RTNL :
 603 * We dont take a reference on found in_device
 604 */
 605struct in_device *inetdev_by_index(struct net *net, int ifindex)
 606{
 607	struct net_device *dev;
 608	struct in_device *in_dev = NULL;
 609
 610	rcu_read_lock();
 611	dev = dev_get_by_index_rcu(net, ifindex);
 612	if (dev)
 613		in_dev = rcu_dereference_rtnl(dev->ip_ptr);
 614	rcu_read_unlock();
 615	return in_dev;
 616}
 617EXPORT_SYMBOL(inetdev_by_index);
 618
 619/* Called only from RTNL semaphored context. No locks. */
 620
 621struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 622				    __be32 mask)
 623{
 624	struct in_ifaddr *ifa;
 625
 626	ASSERT_RTNL();
 627
 628	in_dev_for_each_ifa_rtnl(ifa, in_dev) {
 629		if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 630			return ifa;
 631	}
 632	return NULL;
 633}
 634
 635static int ip_mc_autojoin_config(struct net *net, bool join,
 636				 const struct in_ifaddr *ifa)
 637{
 638#if defined(CONFIG_IP_MULTICAST)
 639	struct ip_mreqn mreq = {
 640		.imr_multiaddr.s_addr = ifa->ifa_address,
 641		.imr_ifindex = ifa->ifa_dev->dev->ifindex,
 642	};
 643	struct sock *sk = net->ipv4.mc_autojoin_sk;
 644	int ret;
 645
 646	ASSERT_RTNL_NET(net);
 647
 648	lock_sock(sk);
 649	if (join)
 650		ret = ip_mc_join_group(sk, &mreq);
 651	else
 652		ret = ip_mc_leave_group(sk, &mreq);
 653	release_sock(sk);
 654
 655	return ret;
 656#else
 657	return -EOPNOTSUPP;
 658#endif
 659}
 660
 661static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
 662			    struct netlink_ext_ack *extack)
 663{
 664	struct net *net = sock_net(skb->sk);
 665	struct in_ifaddr __rcu **ifap;
 666	struct nlattr *tb[IFA_MAX+1];
 667	struct in_device *in_dev;
 668	struct ifaddrmsg *ifm;
 669	struct in_ifaddr *ifa;
 670	int err;
 
 
 671
 672	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 673				     ifa_ipv4_policy, extack);
 674	if (err < 0)
 675		goto out;
 676
 677	ifm = nlmsg_data(nlh);
 678
 679	rtnl_net_lock(net);
 680
 681	in_dev = inetdev_by_index(net, ifm->ifa_index);
 682	if (!in_dev) {
 683		NL_SET_ERR_MSG(extack, "ipv4: Device not found");
 684		err = -ENODEV;
 685		goto unlock;
 686	}
 687
 688	for (ifap = &in_dev->ifa_list;
 689	     (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
 690	     ifap = &ifa->ifa_next) {
 691		if (tb[IFA_LOCAL] &&
 692		    ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
 693			continue;
 694
 695		if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 696			continue;
 697
 698		if (tb[IFA_ADDRESS] &&
 699		    (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 700		    !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
 701			continue;
 702
 703		if (ipv4_is_multicast(ifa->ifa_address))
 704			ip_mc_autojoin_config(net, false, ifa);
 705
 706		__inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
 707		goto unlock;
 708	}
 709
 710	NL_SET_ERR_MSG(extack, "ipv4: Address not found");
 711	err = -EADDRNOTAVAIL;
 712unlock:
 713	rtnl_net_unlock(net);
 714out:
 715	return err;
 716}
 717
 718static void check_lifetime(struct work_struct *work)
 719{
 720	unsigned long now, next, next_sec, next_sched;
 721	struct in_ifaddr *ifa;
 722	struct hlist_node *n;
 723	struct net *net;
 724	int i;
 725
 726	net = container_of(to_delayed_work(work), struct net, ipv4.addr_chk_work);
 727	now = jiffies;
 728	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
 729
 730	for (i = 0; i < IN4_ADDR_HSIZE; i++) {
 731		struct hlist_head *head = &net->ipv4.inet_addr_lst[i];
 732		bool change_needed = false;
 733
 734		rcu_read_lock();
 735		hlist_for_each_entry_rcu(ifa, head, addr_lst) {
 736			unsigned long age, tstamp;
 737			u32 preferred_lft;
 738			u32 valid_lft;
 739			u32 flags;
 740
 741			flags = READ_ONCE(ifa->ifa_flags);
 742			if (flags & IFA_F_PERMANENT)
 743				continue;
 744
 745			preferred_lft = READ_ONCE(ifa->ifa_preferred_lft);
 746			valid_lft = READ_ONCE(ifa->ifa_valid_lft);
 747			tstamp = READ_ONCE(ifa->ifa_tstamp);
 748			/* We try to batch several events at once. */
 749			age = (now - tstamp +
 750			       ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 751
 752			if (valid_lft != INFINITY_LIFE_TIME &&
 753			    age >= valid_lft) {
 754				change_needed = true;
 755			} else if (preferred_lft ==
 756				   INFINITY_LIFE_TIME) {
 757				continue;
 758			} else if (age >= preferred_lft) {
 759				if (time_before(tstamp + valid_lft * HZ, next))
 760					next = tstamp + valid_lft * HZ;
 761
 762				if (!(flags & IFA_F_DEPRECATED))
 763					change_needed = true;
 764			} else if (time_before(tstamp + preferred_lft * HZ,
 765					       next)) {
 766				next = tstamp + preferred_lft * HZ;
 767			}
 768		}
 769		rcu_read_unlock();
 770		if (!change_needed)
 771			continue;
 772
 773		rtnl_net_lock(net);
 774		hlist_for_each_entry_safe(ifa, n, head, addr_lst) {
 775			unsigned long age;
 776
 777			if (ifa->ifa_flags & IFA_F_PERMANENT)
 778				continue;
 779
 780			/* We try to batch several events at once. */
 781			age = (now - ifa->ifa_tstamp +
 782			       ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 783
 784			if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 785			    age >= ifa->ifa_valid_lft) {
 786				struct in_ifaddr __rcu **ifap;
 787				struct in_ifaddr *tmp;
 788
 789				ifap = &ifa->ifa_dev->ifa_list;
 790				tmp = rtnl_net_dereference(net, *ifap);
 791				while (tmp) {
 792					if (tmp == ifa) {
 793						inet_del_ifa(ifa->ifa_dev,
 794							     ifap, 1);
 795						break;
 796					}
 797					ifap = &tmp->ifa_next;
 798					tmp = rtnl_net_dereference(net, *ifap);
 799				}
 800			} else if (ifa->ifa_preferred_lft !=
 801				   INFINITY_LIFE_TIME &&
 802				   age >= ifa->ifa_preferred_lft &&
 803				   !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
 804				ifa->ifa_flags |= IFA_F_DEPRECATED;
 805				rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
 806			}
 807		}
 808		rtnl_net_unlock(net);
 809	}
 810
 811	next_sec = round_jiffies_up(next);
 812	next_sched = next;
 813
 814	/* If rounded timeout is accurate enough, accept it. */
 815	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
 816		next_sched = next_sec;
 817
 818	now = jiffies;
 819	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
 820	if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
 821		next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
 822
 823	queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work,
 824			   next_sched - now);
 825}
 826
 827static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
 828			     __u32 prefered_lft)
 829{
 830	unsigned long timeout;
 831	u32 flags;
 832
 833	flags = ifa->ifa_flags & ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
 834
 835	timeout = addrconf_timeout_fixup(valid_lft, HZ);
 836	if (addrconf_finite_timeout(timeout))
 837		WRITE_ONCE(ifa->ifa_valid_lft, timeout);
 838	else
 839		flags |= IFA_F_PERMANENT;
 840
 841	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
 842	if (addrconf_finite_timeout(timeout)) {
 843		if (timeout == 0)
 844			flags |= IFA_F_DEPRECATED;
 845		WRITE_ONCE(ifa->ifa_preferred_lft, timeout);
 846	}
 847	WRITE_ONCE(ifa->ifa_flags, flags);
 848	WRITE_ONCE(ifa->ifa_tstamp, jiffies);
 849	if (!ifa->ifa_cstamp)
 850		WRITE_ONCE(ifa->ifa_cstamp, ifa->ifa_tstamp);
 851}
 852
 853static int inet_validate_rtm(struct nlmsghdr *nlh, struct nlattr **tb,
 854			     struct netlink_ext_ack *extack,
 855			     __u32 *valid_lft, __u32 *prefered_lft)
 856{
 857	struct ifaddrmsg *ifm = nlmsg_data(nlh);
 858	int err;
 859
 860	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 861				     ifa_ipv4_policy, extack);
 862	if (err < 0)
 863		return err;
 864
 865	if (ifm->ifa_prefixlen > 32) {
 866		NL_SET_ERR_MSG(extack, "ipv4: Invalid prefix length");
 867		return -EINVAL;
 868	}
 869
 870	if (!tb[IFA_LOCAL]) {
 871		NL_SET_ERR_MSG(extack, "ipv4: Local address is not supplied");
 872		return -EINVAL;
 873	}
 874
 875	if (tb[IFA_CACHEINFO]) {
 876		struct ifa_cacheinfo *ci;
 877
 878		ci = nla_data(tb[IFA_CACHEINFO]);
 879		if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
 880			NL_SET_ERR_MSG(extack, "ipv4: address lifetime invalid");
 881			return -EINVAL;
 882		}
 883
 884		*valid_lft = ci->ifa_valid;
 885		*prefered_lft = ci->ifa_prefered;
 886	}
 887
 888	return 0;
 889}
 890
 891static struct in_ifaddr *inet_rtm_to_ifa(struct net *net, struct nlmsghdr *nlh,
 892					 struct nlattr **tb,
 893					 struct netlink_ext_ack *extack)
 894{
 895	struct ifaddrmsg *ifm = nlmsg_data(nlh);
 896	struct in_device *in_dev;
 897	struct net_device *dev;
 898	struct in_ifaddr *ifa;
 899	int err;
 900
 901	dev = __dev_get_by_index(net, ifm->ifa_index);
 902	err = -ENODEV;
 903	if (!dev) {
 904		NL_SET_ERR_MSG(extack, "ipv4: Device not found");
 905		goto errout;
 906	}
 907
 908	in_dev = __in_dev_get_rtnl_net(dev);
 909	err = -ENOBUFS;
 910	if (!in_dev)
 911		goto errout;
 912
 913	ifa = inet_alloc_ifa(in_dev);
 914	if (!ifa)
 915		/*
 916		 * A potential indev allocation can be left alive, it stays
 917		 * assigned to its device and is destroy with it.
 918		 */
 919		goto errout;
 920
 921	ipv4_devconf_setall(in_dev);
 922	neigh_parms_data_state_setall(in_dev->arp_parms);
 923
 924	if (!tb[IFA_ADDRESS])
 925		tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 926
 
 927	ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 928	ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 929	ifa->ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
 930	ifa->ifa_scope = ifm->ifa_scope;
 931	ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
 932	ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
 
 
 933
 934	if (tb[IFA_BROADCAST])
 935		ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
 936
 937	if (tb[IFA_LABEL])
 938		nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 939	else
 940		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 941
 942	if (tb[IFA_RT_PRIORITY])
 943		ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
 944
 945	if (tb[IFA_PROTO])
 946		ifa->ifa_proto = nla_get_u8(tb[IFA_PROTO]);
 947
 948	return ifa;
 949
 950errout:
 951	return ERR_PTR(err);
 952}
 953
 954static struct in_ifaddr *find_matching_ifa(struct net *net, struct in_ifaddr *ifa)
 955{
 956	struct in_device *in_dev = ifa->ifa_dev;
 957	struct in_ifaddr *ifa1;
 958
 959	in_dev_for_each_ifa_rtnl_net(net, ifa1, in_dev) {
 960		if (ifa1->ifa_mask == ifa->ifa_mask &&
 961		    inet_ifa_match(ifa1->ifa_address, ifa) &&
 962		    ifa1->ifa_local == ifa->ifa_local)
 963			return ifa1;
 964	}
 965
 966	return NULL;
 967}
 968
 969static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
 970			    struct netlink_ext_ack *extack)
 971{
 972	__u32 prefered_lft = INFINITY_LIFE_TIME;
 973	__u32 valid_lft = INFINITY_LIFE_TIME;
 974	struct net *net = sock_net(skb->sk);
 975	struct in_ifaddr *ifa_existing;
 976	struct nlattr *tb[IFA_MAX + 1];
 977	struct in_ifaddr *ifa;
 978	int ret;
 979
 980	ret = inet_validate_rtm(nlh, tb, extack, &valid_lft, &prefered_lft);
 981	if (ret < 0)
 982		return ret;
 983
 984	if (!nla_get_in_addr(tb[IFA_LOCAL]))
 985		return 0;
 986
 987	rtnl_net_lock(net);
 988
 989	ifa = inet_rtm_to_ifa(net, nlh, tb, extack);
 990	if (IS_ERR(ifa)) {
 991		ret = PTR_ERR(ifa);
 992		goto unlock;
 993	}
 994
 995	ifa_existing = find_matching_ifa(net, ifa);
 996	if (!ifa_existing) {
 997		/* It would be best to check for !NLM_F_CREATE here but
 998		 * userspace already relies on not having to provide this.
 999		 */
1000		set_ifa_lifetime(ifa, valid_lft, prefered_lft);
1001		if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
1002			ret = ip_mc_autojoin_config(net, true, ifa);
1003			if (ret < 0) {
1004				NL_SET_ERR_MSG(extack, "ipv4: Multicast auto join failed");
1005				inet_free_ifa(ifa);
1006				goto unlock;
1007			}
1008		}
1009
1010		ret = __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid, extack);
1011	} else {
1012		u32 new_metric = ifa->ifa_rt_priority;
1013		u8 new_proto = ifa->ifa_proto;
1014
1015		inet_free_ifa(ifa);
1016
1017		if (nlh->nlmsg_flags & NLM_F_EXCL ||
1018		    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
1019			NL_SET_ERR_MSG(extack, "ipv4: Address already assigned");
1020			ret = -EEXIST;
1021			goto unlock;
1022		}
1023		ifa = ifa_existing;
1024
1025		if (ifa->ifa_rt_priority != new_metric) {
1026			fib_modify_prefix_metric(ifa, new_metric);
1027			ifa->ifa_rt_priority = new_metric;
1028		}
1029
1030		ifa->ifa_proto = new_proto;
1031
1032		set_ifa_lifetime(ifa, valid_lft, prefered_lft);
1033		cancel_delayed_work(&net->ipv4.addr_chk_work);
1034		queue_delayed_work(system_power_efficient_wq,
1035				   &net->ipv4.addr_chk_work, 0);
1036		rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
1037	}
1038
1039unlock:
1040	rtnl_net_unlock(net);
 
1041
1042	return ret;
1043}
1044
1045/*
1046 *	Determine a default network mask, based on the IP address.
1047 */
1048
1049static int inet_abc_len(__be32 addr)
1050{
1051	int rc = -1;	/* Something else, probably a multicast. */
1052
1053	if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
1054		rc = 0;
1055	else {
1056		__u32 haddr = ntohl(addr);
 
1057		if (IN_CLASSA(haddr))
1058			rc = 8;
1059		else if (IN_CLASSB(haddr))
1060			rc = 16;
1061		else if (IN_CLASSC(haddr))
1062			rc = 24;
1063		else if (IN_CLASSE(haddr))
1064			rc = 32;
1065	}
1066
1067	return rc;
1068}
1069
1070
1071int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1072{
 
1073	struct sockaddr_in sin_orig;
1074	struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1075	struct in_ifaddr __rcu **ifap = NULL;
1076	struct in_device *in_dev;
 
1077	struct in_ifaddr *ifa = NULL;
1078	struct net_device *dev;
1079	char *colon;
1080	int ret = -EFAULT;
1081	int tryaddrmatch = 0;
1082
1083	ifr->ifr_name[IFNAMSIZ - 1] = 0;
 
 
 
 
 
 
1084
1085	/* save original address for comparison */
1086	memcpy(&sin_orig, sin, sizeof(*sin));
1087
1088	colon = strchr(ifr->ifr_name, ':');
1089	if (colon)
1090		*colon = 0;
1091
1092	dev_load(net, ifr->ifr_name);
1093
1094	switch (cmd) {
1095	case SIOCGIFADDR:	/* Get interface address */
1096	case SIOCGIFBRDADDR:	/* Get the broadcast address */
1097	case SIOCGIFDSTADDR:	/* Get the destination address */
1098	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
1099		/* Note that these ioctls will not sleep,
1100		   so that we do not impose a lock.
1101		   One day we will be forced to put shlock here (I mean SMP)
1102		 */
1103		tryaddrmatch = (sin_orig.sin_family == AF_INET);
1104		memset(sin, 0, sizeof(*sin));
1105		sin->sin_family = AF_INET;
1106		break;
1107
1108	case SIOCSIFFLAGS:
1109		ret = -EPERM;
1110		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1111			goto out;
1112		break;
1113	case SIOCSIFADDR:	/* Set interface address (and family) */
1114	case SIOCSIFBRDADDR:	/* Set the broadcast address */
1115	case SIOCSIFDSTADDR:	/* Set the destination address */
1116	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
1117		ret = -EPERM;
1118		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1119			goto out;
1120		ret = -EINVAL;
1121		if (sin->sin_family != AF_INET)
1122			goto out;
1123		break;
1124	default:
1125		ret = -EINVAL;
1126		goto out;
1127	}
1128
1129	rtnl_net_lock(net);
1130
1131	ret = -ENODEV;
1132	dev = __dev_get_by_name(net, ifr->ifr_name);
1133	if (!dev)
1134		goto done;
1135
1136	if (colon)
1137		*colon = ':';
1138
1139	in_dev = __in_dev_get_rtnl_net(dev);
1140	if (in_dev) {
1141		if (tryaddrmatch) {
1142			/* Matthias Andree */
1143			/* compare label and address (4.4BSD style) */
1144			/* note: we only do this for a limited set of ioctls
1145			   and only if the original address family was AF_INET.
1146			   This is checked above. */
1147
1148			for (ifap = &in_dev->ifa_list;
1149			     (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
1150			     ifap = &ifa->ifa_next) {
1151				if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1152				    sin_orig.sin_addr.s_addr ==
1153							ifa->ifa_local) {
1154					break; /* found */
1155				}
1156			}
1157		}
1158		/* we didn't get a match, maybe the application is
1159		   4.3BSD-style and passed in junk so we fall back to
1160		   comparing just the label */
1161		if (!ifa) {
1162			for (ifap = &in_dev->ifa_list;
1163			     (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
1164			     ifap = &ifa->ifa_next)
1165				if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1166					break;
1167		}
1168	}
1169
1170	ret = -EADDRNOTAVAIL;
1171	if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1172		goto done;
1173
1174	switch (cmd) {
1175	case SIOCGIFADDR:	/* Get interface address */
1176		ret = 0;
1177		sin->sin_addr.s_addr = ifa->ifa_local;
1178		break;
1179
1180	case SIOCGIFBRDADDR:	/* Get the broadcast address */
1181		ret = 0;
1182		sin->sin_addr.s_addr = ifa->ifa_broadcast;
1183		break;
1184
1185	case SIOCGIFDSTADDR:	/* Get the destination address */
1186		ret = 0;
1187		sin->sin_addr.s_addr = ifa->ifa_address;
1188		break;
1189
1190	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
1191		ret = 0;
1192		sin->sin_addr.s_addr = ifa->ifa_mask;
1193		break;
1194
1195	case SIOCSIFFLAGS:
1196		if (colon) {
1197			ret = -EADDRNOTAVAIL;
1198			if (!ifa)
1199				break;
1200			ret = 0;
1201			if (!(ifr->ifr_flags & IFF_UP))
1202				inet_del_ifa(in_dev, ifap, 1);
1203			break;
1204		}
1205
1206		/* NETDEV_UP/DOWN/CHANGE could touch a peer dev */
1207		ASSERT_RTNL();
1208		ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1209		break;
1210
1211	case SIOCSIFADDR:	/* Set interface address (and family) */
1212		ret = -EINVAL;
1213		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1214			break;
1215
1216		if (!ifa) {
1217			ret = -ENOBUFS;
1218			if (!in_dev)
1219				break;
1220			ifa = inet_alloc_ifa(in_dev);
1221			if (!ifa)
1222				break;
1223
1224			if (colon)
1225				memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1226			else
1227				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1228		} else {
1229			ret = 0;
1230			if (ifa->ifa_local == sin->sin_addr.s_addr)
1231				break;
1232			inet_del_ifa(in_dev, ifap, 0);
1233			ifa->ifa_broadcast = 0;
1234			ifa->ifa_scope = 0;
1235		}
1236
1237		ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1238
1239		if (!(dev->flags & IFF_POINTOPOINT)) {
1240			ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1241			ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1242			if ((dev->flags & IFF_BROADCAST) &&
1243			    ifa->ifa_prefixlen < 31)
1244				ifa->ifa_broadcast = ifa->ifa_address |
1245						     ~ifa->ifa_mask;
1246		} else {
1247			ifa->ifa_prefixlen = 32;
1248			ifa->ifa_mask = inet_make_mask(32);
1249		}
1250		set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1251		ret = inet_set_ifa(dev, ifa);
1252		break;
1253
1254	case SIOCSIFBRDADDR:	/* Set the broadcast address */
1255		ret = 0;
1256		if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1257			inet_del_ifa(in_dev, ifap, 0);
1258			ifa->ifa_broadcast = sin->sin_addr.s_addr;
1259			inet_insert_ifa(ifa);
1260		}
1261		break;
1262
1263	case SIOCSIFDSTADDR:	/* Set the destination address */
1264		ret = 0;
1265		if (ifa->ifa_address == sin->sin_addr.s_addr)
1266			break;
1267		ret = -EINVAL;
1268		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1269			break;
1270		ret = 0;
1271		inet_del_ifa(in_dev, ifap, 0);
1272		ifa->ifa_address = sin->sin_addr.s_addr;
1273		inet_insert_ifa(ifa);
1274		break;
1275
1276	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
1277
1278		/*
1279		 *	The mask we set must be legal.
1280		 */
1281		ret = -EINVAL;
1282		if (bad_mask(sin->sin_addr.s_addr, 0))
1283			break;
1284		ret = 0;
1285		if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1286			__be32 old_mask = ifa->ifa_mask;
1287			inet_del_ifa(in_dev, ifap, 0);
1288			ifa->ifa_mask = sin->sin_addr.s_addr;
1289			ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1290
1291			/* See if current broadcast address matches
1292			 * with current netmask, then recalculate
1293			 * the broadcast address. Otherwise it's a
1294			 * funny address, so don't touch it since
1295			 * the user seems to know what (s)he's doing...
1296			 */
1297			if ((dev->flags & IFF_BROADCAST) &&
1298			    (ifa->ifa_prefixlen < 31) &&
1299			    (ifa->ifa_broadcast ==
1300			     (ifa->ifa_local|~old_mask))) {
1301				ifa->ifa_broadcast = (ifa->ifa_local |
1302						      ~sin->sin_addr.s_addr);
1303			}
1304			inet_insert_ifa(ifa);
1305		}
1306		break;
1307	}
1308done:
1309	rtnl_net_unlock(net);
1310out:
1311	return ret;
 
 
 
 
1312}
1313
1314int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1315{
1316	struct in_device *in_dev = __in_dev_get_rtnl_net(dev);
1317	const struct in_ifaddr *ifa;
1318	struct ifreq ifr;
1319	int done = 0;
1320
1321	if (WARN_ON(size > sizeof(struct ifreq)))
1322		goto out;
1323
1324	if (!in_dev)
1325		goto out;
1326
1327	in_dev_for_each_ifa_rtnl_net(dev_net(dev), ifa, in_dev) {
1328		if (!buf) {
1329			done += size;
1330			continue;
1331		}
1332		if (len < size)
1333			break;
1334		memset(&ifr, 0, sizeof(struct ifreq));
1335		strcpy(ifr.ifr_name, ifa->ifa_label);
 
 
 
1336
1337		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1338		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1339								ifa->ifa_local;
1340
1341		if (copy_to_user(buf + done, &ifr, size)) {
1342			done = -EFAULT;
1343			break;
1344		}
1345		len  -= size;
1346		done += size;
 
1347	}
1348out:
1349	return done;
1350}
1351
1352static __be32 in_dev_select_addr(const struct in_device *in_dev,
1353				 int scope)
1354{
1355	const struct in_ifaddr *ifa;
1356
1357	in_dev_for_each_ifa_rcu(ifa, in_dev) {
1358		if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1359			continue;
1360		if (ifa->ifa_scope != RT_SCOPE_LINK &&
1361		    ifa->ifa_scope <= scope)
1362			return ifa->ifa_local;
1363	}
1364
1365	return 0;
1366}
1367
1368__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1369{
1370	const struct in_ifaddr *ifa;
1371	__be32 addr = 0;
1372	unsigned char localnet_scope = RT_SCOPE_HOST;
1373	struct in_device *in_dev;
1374	struct net *net;
1375	int master_idx;
1376
1377	rcu_read_lock();
1378	net = dev_net_rcu(dev);
1379	in_dev = __in_dev_get_rcu(dev);
1380	if (!in_dev)
1381		goto no_in_dev;
1382
1383	if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1384		localnet_scope = RT_SCOPE_LINK;
1385
1386	in_dev_for_each_ifa_rcu(ifa, in_dev) {
1387		if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1388			continue;
1389		if (min(ifa->ifa_scope, localnet_scope) > scope)
1390			continue;
1391		if (!dst || inet_ifa_match(dst, ifa)) {
1392			addr = ifa->ifa_local;
1393			break;
1394		}
1395		if (!addr)
1396			addr = ifa->ifa_local;
1397	}
1398
1399	if (addr)
1400		goto out_unlock;
1401no_in_dev:
1402	master_idx = l3mdev_master_ifindex_rcu(dev);
1403
1404	/* For VRFs, the VRF device takes the place of the loopback device,
1405	 * with addresses on it being preferred.  Note in such cases the
1406	 * loopback device will be among the devices that fail the master_idx
1407	 * equality check in the loop below.
1408	 */
1409	if (master_idx &&
1410	    (dev = dev_get_by_index_rcu(net, master_idx)) &&
1411	    (in_dev = __in_dev_get_rcu(dev))) {
1412		addr = in_dev_select_addr(in_dev, scope);
1413		if (addr)
1414			goto out_unlock;
1415	}
1416
1417	/* Not loopback addresses on loopback should be preferred
1418	   in this case. It is important that lo is the first interface
1419	   in dev_base list.
1420	 */
1421	for_each_netdev_rcu(net, dev) {
1422		if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1423			continue;
1424
1425		in_dev = __in_dev_get_rcu(dev);
1426		if (!in_dev)
1427			continue;
1428
1429		addr = in_dev_select_addr(in_dev, scope);
1430		if (addr)
1431			goto out_unlock;
 
 
 
 
1432	}
1433out_unlock:
1434	rcu_read_unlock();
1435	return addr;
1436}
1437EXPORT_SYMBOL(inet_select_addr);
1438
1439static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1440			      __be32 local, int scope)
1441{
1442	unsigned char localnet_scope = RT_SCOPE_HOST;
1443	const struct in_ifaddr *ifa;
1444	__be32 addr = 0;
1445	int same = 0;
 
1446
1447	if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1448		localnet_scope = RT_SCOPE_LINK;
1449
1450	in_dev_for_each_ifa_rcu(ifa, in_dev) {
1451		unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1452
1453		if (!addr &&
1454		    (local == ifa->ifa_local || !local) &&
1455		    min_scope <= scope) {
1456			addr = ifa->ifa_local;
1457			if (same)
1458				break;
1459		}
1460		if (!same) {
1461			same = (!local || inet_ifa_match(local, ifa)) &&
1462				(!dst || inet_ifa_match(dst, ifa));
1463			if (same && addr) {
1464				if (local || !dst)
1465					break;
1466				/* Is the selected addr into dst subnet? */
1467				if (inet_ifa_match(addr, ifa))
1468					break;
1469				/* No, then can we use new local src? */
1470				if (min_scope <= scope) {
1471					addr = ifa->ifa_local;
1472					break;
1473				}
1474				/* search for large dst subnet for addr */
1475				same = 0;
1476			}
1477		}
1478	}
1479
1480	return same ? addr : 0;
1481}
1482
1483/*
1484 * Confirm that local IP address exists using wildcards:
1485 * - net: netns to check, cannot be NULL
1486 * - in_dev: only on this interface, NULL=any interface
1487 * - dst: only in the same subnet as dst, 0=any dst
1488 * - local: address, 0=autoselect the local address
1489 * - scope: maximum allowed scope value for the local address
1490 */
1491__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1492			 __be32 dst, __be32 local, int scope)
1493{
1494	__be32 addr = 0;
1495	struct net_device *dev;
 
1496
1497	if (in_dev)
1498		return confirm_addr_indev(in_dev, dst, local, scope);
1499
 
1500	rcu_read_lock();
1501	for_each_netdev_rcu(net, dev) {
1502		in_dev = __in_dev_get_rcu(dev);
1503		if (in_dev) {
1504			addr = confirm_addr_indev(in_dev, dst, local, scope);
1505			if (addr)
1506				break;
1507		}
1508	}
1509	rcu_read_unlock();
1510
1511	return addr;
1512}
1513EXPORT_SYMBOL(inet_confirm_addr);
1514
1515/*
1516 *	Device notifier
1517 */
1518
1519int register_inetaddr_notifier(struct notifier_block *nb)
1520{
1521	return blocking_notifier_chain_register(&inetaddr_chain, nb);
1522}
1523EXPORT_SYMBOL(register_inetaddr_notifier);
1524
1525int unregister_inetaddr_notifier(struct notifier_block *nb)
1526{
1527	return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1528}
1529EXPORT_SYMBOL(unregister_inetaddr_notifier);
1530
1531int register_inetaddr_validator_notifier(struct notifier_block *nb)
1532{
1533	return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1534}
1535EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1536
1537int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1538{
1539	return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1540	    nb);
1541}
1542EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1543
1544/* Rename ifa_labels for a device name change. Make some effort to preserve
1545 * existing alias numbering and to create unique labels if possible.
1546*/
1547static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1548{
1549	struct in_ifaddr *ifa;
1550	int named = 0;
1551
1552	in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1553		char old[IFNAMSIZ], *dot;
1554
1555		memcpy(old, ifa->ifa_label, IFNAMSIZ);
1556		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1557		if (named++ == 0)
1558			goto skip;
1559		dot = strchr(old, ':');
1560		if (!dot) {
1561			sprintf(old, ":%d", named);
1562			dot = old;
1563		}
1564		if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1565			strcat(ifa->ifa_label, dot);
1566		else
1567			strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1568skip:
1569		rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1570	}
1571}
1572
 
 
 
 
 
1573static void inetdev_send_gratuitous_arp(struct net_device *dev,
1574					struct in_device *in_dev)
1575
1576{
1577	const struct in_ifaddr *ifa;
1578
1579	in_dev_for_each_ifa_rtnl(ifa, in_dev) {
 
1580		arp_send(ARPOP_REQUEST, ETH_P_ARP,
1581			 ifa->ifa_local, dev,
1582			 ifa->ifa_local, NULL,
1583			 dev->dev_addr, NULL);
1584	}
1585}
1586
1587/* Called only under RTNL semaphore */
1588
1589static int inetdev_event(struct notifier_block *this, unsigned long event,
1590			 void *ptr)
1591{
1592	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1593	struct in_device *in_dev = __in_dev_get_rtnl(dev);
1594
1595	ASSERT_RTNL();
1596
1597	if (!in_dev) {
1598		if (event == NETDEV_REGISTER) {
1599			in_dev = inetdev_init(dev);
1600			if (IS_ERR(in_dev))
1601				return notifier_from_errno(PTR_ERR(in_dev));
1602			if (dev->flags & IFF_LOOPBACK) {
1603				IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1604				IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1605			}
1606		} else if (event == NETDEV_CHANGEMTU) {
1607			/* Re-enabling IP */
1608			if (inetdev_valid_mtu(dev->mtu))
1609				in_dev = inetdev_init(dev);
1610		}
1611		goto out;
1612	}
1613
1614	switch (event) {
1615	case NETDEV_REGISTER:
1616		pr_debug("%s: bug\n", __func__);
1617		RCU_INIT_POINTER(dev->ip_ptr, NULL);
1618		break;
1619	case NETDEV_UP:
1620		if (!inetdev_valid_mtu(dev->mtu))
1621			break;
1622		if (dev->flags & IFF_LOOPBACK) {
1623			struct in_ifaddr *ifa = inet_alloc_ifa(in_dev);
1624
1625			if (ifa) {
 
1626				ifa->ifa_local =
1627				  ifa->ifa_address = htonl(INADDR_LOOPBACK);
1628				ifa->ifa_prefixlen = 8;
1629				ifa->ifa_mask = inet_make_mask(8);
 
 
1630				ifa->ifa_scope = RT_SCOPE_HOST;
1631				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1632				set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1633						 INFINITY_LIFE_TIME);
1634				ipv4_devconf_setall(in_dev);
1635				neigh_parms_data_state_setall(in_dev->arp_parms);
1636				inet_insert_ifa(ifa);
1637			}
1638		}
1639		ip_mc_up(in_dev);
1640		fallthrough;
1641	case NETDEV_CHANGEADDR:
1642		if (!IN_DEV_ARP_NOTIFY(in_dev))
1643			break;
1644		fallthrough;
1645	case NETDEV_NOTIFY_PEERS:
1646		/* Send gratuitous ARP to notify of link change */
1647		inetdev_send_gratuitous_arp(dev, in_dev);
1648		break;
1649	case NETDEV_DOWN:
1650		ip_mc_down(in_dev);
1651		break;
1652	case NETDEV_PRE_TYPE_CHANGE:
1653		ip_mc_unmap(in_dev);
1654		break;
1655	case NETDEV_POST_TYPE_CHANGE:
1656		ip_mc_remap(in_dev);
1657		break;
1658	case NETDEV_CHANGEMTU:
1659		if (inetdev_valid_mtu(dev->mtu))
1660			break;
1661		/* disable IP when MTU is not enough */
1662		fallthrough;
1663	case NETDEV_UNREGISTER:
1664		inetdev_destroy(in_dev);
1665		break;
1666	case NETDEV_CHANGENAME:
1667		/* Do not notify about label change, this event is
1668		 * not interesting to applications using netlink.
1669		 */
1670		inetdev_changename(dev, in_dev);
1671
1672		devinet_sysctl_unregister(in_dev);
1673		devinet_sysctl_register(in_dev);
1674		break;
1675	}
1676out:
1677	return NOTIFY_DONE;
1678}
1679
1680static struct notifier_block ip_netdev_notifier = {
1681	.notifier_call = inetdev_event,
1682};
1683
1684static size_t inet_nlmsg_size(void)
1685{
1686	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1687	       + nla_total_size(4) /* IFA_ADDRESS */
1688	       + nla_total_size(4) /* IFA_LOCAL */
1689	       + nla_total_size(4) /* IFA_BROADCAST */
1690	       + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1691	       + nla_total_size(4)  /* IFA_FLAGS */
1692	       + nla_total_size(1)  /* IFA_PROTO */
1693	       + nla_total_size(4)  /* IFA_RT_PRIORITY */
1694	       + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1695}
1696
1697static inline u32 cstamp_delta(unsigned long cstamp)
1698{
1699	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1700}
1701
1702static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1703			 unsigned long tstamp, u32 preferred, u32 valid)
1704{
1705	struct ifa_cacheinfo ci;
1706
1707	ci.cstamp = cstamp_delta(cstamp);
1708	ci.tstamp = cstamp_delta(tstamp);
1709	ci.ifa_prefered = preferred;
1710	ci.ifa_valid = valid;
1711
1712	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1713}
1714
1715static int inet_fill_ifaddr(struct sk_buff *skb, const struct in_ifaddr *ifa,
1716			    struct inet_fill_args *args)
1717{
1718	struct ifaddrmsg *ifm;
1719	struct nlmsghdr  *nlh;
1720	unsigned long tstamp;
1721	u32 preferred, valid;
1722	u32 flags;
1723
1724	nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1725			args->flags);
1726	if (!nlh)
1727		return -EMSGSIZE;
1728
1729	ifm = nlmsg_data(nlh);
1730	ifm->ifa_family = AF_INET;
1731	ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1732
1733	flags = READ_ONCE(ifa->ifa_flags);
1734	/* Warning : ifm->ifa_flags is an __u8, it holds only 8 bits.
1735	 * The 32bit value is given in IFA_FLAGS attribute.
1736	 */
1737	ifm->ifa_flags = (__u8)flags;
1738
1739	ifm->ifa_scope = ifa->ifa_scope;
1740	ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1741
1742	if (args->netnsid >= 0 &&
1743	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1744		goto nla_put_failure;
1745
1746	tstamp = READ_ONCE(ifa->ifa_tstamp);
1747	if (!(flags & IFA_F_PERMANENT)) {
1748		preferred = READ_ONCE(ifa->ifa_preferred_lft);
1749		valid = READ_ONCE(ifa->ifa_valid_lft);
1750		if (preferred != INFINITY_LIFE_TIME) {
1751			long tval = (jiffies - tstamp) / HZ;
1752
1753			if (preferred > tval)
1754				preferred -= tval;
1755			else
1756				preferred = 0;
1757			if (valid != INFINITY_LIFE_TIME) {
1758				if (valid > tval)
1759					valid -= tval;
1760				else
1761					valid = 0;
1762			}
1763		}
1764	} else {
1765		preferred = INFINITY_LIFE_TIME;
1766		valid = INFINITY_LIFE_TIME;
1767	}
1768	if ((ifa->ifa_address &&
1769	     nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1770	    (ifa->ifa_local &&
1771	     nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1772	    (ifa->ifa_broadcast &&
1773	     nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1774	    (ifa->ifa_label[0] &&
1775	     nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1776	    (ifa->ifa_proto &&
1777	     nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto)) ||
1778	    nla_put_u32(skb, IFA_FLAGS, flags) ||
1779	    (ifa->ifa_rt_priority &&
1780	     nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1781	    put_cacheinfo(skb, READ_ONCE(ifa->ifa_cstamp), tstamp,
1782			  preferred, valid))
1783		goto nla_put_failure;
1784
1785	nlmsg_end(skb, nlh);
1786	return 0;
1787
1788nla_put_failure:
1789	nlmsg_cancel(skb, nlh);
1790	return -EMSGSIZE;
1791}
1792
1793static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1794				      struct inet_fill_args *fillargs,
1795				      struct net **tgt_net, struct sock *sk,
1796				      struct netlink_callback *cb)
1797{
1798	struct netlink_ext_ack *extack = cb->extack;
1799	struct nlattr *tb[IFA_MAX+1];
1800	struct ifaddrmsg *ifm;
1801	int err, i;
1802
1803	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1804		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1805		return -EINVAL;
1806	}
1807
1808	ifm = nlmsg_data(nlh);
1809	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1810		NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1811		return -EINVAL;
1812	}
1813
1814	fillargs->ifindex = ifm->ifa_index;
1815	if (fillargs->ifindex) {
1816		cb->answer_flags |= NLM_F_DUMP_FILTERED;
1817		fillargs->flags |= NLM_F_DUMP_FILTERED;
1818	}
1819
1820	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1821					    ifa_ipv4_policy, extack);
1822	if (err < 0)
1823		return err;
1824
1825	for (i = 0; i <= IFA_MAX; ++i) {
1826		if (!tb[i])
1827			continue;
1828
1829		if (i == IFA_TARGET_NETNSID) {
1830			struct net *net;
1831
1832			fillargs->netnsid = nla_get_s32(tb[i]);
1833
1834			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1835			if (IS_ERR(net)) {
1836				fillargs->netnsid = -1;
1837				NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1838				return PTR_ERR(net);
1839			}
1840			*tgt_net = net;
1841		} else {
1842			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1843			return -EINVAL;
1844		}
1845	}
1846
1847	return 0;
1848}
1849
1850static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1851			    struct netlink_callback *cb, int *s_ip_idx,
1852			    struct inet_fill_args *fillargs)
1853{
1854	struct in_ifaddr *ifa;
1855	int ip_idx = 0;
1856	int err;
1857
1858	in_dev_for_each_ifa_rcu(ifa, in_dev) {
1859		if (ip_idx < *s_ip_idx) {
1860			ip_idx++;
1861			continue;
1862		}
1863		err = inet_fill_ifaddr(skb, ifa, fillargs);
1864		if (err < 0)
1865			goto done;
1866
1867		nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1868		ip_idx++;
1869	}
1870	err = 0;
1871	ip_idx = 0;
1872done:
1873	*s_ip_idx = ip_idx;
1874
1875	return err;
1876}
1877
1878/* Combine dev_addr_genid and dev_base_seq to detect changes.
1879 */
1880static u32 inet_base_seq(const struct net *net)
1881{
1882	u32 res = atomic_read(&net->ipv4.dev_addr_genid) +
1883		  READ_ONCE(net->dev_base_seq);
1884
1885	/* Must not return 0 (see nl_dump_check_consistent()).
1886	 * Chose a value far away from 0.
1887	 */
1888	if (!res)
1889		res = 0x80000000;
1890	return res;
1891}
1892
1893static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1894{
1895	const struct nlmsghdr *nlh = cb->nlh;
1896	struct inet_fill_args fillargs = {
1897		.portid = NETLINK_CB(cb->skb).portid,
1898		.seq = nlh->nlmsg_seq,
1899		.event = RTM_NEWADDR,
1900		.flags = NLM_F_MULTI,
1901		.netnsid = -1,
1902	};
1903	struct net *net = sock_net(skb->sk);
1904	struct net *tgt_net = net;
1905	struct {
1906		unsigned long ifindex;
1907		int ip_idx;
1908	} *ctx = (void *)cb->ctx;
1909	struct in_device *in_dev;
1910	struct net_device *dev;
1911	int err = 0;
 
 
 
1912
1913	rcu_read_lock();
1914	if (cb->strict_check) {
1915		err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1916						 skb->sk, cb);
1917		if (err < 0)
1918			goto done;
1919
1920		if (fillargs.ifindex) {
1921			dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
1922			if (!dev) {
1923				err = -ENODEV;
1924				goto done;
1925			}
1926			in_dev = __in_dev_get_rcu(dev);
1927			if (!in_dev)
1928				goto done;
1929			err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1930					       &fillargs);
1931			goto done;
 
 
 
 
 
 
 
 
 
 
 
 
1932		}
 
1933	}
1934
1935	cb->seq = inet_base_seq(tgt_net);
1936
1937	for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
1938		in_dev = __in_dev_get_rcu(dev);
1939		if (!in_dev)
1940			continue;
1941		err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1942				       &fillargs);
1943		if (err < 0)
1944			goto done;
1945	}
1946done:
1947	if (fillargs.netnsid >= 0)
1948		put_net(tgt_net);
1949	rcu_read_unlock();
1950	return err;
 
1951}
1952
1953static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1954		      u32 portid)
1955{
1956	struct inet_fill_args fillargs = {
1957		.portid = portid,
1958		.seq = nlh ? nlh->nlmsg_seq : 0,
1959		.event = event,
1960		.flags = 0,
1961		.netnsid = -1,
1962	};
1963	struct sk_buff *skb;
 
1964	int err = -ENOBUFS;
1965	struct net *net;
1966
1967	net = dev_net(ifa->ifa_dev->dev);
1968	skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1969	if (!skb)
1970		goto errout;
1971
1972	err = inet_fill_ifaddr(skb, ifa, &fillargs);
1973	if (err < 0) {
1974		/* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1975		WARN_ON(err == -EMSGSIZE);
1976		kfree_skb(skb);
1977		goto errout;
1978	}
1979	rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1980	return;
1981errout:
1982	rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
 
1983}
1984
1985static size_t inet_get_link_af_size(const struct net_device *dev,
1986				    u32 ext_filter_mask)
1987{
1988	struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1989
1990	if (!in_dev)
1991		return 0;
1992
1993	return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1994}
1995
1996static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1997			     u32 ext_filter_mask)
1998{
1999	struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
2000	struct nlattr *nla;
2001	int i;
2002
2003	if (!in_dev)
2004		return -ENODATA;
2005
2006	nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
2007	if (!nla)
2008		return -EMSGSIZE;
2009
2010	for (i = 0; i < IPV4_DEVCONF_MAX; i++)
2011		((u32 *) nla_data(nla))[i] = READ_ONCE(in_dev->cnf.data[i]);
2012
2013	return 0;
2014}
2015
2016static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
2017	[IFLA_INET_CONF]	= { .type = NLA_NESTED },
2018};
2019
2020static int inet_validate_link_af(const struct net_device *dev,
2021				 const struct nlattr *nla,
2022				 struct netlink_ext_ack *extack)
2023{
2024	struct nlattr *a, *tb[IFLA_INET_MAX+1];
2025	int err, rem;
2026
2027	if (dev && !__in_dev_get_rtnl(dev))
2028		return -EAFNOSUPPORT;
2029
2030	err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
2031					  inet_af_policy, extack);
2032	if (err < 0)
2033		return err;
2034
2035	if (tb[IFLA_INET_CONF]) {
2036		nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
2037			int cfgid = nla_type(a);
2038
2039			if (nla_len(a) < 4)
2040				return -EINVAL;
2041
2042			if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
2043				return -EINVAL;
2044		}
2045	}
2046
2047	return 0;
2048}
2049
2050static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla,
2051			    struct netlink_ext_ack *extack)
2052{
2053	struct in_device *in_dev = __in_dev_get_rtnl(dev);
2054	struct nlattr *a, *tb[IFLA_INET_MAX+1];
2055	int rem;
2056
2057	if (!in_dev)
2058		return -EAFNOSUPPORT;
2059
2060	if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
2061		return -EINVAL;
2062
2063	if (tb[IFLA_INET_CONF]) {
2064		nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
2065			ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
2066	}
2067
2068	return 0;
2069}
2070
2071static int inet_netconf_msgsize_devconf(int type)
2072{
2073	int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2074		   + nla_total_size(4);	/* NETCONFA_IFINDEX */
2075	bool all = false;
2076
2077	if (type == NETCONFA_ALL)
2078		all = true;
2079
2080	if (all || type == NETCONFA_FORWARDING)
2081		size += nla_total_size(4);
2082	if (all || type == NETCONFA_RP_FILTER)
2083		size += nla_total_size(4);
2084	if (all || type == NETCONFA_MC_FORWARDING)
2085		size += nla_total_size(4);
2086	if (all || type == NETCONFA_BC_FORWARDING)
2087		size += nla_total_size(4);
2088	if (all || type == NETCONFA_PROXY_NEIGH)
2089		size += nla_total_size(4);
2090	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2091		size += nla_total_size(4);
2092
2093	return size;
2094}
2095
2096static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2097				     const struct ipv4_devconf *devconf,
2098				     u32 portid, u32 seq, int event,
2099				     unsigned int flags, int type)
2100{
2101	struct nlmsghdr  *nlh;
2102	struct netconfmsg *ncm;
2103	bool all = false;
2104
2105	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2106			flags);
2107	if (!nlh)
2108		return -EMSGSIZE;
2109
2110	if (type == NETCONFA_ALL)
2111		all = true;
2112
2113	ncm = nlmsg_data(nlh);
2114	ncm->ncm_family = AF_INET;
2115
2116	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2117		goto nla_put_failure;
2118
2119	if (!devconf)
2120		goto out;
2121
2122	if ((all || type == NETCONFA_FORWARDING) &&
2123	    nla_put_s32(skb, NETCONFA_FORWARDING,
2124			IPV4_DEVCONF_RO(*devconf, FORWARDING)) < 0)
2125		goto nla_put_failure;
2126	if ((all || type == NETCONFA_RP_FILTER) &&
2127	    nla_put_s32(skb, NETCONFA_RP_FILTER,
2128			IPV4_DEVCONF_RO(*devconf, RP_FILTER)) < 0)
2129		goto nla_put_failure;
2130	if ((all || type == NETCONFA_MC_FORWARDING) &&
2131	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2132			IPV4_DEVCONF_RO(*devconf, MC_FORWARDING)) < 0)
2133		goto nla_put_failure;
2134	if ((all || type == NETCONFA_BC_FORWARDING) &&
2135	    nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2136			IPV4_DEVCONF_RO(*devconf, BC_FORWARDING)) < 0)
2137		goto nla_put_failure;
2138	if ((all || type == NETCONFA_PROXY_NEIGH) &&
2139	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2140			IPV4_DEVCONF_RO(*devconf, PROXY_ARP)) < 0)
2141		goto nla_put_failure;
2142	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2143	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2144			IPV4_DEVCONF_RO(*devconf,
2145					IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2146		goto nla_put_failure;
2147
2148out:
2149	nlmsg_end(skb, nlh);
2150	return 0;
2151
2152nla_put_failure:
2153	nlmsg_cancel(skb, nlh);
2154	return -EMSGSIZE;
2155}
2156
2157void inet_netconf_notify_devconf(struct net *net, int event, int type,
2158				 int ifindex, struct ipv4_devconf *devconf)
2159{
2160	struct sk_buff *skb;
2161	int err = -ENOBUFS;
2162
2163	skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2164	if (!skb)
2165		goto errout;
2166
2167	err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2168					event, 0, type);
2169	if (err < 0) {
2170		/* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2171		WARN_ON(err == -EMSGSIZE);
2172		kfree_skb(skb);
2173		goto errout;
2174	}
2175	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2176	return;
2177errout:
2178	rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2179}
2180
2181static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2182	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
2183	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
2184	[NETCONFA_RP_FILTER]	= { .len = sizeof(int) },
2185	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
2186	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
2187};
2188
2189static int inet_netconf_valid_get_req(struct sk_buff *skb,
2190				      const struct nlmsghdr *nlh,
2191				      struct nlattr **tb,
2192				      struct netlink_ext_ack *extack)
2193{
2194	int i, err;
2195
2196	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2197		NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2198		return -EINVAL;
2199	}
2200
2201	if (!netlink_strict_get_check(skb))
2202		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2203					      tb, NETCONFA_MAX,
2204					      devconf_ipv4_policy, extack);
2205
2206	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2207					    tb, NETCONFA_MAX,
2208					    devconf_ipv4_policy, extack);
2209	if (err)
2210		return err;
2211
2212	for (i = 0; i <= NETCONFA_MAX; i++) {
2213		if (!tb[i])
2214			continue;
2215
2216		switch (i) {
2217		case NETCONFA_IFINDEX:
2218			break;
2219		default:
2220			NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2221			return -EINVAL;
2222		}
2223	}
2224
2225	return 0;
2226}
2227
2228static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2229				    struct nlmsghdr *nlh,
2230				    struct netlink_ext_ack *extack)
2231{
2232	struct net *net = sock_net(in_skb->sk);
2233	struct nlattr *tb[NETCONFA_MAX + 1];
2234	const struct ipv4_devconf *devconf;
2235	struct in_device *in_dev = NULL;
2236	struct net_device *dev = NULL;
2237	struct sk_buff *skb;
2238	int ifindex;
2239	int err;
2240
2241	err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2242	if (err)
2243		return err;
2244
2245	if (!tb[NETCONFA_IFINDEX])
2246		return -EINVAL;
2247
2248	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2249	switch (ifindex) {
2250	case NETCONFA_IFINDEX_ALL:
2251		devconf = net->ipv4.devconf_all;
2252		break;
2253	case NETCONFA_IFINDEX_DEFAULT:
2254		devconf = net->ipv4.devconf_dflt;
2255		break;
2256	default:
2257		err = -ENODEV;
2258		dev = dev_get_by_index(net, ifindex);
2259		if (dev)
2260			in_dev = in_dev_get(dev);
2261		if (!in_dev)
2262			goto errout;
2263		devconf = &in_dev->cnf;
2264		break;
2265	}
2266
2267	err = -ENOBUFS;
2268	skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2269	if (!skb)
2270		goto errout;
2271
2272	err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2273					NETLINK_CB(in_skb).portid,
2274					nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2275					NETCONFA_ALL);
2276	if (err < 0) {
2277		/* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2278		WARN_ON(err == -EMSGSIZE);
2279		kfree_skb(skb);
2280		goto errout;
2281	}
2282	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2283errout:
2284	if (in_dev)
2285		in_dev_put(in_dev);
2286	dev_put(dev);
2287	return err;
2288}
2289
2290static int inet_netconf_dump_devconf(struct sk_buff *skb,
2291				     struct netlink_callback *cb)
2292{
2293	const struct nlmsghdr *nlh = cb->nlh;
2294	struct net *net = sock_net(skb->sk);
2295	struct {
2296		unsigned long ifindex;
2297		unsigned int all_default;
2298	} *ctx = (void *)cb->ctx;
2299	const struct in_device *in_dev;
2300	struct net_device *dev;
2301	int err = 0;
2302
2303	if (cb->strict_check) {
2304		struct netlink_ext_ack *extack = cb->extack;
2305		struct netconfmsg *ncm;
2306
2307		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2308			NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2309			return -EINVAL;
2310		}
2311
2312		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2313			NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2314			return -EINVAL;
2315		}
2316	}
2317
2318	rcu_read_lock();
2319	for_each_netdev_dump(net, dev, ctx->ifindex) {
2320		in_dev = __in_dev_get_rcu(dev);
2321		if (!in_dev)
2322			continue;
2323		err = inet_netconf_fill_devconf(skb, dev->ifindex,
2324						&in_dev->cnf,
2325						NETLINK_CB(cb->skb).portid,
2326						nlh->nlmsg_seq,
2327						RTM_NEWNETCONF, NLM_F_MULTI,
2328						NETCONFA_ALL);
2329		if (err < 0)
2330			goto done;
2331	}
2332	if (ctx->all_default == 0) {
2333		err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2334						net->ipv4.devconf_all,
2335						NETLINK_CB(cb->skb).portid,
2336						nlh->nlmsg_seq,
2337						RTM_NEWNETCONF, NLM_F_MULTI,
2338						NETCONFA_ALL);
2339		if (err < 0)
2340			goto done;
2341		ctx->all_default++;
2342	}
2343	if (ctx->all_default == 1) {
2344		err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2345						net->ipv4.devconf_dflt,
2346						NETLINK_CB(cb->skb).portid,
2347						nlh->nlmsg_seq,
2348						RTM_NEWNETCONF, NLM_F_MULTI,
2349						NETCONFA_ALL);
2350		if (err < 0)
2351			goto done;
2352		ctx->all_default++;
2353	}
2354done:
2355	rcu_read_unlock();
2356	return err;
2357}
2358
2359#ifdef CONFIG_SYSCTL
2360
2361static void devinet_copy_dflt_conf(struct net *net, int i)
2362{
2363	struct net_device *dev;
2364
2365	rcu_read_lock();
2366	for_each_netdev_rcu(net, dev) {
2367		struct in_device *in_dev;
2368
2369		in_dev = __in_dev_get_rcu(dev);
2370		if (in_dev && !test_bit(i, in_dev->cnf.state))
2371			in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2372	}
2373	rcu_read_unlock();
2374}
2375
2376/* called with RTNL locked */
2377static void inet_forward_change(struct net *net)
2378{
2379	struct net_device *dev;
2380	int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2381
2382	IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2383	IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2384	inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2385				    NETCONFA_FORWARDING,
2386				    NETCONFA_IFINDEX_ALL,
2387				    net->ipv4.devconf_all);
2388	inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2389				    NETCONFA_FORWARDING,
2390				    NETCONFA_IFINDEX_DEFAULT,
2391				    net->ipv4.devconf_dflt);
2392
2393	for_each_netdev(net, dev) {
2394		struct in_device *in_dev;
2395
2396		if (on)
2397			dev_disable_lro(dev);
2398
2399		in_dev = __in_dev_get_rtnl_net(dev);
2400		if (in_dev) {
2401			IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2402			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2403						    NETCONFA_FORWARDING,
2404						    dev->ifindex, &in_dev->cnf);
2405		}
2406	}
2407}
2408
2409static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2410{
2411	if (cnf == net->ipv4.devconf_dflt)
2412		return NETCONFA_IFINDEX_DEFAULT;
2413	else if (cnf == net->ipv4.devconf_all)
2414		return NETCONFA_IFINDEX_ALL;
2415	else {
2416		struct in_device *idev
2417			= container_of(cnf, struct in_device, cnf);
2418		return idev->dev->ifindex;
2419	}
2420}
2421
2422static int devinet_conf_proc(const struct ctl_table *ctl, int write,
2423			     void *buffer, size_t *lenp, loff_t *ppos)
 
2424{
2425	int old_value = *(int *)ctl->data;
2426	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2427	int new_value = *(int *)ctl->data;
2428
2429	if (write) {
2430		struct ipv4_devconf *cnf = ctl->extra1;
2431		struct net *net = ctl->extra2;
2432		int i = (int *)ctl->data - cnf->data;
2433		int ifindex;
2434
2435		set_bit(i, cnf->state);
2436
2437		if (cnf == net->ipv4.devconf_dflt)
2438			devinet_copy_dflt_conf(net, i);
2439		if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2440		    i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2441			if ((new_value == 0) && (old_value != 0))
2442				rt_cache_flush(net);
2443
2444		if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2445		    new_value != old_value)
2446			rt_cache_flush(net);
2447
2448		if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2449		    new_value != old_value) {
2450			ifindex = devinet_conf_ifindex(net, cnf);
2451			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2452						    NETCONFA_RP_FILTER,
2453						    ifindex, cnf);
2454		}
2455		if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2456		    new_value != old_value) {
2457			ifindex = devinet_conf_ifindex(net, cnf);
2458			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2459						    NETCONFA_PROXY_NEIGH,
2460						    ifindex, cnf);
2461		}
2462		if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2463		    new_value != old_value) {
2464			ifindex = devinet_conf_ifindex(net, cnf);
2465			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2466						    NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2467						    ifindex, cnf);
2468		}
2469	}
2470
2471	return ret;
2472}
2473
2474static int devinet_sysctl_forward(const struct ctl_table *ctl, int write,
2475				  void *buffer, size_t *lenp, loff_t *ppos)
 
2476{
2477	int *valp = ctl->data;
2478	int val = *valp;
2479	loff_t pos = *ppos;
2480	struct net *net = ctl->extra2;
2481	int ret;
2482
2483	if (write && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2484		return -EPERM;
2485
2486	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2487
2488	if (write && *valp != val) {
 
 
2489		if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2490			if (!rtnl_net_trylock(net)) {
2491				/* Restore the original values before restarting */
2492				*valp = val;
2493				*ppos = pos;
2494				return restart_syscall();
2495			}
2496			if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2497				inet_forward_change(net);
2498			} else {
2499				struct ipv4_devconf *cnf = ctl->extra1;
2500				struct in_device *idev =
2501					container_of(cnf, struct in_device, cnf);
2502				if (*valp)
2503					dev_disable_lro(idev->dev);
2504				inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2505							    NETCONFA_FORWARDING,
2506							    idev->dev->ifindex,
2507							    cnf);
2508			}
2509			rtnl_net_unlock(net);
2510			rt_cache_flush(net);
2511		} else
2512			inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2513						    NETCONFA_FORWARDING,
2514						    NETCONFA_IFINDEX_DEFAULT,
2515						    net->ipv4.devconf_dflt);
2516	}
2517
2518	return ret;
2519}
2520
2521static int ipv4_doint_and_flush(const struct ctl_table *ctl, int write,
2522				void *buffer, size_t *lenp, loff_t *ppos)
 
2523{
2524	int *valp = ctl->data;
2525	int val = *valp;
2526	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2527	struct net *net = ctl->extra2;
2528
2529	if (write && *valp != val)
2530		rt_cache_flush(net);
2531
2532	return ret;
2533}
2534
2535#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2536	{ \
2537		.procname	= name, \
2538		.data		= ipv4_devconf.data + \
2539				  IPV4_DEVCONF_ ## attr - 1, \
2540		.maxlen		= sizeof(int), \
2541		.mode		= mval, \
2542		.proc_handler	= proc, \
2543		.extra1		= &ipv4_devconf, \
2544	}
2545
2546#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2547	DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2548
2549#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2550	DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2551
2552#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2553	DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2554
2555#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2556	DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2557
2558static struct devinet_sysctl_table {
2559	struct ctl_table_header *sysctl_header;
2560	struct ctl_table devinet_vars[IPV4_DEVCONF_MAX];
2561} devinet_sysctl = {
2562	.devinet_vars = {
2563		DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2564					     devinet_sysctl_forward),
2565		DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2566		DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2567
2568		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2569		DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2570		DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2571		DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2572		DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2573		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2574					"accept_source_route"),
2575		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2576		DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2577		DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2578		DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2579		DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2580		DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2581		DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2582		DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2583		DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2584		DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2585		DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2586		DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2587		DEVINET_SYSCTL_RW_ENTRY(ARP_EVICT_NOCARRIER,
2588					"arp_evict_nocarrier"),
2589		DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2590		DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2591					"force_igmp_version"),
2592		DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2593					"igmpv2_unsolicited_report_interval"),
2594		DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2595					"igmpv3_unsolicited_report_interval"),
2596		DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2597					"ignore_routes_with_linkdown"),
2598		DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2599					"drop_gratuitous_arp"),
2600
2601		DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2602		DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
 
 
2603		DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2604					      "promote_secondaries"),
2605		DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2606					      "route_localnet"),
2607		DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2608					      "drop_unicast_in_l2_multicast"),
2609	},
2610};
2611
2612static int __devinet_sysctl_register(struct net *net, char *dev_name,
2613				     int ifindex, struct ipv4_devconf *p)
2614{
2615	int i;
2616	struct devinet_sysctl_table *t;
2617	char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2618
2619	t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL_ACCOUNT);
2620	if (!t)
2621		goto out;
2622
2623	for (i = 0; i < ARRAY_SIZE(t->devinet_vars); i++) {
2624		t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2625		t->devinet_vars[i].extra1 = p;
2626		t->devinet_vars[i].extra2 = net;
2627	}
2628
2629	snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2630
2631	t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2632	if (!t->sysctl_header)
2633		goto free;
2634
2635	p->sysctl = t;
2636
2637	inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2638				    ifindex, p);
2639	return 0;
2640
2641free:
2642	kfree(t);
2643out:
2644	return -ENOMEM;
2645}
2646
2647static void __devinet_sysctl_unregister(struct net *net,
2648					struct ipv4_devconf *cnf, int ifindex)
2649{
2650	struct devinet_sysctl_table *t = cnf->sysctl;
2651
2652	if (t) {
2653		cnf->sysctl = NULL;
2654		unregister_net_sysctl_table(t->sysctl_header);
2655		kfree(t);
2656	}
2657
2658	inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
 
 
2659}
2660
2661static int devinet_sysctl_register(struct in_device *idev)
2662{
2663	int err;
2664
2665	if (!sysctl_dev_name_is_allowed(idev->dev->name))
2666		return -EINVAL;
2667
2668	err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2669	if (err)
2670		return err;
2671	err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2672					idev->dev->ifindex, &idev->cnf);
2673	if (err)
2674		neigh_sysctl_unregister(idev->arp_parms);
2675	return err;
2676}
2677
2678static void devinet_sysctl_unregister(struct in_device *idev)
2679{
2680	struct net *net = dev_net(idev->dev);
2681
2682	__devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2683	neigh_sysctl_unregister(idev->arp_parms);
2684}
2685
2686static struct ctl_table ctl_forward_entry[] = {
2687	{
2688		.procname	= "ip_forward",
2689		.data		= &ipv4_devconf.data[
2690					IPV4_DEVCONF_FORWARDING - 1],
2691		.maxlen		= sizeof(int),
2692		.mode		= 0644,
2693		.proc_handler	= devinet_sysctl_forward,
2694		.extra1		= &ipv4_devconf,
2695		.extra2		= &init_net,
2696	},
 
2697};
2698#endif
2699
2700static __net_init int devinet_init_net(struct net *net)
2701{
 
 
2702#ifdef CONFIG_SYSCTL
 
2703	struct ctl_table_header *forw_hdr;
2704	struct ctl_table *tbl;
2705#endif
2706	struct ipv4_devconf *all, *dflt;
2707	int err;
2708	int i;
2709
2710	err = -ENOMEM;
2711	net->ipv4.inet_addr_lst = kmalloc_array(IN4_ADDR_HSIZE,
2712						sizeof(struct hlist_head),
2713						GFP_KERNEL);
2714	if (!net->ipv4.inet_addr_lst)
2715		goto err_alloc_hash;
2716
2717	all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2718	if (!all)
2719		goto err_alloc_all;
2720
2721	dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2722	if (!dflt)
2723		goto err_alloc_dflt;
2724
2725#ifdef CONFIG_SYSCTL
2726	tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2727	if (!tbl)
2728		goto err_alloc_ctl;
2729
2730	tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2731	tbl[0].extra1 = all;
2732	tbl[0].extra2 = net;
2733#endif
2734
2735	if (!net_eq(net, &init_net)) {
2736		switch (net_inherit_devconf()) {
2737		case 3:
2738			/* copy from the current netns */
2739			memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2740			       sizeof(ipv4_devconf));
2741			memcpy(dflt,
2742			       current->nsproxy->net_ns->ipv4.devconf_dflt,
2743			       sizeof(ipv4_devconf_dflt));
2744			break;
2745		case 0:
2746		case 1:
2747			/* copy from init_net */
2748			memcpy(all, init_net.ipv4.devconf_all,
2749			       sizeof(ipv4_devconf));
2750			memcpy(dflt, init_net.ipv4.devconf_dflt,
2751			       sizeof(ipv4_devconf_dflt));
2752			break;
2753		case 2:
2754			/* use compiled values */
2755			break;
2756		}
2757	}
2758
2759#ifdef CONFIG_SYSCTL
2760	err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2761	if (err < 0)
2762		goto err_reg_all;
2763
2764	err = __devinet_sysctl_register(net, "default",
2765					NETCONFA_IFINDEX_DEFAULT, dflt);
2766	if (err < 0)
2767		goto err_reg_dflt;
2768
2769	err = -ENOMEM;
2770	forw_hdr = register_net_sysctl_sz(net, "net/ipv4", tbl,
2771					  ARRAY_SIZE(ctl_forward_entry));
2772	if (!forw_hdr)
2773		goto err_reg_ctl;
2774	net->ipv4.forw_hdr = forw_hdr;
2775#endif
2776
2777	for (i = 0; i < IN4_ADDR_HSIZE; i++)
2778		INIT_HLIST_HEAD(&net->ipv4.inet_addr_lst[i]);
2779
2780	INIT_DEFERRABLE_WORK(&net->ipv4.addr_chk_work, check_lifetime);
2781
2782	net->ipv4.devconf_all = all;
2783	net->ipv4.devconf_dflt = dflt;
2784	return 0;
2785
2786#ifdef CONFIG_SYSCTL
2787err_reg_ctl:
2788	__devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2789err_reg_dflt:
2790	__devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2791err_reg_all:
2792	kfree(tbl);
 
2793err_alloc_ctl:
2794#endif
2795	kfree(dflt);
 
2796err_alloc_dflt:
2797	kfree(all);
 
2798err_alloc_all:
2799	kfree(net->ipv4.inet_addr_lst);
2800err_alloc_hash:
2801	return err;
2802}
2803
2804static __net_exit void devinet_exit_net(struct net *net)
2805{
2806#ifdef CONFIG_SYSCTL
2807	const struct ctl_table *tbl;
2808#endif
2809
2810	cancel_delayed_work_sync(&net->ipv4.addr_chk_work);
2811
2812#ifdef CONFIG_SYSCTL
2813	tbl = net->ipv4.forw_hdr->ctl_table_arg;
2814	unregister_net_sysctl_table(net->ipv4.forw_hdr);
2815	__devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2816				    NETCONFA_IFINDEX_DEFAULT);
2817	__devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2818				    NETCONFA_IFINDEX_ALL);
2819	kfree(tbl);
2820#endif
2821	kfree(net->ipv4.devconf_dflt);
2822	kfree(net->ipv4.devconf_all);
2823	kfree(net->ipv4.inet_addr_lst);
2824}
2825
2826static __net_initdata struct pernet_operations devinet_ops = {
2827	.init = devinet_init_net,
2828	.exit = devinet_exit_net,
2829};
2830
2831static struct rtnl_af_ops inet_af_ops __read_mostly = {
2832	.family		  = AF_INET,
2833	.fill_link_af	  = inet_fill_link_af,
2834	.get_link_af_size = inet_get_link_af_size,
2835	.validate_link_af = inet_validate_link_af,
2836	.set_link_af	  = inet_set_link_af,
2837};
2838
2839static const struct rtnl_msg_handler devinet_rtnl_msg_handlers[] __initconst = {
2840	{.protocol = PF_INET, .msgtype = RTM_NEWADDR, .doit = inet_rtm_newaddr,
2841	 .flags = RTNL_FLAG_DOIT_PERNET},
2842	{.protocol = PF_INET, .msgtype = RTM_DELADDR, .doit = inet_rtm_deladdr,
2843	 .flags = RTNL_FLAG_DOIT_PERNET},
2844	{.protocol = PF_INET, .msgtype = RTM_GETADDR, .dumpit = inet_dump_ifaddr,
2845	 .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
2846	{.protocol = PF_INET, .msgtype = RTM_GETNETCONF,
2847	 .doit = inet_netconf_get_devconf, .dumpit = inet_netconf_dump_devconf,
2848	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
2849};
2850
2851void __init devinet_init(void)
2852{
 
 
 
 
 
2853	register_pernet_subsys(&devinet_ops);
 
 
2854	register_netdevice_notifier(&ip_netdev_notifier);
2855
2856	if (rtnl_af_register(&inet_af_ops))
2857		panic("Unable to register inet_af_ops\n");
2858
2859	rtnl_register_many(devinet_rtnl_msg_handlers);
 
 
2860}
v3.5.6
 
   1/*
   2 *	NET3	IP device support routines.
   3 *
   4 *		This program is free software; you can redistribute it and/or
   5 *		modify it under the terms of the GNU General Public License
   6 *		as published by the Free Software Foundation; either version
   7 *		2 of the License, or (at your option) any later version.
   8 *
   9 *	Derived from the IP parts of dev.c 1.0.19
  10 * 		Authors:	Ross Biro
  11 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
  13 *
  14 *	Additional Authors:
  15 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
  16 *		Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  17 *
  18 *	Changes:
  19 *		Alexey Kuznetsov:	pa_* fields are replaced with ifaddr
  20 *					lists.
  21 *		Cyrus Durgin:		updated for kmod
  22 *		Matthias Andree:	in devinet_ioctl, compare label and
  23 *					address (4.4BSD alias style support),
  24 *					fall back to comparing just the label
  25 *					if no match found.
  26 */
  27
  28
  29#include <asm/uaccess.h>
  30#include <linux/bitops.h>
  31#include <linux/capability.h>
  32#include <linux/module.h>
  33#include <linux/types.h>
  34#include <linux/kernel.h>
 
  35#include <linux/string.h>
  36#include <linux/mm.h>
  37#include <linux/socket.h>
  38#include <linux/sockios.h>
  39#include <linux/in.h>
  40#include <linux/errno.h>
  41#include <linux/interrupt.h>
  42#include <linux/if_addr.h>
  43#include <linux/if_ether.h>
  44#include <linux/inet.h>
  45#include <linux/netdevice.h>
  46#include <linux/etherdevice.h>
  47#include <linux/skbuff.h>
  48#include <linux/init.h>
  49#include <linux/notifier.h>
  50#include <linux/inetdevice.h>
  51#include <linux/igmp.h>
  52#include <linux/slab.h>
  53#include <linux/hash.h>
  54#ifdef CONFIG_SYSCTL
  55#include <linux/sysctl.h>
  56#endif
  57#include <linux/kmod.h>
 
  58
  59#include <net/arp.h>
  60#include <net/ip.h>
  61#include <net/route.h>
  62#include <net/ip_fib.h>
  63#include <net/rtnetlink.h>
  64#include <net/net_namespace.h>
 
  65
  66#include "fib_lookup.h"
 
 
 
  67
  68static struct ipv4_devconf ipv4_devconf = {
  69	.data = {
  70		[IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  71		[IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  72		[IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  73		[IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
 
 
 
  74	},
  75};
  76
  77static struct ipv4_devconf ipv4_devconf_dflt = {
  78	.data = {
  79		[IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  80		[IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  81		[IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  82		[IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
  83		[IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
 
 
 
  84	},
  85};
  86
  87#define IPV4_DEVCONF_DFLT(net, attr) \
  88	IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
  89
  90static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
  91	[IFA_LOCAL]     	= { .type = NLA_U32 },
  92	[IFA_ADDRESS]   	= { .type = NLA_U32 },
  93	[IFA_BROADCAST] 	= { .type = NLA_U32 },
  94	[IFA_LABEL]     	= { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  95};
  96
  97/* inet_addr_hash's shifting is dependent upon this IN4_ADDR_HSIZE
  98 * value.  So if you change this define, make appropriate changes to
  99 * inet_addr_hash as well.
 100 */
 101#define IN4_ADDR_HSIZE	256
 102static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
 103static DEFINE_SPINLOCK(inet_addr_hash_lock);
 104
 105static inline unsigned int inet_addr_hash(struct net *net, __be32 addr)
 106{
 107	u32 val = (__force u32) addr ^ hash_ptr(net, 8);
 108
 109	return ((val ^ (val >> 8) ^ (val >> 16) ^ (val >> 24)) &
 110		(IN4_ADDR_HSIZE - 1));
 111}
 112
 113static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
 114{
 115	unsigned int hash = inet_addr_hash(net, ifa->ifa_local);
 116
 117	spin_lock(&inet_addr_hash_lock);
 118	hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
 119	spin_unlock(&inet_addr_hash_lock);
 120}
 121
 122static void inet_hash_remove(struct in_ifaddr *ifa)
 123{
 124	spin_lock(&inet_addr_hash_lock);
 125	hlist_del_init_rcu(&ifa->hash);
 126	spin_unlock(&inet_addr_hash_lock);
 127}
 128
 129/**
 130 * __ip_dev_find - find the first device with a given source address.
 131 * @net: the net namespace
 132 * @addr: the source address
 133 * @devref: if true, take a reference on the found device
 134 *
 135 * If a caller uses devref=false, it should be protected by RCU, or RTNL
 136 */
 137struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
 138{
 139	unsigned int hash = inet_addr_hash(net, addr);
 140	struct net_device *result = NULL;
 141	struct in_ifaddr *ifa;
 142	struct hlist_node *node;
 143
 144	rcu_read_lock();
 145	hlist_for_each_entry_rcu(ifa, node, &inet_addr_lst[hash], hash) {
 146		struct net_device *dev = ifa->ifa_dev->dev;
 147
 148		if (!net_eq(dev_net(dev), net))
 149			continue;
 150		if (ifa->ifa_local == addr) {
 151			result = dev;
 152			break;
 153		}
 154	}
 155	if (!result) {
 156		struct flowi4 fl4 = { .daddr = addr };
 157		struct fib_result res = { 0 };
 158		struct fib_table *local;
 159
 160		/* Fallback to FIB local table so that communication
 161		 * over loopback subnets work.
 162		 */
 163		local = fib_get_table(net, RT_TABLE_LOCAL);
 164		if (local &&
 165		    !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
 166		    res.type == RTN_LOCAL)
 167			result = FIB_RES_DEV(res);
 
 
 168	}
 169	if (result && devref)
 170		dev_hold(result);
 171	rcu_read_unlock();
 172	return result;
 173}
 174EXPORT_SYMBOL(__ip_dev_find);
 175
 
 
 
 
 
 
 
 
 
 
 
 
 
 176static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
 177
 178static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
 179static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 
 
 180			 int destroy);
 181#ifdef CONFIG_SYSCTL
 182static void devinet_sysctl_register(struct in_device *idev);
 183static void devinet_sysctl_unregister(struct in_device *idev);
 184#else
 185static inline void devinet_sysctl_register(struct in_device *idev)
 186{
 
 187}
 188static inline void devinet_sysctl_unregister(struct in_device *idev)
 189{
 190}
 191#endif
 192
 193/* Locks all the inet devices. */
 194
 195static struct in_ifaddr *inet_alloc_ifa(void)
 196{
 197	return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
 
 
 
 
 
 
 
 
 
 
 
 198}
 199
 200static void inet_rcu_free_ifa(struct rcu_head *head)
 201{
 202	struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 203	if (ifa->ifa_dev)
 204		in_dev_put(ifa->ifa_dev);
 205	kfree(ifa);
 206}
 207
 208static inline void inet_free_ifa(struct in_ifaddr *ifa)
 
 
 
 
 
 
 
 
 
 209{
 210	call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 
 
 
 211}
 212
 213void in_dev_finish_destroy(struct in_device *idev)
 214{
 215	struct net_device *dev = idev->dev;
 216
 217	WARN_ON(idev->ifa_list);
 218	WARN_ON(idev->mc_list);
 219#ifdef NET_REFCNT_DEBUG
 220	pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
 221#endif
 222	dev_put(dev);
 223	if (!idev->dead)
 224		pr_err("Freeing alive in_device %p\n", idev);
 225	else
 226		kfree(idev);
 227}
 228EXPORT_SYMBOL(in_dev_finish_destroy);
 229
 230static struct in_device *inetdev_init(struct net_device *dev)
 231{
 232	struct in_device *in_dev;
 
 233
 234	ASSERT_RTNL();
 235
 236	in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
 237	if (!in_dev)
 238		goto out;
 239	memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
 240			sizeof(in_dev->cnf));
 241	in_dev->cnf.sysctl = NULL;
 242	in_dev->dev = dev;
 243	in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
 244	if (!in_dev->arp_parms)
 245		goto out_kfree;
 246	if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
 247		dev_disable_lro(dev);
 248	/* Reference in_dev->dev */
 249	dev_hold(dev);
 250	/* Account for reference dev->ip_ptr (below) */
 251	in_dev_hold(in_dev);
 252
 253	devinet_sysctl_register(in_dev);
 254	ip_mc_init_dev(in_dev);
 255	if (dev->flags & IFF_UP)
 256		ip_mc_up(in_dev);
 
 
 
 
 
 
 
 
 
 257
 258	/* we can receive as soon as ip_ptr is set -- do this last */
 259	rcu_assign_pointer(dev->ip_ptr, in_dev);
 260out:
 261	return in_dev;
 262out_kfree:
 263	kfree(in_dev);
 264	in_dev = NULL;
 265	goto out;
 266}
 267
 268static void in_dev_rcu_put(struct rcu_head *head)
 269{
 270	struct in_device *idev = container_of(head, struct in_device, rcu_head);
 271	in_dev_put(idev);
 272}
 273
 274static void inetdev_destroy(struct in_device *in_dev)
 275{
 
 276	struct in_ifaddr *ifa;
 277	struct net_device *dev;
 278
 279	ASSERT_RTNL();
 280
 281	dev = in_dev->dev;
 282
 283	in_dev->dead = 1;
 284
 285	ip_mc_destroy_dev(in_dev);
 286
 287	while ((ifa = in_dev->ifa_list) != NULL) {
 288		inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 289		inet_free_ifa(ifa);
 290	}
 291
 292	RCU_INIT_POINTER(dev->ip_ptr, NULL);
 293
 294	devinet_sysctl_unregister(in_dev);
 295	neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 296	arp_ifdown(dev);
 297
 298	call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 299}
 300
 
 
 
 
 
 
 
 
 
 
 
 
 
 301int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 302{
 
 
 303	rcu_read_lock();
 304	for_primary_ifa(in_dev) {
 305		if (inet_ifa_match(a, ifa)) {
 306			if (!b || inet_ifa_match(b, ifa)) {
 307				rcu_read_unlock();
 308				return 1;
 309			}
 310		}
 311	} endfor_ifa(in_dev);
 312	rcu_read_unlock();
 313	return 0;
 314}
 315
 316static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 317			 int destroy, struct nlmsghdr *nlh, u32 pid)
 
 318{
 319	struct in_ifaddr *promote = NULL;
 320	struct in_ifaddr *ifa, *ifa1 = *ifap;
 321	struct in_ifaddr *last_prim = in_dev->ifa_list;
 322	struct in_ifaddr *prev_prom = NULL;
 323	int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 324
 325	ASSERT_RTNL();
 326
 
 
 
 
 
 327	/* 1. Deleting primary ifaddr forces deletion all secondaries
 328	 * unless alias promotion is set
 329	 **/
 330
 331	if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 332		struct in_ifaddr **ifap1 = &ifa1->ifa_next;
 333
 334		while ((ifa = *ifap1) != NULL) {
 335			if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 336			    ifa1->ifa_scope <= ifa->ifa_scope)
 337				last_prim = ifa;
 338
 339			if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 340			    ifa1->ifa_mask != ifa->ifa_mask ||
 341			    !inet_ifa_match(ifa1->ifa_address, ifa)) {
 342				ifap1 = &ifa->ifa_next;
 343				prev_prom = ifa;
 344				continue;
 345			}
 346
 347			if (!do_promote) {
 348				inet_hash_remove(ifa);
 349				*ifap1 = ifa->ifa_next;
 350
 351				rtmsg_ifa(RTM_DELADDR, ifa, nlh, pid);
 352				blocking_notifier_call_chain(&inetaddr_chain,
 353						NETDEV_DOWN, ifa);
 354				inet_free_ifa(ifa);
 355			} else {
 356				promote = ifa;
 357				break;
 358			}
 359		}
 360	}
 361
 362	/* On promotion all secondaries from subnet are changing
 363	 * the primary IP, we must remove all their routes silently
 364	 * and later to add them back with new prefsrc. Do this
 365	 * while all addresses are on the device list.
 366	 */
 367	for (ifa = promote; ifa; ifa = ifa->ifa_next) {
 368		if (ifa1->ifa_mask == ifa->ifa_mask &&
 369		    inet_ifa_match(ifa1->ifa_address, ifa))
 370			fib_del_ifaddr(ifa, ifa1);
 371	}
 372
 
 373	/* 2. Unlink it */
 374
 375	*ifap = ifa1->ifa_next;
 376	inet_hash_remove(ifa1);
 377
 378	/* 3. Announce address deletion */
 379
 380	/* Send message first, then call notifier.
 381	   At first sight, FIB update triggered by notifier
 382	   will refer to already deleted ifaddr, that could confuse
 383	   netlink listeners. It is not true: look, gated sees
 384	   that route deleted and if it still thinks that ifaddr
 385	   is valid, it will try to restore deleted routes... Grr.
 386	   So that, this order is correct.
 387	 */
 388	rtmsg_ifa(RTM_DELADDR, ifa1, nlh, pid);
 389	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 390
 391	if (promote) {
 392		struct in_ifaddr *next_sec = promote->ifa_next;
 393
 
 394		if (prev_prom) {
 395			prev_prom->ifa_next = promote->ifa_next;
 396			promote->ifa_next = last_prim->ifa_next;
 397			last_prim->ifa_next = promote;
 
 
 
 
 398		}
 399
 400		promote->ifa_flags &= ~IFA_F_SECONDARY;
 401		rtmsg_ifa(RTM_NEWADDR, promote, nlh, pid);
 402		blocking_notifier_call_chain(&inetaddr_chain,
 403				NETDEV_UP, promote);
 404		for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
 
 405			if (ifa1->ifa_mask != ifa->ifa_mask ||
 406			    !inet_ifa_match(ifa1->ifa_address, ifa))
 407					continue;
 408			fib_add_ifaddr(ifa);
 409		}
 410
 411	}
 412	if (destroy)
 413		inet_free_ifa(ifa1);
 414}
 415
 416static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 
 417			 int destroy)
 418{
 419	__inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 420}
 421
 422static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 423			     u32 pid)
 424{
 
 425	struct in_device *in_dev = ifa->ifa_dev;
 426	struct in_ifaddr *ifa1, **ifap, **last_primary;
 
 
 
 427
 428	ASSERT_RTNL();
 429
 430	if (!ifa->ifa_local) {
 431		inet_free_ifa(ifa);
 432		return 0;
 433	}
 434
 435	ifa->ifa_flags &= ~IFA_F_SECONDARY;
 436	last_primary = &in_dev->ifa_list;
 437
 438	for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 439	     ifap = &ifa1->ifa_next) {
 
 
 
 
 
 440		if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 441		    ifa->ifa_scope <= ifa1->ifa_scope)
 442			last_primary = &ifa1->ifa_next;
 443		if (ifa1->ifa_mask == ifa->ifa_mask &&
 444		    inet_ifa_match(ifa1->ifa_address, ifa)) {
 445			if (ifa1->ifa_local == ifa->ifa_local) {
 446				inet_free_ifa(ifa);
 447				return -EEXIST;
 448			}
 449			if (ifa1->ifa_scope != ifa->ifa_scope) {
 
 450				inet_free_ifa(ifa);
 451				return -EINVAL;
 452			}
 453			ifa->ifa_flags |= IFA_F_SECONDARY;
 454		}
 
 
 
 455	}
 456
 457	if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 458		net_srandom(ifa->ifa_local);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 459		ifap = last_primary;
 460	}
 461
 462	ifa->ifa_next = *ifap;
 463	*ifap = ifa;
 464
 465	inet_hash_insert(dev_net(in_dev->dev), ifa);
 466
 
 
 
 467	/* Send message first, then call notifier.
 468	   Notifier will trigger FIB update, so that
 469	   listeners of netlink will know about new ifaddr */
 470	rtmsg_ifa(RTM_NEWADDR, ifa, nlh, pid);
 471	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 472
 473	return 0;
 474}
 475
 476static int inet_insert_ifa(struct in_ifaddr *ifa)
 477{
 478	return __inet_insert_ifa(ifa, NULL, 0);
 
 
 
 
 
 479}
 480
 481static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 482{
 483	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 484
 485	ASSERT_RTNL();
 
 486
 487	if (!in_dev) {
 488		inet_free_ifa(ifa);
 489		return -ENOBUFS;
 490	}
 491	ipv4_devconf_setall(in_dev);
 492	if (ifa->ifa_dev != in_dev) {
 493		WARN_ON(ifa->ifa_dev);
 494		in_dev_hold(in_dev);
 495		ifa->ifa_dev = in_dev;
 496	}
 497	if (ipv4_is_loopback(ifa->ifa_local))
 498		ifa->ifa_scope = RT_SCOPE_HOST;
 499	return inet_insert_ifa(ifa);
 500}
 501
 502/* Caller must hold RCU or RTNL :
 503 * We dont take a reference on found in_device
 504 */
 505struct in_device *inetdev_by_index(struct net *net, int ifindex)
 506{
 507	struct net_device *dev;
 508	struct in_device *in_dev = NULL;
 509
 510	rcu_read_lock();
 511	dev = dev_get_by_index_rcu(net, ifindex);
 512	if (dev)
 513		in_dev = rcu_dereference_rtnl(dev->ip_ptr);
 514	rcu_read_unlock();
 515	return in_dev;
 516}
 517EXPORT_SYMBOL(inetdev_by_index);
 518
 519/* Called only from RTNL semaphored context. No locks. */
 520
 521struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 522				    __be32 mask)
 523{
 
 
 524	ASSERT_RTNL();
 525
 526	for_primary_ifa(in_dev) {
 527		if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 528			return ifa;
 529	} endfor_ifa(in_dev);
 530	return NULL;
 531}
 532
 533static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 534{
 535	struct net *net = sock_net(skb->sk);
 
 536	struct nlattr *tb[IFA_MAX+1];
 537	struct in_device *in_dev;
 538	struct ifaddrmsg *ifm;
 539	struct in_ifaddr *ifa, **ifap;
 540	int err = -EINVAL;
 541
 542	ASSERT_RTNL();
 543
 544	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 
 545	if (err < 0)
 546		goto errout;
 547
 548	ifm = nlmsg_data(nlh);
 
 
 
 549	in_dev = inetdev_by_index(net, ifm->ifa_index);
 550	if (in_dev == NULL) {
 
 551		err = -ENODEV;
 552		goto errout;
 553	}
 554
 555	for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 
 556	     ifap = &ifa->ifa_next) {
 557		if (tb[IFA_LOCAL] &&
 558		    ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
 559			continue;
 560
 561		if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 562			continue;
 563
 564		if (tb[IFA_ADDRESS] &&
 565		    (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 566		    !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
 567			continue;
 568
 569		__inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).pid);
 570		return 0;
 
 
 
 571	}
 572
 
 573	err = -EADDRNOTAVAIL;
 574errout:
 
 
 575	return err;
 576}
 577
 578static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh)
 579{
 580	struct nlattr *tb[IFA_MAX+1];
 581	struct in_ifaddr *ifa;
 582	struct ifaddrmsg *ifm;
 583	struct net_device *dev;
 584	struct in_device *in_dev;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 585	int err;
 586
 587	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 
 588	if (err < 0)
 589		goto errout;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 590
 591	ifm = nlmsg_data(nlh);
 592	err = -EINVAL;
 593	if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
 594		goto errout;
 
 
 
 
 
 595
 596	dev = __dev_get_by_index(net, ifm->ifa_index);
 597	err = -ENODEV;
 598	if (dev == NULL)
 
 599		goto errout;
 
 600
 601	in_dev = __in_dev_get_rtnl(dev);
 602	err = -ENOBUFS;
 603	if (in_dev == NULL)
 604		goto errout;
 605
 606	ifa = inet_alloc_ifa();
 607	if (ifa == NULL)
 608		/*
 609		 * A potential indev allocation can be left alive, it stays
 610		 * assigned to its device and is destroy with it.
 611		 */
 612		goto errout;
 613
 614	ipv4_devconf_setall(in_dev);
 615	in_dev_hold(in_dev);
 616
 617	if (tb[IFA_ADDRESS] == NULL)
 618		tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 619
 620	INIT_HLIST_NODE(&ifa->hash);
 621	ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 622	ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 623	ifa->ifa_flags = ifm->ifa_flags;
 624	ifa->ifa_scope = ifm->ifa_scope;
 625	ifa->ifa_dev = in_dev;
 626
 627	ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
 628	ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
 629
 630	if (tb[IFA_BROADCAST])
 631		ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
 632
 633	if (tb[IFA_LABEL])
 634		nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 635	else
 636		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 637
 
 
 
 
 
 
 638	return ifa;
 639
 640errout:
 641	return ERR_PTR(err);
 642}
 643
 644static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 645{
 
 
 646	struct net *net = sock_net(skb->sk);
 
 
 647	struct in_ifaddr *ifa;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 648
 649	ASSERT_RTNL();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 650
 651	ifa = rtm_to_ifaddr(net, nlh);
 652	if (IS_ERR(ifa))
 653		return PTR_ERR(ifa);
 654
 655	return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).pid);
 656}
 657
 658/*
 659 *	Determine a default network mask, based on the IP address.
 660 */
 661
 662static inline int inet_abc_len(__be32 addr)
 663{
 664	int rc = -1;	/* Something else, probably a multicast. */
 665
 666	if (ipv4_is_zeronet(addr))
 667		rc = 0;
 668	else {
 669		__u32 haddr = ntohl(addr);
 670
 671		if (IN_CLASSA(haddr))
 672			rc = 8;
 673		else if (IN_CLASSB(haddr))
 674			rc = 16;
 675		else if (IN_CLASSC(haddr))
 676			rc = 24;
 
 
 677	}
 678
 679	return rc;
 680}
 681
 682
 683int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 684{
 685	struct ifreq ifr;
 686	struct sockaddr_in sin_orig;
 687	struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
 
 688	struct in_device *in_dev;
 689	struct in_ifaddr **ifap = NULL;
 690	struct in_ifaddr *ifa = NULL;
 691	struct net_device *dev;
 692	char *colon;
 693	int ret = -EFAULT;
 694	int tryaddrmatch = 0;
 695
 696	/*
 697	 *	Fetch the caller's info block into kernel space
 698	 */
 699
 700	if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
 701		goto out;
 702	ifr.ifr_name[IFNAMSIZ - 1] = 0;
 703
 704	/* save original address for comparison */
 705	memcpy(&sin_orig, sin, sizeof(*sin));
 706
 707	colon = strchr(ifr.ifr_name, ':');
 708	if (colon)
 709		*colon = 0;
 710
 711	dev_load(net, ifr.ifr_name);
 712
 713	switch (cmd) {
 714	case SIOCGIFADDR:	/* Get interface address */
 715	case SIOCGIFBRDADDR:	/* Get the broadcast address */
 716	case SIOCGIFDSTADDR:	/* Get the destination address */
 717	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
 718		/* Note that these ioctls will not sleep,
 719		   so that we do not impose a lock.
 720		   One day we will be forced to put shlock here (I mean SMP)
 721		 */
 722		tryaddrmatch = (sin_orig.sin_family == AF_INET);
 723		memset(sin, 0, sizeof(*sin));
 724		sin->sin_family = AF_INET;
 725		break;
 726
 727	case SIOCSIFFLAGS:
 728		ret = -EACCES;
 729		if (!capable(CAP_NET_ADMIN))
 730			goto out;
 731		break;
 732	case SIOCSIFADDR:	/* Set interface address (and family) */
 733	case SIOCSIFBRDADDR:	/* Set the broadcast address */
 734	case SIOCSIFDSTADDR:	/* Set the destination address */
 735	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
 736		ret = -EACCES;
 737		if (!capable(CAP_NET_ADMIN))
 738			goto out;
 739		ret = -EINVAL;
 740		if (sin->sin_family != AF_INET)
 741			goto out;
 742		break;
 743	default:
 744		ret = -EINVAL;
 745		goto out;
 746	}
 747
 748	rtnl_lock();
 749
 750	ret = -ENODEV;
 751	dev = __dev_get_by_name(net, ifr.ifr_name);
 752	if (!dev)
 753		goto done;
 754
 755	if (colon)
 756		*colon = ':';
 757
 758	in_dev = __in_dev_get_rtnl(dev);
 759	if (in_dev) {
 760		if (tryaddrmatch) {
 761			/* Matthias Andree */
 762			/* compare label and address (4.4BSD style) */
 763			/* note: we only do this for a limited set of ioctls
 764			   and only if the original address family was AF_INET.
 765			   This is checked above. */
 766			for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 
 
 767			     ifap = &ifa->ifa_next) {
 768				if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
 769				    sin_orig.sin_addr.s_addr ==
 770							ifa->ifa_local) {
 771					break; /* found */
 772				}
 773			}
 774		}
 775		/* we didn't get a match, maybe the application is
 776		   4.3BSD-style and passed in junk so we fall back to
 777		   comparing just the label */
 778		if (!ifa) {
 779			for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 
 780			     ifap = &ifa->ifa_next)
 781				if (!strcmp(ifr.ifr_name, ifa->ifa_label))
 782					break;
 783		}
 784	}
 785
 786	ret = -EADDRNOTAVAIL;
 787	if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
 788		goto done;
 789
 790	switch (cmd) {
 791	case SIOCGIFADDR:	/* Get interface address */
 
 792		sin->sin_addr.s_addr = ifa->ifa_local;
 793		goto rarok;
 794
 795	case SIOCGIFBRDADDR:	/* Get the broadcast address */
 
 796		sin->sin_addr.s_addr = ifa->ifa_broadcast;
 797		goto rarok;
 798
 799	case SIOCGIFDSTADDR:	/* Get the destination address */
 
 800		sin->sin_addr.s_addr = ifa->ifa_address;
 801		goto rarok;
 802
 803	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
 
 804		sin->sin_addr.s_addr = ifa->ifa_mask;
 805		goto rarok;
 806
 807	case SIOCSIFFLAGS:
 808		if (colon) {
 809			ret = -EADDRNOTAVAIL;
 810			if (!ifa)
 811				break;
 812			ret = 0;
 813			if (!(ifr.ifr_flags & IFF_UP))
 814				inet_del_ifa(in_dev, ifap, 1);
 815			break;
 816		}
 817		ret = dev_change_flags(dev, ifr.ifr_flags);
 
 
 
 818		break;
 819
 820	case SIOCSIFADDR:	/* Set interface address (and family) */
 821		ret = -EINVAL;
 822		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 823			break;
 824
 825		if (!ifa) {
 826			ret = -ENOBUFS;
 827			ifa = inet_alloc_ifa();
 828			INIT_HLIST_NODE(&ifa->hash);
 
 829			if (!ifa)
 830				break;
 
 831			if (colon)
 832				memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
 833			else
 834				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 835		} else {
 836			ret = 0;
 837			if (ifa->ifa_local == sin->sin_addr.s_addr)
 838				break;
 839			inet_del_ifa(in_dev, ifap, 0);
 840			ifa->ifa_broadcast = 0;
 841			ifa->ifa_scope = 0;
 842		}
 843
 844		ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
 845
 846		if (!(dev->flags & IFF_POINTOPOINT)) {
 847			ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
 848			ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
 849			if ((dev->flags & IFF_BROADCAST) &&
 850			    ifa->ifa_prefixlen < 31)
 851				ifa->ifa_broadcast = ifa->ifa_address |
 852						     ~ifa->ifa_mask;
 853		} else {
 854			ifa->ifa_prefixlen = 32;
 855			ifa->ifa_mask = inet_make_mask(32);
 856		}
 
 857		ret = inet_set_ifa(dev, ifa);
 858		break;
 859
 860	case SIOCSIFBRDADDR:	/* Set the broadcast address */
 861		ret = 0;
 862		if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
 863			inet_del_ifa(in_dev, ifap, 0);
 864			ifa->ifa_broadcast = sin->sin_addr.s_addr;
 865			inet_insert_ifa(ifa);
 866		}
 867		break;
 868
 869	case SIOCSIFDSTADDR:	/* Set the destination address */
 870		ret = 0;
 871		if (ifa->ifa_address == sin->sin_addr.s_addr)
 872			break;
 873		ret = -EINVAL;
 874		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 875			break;
 876		ret = 0;
 877		inet_del_ifa(in_dev, ifap, 0);
 878		ifa->ifa_address = sin->sin_addr.s_addr;
 879		inet_insert_ifa(ifa);
 880		break;
 881
 882	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
 883
 884		/*
 885		 *	The mask we set must be legal.
 886		 */
 887		ret = -EINVAL;
 888		if (bad_mask(sin->sin_addr.s_addr, 0))
 889			break;
 890		ret = 0;
 891		if (ifa->ifa_mask != sin->sin_addr.s_addr) {
 892			__be32 old_mask = ifa->ifa_mask;
 893			inet_del_ifa(in_dev, ifap, 0);
 894			ifa->ifa_mask = sin->sin_addr.s_addr;
 895			ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
 896
 897			/* See if current broadcast address matches
 898			 * with current netmask, then recalculate
 899			 * the broadcast address. Otherwise it's a
 900			 * funny address, so don't touch it since
 901			 * the user seems to know what (s)he's doing...
 902			 */
 903			if ((dev->flags & IFF_BROADCAST) &&
 904			    (ifa->ifa_prefixlen < 31) &&
 905			    (ifa->ifa_broadcast ==
 906			     (ifa->ifa_local|~old_mask))) {
 907				ifa->ifa_broadcast = (ifa->ifa_local |
 908						      ~sin->sin_addr.s_addr);
 909			}
 910			inet_insert_ifa(ifa);
 911		}
 912		break;
 913	}
 914done:
 915	rtnl_unlock();
 916out:
 917	return ret;
 918rarok:
 919	rtnl_unlock();
 920	ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
 921	goto out;
 922}
 923
 924static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
 925{
 926	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 927	struct in_ifaddr *ifa;
 928	struct ifreq ifr;
 929	int done = 0;
 930
 
 
 
 931	if (!in_dev)
 932		goto out;
 933
 934	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 935		if (!buf) {
 936			done += sizeof(ifr);
 937			continue;
 938		}
 939		if (len < (int) sizeof(ifr))
 940			break;
 941		memset(&ifr, 0, sizeof(struct ifreq));
 942		if (ifa->ifa_label)
 943			strcpy(ifr.ifr_name, ifa->ifa_label);
 944		else
 945			strcpy(ifr.ifr_name, dev->name);
 946
 947		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
 948		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
 949								ifa->ifa_local;
 950
 951		if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
 952			done = -EFAULT;
 953			break;
 954		}
 955		buf  += sizeof(struct ifreq);
 956		len  -= sizeof(struct ifreq);
 957		done += sizeof(struct ifreq);
 958	}
 959out:
 960	return done;
 961}
 962
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 963__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
 964{
 
 965	__be32 addr = 0;
 
 966	struct in_device *in_dev;
 967	struct net *net = dev_net(dev);
 
 968
 969	rcu_read_lock();
 
 970	in_dev = __in_dev_get_rcu(dev);
 971	if (!in_dev)
 972		goto no_in_dev;
 973
 974	for_primary_ifa(in_dev) {
 975		if (ifa->ifa_scope > scope)
 
 
 
 
 
 976			continue;
 977		if (!dst || inet_ifa_match(dst, ifa)) {
 978			addr = ifa->ifa_local;
 979			break;
 980		}
 981		if (!addr)
 982			addr = ifa->ifa_local;
 983	} endfor_ifa(in_dev);
 984
 985	if (addr)
 986		goto out_unlock;
 987no_in_dev:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 988
 989	/* Not loopback addresses on loopback should be preferred
 990	   in this case. It is importnat that lo is the first interface
 991	   in dev_base list.
 992	 */
 993	for_each_netdev_rcu(net, dev) {
 
 
 
 994		in_dev = __in_dev_get_rcu(dev);
 995		if (!in_dev)
 996			continue;
 997
 998		for_primary_ifa(in_dev) {
 999			if (ifa->ifa_scope != RT_SCOPE_LINK &&
1000			    ifa->ifa_scope <= scope) {
1001				addr = ifa->ifa_local;
1002				goto out_unlock;
1003			}
1004		} endfor_ifa(in_dev);
1005	}
1006out_unlock:
1007	rcu_read_unlock();
1008	return addr;
1009}
1010EXPORT_SYMBOL(inet_select_addr);
1011
1012static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1013			      __be32 local, int scope)
1014{
 
 
 
1015	int same = 0;
1016	__be32 addr = 0;
1017
1018	for_ifa(in_dev) {
 
 
 
 
 
1019		if (!addr &&
1020		    (local == ifa->ifa_local || !local) &&
1021		    ifa->ifa_scope <= scope) {
1022			addr = ifa->ifa_local;
1023			if (same)
1024				break;
1025		}
1026		if (!same) {
1027			same = (!local || inet_ifa_match(local, ifa)) &&
1028				(!dst || inet_ifa_match(dst, ifa));
1029			if (same && addr) {
1030				if (local || !dst)
1031					break;
1032				/* Is the selected addr into dst subnet? */
1033				if (inet_ifa_match(addr, ifa))
1034					break;
1035				/* No, then can we use new local src? */
1036				if (ifa->ifa_scope <= scope) {
1037					addr = ifa->ifa_local;
1038					break;
1039				}
1040				/* search for large dst subnet for addr */
1041				same = 0;
1042			}
1043		}
1044	} endfor_ifa(in_dev);
1045
1046	return same ? addr : 0;
1047}
1048
1049/*
1050 * Confirm that local IP address exists using wildcards:
1051 * - in_dev: only on this interface, 0=any interface
 
1052 * - dst: only in the same subnet as dst, 0=any dst
1053 * - local: address, 0=autoselect the local address
1054 * - scope: maximum allowed scope value for the local address
1055 */
1056__be32 inet_confirm_addr(struct in_device *in_dev,
1057			 __be32 dst, __be32 local, int scope)
1058{
1059	__be32 addr = 0;
1060	struct net_device *dev;
1061	struct net *net;
1062
1063	if (scope != RT_SCOPE_LINK)
1064		return confirm_addr_indev(in_dev, dst, local, scope);
1065
1066	net = dev_net(in_dev->dev);
1067	rcu_read_lock();
1068	for_each_netdev_rcu(net, dev) {
1069		in_dev = __in_dev_get_rcu(dev);
1070		if (in_dev) {
1071			addr = confirm_addr_indev(in_dev, dst, local, scope);
1072			if (addr)
1073				break;
1074		}
1075	}
1076	rcu_read_unlock();
1077
1078	return addr;
1079}
1080EXPORT_SYMBOL(inet_confirm_addr);
1081
1082/*
1083 *	Device notifier
1084 */
1085
1086int register_inetaddr_notifier(struct notifier_block *nb)
1087{
1088	return blocking_notifier_chain_register(&inetaddr_chain, nb);
1089}
1090EXPORT_SYMBOL(register_inetaddr_notifier);
1091
1092int unregister_inetaddr_notifier(struct notifier_block *nb)
1093{
1094	return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1095}
1096EXPORT_SYMBOL(unregister_inetaddr_notifier);
1097
 
 
 
 
 
 
 
 
 
 
 
 
 
1098/* Rename ifa_labels for a device name change. Make some effort to preserve
1099 * existing alias numbering and to create unique labels if possible.
1100*/
1101static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1102{
1103	struct in_ifaddr *ifa;
1104	int named = 0;
1105
1106	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1107		char old[IFNAMSIZ], *dot;
1108
1109		memcpy(old, ifa->ifa_label, IFNAMSIZ);
1110		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1111		if (named++ == 0)
1112			goto skip;
1113		dot = strchr(old, ':');
1114		if (dot == NULL) {
1115			sprintf(old, ":%d", named);
1116			dot = old;
1117		}
1118		if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1119			strcat(ifa->ifa_label, dot);
1120		else
1121			strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1122skip:
1123		rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1124	}
1125}
1126
1127static inline bool inetdev_valid_mtu(unsigned int mtu)
1128{
1129	return mtu >= 68;
1130}
1131
1132static void inetdev_send_gratuitous_arp(struct net_device *dev,
1133					struct in_device *in_dev)
1134
1135{
1136	struct in_ifaddr *ifa;
1137
1138	for (ifa = in_dev->ifa_list; ifa;
1139	     ifa = ifa->ifa_next) {
1140		arp_send(ARPOP_REQUEST, ETH_P_ARP,
1141			 ifa->ifa_local, dev,
1142			 ifa->ifa_local, NULL,
1143			 dev->dev_addr, NULL);
1144	}
1145}
1146
1147/* Called only under RTNL semaphore */
1148
1149static int inetdev_event(struct notifier_block *this, unsigned long event,
1150			 void *ptr)
1151{
1152	struct net_device *dev = ptr;
1153	struct in_device *in_dev = __in_dev_get_rtnl(dev);
1154
1155	ASSERT_RTNL();
1156
1157	if (!in_dev) {
1158		if (event == NETDEV_REGISTER) {
1159			in_dev = inetdev_init(dev);
1160			if (!in_dev)
1161				return notifier_from_errno(-ENOMEM);
1162			if (dev->flags & IFF_LOOPBACK) {
1163				IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1164				IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1165			}
1166		} else if (event == NETDEV_CHANGEMTU) {
1167			/* Re-enabling IP */
1168			if (inetdev_valid_mtu(dev->mtu))
1169				in_dev = inetdev_init(dev);
1170		}
1171		goto out;
1172	}
1173
1174	switch (event) {
1175	case NETDEV_REGISTER:
1176		pr_debug("%s: bug\n", __func__);
1177		RCU_INIT_POINTER(dev->ip_ptr, NULL);
1178		break;
1179	case NETDEV_UP:
1180		if (!inetdev_valid_mtu(dev->mtu))
1181			break;
1182		if (dev->flags & IFF_LOOPBACK) {
1183			struct in_ifaddr *ifa = inet_alloc_ifa();
1184
1185			if (ifa) {
1186				INIT_HLIST_NODE(&ifa->hash);
1187				ifa->ifa_local =
1188				  ifa->ifa_address = htonl(INADDR_LOOPBACK);
1189				ifa->ifa_prefixlen = 8;
1190				ifa->ifa_mask = inet_make_mask(8);
1191				in_dev_hold(in_dev);
1192				ifa->ifa_dev = in_dev;
1193				ifa->ifa_scope = RT_SCOPE_HOST;
1194				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 
 
 
 
1195				inet_insert_ifa(ifa);
1196			}
1197		}
1198		ip_mc_up(in_dev);
1199		/* fall through */
1200	case NETDEV_CHANGEADDR:
1201		if (!IN_DEV_ARP_NOTIFY(in_dev))
1202			break;
1203		/* fall through */
1204	case NETDEV_NOTIFY_PEERS:
1205		/* Send gratuitous ARP to notify of link change */
1206		inetdev_send_gratuitous_arp(dev, in_dev);
1207		break;
1208	case NETDEV_DOWN:
1209		ip_mc_down(in_dev);
1210		break;
1211	case NETDEV_PRE_TYPE_CHANGE:
1212		ip_mc_unmap(in_dev);
1213		break;
1214	case NETDEV_POST_TYPE_CHANGE:
1215		ip_mc_remap(in_dev);
1216		break;
1217	case NETDEV_CHANGEMTU:
1218		if (inetdev_valid_mtu(dev->mtu))
1219			break;
1220		/* disable IP when MTU is not enough */
 
1221	case NETDEV_UNREGISTER:
1222		inetdev_destroy(in_dev);
1223		break;
1224	case NETDEV_CHANGENAME:
1225		/* Do not notify about label change, this event is
1226		 * not interesting to applications using netlink.
1227		 */
1228		inetdev_changename(dev, in_dev);
1229
1230		devinet_sysctl_unregister(in_dev);
1231		devinet_sysctl_register(in_dev);
1232		break;
1233	}
1234out:
1235	return NOTIFY_DONE;
1236}
1237
1238static struct notifier_block ip_netdev_notifier = {
1239	.notifier_call = inetdev_event,
1240};
1241
1242static inline size_t inet_nlmsg_size(void)
1243{
1244	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1245	       + nla_total_size(4) /* IFA_ADDRESS */
1246	       + nla_total_size(4) /* IFA_LOCAL */
1247	       + nla_total_size(4) /* IFA_BROADCAST */
1248	       + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
 
 
 
 
1249}
1250
1251static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1252			    u32 pid, u32 seq, int event, unsigned int flags)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1253{
1254	struct ifaddrmsg *ifm;
1255	struct nlmsghdr  *nlh;
1256
1257	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*ifm), flags);
1258	if (nlh == NULL)
 
 
 
 
1259		return -EMSGSIZE;
1260
1261	ifm = nlmsg_data(nlh);
1262	ifm->ifa_family = AF_INET;
1263	ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1264	ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
 
 
 
 
 
 
1265	ifm->ifa_scope = ifa->ifa_scope;
1266	ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1267
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1268	if ((ifa->ifa_address &&
1269	     nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1270	    (ifa->ifa_local &&
1271	     nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1272	    (ifa->ifa_broadcast &&
1273	     nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1274	    (ifa->ifa_label[0] &&
1275	     nla_put_string(skb, IFA_LABEL, ifa->ifa_label)))
 
 
 
 
 
 
 
1276		goto nla_put_failure;
1277
1278	return nlmsg_end(skb, nlh);
 
1279
1280nla_put_failure:
1281	nlmsg_cancel(skb, nlh);
1282	return -EMSGSIZE;
1283}
1284
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1285static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1286{
 
 
 
 
 
 
 
 
1287	struct net *net = sock_net(skb->sk);
1288	int h, s_h;
1289	int idx, s_idx;
1290	int ip_idx, s_ip_idx;
 
 
 
1291	struct net_device *dev;
1292	struct in_device *in_dev;
1293	struct in_ifaddr *ifa;
1294	struct hlist_head *head;
1295	struct hlist_node *node;
1296
1297	s_h = cb->args[0];
1298	s_idx = idx = cb->args[1];
1299	s_ip_idx = ip_idx = cb->args[2];
1300
1301	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1302		idx = 0;
1303		head = &net->dev_index_head[h];
1304		rcu_read_lock();
1305		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1306			if (idx < s_idx)
1307				goto cont;
1308			if (h > s_h || idx > s_idx)
1309				s_ip_idx = 0;
1310			in_dev = __in_dev_get_rcu(dev);
1311			if (!in_dev)
1312				goto cont;
1313
1314			for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1315			     ifa = ifa->ifa_next, ip_idx++) {
1316				if (ip_idx < s_ip_idx)
1317					continue;
1318				if (inet_fill_ifaddr(skb, ifa,
1319					     NETLINK_CB(cb->skb).pid,
1320					     cb->nlh->nlmsg_seq,
1321					     RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1322					rcu_read_unlock();
1323					goto done;
1324				}
1325			}
1326cont:
1327			idx++;
1328		}
1329		rcu_read_unlock();
1330	}
1331
 
 
 
 
 
 
 
 
 
 
 
1332done:
1333	cb->args[0] = h;
1334	cb->args[1] = idx;
1335	cb->args[2] = ip_idx;
1336
1337	return skb->len;
1338}
1339
1340static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1341		      u32 pid)
1342{
 
 
 
 
 
 
 
1343	struct sk_buff *skb;
1344	u32 seq = nlh ? nlh->nlmsg_seq : 0;
1345	int err = -ENOBUFS;
1346	struct net *net;
1347
1348	net = dev_net(ifa->ifa_dev->dev);
1349	skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1350	if (skb == NULL)
1351		goto errout;
1352
1353	err = inet_fill_ifaddr(skb, ifa, pid, seq, event, 0);
1354	if (err < 0) {
1355		/* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1356		WARN_ON(err == -EMSGSIZE);
1357		kfree_skb(skb);
1358		goto errout;
1359	}
1360	rtnl_notify(skb, net, pid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1361	return;
1362errout:
1363	if (err < 0)
1364		rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1365}
1366
1367static size_t inet_get_link_af_size(const struct net_device *dev)
 
1368{
1369	struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1370
1371	if (!in_dev)
1372		return 0;
1373
1374	return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1375}
1376
1377static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
 
1378{
1379	struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1380	struct nlattr *nla;
1381	int i;
1382
1383	if (!in_dev)
1384		return -ENODATA;
1385
1386	nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1387	if (nla == NULL)
1388		return -EMSGSIZE;
1389
1390	for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1391		((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1392
1393	return 0;
1394}
1395
1396static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1397	[IFLA_INET_CONF]	= { .type = NLA_NESTED },
1398};
1399
1400static int inet_validate_link_af(const struct net_device *dev,
1401				 const struct nlattr *nla)
 
1402{
1403	struct nlattr *a, *tb[IFLA_INET_MAX+1];
1404	int err, rem;
1405
1406	if (dev && !__in_dev_get_rtnl(dev))
1407		return -EAFNOSUPPORT;
1408
1409	err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
 
1410	if (err < 0)
1411		return err;
1412
1413	if (tb[IFLA_INET_CONF]) {
1414		nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1415			int cfgid = nla_type(a);
1416
1417			if (nla_len(a) < 4)
1418				return -EINVAL;
1419
1420			if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1421				return -EINVAL;
1422		}
1423	}
1424
1425	return 0;
1426}
1427
1428static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
 
1429{
1430	struct in_device *in_dev = __in_dev_get_rtnl(dev);
1431	struct nlattr *a, *tb[IFLA_INET_MAX+1];
1432	int rem;
1433
1434	if (!in_dev)
1435		return -EAFNOSUPPORT;
1436
1437	if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1438		BUG();
1439
1440	if (tb[IFLA_INET_CONF]) {
1441		nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1442			ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1443	}
1444
1445	return 0;
1446}
1447
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1448#ifdef CONFIG_SYSCTL
1449
1450static void devinet_copy_dflt_conf(struct net *net, int i)
1451{
1452	struct net_device *dev;
1453
1454	rcu_read_lock();
1455	for_each_netdev_rcu(net, dev) {
1456		struct in_device *in_dev;
1457
1458		in_dev = __in_dev_get_rcu(dev);
1459		if (in_dev && !test_bit(i, in_dev->cnf.state))
1460			in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1461	}
1462	rcu_read_unlock();
1463}
1464
1465/* called with RTNL locked */
1466static void inet_forward_change(struct net *net)
1467{
1468	struct net_device *dev;
1469	int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1470
1471	IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1472	IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
 
 
 
 
 
 
 
 
1473
1474	for_each_netdev(net, dev) {
1475		struct in_device *in_dev;
 
1476		if (on)
1477			dev_disable_lro(dev);
1478		rcu_read_lock();
1479		in_dev = __in_dev_get_rcu(dev);
1480		if (in_dev)
1481			IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1482		rcu_read_unlock();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1483	}
1484}
1485
1486static int devinet_conf_proc(ctl_table *ctl, int write,
1487			     void __user *buffer,
1488			     size_t *lenp, loff_t *ppos)
1489{
1490	int old_value = *(int *)ctl->data;
1491	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1492	int new_value = *(int *)ctl->data;
1493
1494	if (write) {
1495		struct ipv4_devconf *cnf = ctl->extra1;
1496		struct net *net = ctl->extra2;
1497		int i = (int *)ctl->data - cnf->data;
 
1498
1499		set_bit(i, cnf->state);
1500
1501		if (cnf == net->ipv4.devconf_dflt)
1502			devinet_copy_dflt_conf(net, i);
1503		if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1)
 
1504			if ((new_value == 0) && (old_value != 0))
1505				rt_cache_flush(net, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1506	}
1507
1508	return ret;
1509}
1510
1511static int devinet_sysctl_forward(ctl_table *ctl, int write,
1512				  void __user *buffer,
1513				  size_t *lenp, loff_t *ppos)
1514{
1515	int *valp = ctl->data;
1516	int val = *valp;
1517	loff_t pos = *ppos;
1518	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 
 
 
 
 
 
1519
1520	if (write && *valp != val) {
1521		struct net *net = ctl->extra2;
1522
1523		if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
1524			if (!rtnl_trylock()) {
1525				/* Restore the original values before restarting */
1526				*valp = val;
1527				*ppos = pos;
1528				return restart_syscall();
1529			}
1530			if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
1531				inet_forward_change(net);
1532			} else if (*valp) {
1533				struct ipv4_devconf *cnf = ctl->extra1;
1534				struct in_device *idev =
1535					container_of(cnf, struct in_device, cnf);
1536				dev_disable_lro(idev->dev);
 
 
 
 
 
1537			}
1538			rtnl_unlock();
1539			rt_cache_flush(net, 0);
1540		}
 
 
 
 
1541	}
1542
1543	return ret;
1544}
1545
1546static int ipv4_doint_and_flush(ctl_table *ctl, int write,
1547				void __user *buffer,
1548				size_t *lenp, loff_t *ppos)
1549{
1550	int *valp = ctl->data;
1551	int val = *valp;
1552	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1553	struct net *net = ctl->extra2;
1554
1555	if (write && *valp != val)
1556		rt_cache_flush(net, 0);
1557
1558	return ret;
1559}
1560
1561#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
1562	{ \
1563		.procname	= name, \
1564		.data		= ipv4_devconf.data + \
1565				  IPV4_DEVCONF_ ## attr - 1, \
1566		.maxlen		= sizeof(int), \
1567		.mode		= mval, \
1568		.proc_handler	= proc, \
1569		.extra1		= &ipv4_devconf, \
1570	}
1571
1572#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1573	DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
1574
1575#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1576	DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
1577
1578#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
1579	DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
1580
1581#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1582	DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
1583
1584static struct devinet_sysctl_table {
1585	struct ctl_table_header *sysctl_header;
1586	struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
1587} devinet_sysctl = {
1588	.devinet_vars = {
1589		DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
1590					     devinet_sysctl_forward),
1591		DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
 
1592
1593		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1594		DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1595		DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1596		DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1597		DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1598		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1599					"accept_source_route"),
1600		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
1601		DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
1602		DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1603		DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1604		DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1605		DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1606		DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1607		DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1608		DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1609		DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1610		DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
1611		DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
 
 
1612		DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
 
 
 
 
 
 
 
 
 
 
1613
1614		DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1615		DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1616		DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1617					      "force_igmp_version"),
1618		DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
1619					      "promote_secondaries"),
 
 
 
 
1620	},
1621};
1622
1623static int __devinet_sysctl_register(struct net *net, char *dev_name,
1624					struct ipv4_devconf *p)
1625{
1626	int i;
1627	struct devinet_sysctl_table *t;
1628	char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
1629
1630	t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1631	if (!t)
1632		goto out;
1633
1634	for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1635		t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
1636		t->devinet_vars[i].extra1 = p;
1637		t->devinet_vars[i].extra2 = net;
1638	}
1639
1640	snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
1641
1642	t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
1643	if (!t->sysctl_header)
1644		goto free;
1645
1646	p->sysctl = t;
 
 
 
1647	return 0;
1648
1649free:
1650	kfree(t);
1651out:
1652	return -ENOBUFS;
1653}
1654
1655static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
 
1656{
1657	struct devinet_sysctl_table *t = cnf->sysctl;
1658
1659	if (t == NULL)
1660		return;
 
 
 
1661
1662	cnf->sysctl = NULL;
1663	unregister_net_sysctl_table(t->sysctl_header);
1664	kfree(t);
1665}
1666
1667static void devinet_sysctl_register(struct in_device *idev)
1668{
1669	neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL);
1670	__devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
1671					&idev->cnf);
 
 
 
 
 
 
 
 
 
 
1672}
1673
1674static void devinet_sysctl_unregister(struct in_device *idev)
1675{
1676	__devinet_sysctl_unregister(&idev->cnf);
 
 
1677	neigh_sysctl_unregister(idev->arp_parms);
1678}
1679
1680static struct ctl_table ctl_forward_entry[] = {
1681	{
1682		.procname	= "ip_forward",
1683		.data		= &ipv4_devconf.data[
1684					IPV4_DEVCONF_FORWARDING - 1],
1685		.maxlen		= sizeof(int),
1686		.mode		= 0644,
1687		.proc_handler	= devinet_sysctl_forward,
1688		.extra1		= &ipv4_devconf,
1689		.extra2		= &init_net,
1690	},
1691	{ },
1692};
1693#endif
1694
1695static __net_init int devinet_init_net(struct net *net)
1696{
1697	int err;
1698	struct ipv4_devconf *all, *dflt;
1699#ifdef CONFIG_SYSCTL
1700	struct ctl_table *tbl = ctl_forward_entry;
1701	struct ctl_table_header *forw_hdr;
 
1702#endif
 
 
 
1703
1704	err = -ENOMEM;
1705	all = &ipv4_devconf;
1706	dflt = &ipv4_devconf_dflt;
1707
1708	if (!net_eq(net, &init_net)) {
1709		all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
1710		if (all == NULL)
1711			goto err_alloc_all;
1712
1713		dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
1714		if (dflt == NULL)
1715			goto err_alloc_dflt;
 
 
1716
1717#ifdef CONFIG_SYSCTL
1718		tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
1719		if (tbl == NULL)
1720			goto err_alloc_ctl;
1721
1722		tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
1723		tbl[0].extra1 = all;
1724		tbl[0].extra2 = net;
1725#endif
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1726	}
1727
1728#ifdef CONFIG_SYSCTL
1729	err = __devinet_sysctl_register(net, "all", all);
1730	if (err < 0)
1731		goto err_reg_all;
1732
1733	err = __devinet_sysctl_register(net, "default", dflt);
 
1734	if (err < 0)
1735		goto err_reg_dflt;
1736
1737	err = -ENOMEM;
1738	forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
1739	if (forw_hdr == NULL)
 
1740		goto err_reg_ctl;
1741	net->ipv4.forw_hdr = forw_hdr;
1742#endif
1743
 
 
 
 
 
1744	net->ipv4.devconf_all = all;
1745	net->ipv4.devconf_dflt = dflt;
1746	return 0;
1747
1748#ifdef CONFIG_SYSCTL
1749err_reg_ctl:
1750	__devinet_sysctl_unregister(dflt);
1751err_reg_dflt:
1752	__devinet_sysctl_unregister(all);
1753err_reg_all:
1754	if (tbl != ctl_forward_entry)
1755		kfree(tbl);
1756err_alloc_ctl:
1757#endif
1758	if (dflt != &ipv4_devconf_dflt)
1759		kfree(dflt);
1760err_alloc_dflt:
1761	if (all != &ipv4_devconf)
1762		kfree(all);
1763err_alloc_all:
 
 
1764	return err;
1765}
1766
1767static __net_exit void devinet_exit_net(struct net *net)
1768{
1769#ifdef CONFIG_SYSCTL
1770	struct ctl_table *tbl;
 
 
 
1771
 
1772	tbl = net->ipv4.forw_hdr->ctl_table_arg;
1773	unregister_net_sysctl_table(net->ipv4.forw_hdr);
1774	__devinet_sysctl_unregister(net->ipv4.devconf_dflt);
1775	__devinet_sysctl_unregister(net->ipv4.devconf_all);
 
 
1776	kfree(tbl);
1777#endif
1778	kfree(net->ipv4.devconf_dflt);
1779	kfree(net->ipv4.devconf_all);
 
1780}
1781
1782static __net_initdata struct pernet_operations devinet_ops = {
1783	.init = devinet_init_net,
1784	.exit = devinet_exit_net,
1785};
1786
1787static struct rtnl_af_ops inet_af_ops = {
1788	.family		  = AF_INET,
1789	.fill_link_af	  = inet_fill_link_af,
1790	.get_link_af_size = inet_get_link_af_size,
1791	.validate_link_af = inet_validate_link_af,
1792	.set_link_af	  = inet_set_link_af,
1793};
1794
 
 
 
 
 
 
 
 
 
 
 
 
1795void __init devinet_init(void)
1796{
1797	int i;
1798
1799	for (i = 0; i < IN4_ADDR_HSIZE; i++)
1800		INIT_HLIST_HEAD(&inet_addr_lst[i]);
1801
1802	register_pernet_subsys(&devinet_ops);
1803
1804	register_gifconf(PF_INET, inet_gifconf);
1805	register_netdevice_notifier(&ip_netdev_notifier);
1806
1807	rtnl_af_register(&inet_af_ops);
 
1808
1809	rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
1810	rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
1811	rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
1812}
1813