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/sysctl.h>
10#include <linux/seqlock.h>
11#include <linux/init.h>
12#include <linux/slab.h>
13#include <net/icmp.h>
14#include <net/ip.h>
15#include <net/ip_fib.h>
16#include <net/tcp.h>
17#include <net/udp.h>
18#include <net/cipso_ipv4.h>
19#include <net/ping.h>
20#include <net/protocol.h>
21#include <net/netevent.h>
22
23static int tcp_retr1_max = 255;
24static int ip_local_port_range_min[] = { 1, 1 };
25static int ip_local_port_range_max[] = { 65535, 65535 };
26static int tcp_adv_win_scale_min = -31;
27static int tcp_adv_win_scale_max = 31;
28static int tcp_app_win_max = 31;
29static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
30static int tcp_min_snd_mss_max = 65535;
31static int ip_privileged_port_min;
32static int ip_privileged_port_max = 65535;
33static int ip_ttl_min = 1;
34static int ip_ttl_max = 255;
35static int tcp_syn_retries_min = 1;
36static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
37static int tcp_syn_linear_timeouts_max = MAX_TCP_SYNCNT;
38static unsigned long ip_ping_group_range_min[] = { 0, 0 };
39static unsigned long ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
40static u32 u32_max_div_HZ = UINT_MAX / HZ;
41static int one_day_secs = 24 * 3600;
42static u32 fib_multipath_hash_fields_all_mask __maybe_unused =
43	FIB_MULTIPATH_HASH_FIELD_ALL_MASK;
44static unsigned int tcp_child_ehash_entries_max = 16 * 1024 * 1024;
45static unsigned int udp_child_hash_entries_max = UDP_HTABLE_SIZE_MAX;
46static int tcp_plb_max_rounds = 31;
47static int tcp_plb_max_cong_thresh = 256;
48
49/* obsolete */
50static int sysctl_tcp_low_latency __read_mostly;
51
52/* Update system visible IP port range */
53static void set_local_port_range(struct net *net, int range[2])
54{
55	bool same_parity = !((range[0] ^ range[1]) & 1);
56
57	write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
58	if (same_parity && !net->ipv4.ip_local_ports.warned) {
59		net->ipv4.ip_local_ports.warned = true;
60		pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
61	}
62	net->ipv4.ip_local_ports.range[0] = range[0];
63	net->ipv4.ip_local_ports.range[1] = range[1];
64	write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
65}
66
67/* Validate changes from /proc interface. */
68static int ipv4_local_port_range(struct ctl_table *table, int write,
69				 void *buffer, size_t *lenp, loff_t *ppos)
70{
71	struct net *net =
72		container_of(table->data, struct net, ipv4.ip_local_ports.range);
73	int ret;
74	int range[2];
75	struct ctl_table tmp = {
76		.data = &range,
77		.maxlen = sizeof(range),
78		.mode = table->mode,
79		.extra1 = &ip_local_port_range_min,
80		.extra2 = &ip_local_port_range_max,
81	};
82
83	inet_get_local_port_range(net, &range[0], &range[1]);
84
85	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
86
87	if (write && ret == 0) {
88		/* Ensure that the upper limit is not smaller than the lower,
89		 * and that the lower does not encroach upon the privileged
90		 * port limit.
91		 */
92		if ((range[1] < range[0]) ||
93		    (range[0] < READ_ONCE(net->ipv4.sysctl_ip_prot_sock)))
94			ret = -EINVAL;
95		else
96			set_local_port_range(net, range);
97	}
98
99	return ret;
100}
101
102/* Validate changes from /proc interface. */
103static int ipv4_privileged_ports(struct ctl_table *table, int write,
104				void *buffer, size_t *lenp, loff_t *ppos)
105{
106	struct net *net = container_of(table->data, struct net,
107	    ipv4.sysctl_ip_prot_sock);
108	int ret;
109	int pports;
110	int range[2];
111	struct ctl_table tmp = {
112		.data = &pports,
113		.maxlen = sizeof(pports),
114		.mode = table->mode,
115		.extra1 = &ip_privileged_port_min,
116		.extra2 = &ip_privileged_port_max,
117	};
118
119	pports = READ_ONCE(net->ipv4.sysctl_ip_prot_sock);
120
121	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
122
123	if (write && ret == 0) {
124		inet_get_local_port_range(net, &range[0], &range[1]);
125		/* Ensure that the local port range doesn't overlap with the
126		 * privileged port range.
127		 */
128		if (range[0] < pports)
129			ret = -EINVAL;
130		else
131			WRITE_ONCE(net->ipv4.sysctl_ip_prot_sock, pports);
132	}
133
134	return ret;
135}
136
137static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
138{
139	kgid_t *data = table->data;
140	struct net *net =
141		container_of(table->data, struct net, ipv4.ping_group_range.range);
142	unsigned int seq;
143	do {
144		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
145
146		*low = data[0];
147		*high = data[1];
148	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
149}
150
151/* Update system visible IP port range */
152static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
153{
154	kgid_t *data = table->data;
155	struct net *net =
156		container_of(table->data, struct net, ipv4.ping_group_range.range);
157	write_seqlock(&net->ipv4.ping_group_range.lock);
158	data[0] = low;
159	data[1] = high;
160	write_sequnlock(&net->ipv4.ping_group_range.lock);
161}
162
163/* Validate changes from /proc interface. */
164static int ipv4_ping_group_range(struct ctl_table *table, int write,
165				 void *buffer, size_t *lenp, loff_t *ppos)
166{
167	struct user_namespace *user_ns = current_user_ns();
168	int ret;
169	unsigned long urange[2];
170	kgid_t low, high;
171	struct ctl_table tmp = {
172		.data = &urange,
173		.maxlen = sizeof(urange),
174		.mode = table->mode,
175		.extra1 = &ip_ping_group_range_min,
176		.extra2 = &ip_ping_group_range_max,
177	};
178
179	inet_get_ping_group_range_table(table, &low, &high);
180	urange[0] = from_kgid_munged(user_ns, low);
181	urange[1] = from_kgid_munged(user_ns, high);
182	ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
183
184	if (write && ret == 0) {
185		low = make_kgid(user_ns, urange[0]);
186		high = make_kgid(user_ns, urange[1]);
187		if (!gid_valid(low) || !gid_valid(high))
188			return -EINVAL;
189		if (urange[1] < urange[0] || gid_lt(high, low)) {
190			low = make_kgid(&init_user_ns, 1);
191			high = make_kgid(&init_user_ns, 0);
192		}
193		set_ping_group_range(table, low, high);
194	}
195
196	return ret;
197}
198
199static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
200				    void *buffer, size_t *lenp, loff_t *ppos)
201{
202	struct net *net;
203	int ret;
204
205	net = container_of(table->data, struct net,
206			   ipv4.sysctl_ip_fwd_update_priority);
207	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
208	if (write && ret == 0)
209		call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
210					net);
211
212	return ret;
213}
214
215static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
216				       void *buffer, size_t *lenp, loff_t *ppos)
217{
218	struct net *net = container_of(ctl->data, struct net,
219				       ipv4.tcp_congestion_control);
220	char val[TCP_CA_NAME_MAX];
221	struct ctl_table tbl = {
222		.data = val,
223		.maxlen = TCP_CA_NAME_MAX,
224	};
225	int ret;
226
227	tcp_get_default_congestion_control(net, val);
228
229	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
230	if (write && ret == 0)
231		ret = tcp_set_default_congestion_control(net, val);
232	return ret;
233}
234
235static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
236						 int write, void *buffer,
237						 size_t *lenp, loff_t *ppos)
238{
239	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
240	int ret;
241
242	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
243	if (!tbl.data)
244		return -ENOMEM;
245	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
246	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
247	kfree(tbl.data);
248	return ret;
249}
250
251static int proc_allowed_congestion_control(struct ctl_table *ctl,
252					   int write, void *buffer,
253					   size_t *lenp, loff_t *ppos)
254{
255	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
256	int ret;
257
258	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
259	if (!tbl.data)
260		return -ENOMEM;
261
262	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
263	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
264	if (write && ret == 0)
265		ret = tcp_set_allowed_congestion_control(tbl.data);
266	kfree(tbl.data);
267	return ret;
268}
269
270static int sscanf_key(char *buf, __le32 *key)
271{
272	u32 user_key[4];
273	int i, ret = 0;
274
275	if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
276		   user_key + 2, user_key + 3) != 4) {
277		ret = -EINVAL;
278	} else {
279		for (i = 0; i < ARRAY_SIZE(user_key); i++)
280			key[i] = cpu_to_le32(user_key[i]);
281	}
282	pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
283		 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
284
285	return ret;
286}
287
288static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
289				 void *buffer, size_t *lenp, loff_t *ppos)
290{
291	struct net *net = container_of(table->data, struct net,
292	    ipv4.sysctl_tcp_fastopen);
293	/* maxlen to print the list of keys in hex (*2), with dashes
294	 * separating doublewords and a comma in between keys.
295	 */
296	struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
297					    2 * TCP_FASTOPEN_KEY_MAX) +
298					    (TCP_FASTOPEN_KEY_MAX * 5)) };
299	u32 user_key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(u32)];
300	__le32 key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(__le32)];
301	char *backup_data;
302	int ret, i = 0, off = 0, n_keys;
303
304	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
305	if (!tbl.data)
306		return -ENOMEM;
307
308	n_keys = tcp_fastopen_get_cipher(net, NULL, (u64 *)key);
309	if (!n_keys) {
310		memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
311		n_keys = 1;
312	}
313
314	for (i = 0; i < n_keys * 4; i++)
315		user_key[i] = le32_to_cpu(key[i]);
316
317	for (i = 0; i < n_keys; i++) {
318		off += snprintf(tbl.data + off, tbl.maxlen - off,
319				"%08x-%08x-%08x-%08x",
320				user_key[i * 4],
321				user_key[i * 4 + 1],
322				user_key[i * 4 + 2],
323				user_key[i * 4 + 3]);
324
325		if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
326			break;
327
328		if (i + 1 < n_keys)
329			off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
330	}
331
332	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
333
334	if (write && ret == 0) {
335		backup_data = strchr(tbl.data, ',');
336		if (backup_data) {
337			*backup_data = '\0';
338			backup_data++;
339		}
340		if (sscanf_key(tbl.data, key)) {
341			ret = -EINVAL;
342			goto bad_key;
343		}
344		if (backup_data) {
345			if (sscanf_key(backup_data, key + 4)) {
346				ret = -EINVAL;
347				goto bad_key;
348			}
349		}
350		tcp_fastopen_reset_cipher(net, NULL, key,
351					  backup_data ? key + 4 : NULL);
352	}
353
354bad_key:
355	kfree(tbl.data);
356	return ret;
357}
358
359static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
360					     int write, void *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, void *buffer, size_t *lenp,
376				  loff_t *ppos)
377{
378	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
379	int ret;
380
381	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
382	if (!tbl.data)
383		return -ENOMEM;
384	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
385	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
386	kfree(tbl.data);
387
388	return ret;
389}
390
391static int proc_tcp_ehash_entries(struct ctl_table *table, int write,
392				  void *buffer, size_t *lenp, loff_t *ppos)
393{
394	struct net *net = container_of(table->data, struct net,
395				       ipv4.sysctl_tcp_child_ehash_entries);
396	struct inet_hashinfo *hinfo = net->ipv4.tcp_death_row.hashinfo;
397	int tcp_ehash_entries;
398	struct ctl_table tbl;
399
400	tcp_ehash_entries = hinfo->ehash_mask + 1;
401
402	/* A negative number indicates that the child netns
403	 * shares the global ehash.
404	 */
405	if (!net_eq(net, &init_net) && !hinfo->pernet)
406		tcp_ehash_entries *= -1;
407
408	memset(&tbl, 0, sizeof(tbl));
409	tbl.data = &tcp_ehash_entries;
410	tbl.maxlen = sizeof(int);
411
412	return proc_dointvec(&tbl, write, buffer, lenp, ppos);
413}
414
415static int proc_udp_hash_entries(struct ctl_table *table, int write,
416				 void *buffer, size_t *lenp, loff_t *ppos)
417{
418	struct net *net = container_of(table->data, struct net,
419				       ipv4.sysctl_udp_child_hash_entries);
420	int udp_hash_entries;
421	struct ctl_table tbl;
422
423	udp_hash_entries = net->ipv4.udp_table->mask + 1;
424
425	/* A negative number indicates that the child netns
426	 * shares the global udp_table.
427	 */
428	if (!net_eq(net, &init_net) && net->ipv4.udp_table == &udp_table)
429		udp_hash_entries *= -1;
430
431	memset(&tbl, 0, sizeof(tbl));
432	tbl.data = &udp_hash_entries;
433	tbl.maxlen = sizeof(int);
434
435	return proc_dointvec(&tbl, write, buffer, lenp, ppos);
436}
437
438#ifdef CONFIG_IP_ROUTE_MULTIPATH
439static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
440					  void *buffer, size_t *lenp,
441					  loff_t *ppos)
442{
443	struct net *net = container_of(table->data, struct net,
444	    ipv4.sysctl_fib_multipath_hash_policy);
445	int ret;
446
447	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
448	if (write && ret == 0)
449		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
450
451	return ret;
452}
453
454static int proc_fib_multipath_hash_fields(struct ctl_table *table, int write,
455					  void *buffer, size_t *lenp,
456					  loff_t *ppos)
457{
458	struct net *net;
459	int ret;
460
461	net = container_of(table->data, struct net,
462			   ipv4.sysctl_fib_multipath_hash_fields);
463	ret = proc_douintvec_minmax(table, write, buffer, lenp, ppos);
464	if (write && ret == 0)
465		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
466
467	return ret;
468}
469#endif
470
471static struct ctl_table ipv4_table[] = {
472	{
473		.procname	= "tcp_max_orphans",
474		.data		= &sysctl_tcp_max_orphans,
475		.maxlen		= sizeof(int),
476		.mode		= 0644,
477		.proc_handler	= proc_dointvec
478	},
479	{
480		.procname	= "inet_peer_threshold",
481		.data		= &inet_peer_threshold,
482		.maxlen		= sizeof(int),
483		.mode		= 0644,
484		.proc_handler	= proc_dointvec
485	},
486	{
487		.procname	= "inet_peer_minttl",
488		.data		= &inet_peer_minttl,
489		.maxlen		= sizeof(int),
490		.mode		= 0644,
491		.proc_handler	= proc_dointvec_jiffies,
492	},
493	{
494		.procname	= "inet_peer_maxttl",
495		.data		= &inet_peer_maxttl,
496		.maxlen		= sizeof(int),
497		.mode		= 0644,
498		.proc_handler	= proc_dointvec_jiffies,
499	},
500	{
501		.procname	= "tcp_mem",
502		.maxlen		= sizeof(sysctl_tcp_mem),
503		.data		= &sysctl_tcp_mem,
504		.mode		= 0644,
505		.proc_handler	= proc_doulongvec_minmax,
506	},
507	{
508		.procname	= "tcp_low_latency",
509		.data		= &sysctl_tcp_low_latency,
510		.maxlen		= sizeof(int),
511		.mode		= 0644,
512		.proc_handler	= proc_dointvec
513	},
514#ifdef CONFIG_NETLABEL
515	{
516		.procname	= "cipso_cache_enable",
517		.data		= &cipso_v4_cache_enabled,
518		.maxlen		= sizeof(int),
519		.mode		= 0644,
520		.proc_handler	= proc_dointvec,
521	},
522	{
523		.procname	= "cipso_cache_bucket_size",
524		.data		= &cipso_v4_cache_bucketsize,
525		.maxlen		= sizeof(int),
526		.mode		= 0644,
527		.proc_handler	= proc_dointvec,
528	},
529	{
530		.procname	= "cipso_rbm_optfmt",
531		.data		= &cipso_v4_rbm_optfmt,
532		.maxlen		= sizeof(int),
533		.mode		= 0644,
534		.proc_handler	= proc_dointvec,
535	},
536	{
537		.procname	= "cipso_rbm_strictvalid",
538		.data		= &cipso_v4_rbm_strictvalid,
539		.maxlen		= sizeof(int),
540		.mode		= 0644,
541		.proc_handler	= proc_dointvec,
542	},
543#endif /* CONFIG_NETLABEL */
544	{
545		.procname	= "tcp_available_ulp",
546		.maxlen		= TCP_ULP_BUF_MAX,
547		.mode		= 0444,
548		.proc_handler   = proc_tcp_available_ulp,
549	},
550	{
551		.procname	= "icmp_msgs_per_sec",
552		.data		= &sysctl_icmp_msgs_per_sec,
553		.maxlen		= sizeof(int),
554		.mode		= 0644,
555		.proc_handler	= proc_dointvec_minmax,
556		.extra1		= SYSCTL_ZERO,
557	},
558	{
559		.procname	= "icmp_msgs_burst",
560		.data		= &sysctl_icmp_msgs_burst,
561		.maxlen		= sizeof(int),
562		.mode		= 0644,
563		.proc_handler	= proc_dointvec_minmax,
564		.extra1		= SYSCTL_ZERO,
565	},
566	{
567		.procname	= "udp_mem",
568		.data		= &sysctl_udp_mem,
569		.maxlen		= sizeof(sysctl_udp_mem),
570		.mode		= 0644,
571		.proc_handler	= proc_doulongvec_minmax,
572	},
573	{
574		.procname	= "fib_sync_mem",
575		.data		= &sysctl_fib_sync_mem,
576		.maxlen		= sizeof(sysctl_fib_sync_mem),
577		.mode		= 0644,
578		.proc_handler	= proc_douintvec_minmax,
579		.extra1		= &sysctl_fib_sync_mem_min,
580		.extra2		= &sysctl_fib_sync_mem_max,
581	},
582	{ }
583};
584
585static struct ctl_table ipv4_net_table[] = {
586	{
587		.procname	= "tcp_max_tw_buckets",
588		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
589		.maxlen		= sizeof(int),
590		.mode		= 0644,
591		.proc_handler	= proc_dointvec
592	},
593	{
594		.procname	= "icmp_echo_ignore_all",
595		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
596		.maxlen		= sizeof(u8),
597		.mode		= 0644,
598		.proc_handler	= proc_dou8vec_minmax,
599		.extra1		= SYSCTL_ZERO,
600		.extra2		= SYSCTL_ONE
601	},
602	{
603		.procname	= "icmp_echo_enable_probe",
604		.data		= &init_net.ipv4.sysctl_icmp_echo_enable_probe,
605		.maxlen		= sizeof(u8),
606		.mode		= 0644,
607		.proc_handler	= proc_dou8vec_minmax,
608		.extra1		= SYSCTL_ZERO,
609		.extra2		= SYSCTL_ONE
610	},
611	{
612		.procname	= "icmp_echo_ignore_broadcasts",
613		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
614		.maxlen		= sizeof(u8),
615		.mode		= 0644,
616		.proc_handler	= proc_dou8vec_minmax,
617		.extra1		= SYSCTL_ZERO,
618		.extra2		= SYSCTL_ONE
619	},
620	{
621		.procname	= "icmp_ignore_bogus_error_responses",
622		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
623		.maxlen		= sizeof(u8),
624		.mode		= 0644,
625		.proc_handler	= proc_dou8vec_minmax,
626		.extra1		= SYSCTL_ZERO,
627		.extra2		= SYSCTL_ONE
628	},
629	{
630		.procname	= "icmp_errors_use_inbound_ifaddr",
631		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
632		.maxlen		= sizeof(u8),
633		.mode		= 0644,
634		.proc_handler	= proc_dou8vec_minmax,
635		.extra1		= SYSCTL_ZERO,
636		.extra2		= SYSCTL_ONE
637	},
638	{
639		.procname	= "icmp_ratelimit",
640		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
641		.maxlen		= sizeof(int),
642		.mode		= 0644,
643		.proc_handler	= proc_dointvec_ms_jiffies,
644	},
645	{
646		.procname	= "icmp_ratemask",
647		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
648		.maxlen		= sizeof(int),
649		.mode		= 0644,
650		.proc_handler	= proc_dointvec
651	},
652	{
653		.procname	= "ping_group_range",
654		.data		= &init_net.ipv4.ping_group_range.range,
655		.maxlen		= sizeof(gid_t)*2,
656		.mode		= 0644,
657		.proc_handler	= ipv4_ping_group_range,
658	},
659#ifdef CONFIG_NET_L3_MASTER_DEV
660	{
661		.procname	= "raw_l3mdev_accept",
662		.data		= &init_net.ipv4.sysctl_raw_l3mdev_accept,
663		.maxlen		= sizeof(u8),
664		.mode		= 0644,
665		.proc_handler	= proc_dou8vec_minmax,
666		.extra1		= SYSCTL_ZERO,
667		.extra2		= SYSCTL_ONE,
668	},
669#endif
670	{
671		.procname	= "tcp_ecn",
672		.data		= &init_net.ipv4.sysctl_tcp_ecn,
673		.maxlen		= sizeof(u8),
674		.mode		= 0644,
675		.proc_handler	= proc_dou8vec_minmax,
676		.extra1		= SYSCTL_ZERO,
677		.extra2		= SYSCTL_TWO,
678	},
679	{
680		.procname	= "tcp_ecn_fallback",
681		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
682		.maxlen		= sizeof(u8),
683		.mode		= 0644,
684		.proc_handler	= proc_dou8vec_minmax,
685		.extra1		= SYSCTL_ZERO,
686		.extra2		= SYSCTL_ONE,
687	},
688	{
689		.procname	= "ip_dynaddr",
690		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
691		.maxlen		= sizeof(u8),
692		.mode		= 0644,
693		.proc_handler	= proc_dou8vec_minmax,
694	},
695	{
696		.procname	= "ip_early_demux",
697		.data		= &init_net.ipv4.sysctl_ip_early_demux,
698		.maxlen		= sizeof(u8),
699		.mode		= 0644,
700		.proc_handler	= proc_dou8vec_minmax,
701	},
702	{
703		.procname       = "udp_early_demux",
704		.data           = &init_net.ipv4.sysctl_udp_early_demux,
705		.maxlen         = sizeof(u8),
706		.mode           = 0644,
707		.proc_handler   = proc_dou8vec_minmax,
708	},
709	{
710		.procname       = "tcp_early_demux",
711		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
712		.maxlen         = sizeof(u8),
713		.mode           = 0644,
714		.proc_handler   = proc_dou8vec_minmax,
715	},
716	{
717		.procname       = "nexthop_compat_mode",
718		.data           = &init_net.ipv4.sysctl_nexthop_compat_mode,
719		.maxlen         = sizeof(u8),
720		.mode           = 0644,
721		.proc_handler   = proc_dou8vec_minmax,
722		.extra1		= SYSCTL_ZERO,
723		.extra2		= SYSCTL_ONE,
724	},
725	{
726		.procname	= "ip_default_ttl",
727		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
728		.maxlen		= sizeof(u8),
729		.mode		= 0644,
730		.proc_handler	= proc_dou8vec_minmax,
731		.extra1		= &ip_ttl_min,
732		.extra2		= &ip_ttl_max,
733	},
734	{
735		.procname	= "ip_local_port_range",
736		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
737		.data		= &init_net.ipv4.ip_local_ports.range,
738		.mode		= 0644,
739		.proc_handler	= ipv4_local_port_range,
740	},
741	{
742		.procname	= "ip_local_reserved_ports",
743		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
744		.maxlen		= 65536,
745		.mode		= 0644,
746		.proc_handler	= proc_do_large_bitmap,
747	},
748	{
749		.procname	= "ip_no_pmtu_disc",
750		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
751		.maxlen		= sizeof(u8),
752		.mode		= 0644,
753		.proc_handler	= proc_dou8vec_minmax,
754	},
755	{
756		.procname	= "ip_forward_use_pmtu",
757		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
758		.maxlen		= sizeof(u8),
759		.mode		= 0644,
760		.proc_handler	= proc_dou8vec_minmax,
761	},
762	{
763		.procname	= "ip_forward_update_priority",
764		.data		= &init_net.ipv4.sysctl_ip_fwd_update_priority,
765		.maxlen		= sizeof(u8),
766		.mode		= 0644,
767		.proc_handler   = ipv4_fwd_update_priority,
768		.extra1		= SYSCTL_ZERO,
769		.extra2		= SYSCTL_ONE,
770	},
771	{
772		.procname	= "ip_nonlocal_bind",
773		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
774		.maxlen		= sizeof(u8),
775		.mode		= 0644,
776		.proc_handler	= proc_dou8vec_minmax,
777	},
778	{
779		.procname	= "ip_autobind_reuse",
780		.data		= &init_net.ipv4.sysctl_ip_autobind_reuse,
781		.maxlen		= sizeof(u8),
782		.mode		= 0644,
783		.proc_handler	= proc_dou8vec_minmax,
784		.extra1         = SYSCTL_ZERO,
785		.extra2         = SYSCTL_ONE,
786	},
787	{
788		.procname	= "fwmark_reflect",
789		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
790		.maxlen		= sizeof(u8),
791		.mode		= 0644,
792		.proc_handler	= proc_dou8vec_minmax,
793	},
794	{
795		.procname	= "tcp_fwmark_accept",
796		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
797		.maxlen		= sizeof(u8),
798		.mode		= 0644,
799		.proc_handler	= proc_dou8vec_minmax,
800	},
801#ifdef CONFIG_NET_L3_MASTER_DEV
802	{
803		.procname	= "tcp_l3mdev_accept",
804		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
805		.maxlen		= sizeof(u8),
806		.mode		= 0644,
807		.proc_handler	= proc_dou8vec_minmax,
808		.extra1		= SYSCTL_ZERO,
809		.extra2		= SYSCTL_ONE,
810	},
811#endif
812	{
813		.procname	= "tcp_mtu_probing",
814		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
815		.maxlen		= sizeof(u8),
816		.mode		= 0644,
817		.proc_handler	= proc_dou8vec_minmax,
818	},
819	{
820		.procname	= "tcp_base_mss",
821		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
822		.maxlen		= sizeof(int),
823		.mode		= 0644,
824		.proc_handler	= proc_dointvec,
825	},
826	{
827		.procname	= "tcp_min_snd_mss",
828		.data		= &init_net.ipv4.sysctl_tcp_min_snd_mss,
829		.maxlen		= sizeof(int),
830		.mode		= 0644,
831		.proc_handler	= proc_dointvec_minmax,
832		.extra1		= &tcp_min_snd_mss_min,
833		.extra2		= &tcp_min_snd_mss_max,
834	},
835	{
836		.procname	= "tcp_mtu_probe_floor",
837		.data		= &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
838		.maxlen		= sizeof(int),
839		.mode		= 0644,
840		.proc_handler	= proc_dointvec_minmax,
841		.extra1		= &tcp_min_snd_mss_min,
842		.extra2		= &tcp_min_snd_mss_max,
843	},
844	{
845		.procname	= "tcp_probe_threshold",
846		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
847		.maxlen		= sizeof(int),
848		.mode		= 0644,
849		.proc_handler	= proc_dointvec,
850	},
851	{
852		.procname	= "tcp_probe_interval",
853		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
854		.maxlen		= sizeof(u32),
855		.mode		= 0644,
856		.proc_handler	= proc_douintvec_minmax,
857		.extra2		= &u32_max_div_HZ,
858	},
859	{
860		.procname	= "igmp_link_local_mcast_reports",
861		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
862		.maxlen		= sizeof(u8),
863		.mode		= 0644,
864		.proc_handler	= proc_dou8vec_minmax,
865	},
866	{
867		.procname	= "igmp_max_memberships",
868		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
869		.maxlen		= sizeof(int),
870		.mode		= 0644,
871		.proc_handler	= proc_dointvec
872	},
873	{
874		.procname	= "igmp_max_msf",
875		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
876		.maxlen		= sizeof(int),
877		.mode		= 0644,
878		.proc_handler	= proc_dointvec
879	},
880#ifdef CONFIG_IP_MULTICAST
881	{
882		.procname	= "igmp_qrv",
883		.data		= &init_net.ipv4.sysctl_igmp_qrv,
884		.maxlen		= sizeof(int),
885		.mode		= 0644,
886		.proc_handler	= proc_dointvec_minmax,
887		.extra1		= SYSCTL_ONE
888	},
889#endif
890	{
891		.procname	= "tcp_congestion_control",
892		.data		= &init_net.ipv4.tcp_congestion_control,
893		.mode		= 0644,
894		.maxlen		= TCP_CA_NAME_MAX,
895		.proc_handler	= proc_tcp_congestion_control,
896	},
897	{
898		.procname	= "tcp_available_congestion_control",
899		.maxlen		= TCP_CA_BUF_MAX,
900		.mode		= 0444,
901		.proc_handler   = proc_tcp_available_congestion_control,
902	},
903	{
904		.procname	= "tcp_allowed_congestion_control",
905		.maxlen		= TCP_CA_BUF_MAX,
906		.mode		= 0644,
907		.proc_handler   = proc_allowed_congestion_control,
908	},
909	{
910		.procname	= "tcp_keepalive_time",
911		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
912		.maxlen		= sizeof(int),
913		.mode		= 0644,
914		.proc_handler	= proc_dointvec_jiffies,
915	},
916	{
917		.procname	= "tcp_keepalive_probes",
918		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
919		.maxlen		= sizeof(u8),
920		.mode		= 0644,
921		.proc_handler	= proc_dou8vec_minmax,
922	},
923	{
924		.procname	= "tcp_keepalive_intvl",
925		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
926		.maxlen		= sizeof(int),
927		.mode		= 0644,
928		.proc_handler	= proc_dointvec_jiffies,
929	},
930	{
931		.procname	= "tcp_syn_retries",
932		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
933		.maxlen		= sizeof(u8),
934		.mode		= 0644,
935		.proc_handler	= proc_dou8vec_minmax,
936		.extra1		= &tcp_syn_retries_min,
937		.extra2		= &tcp_syn_retries_max
938	},
939	{
940		.procname	= "tcp_synack_retries",
941		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
942		.maxlen		= sizeof(u8),
943		.mode		= 0644,
944		.proc_handler	= proc_dou8vec_minmax,
945	},
946#ifdef CONFIG_SYN_COOKIES
947	{
948		.procname	= "tcp_syncookies",
949		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
950		.maxlen		= sizeof(u8),
951		.mode		= 0644,
952		.proc_handler	= proc_dou8vec_minmax,
953	},
954#endif
955	{
956		.procname	= "tcp_migrate_req",
957		.data		= &init_net.ipv4.sysctl_tcp_migrate_req,
958		.maxlen		= sizeof(u8),
959		.mode		= 0644,
960		.proc_handler	= proc_dou8vec_minmax,
961		.extra1		= SYSCTL_ZERO,
962		.extra2		= SYSCTL_ONE
963	},
964	{
965		.procname	= "tcp_reordering",
966		.data		= &init_net.ipv4.sysctl_tcp_reordering,
967		.maxlen		= sizeof(int),
968		.mode		= 0644,
969		.proc_handler	= proc_dointvec
970	},
971	{
972		.procname	= "tcp_retries1",
973		.data		= &init_net.ipv4.sysctl_tcp_retries1,
974		.maxlen		= sizeof(u8),
975		.mode		= 0644,
976		.proc_handler	= proc_dou8vec_minmax,
977		.extra2		= &tcp_retr1_max
978	},
979	{
980		.procname	= "tcp_retries2",
981		.data		= &init_net.ipv4.sysctl_tcp_retries2,
982		.maxlen		= sizeof(u8),
983		.mode		= 0644,
984		.proc_handler	= proc_dou8vec_minmax,
985	},
986	{
987		.procname	= "tcp_orphan_retries",
988		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
989		.maxlen		= sizeof(u8),
990		.mode		= 0644,
991		.proc_handler	= proc_dou8vec_minmax,
992	},
993	{
994		.procname	= "tcp_fin_timeout",
995		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
996		.maxlen		= sizeof(int),
997		.mode		= 0644,
998		.proc_handler	= proc_dointvec_jiffies,
999	},
1000	{
1001		.procname	= "tcp_notsent_lowat",
1002		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
1003		.maxlen		= sizeof(unsigned int),
1004		.mode		= 0644,
1005		.proc_handler	= proc_douintvec,
1006	},
1007	{
1008		.procname	= "tcp_tw_reuse",
1009		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
1010		.maxlen		= sizeof(u8),
1011		.mode		= 0644,
1012		.proc_handler	= proc_dou8vec_minmax,
1013		.extra1		= SYSCTL_ZERO,
1014		.extra2		= SYSCTL_TWO,
1015	},
1016	{
1017		.procname	= "tcp_max_syn_backlog",
1018		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
1019		.maxlen		= sizeof(int),
1020		.mode		= 0644,
1021		.proc_handler	= proc_dointvec
1022	},
1023	{
1024		.procname	= "tcp_fastopen",
1025		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1026		.maxlen		= sizeof(int),
1027		.mode		= 0644,
1028		.proc_handler	= proc_dointvec,
1029	},
1030	{
1031		.procname	= "tcp_fastopen_key",
1032		.mode		= 0600,
1033		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1034		/* maxlen to print the list of keys in hex (*2), with dashes
1035		 * separating doublewords and a comma in between keys.
1036		 */
1037		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH *
1038				   2 * TCP_FASTOPEN_KEY_MAX) +
1039				   (TCP_FASTOPEN_KEY_MAX * 5)),
1040		.proc_handler	= proc_tcp_fastopen_key,
1041	},
1042	{
1043		.procname	= "tcp_fastopen_blackhole_timeout_sec",
1044		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1045		.maxlen		= sizeof(int),
1046		.mode		= 0644,
1047		.proc_handler	= proc_tfo_blackhole_detect_timeout,
1048		.extra1		= SYSCTL_ZERO,
1049	},
1050#ifdef CONFIG_IP_ROUTE_MULTIPATH
1051	{
1052		.procname	= "fib_multipath_use_neigh",
1053		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1054		.maxlen		= sizeof(u8),
1055		.mode		= 0644,
1056		.proc_handler	= proc_dou8vec_minmax,
1057		.extra1		= SYSCTL_ZERO,
1058		.extra2		= SYSCTL_ONE,
1059	},
1060	{
1061		.procname	= "fib_multipath_hash_policy",
1062		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1063		.maxlen		= sizeof(u8),
1064		.mode		= 0644,
1065		.proc_handler	= proc_fib_multipath_hash_policy,
1066		.extra1		= SYSCTL_ZERO,
1067		.extra2		= SYSCTL_THREE,
1068	},
1069	{
1070		.procname	= "fib_multipath_hash_fields",
1071		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_fields,
1072		.maxlen		= sizeof(u32),
1073		.mode		= 0644,
1074		.proc_handler	= proc_fib_multipath_hash_fields,
1075		.extra1		= SYSCTL_ONE,
1076		.extra2		= &fib_multipath_hash_fields_all_mask,
1077	},
1078#endif
1079	{
1080		.procname	= "ip_unprivileged_port_start",
1081		.maxlen		= sizeof(int),
1082		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
1083		.mode		= 0644,
1084		.proc_handler	= ipv4_privileged_ports,
1085	},
1086#ifdef CONFIG_NET_L3_MASTER_DEV
1087	{
1088		.procname	= "udp_l3mdev_accept",
1089		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
1090		.maxlen		= sizeof(u8),
1091		.mode		= 0644,
1092		.proc_handler	= proc_dou8vec_minmax,
1093		.extra1		= SYSCTL_ZERO,
1094		.extra2		= SYSCTL_ONE,
1095	},
1096#endif
1097	{
1098		.procname	= "tcp_sack",
1099		.data		= &init_net.ipv4.sysctl_tcp_sack,
1100		.maxlen		= sizeof(u8),
1101		.mode		= 0644,
1102		.proc_handler	= proc_dou8vec_minmax,
1103	},
1104	{
1105		.procname	= "tcp_window_scaling",
1106		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
1107		.maxlen		= sizeof(u8),
1108		.mode		= 0644,
1109		.proc_handler	= proc_dou8vec_minmax,
1110	},
1111	{
1112		.procname	= "tcp_timestamps",
1113		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
1114		.maxlen		= sizeof(u8),
1115		.mode		= 0644,
1116		.proc_handler	= proc_dou8vec_minmax,
1117	},
1118	{
1119		.procname	= "tcp_early_retrans",
1120		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
1121		.maxlen		= sizeof(u8),
1122		.mode		= 0644,
1123		.proc_handler	= proc_dou8vec_minmax,
1124		.extra1		= SYSCTL_ZERO,
1125		.extra2		= SYSCTL_FOUR,
1126	},
1127	{
1128		.procname	= "tcp_recovery",
1129		.data		= &init_net.ipv4.sysctl_tcp_recovery,
1130		.maxlen		= sizeof(u8),
1131		.mode		= 0644,
1132		.proc_handler	= proc_dou8vec_minmax,
1133	},
1134	{
1135		.procname       = "tcp_thin_linear_timeouts",
1136		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1137		.maxlen         = sizeof(u8),
1138		.mode           = 0644,
1139		.proc_handler   = proc_dou8vec_minmax,
1140	},
1141	{
1142		.procname	= "tcp_slow_start_after_idle",
1143		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1144		.maxlen		= sizeof(u8),
1145		.mode		= 0644,
1146		.proc_handler	= proc_dou8vec_minmax,
1147	},
1148	{
1149		.procname	= "tcp_retrans_collapse",
1150		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
1151		.maxlen		= sizeof(u8),
1152		.mode		= 0644,
1153		.proc_handler	= proc_dou8vec_minmax,
1154	},
1155	{
1156		.procname	= "tcp_stdurg",
1157		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
1158		.maxlen		= sizeof(u8),
1159		.mode		= 0644,
1160		.proc_handler	= proc_dou8vec_minmax,
1161	},
1162	{
1163		.procname	= "tcp_rfc1337",
1164		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
1165		.maxlen		= sizeof(u8),
1166		.mode		= 0644,
1167		.proc_handler	= proc_dou8vec_minmax,
1168	},
1169	{
1170		.procname	= "tcp_abort_on_overflow",
1171		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1172		.maxlen		= sizeof(u8),
1173		.mode		= 0644,
1174		.proc_handler	= proc_dou8vec_minmax,
1175	},
1176	{
1177		.procname	= "tcp_fack",
1178		.data		= &init_net.ipv4.sysctl_tcp_fack,
1179		.maxlen		= sizeof(u8),
1180		.mode		= 0644,
1181		.proc_handler	= proc_dou8vec_minmax,
1182	},
1183	{
1184		.procname	= "tcp_max_reordering",
1185		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1186		.maxlen		= sizeof(int),
1187		.mode		= 0644,
1188		.proc_handler	= proc_dointvec
1189	},
1190	{
1191		.procname	= "tcp_dsack",
1192		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1193		.maxlen		= sizeof(u8),
1194		.mode		= 0644,
1195		.proc_handler	= proc_dou8vec_minmax,
1196	},
1197	{
1198		.procname	= "tcp_app_win",
1199		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1200		.maxlen		= sizeof(u8),
1201		.mode		= 0644,
1202		.proc_handler	= proc_dou8vec_minmax,
1203		.extra1		= SYSCTL_ZERO,
1204		.extra2		= &tcp_app_win_max,
1205	},
1206	{
1207		.procname	= "tcp_adv_win_scale",
1208		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1209		.maxlen		= sizeof(int),
1210		.mode		= 0644,
1211		.proc_handler	= proc_dointvec_minmax,
1212		.extra1		= &tcp_adv_win_scale_min,
1213		.extra2		= &tcp_adv_win_scale_max,
1214	},
1215	{
1216		.procname	= "tcp_frto",
1217		.data		= &init_net.ipv4.sysctl_tcp_frto,
1218		.maxlen		= sizeof(u8),
1219		.mode		= 0644,
1220		.proc_handler	= proc_dou8vec_minmax,
1221	},
1222	{
1223		.procname	= "tcp_no_metrics_save",
1224		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1225		.maxlen		= sizeof(u8),
1226		.mode		= 0644,
1227		.proc_handler	= proc_dou8vec_minmax,
1228	},
1229	{
1230		.procname	= "tcp_no_ssthresh_metrics_save",
1231		.data		= &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1232		.maxlen		= sizeof(u8),
1233		.mode		= 0644,
1234		.proc_handler	= proc_dou8vec_minmax,
1235		.extra1		= SYSCTL_ZERO,
1236		.extra2		= SYSCTL_ONE,
1237	},
1238	{
1239		.procname	= "tcp_moderate_rcvbuf",
1240		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1241		.maxlen		= sizeof(u8),
1242		.mode		= 0644,
1243		.proc_handler	= proc_dou8vec_minmax,
1244	},
1245	{
1246		.procname	= "tcp_tso_win_divisor",
1247		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1248		.maxlen		= sizeof(u8),
1249		.mode		= 0644,
1250		.proc_handler	= proc_dou8vec_minmax,
1251	},
1252	{
1253		.procname	= "tcp_workaround_signed_windows",
1254		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1255		.maxlen		= sizeof(u8),
1256		.mode		= 0644,
1257		.proc_handler	= proc_dou8vec_minmax,
1258	},
1259	{
1260		.procname	= "tcp_limit_output_bytes",
1261		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1262		.maxlen		= sizeof(int),
1263		.mode		= 0644,
1264		.proc_handler	= proc_dointvec
1265	},
1266	{
1267		.procname	= "tcp_challenge_ack_limit",
1268		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1269		.maxlen		= sizeof(int),
1270		.mode		= 0644,
1271		.proc_handler	= proc_dointvec
1272	},
1273	{
1274		.procname	= "tcp_min_tso_segs",
1275		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1276		.maxlen		= sizeof(u8),
1277		.mode		= 0644,
1278		.proc_handler	= proc_dou8vec_minmax,
1279		.extra1		= SYSCTL_ONE,
1280	},
1281	{
1282		.procname	= "tcp_tso_rtt_log",
1283		.data		= &init_net.ipv4.sysctl_tcp_tso_rtt_log,
1284		.maxlen		= sizeof(u8),
1285		.mode		= 0644,
1286		.proc_handler	= proc_dou8vec_minmax,
1287	},
1288	{
1289		.procname	= "tcp_min_rtt_wlen",
1290		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1291		.maxlen		= sizeof(int),
1292		.mode		= 0644,
1293		.proc_handler	= proc_dointvec_minmax,
1294		.extra1		= SYSCTL_ZERO,
1295		.extra2		= &one_day_secs
1296	},
1297	{
1298		.procname	= "tcp_autocorking",
1299		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1300		.maxlen		= sizeof(u8),
1301		.mode		= 0644,
1302		.proc_handler	= proc_dou8vec_minmax,
1303		.extra1		= SYSCTL_ZERO,
1304		.extra2		= SYSCTL_ONE,
1305	},
1306	{
1307		.procname	= "tcp_invalid_ratelimit",
1308		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1309		.maxlen		= sizeof(int),
1310		.mode		= 0644,
1311		.proc_handler	= proc_dointvec_ms_jiffies,
1312	},
1313	{
1314		.procname	= "tcp_pacing_ss_ratio",
1315		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1316		.maxlen		= sizeof(int),
1317		.mode		= 0644,
1318		.proc_handler	= proc_dointvec_minmax,
1319		.extra1		= SYSCTL_ZERO,
1320		.extra2		= SYSCTL_ONE_THOUSAND,
1321	},
1322	{
1323		.procname	= "tcp_pacing_ca_ratio",
1324		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1325		.maxlen		= sizeof(int),
1326		.mode		= 0644,
1327		.proc_handler	= proc_dointvec_minmax,
1328		.extra1		= SYSCTL_ZERO,
1329		.extra2		= SYSCTL_ONE_THOUSAND,
1330	},
1331	{
1332		.procname	= "tcp_wmem",
1333		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1334		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1335		.mode		= 0644,
1336		.proc_handler	= proc_dointvec_minmax,
1337		.extra1		= SYSCTL_ONE,
1338	},
1339	{
1340		.procname	= "tcp_rmem",
1341		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1342		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1343		.mode		= 0644,
1344		.proc_handler	= proc_dointvec_minmax,
1345		.extra1		= SYSCTL_ONE,
1346	},
1347	{
1348		.procname	= "tcp_comp_sack_delay_ns",
1349		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1350		.maxlen		= sizeof(unsigned long),
1351		.mode		= 0644,
1352		.proc_handler	= proc_doulongvec_minmax,
1353	},
1354	{
1355		.procname	= "tcp_comp_sack_slack_ns",
1356		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1357		.maxlen		= sizeof(unsigned long),
1358		.mode		= 0644,
1359		.proc_handler	= proc_doulongvec_minmax,
1360	},
1361	{
1362		.procname	= "tcp_comp_sack_nr",
1363		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1364		.maxlen		= sizeof(u8),
1365		.mode		= 0644,
1366		.proc_handler	= proc_dou8vec_minmax,
1367		.extra1		= SYSCTL_ZERO,
1368	},
1369	{
1370		.procname       = "tcp_reflect_tos",
1371		.data           = &init_net.ipv4.sysctl_tcp_reflect_tos,
1372		.maxlen         = sizeof(u8),
1373		.mode           = 0644,
1374		.proc_handler   = proc_dou8vec_minmax,
1375		.extra1         = SYSCTL_ZERO,
1376		.extra2         = SYSCTL_ONE,
1377	},
1378	{
1379		.procname	= "tcp_ehash_entries",
1380		.data		= &init_net.ipv4.sysctl_tcp_child_ehash_entries,
1381		.mode		= 0444,
1382		.proc_handler	= proc_tcp_ehash_entries,
1383	},
1384	{
1385		.procname	= "tcp_child_ehash_entries",
1386		.data		= &init_net.ipv4.sysctl_tcp_child_ehash_entries,
1387		.maxlen		= sizeof(unsigned int),
1388		.mode		= 0644,
1389		.proc_handler	= proc_douintvec_minmax,
1390		.extra1		= SYSCTL_ZERO,
1391		.extra2		= &tcp_child_ehash_entries_max,
1392	},
1393	{
1394		.procname	= "udp_hash_entries",
1395		.data		= &init_net.ipv4.sysctl_udp_child_hash_entries,
1396		.mode		= 0444,
1397		.proc_handler	= proc_udp_hash_entries,
1398	},
1399	{
1400		.procname	= "udp_child_hash_entries",
1401		.data		= &init_net.ipv4.sysctl_udp_child_hash_entries,
1402		.maxlen		= sizeof(unsigned int),
1403		.mode		= 0644,
1404		.proc_handler	= proc_douintvec_minmax,
1405		.extra1		= SYSCTL_ZERO,
1406		.extra2		= &udp_child_hash_entries_max,
1407	},
1408	{
1409		.procname	= "udp_rmem_min",
1410		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1411		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1412		.mode		= 0644,
1413		.proc_handler	= proc_dointvec_minmax,
1414		.extra1		= SYSCTL_ONE
1415	},
1416	{
1417		.procname	= "udp_wmem_min",
1418		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1419		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1420		.mode		= 0644,
1421		.proc_handler	= proc_dointvec_minmax,
1422		.extra1		= SYSCTL_ONE
1423	},
1424	{
1425		.procname	= "fib_notify_on_flag_change",
1426		.data		= &init_net.ipv4.sysctl_fib_notify_on_flag_change,
1427		.maxlen		= sizeof(u8),
1428		.mode		= 0644,
1429		.proc_handler	= proc_dou8vec_minmax,
1430		.extra1		= SYSCTL_ZERO,
1431		.extra2		= SYSCTL_TWO,
1432	},
1433	{
1434		.procname       = "tcp_plb_enabled",
1435		.data           = &init_net.ipv4.sysctl_tcp_plb_enabled,
1436		.maxlen         = sizeof(u8),
1437		.mode           = 0644,
1438		.proc_handler   = proc_dou8vec_minmax,
1439		.extra1         = SYSCTL_ZERO,
1440		.extra2         = SYSCTL_ONE,
1441	},
1442	{
1443		.procname       = "tcp_plb_idle_rehash_rounds",
1444		.data           = &init_net.ipv4.sysctl_tcp_plb_idle_rehash_rounds,
1445		.maxlen         = sizeof(u8),
1446		.mode           = 0644,
1447		.proc_handler   = proc_dou8vec_minmax,
1448		.extra2		= &tcp_plb_max_rounds,
1449	},
1450	{
1451		.procname       = "tcp_plb_rehash_rounds",
1452		.data           = &init_net.ipv4.sysctl_tcp_plb_rehash_rounds,
1453		.maxlen         = sizeof(u8),
1454		.mode           = 0644,
1455		.proc_handler   = proc_dou8vec_minmax,
1456		.extra2         = &tcp_plb_max_rounds,
1457	},
1458	{
1459		.procname       = "tcp_plb_suspend_rto_sec",
1460		.data           = &init_net.ipv4.sysctl_tcp_plb_suspend_rto_sec,
1461		.maxlen         = sizeof(u8),
1462		.mode           = 0644,
1463		.proc_handler   = proc_dou8vec_minmax,
1464	},
1465	{
1466		.procname       = "tcp_plb_cong_thresh",
1467		.data           = &init_net.ipv4.sysctl_tcp_plb_cong_thresh,
1468		.maxlen         = sizeof(int),
1469		.mode           = 0644,
1470		.proc_handler   = proc_dointvec_minmax,
1471		.extra1         = SYSCTL_ZERO,
1472		.extra2         = &tcp_plb_max_cong_thresh,
1473	},
1474	{
1475		.procname	= "tcp_syn_linear_timeouts",
1476		.data		= &init_net.ipv4.sysctl_tcp_syn_linear_timeouts,
1477		.maxlen		= sizeof(u8),
1478		.mode		= 0644,
1479		.proc_handler	= proc_dou8vec_minmax,
1480		.extra1		= SYSCTL_ZERO,
1481		.extra2		= &tcp_syn_linear_timeouts_max,
1482	},
1483	{
1484		.procname	= "tcp_shrink_window",
1485		.data		= &init_net.ipv4.sysctl_tcp_shrink_window,
1486		.maxlen		= sizeof(u8),
1487		.mode		= 0644,
1488		.proc_handler	= proc_dou8vec_minmax,
1489		.extra1		= SYSCTL_ZERO,
1490		.extra2		= SYSCTL_ONE,
1491	},
1492	{ }
1493};
1494
1495static __net_init int ipv4_sysctl_init_net(struct net *net)
1496{
1497	struct ctl_table *table;
1498
1499	table = ipv4_net_table;
1500	if (!net_eq(net, &init_net)) {
1501		int i;
1502
1503		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1504		if (!table)
1505			goto err_alloc;
1506
1507		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) {
1508			if (table[i].data) {
1509				/* Update the variables to point into
1510				 * the current struct net
1511				 */
1512				table[i].data += (void *)net - (void *)&init_net;
1513			} else {
1514				/* Entries without data pointer are global;
1515				 * Make them read-only in non-init_net ns
1516				 */
1517				table[i].mode &= ~0222;
1518			}
1519		}
1520	}
1521
1522	net->ipv4.ipv4_hdr = register_net_sysctl_sz(net, "net/ipv4", table,
1523						    ARRAY_SIZE(ipv4_net_table));
1524	if (!net->ipv4.ipv4_hdr)
1525		goto err_reg;
1526
1527	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1528	if (!net->ipv4.sysctl_local_reserved_ports)
1529		goto err_ports;
1530
1531	return 0;
1532
1533err_ports:
1534	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1535err_reg:
1536	if (!net_eq(net, &init_net))
1537		kfree(table);
1538err_alloc:
1539	return -ENOMEM;
1540}
1541
1542static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1543{
1544	struct ctl_table *table;
1545
1546	kfree(net->ipv4.sysctl_local_reserved_ports);
1547	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1548	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1549	kfree(table);
1550}
1551
1552static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1553	.init = ipv4_sysctl_init_net,
1554	.exit = ipv4_sysctl_exit_net,
1555};
1556
1557static __init int sysctl_ipv4_init(void)
1558{
1559	struct ctl_table_header *hdr;
1560
1561	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1562	if (!hdr)
1563		return -ENOMEM;
1564
1565	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1566		unregister_net_sysctl_table(hdr);
1567		return -ENOMEM;
1568	}
1569
1570	return 0;
1571}
1572
1573__initcall(sysctl_ipv4_init);
1574