1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   4 *
   5 * Begun April 1, 1996, Mike Shaver.
   6 * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   7 */
   8
   9#include <linux/mm.h>
  10#include <linux/module.h>
  11#include <linux/sysctl.h>
  12#include <linux/igmp.h>
  13#include <linux/inetdevice.h>
  14#include <linux/seqlock.h>
  15#include <linux/init.h>
  16#include <linux/slab.h>
  17#include <linux/nsproxy.h>
  18#include <linux/swap.h>
  19#include <net/snmp.h>
  20#include <net/icmp.h>
  21#include <net/ip.h>
  22#include <net/route.h>
  23#include <net/tcp.h>
  24#include <net/udp.h>
  25#include <net/cipso_ipv4.h>
  26#include <net/inet_frag.h>
  27#include <net/ping.h>
  28#include <net/protocol.h>
  29#include <net/netevent.h>
  30
  31static int zero;
  32static int one = 1;
  33static int four = 4;
  34static int thousand = 1000;
  35static int gso_max_segs = GSO_MAX_SEGS;
  36static int tcp_retr1_max = 255;
  37static int ip_local_port_range_min[] = { 1, 1 };
  38static int ip_local_port_range_max[] = { 65535, 65535 };
  39static int tcp_adv_win_scale_min = -31;
  40static int tcp_adv_win_scale_max = 31;
  41static int ip_privileged_port_min;
  42static int ip_privileged_port_max = 65535;
  43static int ip_ttl_min = 1;
  44static int ip_ttl_max = 255;
  45static int tcp_syn_retries_min = 1;
  46static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  47static int ip_ping_group_range_min[] = { 0, 0 };
  48static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  49
  50/* obsolete */
  51static int sysctl_tcp_low_latency __read_mostly;
  52
  53/* Update system visible IP port range */
  54static void set_local_port_range(struct net *net, int range[2])
  55{
  56	bool same_parity = !((range[0] ^ range[1]) & 1);
  57
  58	write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  59	if (same_parity && !net->ipv4.ip_local_ports.warned) {
  60		net->ipv4.ip_local_ports.warned = true;
  61		pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  62	}
  63	net->ipv4.ip_local_ports.range[0] = range[0];
  64	net->ipv4.ip_local_ports.range[1] = range[1];
  65	write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  66}
  67
  68/* Validate changes from /proc interface. */
  69static int ipv4_local_port_range(struct ctl_table *table, int write,
  70				 void __user *buffer,
  71				 size_t *lenp, loff_t *ppos)
  72{
  73	struct net *net =
  74		container_of(table->data, struct net, ipv4.ip_local_ports.range);
  75	int ret;
  76	int range[2];
  77	struct ctl_table tmp = {
  78		.data = &range,
  79		.maxlen = sizeof(range),
  80		.mode = table->mode,
  81		.extra1 = &ip_local_port_range_min,
  82		.extra2 = &ip_local_port_range_max,
  83	};
  84
  85	inet_get_local_port_range(net, &range[0], &range[1]);
  86
  87	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  88
  89	if (write && ret == 0) {
  90		/* Ensure that the upper limit is not smaller than the lower,
  91		 * and that the lower does not encroach upon the privileged
  92		 * port limit.
  93		 */
  94		if ((range[1] < range[0]) ||
  95		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
  96			ret = -EINVAL;
  97		else
  98			set_local_port_range(net, range);
  99	}
 100
 101	return ret;
 102}
 103
 104/* Validate changes from /proc interface. */
 105static int ipv4_privileged_ports(struct ctl_table *table, int write,
 106				void __user *buffer, size_t *lenp, loff_t *ppos)
 107{
 108	struct net *net = container_of(table->data, struct net,
 109	    ipv4.sysctl_ip_prot_sock);
 110	int ret;
 111	int pports;
 112	int range[2];
 113	struct ctl_table tmp = {
 114		.data = &pports,
 115		.maxlen = sizeof(pports),
 116		.mode = table->mode,
 117		.extra1 = &ip_privileged_port_min,
 118		.extra2 = &ip_privileged_port_max,
 119	};
 120
 121	pports = net->ipv4.sysctl_ip_prot_sock;
 122
 123	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 124
 125	if (write && ret == 0) {
 126		inet_get_local_port_range(net, &range[0], &range[1]);
 127		/* Ensure that the local port range doesn't overlap with the
 128		 * privileged port range.
 129		 */
 130		if (range[0] < pports)
 131			ret = -EINVAL;
 132		else
 133			net->ipv4.sysctl_ip_prot_sock = pports;
 134	}
 135
 136	return ret;
 137}
 138
 139static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
 140{
 141	kgid_t *data = table->data;
 142	struct net *net =
 143		container_of(table->data, struct net, ipv4.ping_group_range.range);
 144	unsigned int seq;
 145	do {
 146		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 147
 148		*low = data[0];
 149		*high = data[1];
 150	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 151}
 152
 153/* Update system visible IP port range */
 154static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 155{
 156	kgid_t *data = table->data;
 157	struct net *net =
 158		container_of(table->data, struct net, ipv4.ping_group_range.range);
 159	write_seqlock(&net->ipv4.ping_group_range.lock);
 160	data[0] = low;
 161	data[1] = high;
 162	write_sequnlock(&net->ipv4.ping_group_range.lock);
 163}
 164
 165/* Validate changes from /proc interface. */
 166static int ipv4_ping_group_range(struct ctl_table *table, int write,
 167				 void __user *buffer,
 168				 size_t *lenp, loff_t *ppos)
 169{
 170	struct user_namespace *user_ns = current_user_ns();
 171	int ret;
 172	gid_t urange[2];
 173	kgid_t low, high;
 174	struct ctl_table tmp = {
 175		.data = &urange,
 176		.maxlen = sizeof(urange),
 177		.mode = table->mode,
 178		.extra1 = &ip_ping_group_range_min,
 179		.extra2 = &ip_ping_group_range_max,
 180	};
 181
 182	inet_get_ping_group_range_table(table, &low, &high);
 183	urange[0] = from_kgid_munged(user_ns, low);
 184	urange[1] = from_kgid_munged(user_ns, high);
 185	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 186
 187	if (write && ret == 0) {
 188		low = make_kgid(user_ns, urange[0]);
 189		high = make_kgid(user_ns, urange[1]);
 190		if (!gid_valid(low) || !gid_valid(high) ||
 191		    (urange[1] < urange[0]) || gid_lt(high, low)) {
 192			low = make_kgid(&init_user_ns, 1);
 193			high = make_kgid(&init_user_ns, 0);
 194		}
 195		set_ping_group_range(table, low, high);
 196	}
 197
 198	return ret;
 199}
 200
 201static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 202				       void __user *buffer, size_t *lenp, loff_t *ppos)
 203{
 204	struct net *net = container_of(ctl->data, struct net,
 205				       ipv4.tcp_congestion_control);
 206	char val[TCP_CA_NAME_MAX];
 207	struct ctl_table tbl = {
 208		.data = val,
 209		.maxlen = TCP_CA_NAME_MAX,
 210	};
 211	int ret;
 212
 213	tcp_get_default_congestion_control(net, val);
 214
 215	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 216	if (write && ret == 0)
 217		ret = tcp_set_default_congestion_control(net, val);
 218	return ret;
 219}
 220
 221static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 222						 int write,
 223						 void __user *buffer, size_t *lenp,
 224						 loff_t *ppos)
 225{
 226	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 227	int ret;
 228
 229	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 230	if (!tbl.data)
 231		return -ENOMEM;
 232	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 233	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 234	kfree(tbl.data);
 235	return ret;
 236}
 237
 238static int proc_allowed_congestion_control(struct ctl_table *ctl,
 239					   int write,
 240					   void __user *buffer, size_t *lenp,
 241					   loff_t *ppos)
 242{
 243	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 244	int ret;
 245
 246	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 247	if (!tbl.data)
 248		return -ENOMEM;
 249
 250	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 251	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 252	if (write && ret == 0)
 253		ret = tcp_set_allowed_congestion_control(tbl.data);
 254	kfree(tbl.data);
 255	return ret;
 256}
 257
 258static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
 259				 void __user *buffer, size_t *lenp,
 260				 loff_t *ppos)
 261{
 262	struct net *net = container_of(table->data, struct net,
 263	    ipv4.sysctl_tcp_fastopen);
 264	struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 265	struct tcp_fastopen_context *ctxt;
 266	int ret;
 267	u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 268
 269	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 270	if (!tbl.data)
 271		return -ENOMEM;
 272
 273	rcu_read_lock();
 274	ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
 275	if (ctxt)
 276		memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 277	else
 278		memset(user_key, 0, sizeof(user_key));
 279	rcu_read_unlock();
 280
 281	snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 282		user_key[0], user_key[1], user_key[2], user_key[3]);
 283	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 284
 285	if (write && ret == 0) {
 286		if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 287			   user_key + 2, user_key + 3) != 4) {
 288			ret = -EINVAL;
 289			goto bad_key;
 290		}
 291		tcp_fastopen_reset_cipher(net, NULL, user_key,
 292					  TCP_FASTOPEN_KEY_LENGTH);
 
 
 
 
 293	}
 294
 295bad_key:
 296	pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 297	       user_key[0], user_key[1], user_key[2], user_key[3],
 298	       (char *)tbl.data, ret);
 299	kfree(tbl.data);
 300	return ret;
 301}
 302
 303static void proc_configure_early_demux(int enabled, int protocol)
 304{
 305	struct net_protocol *ipprot;
 306#if IS_ENABLED(CONFIG_IPV6)
 307	struct inet6_protocol *ip6prot;
 308#endif
 309
 310	rcu_read_lock();
 311
 312	ipprot = rcu_dereference(inet_protos[protocol]);
 313	if (ipprot)
 314		ipprot->early_demux = enabled ? ipprot->early_demux_handler :
 315						NULL;
 316
 317#if IS_ENABLED(CONFIG_IPV6)
 318	ip6prot = rcu_dereference(inet6_protos[protocol]);
 319	if (ip6prot)
 320		ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
 321						 NULL;
 322#endif
 323	rcu_read_unlock();
 324}
 325
 326static int proc_tcp_early_demux(struct ctl_table *table, int write,
 327				void __user *buffer, size_t *lenp, loff_t *ppos)
 328{
 329	int ret = 0;
 330
 331	ret = proc_dointvec(table, write, buffer, lenp, ppos);
 332
 333	if (write && !ret) {
 334		int enabled = init_net.ipv4.sysctl_tcp_early_demux;
 335
 336		proc_configure_early_demux(enabled, IPPROTO_TCP);
 337	}
 338
 339	return ret;
 340}
 341
 342static int proc_udp_early_demux(struct ctl_table *table, int write,
 343				void __user *buffer, size_t *lenp, loff_t *ppos)
 344{
 345	int ret = 0;
 346
 347	ret = proc_dointvec(table, write, buffer, lenp, ppos);
 348
 349	if (write && !ret) {
 350		int enabled = init_net.ipv4.sysctl_udp_early_demux;
 351
 352		proc_configure_early_demux(enabled, IPPROTO_UDP);
 353	}
 354
 355	return ret;
 356}
 357
 358static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
 359					     int write,
 360					     void __user *buffer,
 361					     size_t *lenp, loff_t *ppos)
 362{
 363	struct net *net = container_of(table->data, struct net,
 364	    ipv4.sysctl_tcp_fastopen_blackhole_timeout);
 365	int ret;
 366
 367	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 368	if (write && ret == 0)
 369		atomic_set(&net->ipv4.tfo_active_disable_times, 0);
 370
 371	return ret;
 372}
 373
 374static int proc_tcp_available_ulp(struct ctl_table *ctl,
 375				  int write,
 376				  void __user *buffer, size_t *lenp,
 377				  loff_t *ppos)
 378{
 379	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
 380	int ret;
 381
 382	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 383	if (!tbl.data)
 384		return -ENOMEM;
 385	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
 386	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 387	kfree(tbl.data);
 388
 389	return ret;
 390}
 391
 392#ifdef CONFIG_IP_ROUTE_MULTIPATH
 393static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
 394					  void __user *buffer, size_t *lenp,
 395					  loff_t *ppos)
 396{
 397	struct net *net = container_of(table->data, struct net,
 398	    ipv4.sysctl_fib_multipath_hash_policy);
 399	int ret;
 400
 401	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 402	if (write && ret == 0)
 403		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
 404
 405	return ret;
 406}
 407#endif
 408
 409static struct ctl_table ipv4_table[] = {
 410	{
 411		.procname	= "tcp_max_orphans",
 412		.data		= &sysctl_tcp_max_orphans,
 413		.maxlen		= sizeof(int),
 414		.mode		= 0644,
 415		.proc_handler	= proc_dointvec
 416	},
 417	{
 418		.procname	= "inet_peer_threshold",
 419		.data		= &inet_peer_threshold,
 420		.maxlen		= sizeof(int),
 421		.mode		= 0644,
 422		.proc_handler	= proc_dointvec
 423	},
 424	{
 425		.procname	= "inet_peer_minttl",
 426		.data		= &inet_peer_minttl,
 427		.maxlen		= sizeof(int),
 428		.mode		= 0644,
 429		.proc_handler	= proc_dointvec_jiffies,
 430	},
 431	{
 432		.procname	= "inet_peer_maxttl",
 433		.data		= &inet_peer_maxttl,
 434		.maxlen		= sizeof(int),
 435		.mode		= 0644,
 436		.proc_handler	= proc_dointvec_jiffies,
 437	},
 438	{
 439		.procname	= "tcp_mem",
 440		.maxlen		= sizeof(sysctl_tcp_mem),
 441		.data		= &sysctl_tcp_mem,
 442		.mode		= 0644,
 443		.proc_handler	= proc_doulongvec_minmax,
 444	},
 445	{
 446		.procname	= "tcp_low_latency",
 447		.data		= &sysctl_tcp_low_latency,
 448		.maxlen		= sizeof(int),
 449		.mode		= 0644,
 450		.proc_handler	= proc_dointvec
 451	},
 452#ifdef CONFIG_NETLABEL
 453	{
 454		.procname	= "cipso_cache_enable",
 455		.data		= &cipso_v4_cache_enabled,
 456		.maxlen		= sizeof(int),
 457		.mode		= 0644,
 458		.proc_handler	= proc_dointvec,
 459	},
 460	{
 461		.procname	= "cipso_cache_bucket_size",
 462		.data		= &cipso_v4_cache_bucketsize,
 463		.maxlen		= sizeof(int),
 464		.mode		= 0644,
 465		.proc_handler	= proc_dointvec,
 466	},
 467	{
 468		.procname	= "cipso_rbm_optfmt",
 469		.data		= &cipso_v4_rbm_optfmt,
 470		.maxlen		= sizeof(int),
 471		.mode		= 0644,
 472		.proc_handler	= proc_dointvec,
 473	},
 474	{
 475		.procname	= "cipso_rbm_strictvalid",
 476		.data		= &cipso_v4_rbm_strictvalid,
 477		.maxlen		= sizeof(int),
 478		.mode		= 0644,
 479		.proc_handler	= proc_dointvec,
 480	},
 481#endif /* CONFIG_NETLABEL */
 482	{
 483		.procname	= "tcp_available_congestion_control",
 484		.maxlen		= TCP_CA_BUF_MAX,
 485		.mode		= 0444,
 486		.proc_handler   = proc_tcp_available_congestion_control,
 487	},
 488	{
 489		.procname	= "tcp_allowed_congestion_control",
 490		.maxlen		= TCP_CA_BUF_MAX,
 491		.mode		= 0644,
 492		.proc_handler   = proc_allowed_congestion_control,
 493	},
 494	{
 495		.procname	= "tcp_available_ulp",
 496		.maxlen		= TCP_ULP_BUF_MAX,
 497		.mode		= 0444,
 498		.proc_handler   = proc_tcp_available_ulp,
 499	},
 500	{
 501		.procname	= "icmp_msgs_per_sec",
 502		.data		= &sysctl_icmp_msgs_per_sec,
 503		.maxlen		= sizeof(int),
 504		.mode		= 0644,
 505		.proc_handler	= proc_dointvec_minmax,
 506		.extra1		= &zero,
 507	},
 508	{
 509		.procname	= "icmp_msgs_burst",
 510		.data		= &sysctl_icmp_msgs_burst,
 511		.maxlen		= sizeof(int),
 512		.mode		= 0644,
 513		.proc_handler	= proc_dointvec_minmax,
 514		.extra1		= &zero,
 515	},
 516	{
 517		.procname	= "udp_mem",
 518		.data		= &sysctl_udp_mem,
 519		.maxlen		= sizeof(sysctl_udp_mem),
 520		.mode		= 0644,
 521		.proc_handler	= proc_doulongvec_minmax,
 522	},
 523	{ }
 524};
 525
 526static struct ctl_table ipv4_net_table[] = {
 527	{
 528		.procname	= "icmp_echo_ignore_all",
 529		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 530		.maxlen		= sizeof(int),
 531		.mode		= 0644,
 532		.proc_handler	= proc_dointvec
 533	},
 534	{
 535		.procname	= "icmp_echo_ignore_broadcasts",
 536		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 537		.maxlen		= sizeof(int),
 538		.mode		= 0644,
 539		.proc_handler	= proc_dointvec
 
 
 540	},
 541	{
 542		.procname	= "icmp_ignore_bogus_error_responses",
 543		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 544		.maxlen		= sizeof(int),
 545		.mode		= 0644,
 546		.proc_handler	= proc_dointvec
 547	},
 548	{
 549		.procname	= "icmp_errors_use_inbound_ifaddr",
 550		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 551		.maxlen		= sizeof(int),
 552		.mode		= 0644,
 553		.proc_handler	= proc_dointvec
 554	},
 555	{
 556		.procname	= "icmp_ratelimit",
 557		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
 558		.maxlen		= sizeof(int),
 559		.mode		= 0644,
 560		.proc_handler	= proc_dointvec_ms_jiffies,
 561	},
 562	{
 563		.procname	= "icmp_ratemask",
 564		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
 565		.maxlen		= sizeof(int),
 566		.mode		= 0644,
 567		.proc_handler	= proc_dointvec
 568	},
 569	{
 570		.procname	= "ping_group_range",
 571		.data		= &init_net.ipv4.ping_group_range.range,
 572		.maxlen		= sizeof(gid_t)*2,
 573		.mode		= 0644,
 574		.proc_handler	= ipv4_ping_group_range,
 575	},
 576	{
 577		.procname	= "tcp_ecn",
 578		.data		= &init_net.ipv4.sysctl_tcp_ecn,
 579		.maxlen		= sizeof(int),
 580		.mode		= 0644,
 581		.proc_handler	= proc_dointvec
 582	},
 583	{
 584		.procname	= "tcp_ecn_fallback",
 585		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
 586		.maxlen		= sizeof(int),
 587		.mode		= 0644,
 588		.proc_handler	= proc_dointvec
 589	},
 590	{
 591		.procname	= "ip_dynaddr",
 592		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
 593		.maxlen		= sizeof(int),
 594		.mode		= 0644,
 595		.proc_handler	= proc_dointvec
 596	},
 597	{
 598		.procname	= "ip_early_demux",
 599		.data		= &init_net.ipv4.sysctl_ip_early_demux,
 600		.maxlen		= sizeof(int),
 601		.mode		= 0644,
 602		.proc_handler	= proc_dointvec
 603	},
 604	{
 605		.procname       = "udp_early_demux",
 606		.data           = &init_net.ipv4.sysctl_udp_early_demux,
 607		.maxlen         = sizeof(int),
 608		.mode           = 0644,
 609		.proc_handler   = proc_udp_early_demux
 610	},
 611	{
 612		.procname       = "tcp_early_demux",
 613		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
 614		.maxlen         = sizeof(int),
 615		.mode           = 0644,
 616		.proc_handler   = proc_tcp_early_demux
 617	},
 618	{
 619		.procname	= "ip_default_ttl",
 620		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
 621		.maxlen		= sizeof(int),
 622		.mode		= 0644,
 623		.proc_handler	= proc_dointvec_minmax,
 624		.extra1		= &ip_ttl_min,
 625		.extra2		= &ip_ttl_max,
 626	},
 627	{
 628		.procname	= "ip_local_port_range",
 629		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
 630		.data		= &init_net.ipv4.ip_local_ports.range,
 631		.mode		= 0644,
 632		.proc_handler	= ipv4_local_port_range,
 633	},
 634	{
 635		.procname	= "ip_local_reserved_ports",
 636		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
 637		.maxlen		= 65536,
 638		.mode		= 0644,
 639		.proc_handler	= proc_do_large_bitmap,
 640	},
 641	{
 642		.procname	= "ip_no_pmtu_disc",
 643		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 644		.maxlen		= sizeof(int),
 645		.mode		= 0644,
 646		.proc_handler	= proc_dointvec
 647	},
 648	{
 649		.procname	= "ip_forward_use_pmtu",
 650		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 651		.maxlen		= sizeof(int),
 652		.mode		= 0644,
 653		.proc_handler	= proc_dointvec,
 654	},
 
 655	{
 656		.procname	= "ip_nonlocal_bind",
 657		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
 658		.maxlen		= sizeof(int),
 659		.mode		= 0644,
 660		.proc_handler	= proc_dointvec
 661	},
 
 662	{
 663		.procname	= "fwmark_reflect",
 664		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
 665		.maxlen		= sizeof(int),
 666		.mode		= 0644,
 667		.proc_handler	= proc_dointvec,
 668	},
 669	{
 670		.procname	= "tcp_fwmark_accept",
 671		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
 672		.maxlen		= sizeof(int),
 673		.mode		= 0644,
 674		.proc_handler	= proc_dointvec,
 675	},
 676#ifdef CONFIG_NET_L3_MASTER_DEV
 677	{
 678		.procname	= "tcp_l3mdev_accept",
 679		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 680		.maxlen		= sizeof(int),
 681		.mode		= 0644,
 682		.proc_handler	= proc_dointvec_minmax,
 683		.extra1		= &zero,
 684		.extra2		= &one,
 685	},
 686#endif
 687	{
 688		.procname	= "tcp_mtu_probing",
 689		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
 690		.maxlen		= sizeof(int),
 691		.mode		= 0644,
 692		.proc_handler	= proc_dointvec,
 693	},
 694	{
 695		.procname	= "tcp_base_mss",
 696		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
 697		.maxlen		= sizeof(int),
 698		.mode		= 0644,
 699		.proc_handler	= proc_dointvec,
 700	},
 701	{
 702		.procname	= "tcp_probe_threshold",
 703		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
 704		.maxlen		= sizeof(int),
 705		.mode		= 0644,
 706		.proc_handler	= proc_dointvec,
 707	},
 708	{
 709		.procname	= "tcp_probe_interval",
 710		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
 711		.maxlen		= sizeof(int),
 712		.mode		= 0644,
 713		.proc_handler	= proc_dointvec,
 714	},
 715	{
 716		.procname	= "igmp_link_local_mcast_reports",
 717		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
 718		.maxlen		= sizeof(int),
 719		.mode		= 0644,
 720		.proc_handler	= proc_dointvec
 721	},
 722	{
 723		.procname	= "igmp_max_memberships",
 724		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
 725		.maxlen		= sizeof(int),
 726		.mode		= 0644,
 727		.proc_handler	= proc_dointvec
 728	},
 729	{
 730		.procname	= "igmp_max_msf",
 731		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
 732		.maxlen		= sizeof(int),
 733		.mode		= 0644,
 734		.proc_handler	= proc_dointvec
 735	},
 736#ifdef CONFIG_IP_MULTICAST
 737	{
 738		.procname	= "igmp_qrv",
 739		.data		= &init_net.ipv4.sysctl_igmp_qrv,
 740		.maxlen		= sizeof(int),
 741		.mode		= 0644,
 742		.proc_handler	= proc_dointvec_minmax,
 743		.extra1		= &one
 744	},
 745#endif
 746	{
 747		.procname	= "tcp_congestion_control",
 748		.data		= &init_net.ipv4.tcp_congestion_control,
 749		.mode		= 0644,
 750		.maxlen		= TCP_CA_NAME_MAX,
 751		.proc_handler	= proc_tcp_congestion_control,
 752	},
 753	{
 754		.procname	= "tcp_keepalive_time",
 755		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
 756		.maxlen		= sizeof(int),
 757		.mode		= 0644,
 758		.proc_handler	= proc_dointvec_jiffies,
 759	},
 760	{
 761		.procname	= "tcp_keepalive_probes",
 762		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
 763		.maxlen		= sizeof(int),
 764		.mode		= 0644,
 765		.proc_handler	= proc_dointvec
 766	},
 767	{
 768		.procname	= "tcp_keepalive_intvl",
 769		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 770		.maxlen		= sizeof(int),
 771		.mode		= 0644,
 772		.proc_handler	= proc_dointvec_jiffies,
 773	},
 774	{
 775		.procname	= "tcp_syn_retries",
 776		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
 777		.maxlen		= sizeof(int),
 778		.mode		= 0644,
 779		.proc_handler	= proc_dointvec_minmax,
 780		.extra1		= &tcp_syn_retries_min,
 781		.extra2		= &tcp_syn_retries_max
 782	},
 783	{
 784		.procname	= "tcp_synack_retries",
 785		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
 786		.maxlen		= sizeof(int),
 787		.mode		= 0644,
 788		.proc_handler	= proc_dointvec
 789	},
 790#ifdef CONFIG_SYN_COOKIES
 791	{
 792		.procname	= "tcp_syncookies",
 793		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
 794		.maxlen		= sizeof(int),
 795		.mode		= 0644,
 796		.proc_handler	= proc_dointvec
 797	},
 798#endif
 799	{
 800		.procname	= "tcp_reordering",
 801		.data		= &init_net.ipv4.sysctl_tcp_reordering,
 802		.maxlen		= sizeof(int),
 803		.mode		= 0644,
 804		.proc_handler	= proc_dointvec
 805	},
 806	{
 807		.procname	= "tcp_retries1",
 808		.data		= &init_net.ipv4.sysctl_tcp_retries1,
 809		.maxlen		= sizeof(int),
 810		.mode		= 0644,
 811		.proc_handler	= proc_dointvec_minmax,
 812		.extra2		= &tcp_retr1_max
 813	},
 814	{
 815		.procname	= "tcp_retries2",
 816		.data		= &init_net.ipv4.sysctl_tcp_retries2,
 817		.maxlen		= sizeof(int),
 818		.mode		= 0644,
 819		.proc_handler	= proc_dointvec
 820	},
 821	{
 822		.procname	= "tcp_orphan_retries",
 823		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
 824		.maxlen		= sizeof(int),
 825		.mode		= 0644,
 826		.proc_handler	= proc_dointvec
 
 827	},
 828	{
 829		.procname	= "tcp_fin_timeout",
 830		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
 831		.maxlen		= sizeof(int),
 832		.mode		= 0644,
 833		.proc_handler	= proc_dointvec_jiffies,
 834	},
 835	{
 836		.procname	= "tcp_notsent_lowat",
 837		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
 838		.maxlen		= sizeof(unsigned int),
 839		.mode		= 0644,
 840		.proc_handler	= proc_douintvec,
 
 
 841	},
 842	{
 843		.procname	= "tcp_tw_reuse",
 844		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
 845		.maxlen		= sizeof(int),
 846		.mode		= 0644,
 847		.proc_handler	= proc_dointvec
 848	},
 849	{
 850		.procname	= "tcp_max_tw_buckets",
 851		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
 852		.maxlen		= sizeof(int),
 853		.mode		= 0644,
 854		.proc_handler	= proc_dointvec
 855	},
 856	{
 857		.procname	= "tcp_max_syn_backlog",
 858		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
 859		.maxlen		= sizeof(int),
 860		.mode		= 0644,
 861		.proc_handler	= proc_dointvec
 862	},
 863	{
 864		.procname	= "tcp_fastopen",
 865		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
 866		.maxlen		= sizeof(int),
 867		.mode		= 0644,
 868		.proc_handler	= proc_dointvec,
 869	},
 870	{
 871		.procname	= "tcp_fastopen_key",
 872		.mode		= 0600,
 873		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
 874		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 875		.proc_handler	= proc_tcp_fastopen_key,
 876	},
 877	{
 878		.procname	= "tcp_fastopen_blackhole_timeout_sec",
 879		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
 880		.maxlen		= sizeof(int),
 881		.mode		= 0644,
 882		.proc_handler	= proc_tfo_blackhole_detect_timeout,
 883		.extra1		= &zero,
 884	},
 885#ifdef CONFIG_IP_ROUTE_MULTIPATH
 886	{
 887		.procname	= "fib_multipath_use_neigh",
 888		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
 889		.maxlen		= sizeof(int),
 890		.mode		= 0644,
 891		.proc_handler	= proc_dointvec_minmax,
 892		.extra1		= &zero,
 893		.extra2		= &one,
 894	},
 895	{
 896		.procname	= "fib_multipath_hash_policy",
 897		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
 898		.maxlen		= sizeof(int),
 899		.mode		= 0644,
 900		.proc_handler	= proc_fib_multipath_hash_policy,
 901		.extra1		= &zero,
 902		.extra2		= &one,
 903	},
 904#endif
 905	{
 906		.procname	= "ip_unprivileged_port_start",
 
 907		.maxlen		= sizeof(int),
 908		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
 909		.mode		= 0644,
 910		.proc_handler	= ipv4_privileged_ports,
 911	},
 912#ifdef CONFIG_NET_L3_MASTER_DEV
 913	{
 914		.procname	= "udp_l3mdev_accept",
 915		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
 916		.maxlen		= sizeof(int),
 917		.mode		= 0644,
 918		.proc_handler	= proc_dointvec_minmax,
 919		.extra1		= &zero,
 920		.extra2		= &one,
 921	},
 922#endif
 923	{
 924		.procname	= "tcp_sack",
 925		.data		= &init_net.ipv4.sysctl_tcp_sack,
 926		.maxlen		= sizeof(int),
 927		.mode		= 0644,
 928		.proc_handler	= proc_dointvec
 929	},
 930	{
 931		.procname	= "tcp_window_scaling",
 932		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
 933		.maxlen		= sizeof(int),
 934		.mode		= 0644,
 935		.proc_handler	= proc_dointvec
 936	},
 937	{
 938		.procname	= "tcp_timestamps",
 939		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
 940		.maxlen		= sizeof(int),
 941		.mode		= 0644,
 942		.proc_handler	= proc_dointvec
 943	},
 
 944	{
 945		.procname	= "tcp_early_retrans",
 946		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
 947		.maxlen		= sizeof(int),
 948		.mode		= 0644,
 949		.proc_handler	= proc_dointvec_minmax,
 950		.extra1		= &zero,
 951		.extra2		= &four,
 952	},
 953	{
 954		.procname	= "tcp_recovery",
 955		.data		= &init_net.ipv4.sysctl_tcp_recovery,
 956		.maxlen		= sizeof(int),
 957		.mode		= 0644,
 958		.proc_handler	= proc_dointvec,
 959	},
 960	{
 961		.procname       = "tcp_thin_linear_timeouts",
 962		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
 963		.maxlen         = sizeof(int),
 964		.mode           = 0644,
 965		.proc_handler   = proc_dointvec
 966	},
 
 967	{
 968		.procname	= "tcp_slow_start_after_idle",
 969		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
 970		.maxlen		= sizeof(int),
 971		.mode		= 0644,
 972		.proc_handler	= proc_dointvec
 973	},
 
 974	{
 975		.procname	= "tcp_retrans_collapse",
 976		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
 977		.maxlen		= sizeof(int),
 978		.mode		= 0644,
 979		.proc_handler	= proc_dointvec
 980	},
 981	{
 982		.procname	= "tcp_stdurg",
 983		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
 984		.maxlen		= sizeof(int),
 985		.mode		= 0644,
 986		.proc_handler	= proc_dointvec
 987	},
 988	{
 989		.procname	= "tcp_rfc1337",
 990		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
 991		.maxlen		= sizeof(int),
 992		.mode		= 0644,
 993		.proc_handler	= proc_dointvec
 994	},
 995	{
 996		.procname	= "tcp_abort_on_overflow",
 997		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
 998		.maxlen		= sizeof(int),
 999		.mode		= 0644,
1000		.proc_handler	= proc_dointvec
1001	},
 
1002	{
1003		.procname	= "tcp_fack",
1004		.data		= &init_net.ipv4.sysctl_tcp_fack,
1005		.maxlen		= sizeof(int),
1006		.mode		= 0644,
1007		.proc_handler	= proc_dointvec
1008	},
1009	{
1010		.procname	= "tcp_max_reordering",
1011		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1012		.maxlen		= sizeof(int),
1013		.mode		= 0644,
1014		.proc_handler	= proc_dointvec
1015	},
1016	{
1017		.procname	= "tcp_dsack",
1018		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1019		.maxlen		= sizeof(int),
1020		.mode		= 0644,
1021		.proc_handler	= proc_dointvec
1022	},
1023	{
1024		.procname	= "tcp_app_win",
1025		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1026		.maxlen		= sizeof(int),
1027		.mode		= 0644,
1028		.proc_handler	= proc_dointvec
1029	},
1030	{
1031		.procname	= "tcp_adv_win_scale",
1032		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1033		.maxlen		= sizeof(int),
1034		.mode		= 0644,
1035		.proc_handler	= proc_dointvec_minmax,
1036		.extra1		= &tcp_adv_win_scale_min,
1037		.extra2		= &tcp_adv_win_scale_max,
1038	},
1039	{
1040		.procname	= "tcp_frto",
1041		.data		= &init_net.ipv4.sysctl_tcp_frto,
1042		.maxlen		= sizeof(int),
1043		.mode		= 0644,
1044		.proc_handler	= proc_dointvec
 
 
1045	},
1046	{
1047		.procname	= "tcp_no_metrics_save",
1048		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1049		.maxlen		= sizeof(int),
1050		.mode		= 0644,
1051		.proc_handler	= proc_dointvec,
 
 
1052	},
1053	{
1054		.procname	= "tcp_moderate_rcvbuf",
1055		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1056		.maxlen		= sizeof(int),
1057		.mode		= 0644,
1058		.proc_handler	= proc_dointvec,
1059	},
1060	{
1061		.procname	= "tcp_tso_win_divisor",
1062		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1063		.maxlen		= sizeof(int),
1064		.mode		= 0644,
1065		.proc_handler	= proc_dointvec,
 
1066	},
1067	{
1068		.procname	= "tcp_workaround_signed_windows",
1069		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1070		.maxlen		= sizeof(int),
1071		.mode		= 0644,
1072		.proc_handler	= proc_dointvec
 
1073	},
 
 
 
 
1074	{
1075		.procname	= "tcp_limit_output_bytes",
1076		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1077		.maxlen		= sizeof(int),
1078		.mode		= 0644,
1079		.proc_handler	= proc_dointvec
1080	},
1081	{
1082		.procname	= "tcp_challenge_ack_limit",
1083		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1084		.maxlen		= sizeof(int),
1085		.mode		= 0644,
1086		.proc_handler	= proc_dointvec
1087	},
1088	{
1089		.procname	= "tcp_min_tso_segs",
1090		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1091		.maxlen		= sizeof(int),
1092		.mode		= 0644,
1093		.proc_handler	= proc_dointvec_minmax,
1094		.extra1		= &one,
1095		.extra2		= &gso_max_segs,
1096	},
1097	{
1098		.procname	= "tcp_min_rtt_wlen",
1099		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1100		.maxlen		= sizeof(int),
1101		.mode		= 0644,
1102		.proc_handler	= proc_dointvec
1103	},
1104	{
1105		.procname	= "tcp_autocorking",
1106		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1107		.maxlen		= sizeof(int),
1108		.mode		= 0644,
1109		.proc_handler	= proc_dointvec_minmax,
1110		.extra1		= &zero,
1111		.extra2		= &one,
1112	},
1113	{
1114		.procname	= "tcp_invalid_ratelimit",
1115		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1116		.maxlen		= sizeof(int),
1117		.mode		= 0644,
1118		.proc_handler	= proc_dointvec_ms_jiffies,
1119	},
1120	{
1121		.procname	= "tcp_pacing_ss_ratio",
1122		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1123		.maxlen		= sizeof(int),
1124		.mode		= 0644,
1125		.proc_handler	= proc_dointvec_minmax,
1126		.extra1		= &zero,
1127		.extra2		= &thousand,
1128	},
1129	{
1130		.procname	= "tcp_pacing_ca_ratio",
1131		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1132		.maxlen		= sizeof(int),
1133		.mode		= 0644,
1134		.proc_handler	= proc_dointvec_minmax,
1135		.extra1		= &zero,
1136		.extra2		= &thousand,
1137	},
1138	{
1139		.procname	= "tcp_wmem",
1140		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1141		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1142		.mode		= 0644,
1143		.proc_handler	= proc_dointvec_minmax,
1144		.extra1		= &one,
1145	},
1146	{
1147		.procname	= "tcp_rmem",
1148		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1149		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1150		.mode		= 0644,
1151		.proc_handler	= proc_dointvec_minmax,
1152		.extra1		= &one,
1153	},
1154	{
1155		.procname	= "udp_rmem_min",
1156		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1157		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1158		.mode		= 0644,
1159		.proc_handler	= proc_dointvec_minmax,
1160		.extra1		= &one
1161	},
1162	{
1163		.procname	= "udp_wmem_min",
1164		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1165		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1166		.mode		= 0644,
1167		.proc_handler	= proc_dointvec_minmax,
1168		.extra1		= &one
1169	},
1170	{ }
1171};
1172
1173static __net_init int ipv4_sysctl_init_net(struct net *net)
1174{
1175	struct ctl_table *table;
1176
1177	table = ipv4_net_table;
1178	if (!net_eq(net, &init_net)) {
1179		int i;
1180
1181		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1182		if (!table)
1183			goto err_alloc;
1184
1185		/* Update the variables to point into the current struct net */
1186		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1187			table[i].data += (void *)net - (void *)&init_net;
1188	}
1189
1190	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1191	if (!net->ipv4.ipv4_hdr)
1192		goto err_reg;
1193
1194	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1195	if (!net->ipv4.sysctl_local_reserved_ports)
1196		goto err_ports;
1197
1198	return 0;
1199
1200err_ports:
1201	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1202err_reg:
1203	if (!net_eq(net, &init_net))
1204		kfree(table);
1205err_alloc:
1206	return -ENOMEM;
1207}
1208
1209static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1210{
1211	struct ctl_table *table;
1212
1213	kfree(net->ipv4.sysctl_local_reserved_ports);
1214	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1215	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1216	kfree(table);
1217}
1218
1219static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1220	.init = ipv4_sysctl_init_net,
1221	.exit = ipv4_sysctl_exit_net,
1222};
1223
1224static __init int sysctl_ipv4_init(void)
1225{
1226	struct ctl_table_header *hdr;
 
 
 
 
 
 
 
 
 
 
1227
1228	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1229	if (!hdr)
1230		return -ENOMEM;
1231
1232	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1233		unregister_net_sysctl_table(hdr);
1234		return -ENOMEM;
1235	}
1236
1237	return 0;
1238}
1239
1240__initcall(sysctl_ipv4_init);