xref: /openbmc/linux/net/ipv4/sysctl_net_ipv4.c (revision 94588c1b)
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 
31 static int two = 2;
32 static int four = 4;
33 static int thousand = 1000;
34 static int tcp_retr1_max = 255;
35 static int ip_local_port_range_min[] = { 1, 1 };
36 static int ip_local_port_range_max[] = { 65535, 65535 };
37 static int tcp_adv_win_scale_min = -31;
38 static int tcp_adv_win_scale_max = 31;
39 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
40 static int tcp_min_snd_mss_max = 65535;
41 static int ip_privileged_port_min;
42 static int ip_privileged_port_max = 65535;
43 static int ip_ttl_min = 1;
44 static int ip_ttl_max = 255;
45 static int tcp_syn_retries_min = 1;
46 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
47 static int ip_ping_group_range_min[] = { 0, 0 };
48 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
49 static u32 u32_max_div_HZ = UINT_MAX / HZ;
50 static int one_day_secs = 24 * 3600;
51 
52 /* obsolete */
53 static int sysctl_tcp_low_latency __read_mostly;
54 
55 /* Update system visible IP port range */
56 static void set_local_port_range(struct net *net, int range[2])
57 {
58 	bool same_parity = !((range[0] ^ range[1]) & 1);
59 
60 	write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
61 	if (same_parity && !net->ipv4.ip_local_ports.warned) {
62 		net->ipv4.ip_local_ports.warned = true;
63 		pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
64 	}
65 	net->ipv4.ip_local_ports.range[0] = range[0];
66 	net->ipv4.ip_local_ports.range[1] = range[1];
67 	write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
68 }
69 
70 /* Validate changes from /proc interface. */
71 static int ipv4_local_port_range(struct ctl_table *table, int write,
72 				 void *buffer, size_t *lenp, loff_t *ppos)
73 {
74 	struct net *net =
75 		container_of(table->data, struct net, ipv4.ip_local_ports.range);
76 	int ret;
77 	int range[2];
78 	struct ctl_table tmp = {
79 		.data = &range,
80 		.maxlen = sizeof(range),
81 		.mode = table->mode,
82 		.extra1 = &ip_local_port_range_min,
83 		.extra2 = &ip_local_port_range_max,
84 	};
85 
86 	inet_get_local_port_range(net, &range[0], &range[1]);
87 
88 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
89 
90 	if (write && ret == 0) {
91 		/* Ensure that the upper limit is not smaller than the lower,
92 		 * and that the lower does not encroach upon the privileged
93 		 * port limit.
94 		 */
95 		if ((range[1] < range[0]) ||
96 		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
97 			ret = -EINVAL;
98 		else
99 			set_local_port_range(net, range);
100 	}
101 
102 	return ret;
103 }
104 
105 /* Validate changes from /proc interface. */
106 static int ipv4_privileged_ports(struct ctl_table *table, int write,
107 				void *buffer, size_t *lenp, loff_t *ppos)
108 {
109 	struct net *net = container_of(table->data, struct net,
110 	    ipv4.sysctl_ip_prot_sock);
111 	int ret;
112 	int pports;
113 	int range[2];
114 	struct ctl_table tmp = {
115 		.data = &pports,
116 		.maxlen = sizeof(pports),
117 		.mode = table->mode,
118 		.extra1 = &ip_privileged_port_min,
119 		.extra2 = &ip_privileged_port_max,
120 	};
121 
122 	pports = net->ipv4.sysctl_ip_prot_sock;
123 
124 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
125 
126 	if (write && ret == 0) {
127 		inet_get_local_port_range(net, &range[0], &range[1]);
128 		/* Ensure that the local port range doesn't overlap with the
129 		 * privileged port range.
130 		 */
131 		if (range[0] < pports)
132 			ret = -EINVAL;
133 		else
134 			net->ipv4.sysctl_ip_prot_sock = pports;
135 	}
136 
137 	return ret;
138 }
139 
140 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
141 {
142 	kgid_t *data = table->data;
143 	struct net *net =
144 		container_of(table->data, struct net, ipv4.ping_group_range.range);
145 	unsigned int seq;
146 	do {
147 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
148 
149 		*low = data[0];
150 		*high = data[1];
151 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
152 }
153 
154 /* Update system visible IP port range */
155 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
156 {
157 	kgid_t *data = table->data;
158 	struct net *net =
159 		container_of(table->data, struct net, ipv4.ping_group_range.range);
160 	write_seqlock(&net->ipv4.ping_group_range.lock);
161 	data[0] = low;
162 	data[1] = high;
163 	write_sequnlock(&net->ipv4.ping_group_range.lock);
164 }
165 
166 /* Validate changes from /proc interface. */
167 static int ipv4_ping_group_range(struct ctl_table *table, int write,
168 				 void *buffer, 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 			return -EINVAL;
192 		if (urange[1] < urange[0] || gid_lt(high, low)) {
193 			low = make_kgid(&init_user_ns, 1);
194 			high = make_kgid(&init_user_ns, 0);
195 		}
196 		set_ping_group_range(table, low, high);
197 	}
198 
199 	return ret;
200 }
201 
202 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
203 				    void *buffer, size_t *lenp, loff_t *ppos)
204 {
205 	struct net *net;
206 	int ret;
207 
208 	net = container_of(table->data, struct net,
209 			   ipv4.sysctl_ip_fwd_update_priority);
210 	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
211 	if (write && ret == 0)
212 		call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
213 					net);
214 
215 	return ret;
216 }
217 
218 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
219 				       void *buffer, size_t *lenp, loff_t *ppos)
220 {
221 	struct net *net = container_of(ctl->data, struct net,
222 				       ipv4.tcp_congestion_control);
223 	char val[TCP_CA_NAME_MAX];
224 	struct ctl_table tbl = {
225 		.data = val,
226 		.maxlen = TCP_CA_NAME_MAX,
227 	};
228 	int ret;
229 
230 	tcp_get_default_congestion_control(net, val);
231 
232 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
233 	if (write && ret == 0)
234 		ret = tcp_set_default_congestion_control(net, val);
235 	return ret;
236 }
237 
238 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
239 						 int write, void *buffer,
240 						 size_t *lenp, loff_t *ppos)
241 {
242 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
243 	int ret;
244 
245 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
246 	if (!tbl.data)
247 		return -ENOMEM;
248 	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
249 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
250 	kfree(tbl.data);
251 	return ret;
252 }
253 
254 static int proc_allowed_congestion_control(struct ctl_table *ctl,
255 					   int write, void *buffer,
256 					   size_t *lenp, loff_t *ppos)
257 {
258 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
259 	int ret;
260 
261 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
262 	if (!tbl.data)
263 		return -ENOMEM;
264 
265 	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
266 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
267 	if (write && ret == 0)
268 		ret = tcp_set_allowed_congestion_control(tbl.data);
269 	kfree(tbl.data);
270 	return ret;
271 }
272 
273 static int sscanf_key(char *buf, __le32 *key)
274 {
275 	u32 user_key[4];
276 	int i, ret = 0;
277 
278 	if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
279 		   user_key + 2, user_key + 3) != 4) {
280 		ret = -EINVAL;
281 	} else {
282 		for (i = 0; i < ARRAY_SIZE(user_key); i++)
283 			key[i] = cpu_to_le32(user_key[i]);
284 	}
285 	pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
286 		 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
287 
288 	return ret;
289 }
290 
291 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
292 				 void *buffer, size_t *lenp, loff_t *ppos)
293 {
294 	struct net *net = container_of(table->data, struct net,
295 	    ipv4.sysctl_tcp_fastopen);
296 	/* maxlen to print the list of keys in hex (*2), with dashes
297 	 * separating doublewords and a comma in between keys.
298 	 */
299 	struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
300 					    2 * TCP_FASTOPEN_KEY_MAX) +
301 					    (TCP_FASTOPEN_KEY_MAX * 5)) };
302 	u32 user_key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(u32)];
303 	__le32 key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(__le32)];
304 	char *backup_data;
305 	int ret, i = 0, off = 0, n_keys;
306 
307 	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
308 	if (!tbl.data)
309 		return -ENOMEM;
310 
311 	n_keys = tcp_fastopen_get_cipher(net, NULL, (u64 *)key);
312 	if (!n_keys) {
313 		memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
314 		n_keys = 1;
315 	}
316 
317 	for (i = 0; i < n_keys * 4; i++)
318 		user_key[i] = le32_to_cpu(key[i]);
319 
320 	for (i = 0; i < n_keys; i++) {
321 		off += snprintf(tbl.data + off, tbl.maxlen - off,
322 				"%08x-%08x-%08x-%08x",
323 				user_key[i * 4],
324 				user_key[i * 4 + 1],
325 				user_key[i * 4 + 2],
326 				user_key[i * 4 + 3]);
327 
328 		if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
329 			break;
330 
331 		if (i + 1 < n_keys)
332 			off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
333 	}
334 
335 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
336 
337 	if (write && ret == 0) {
338 		backup_data = strchr(tbl.data, ',');
339 		if (backup_data) {
340 			*backup_data = '\0';
341 			backup_data++;
342 		}
343 		if (sscanf_key(tbl.data, key)) {
344 			ret = -EINVAL;
345 			goto bad_key;
346 		}
347 		if (backup_data) {
348 			if (sscanf_key(backup_data, key + 4)) {
349 				ret = -EINVAL;
350 				goto bad_key;
351 			}
352 		}
353 		tcp_fastopen_reset_cipher(net, NULL, key,
354 					  backup_data ? key + 4 : NULL);
355 	}
356 
357 bad_key:
358 	kfree(tbl.data);
359 	return ret;
360 }
361 
362 static void proc_configure_early_demux(int enabled, int protocol)
363 {
364 	struct net_protocol *ipprot;
365 #if IS_ENABLED(CONFIG_IPV6)
366 	struct inet6_protocol *ip6prot;
367 #endif
368 
369 	rcu_read_lock();
370 
371 	ipprot = rcu_dereference(inet_protos[protocol]);
372 	if (ipprot)
373 		ipprot->early_demux = enabled ? ipprot->early_demux_handler :
374 						NULL;
375 
376 #if IS_ENABLED(CONFIG_IPV6)
377 	ip6prot = rcu_dereference(inet6_protos[protocol]);
378 	if (ip6prot)
379 		ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
380 						 NULL;
381 #endif
382 	rcu_read_unlock();
383 }
384 
385 static int proc_tcp_early_demux(struct ctl_table *table, int write,
386 				void *buffer, size_t *lenp, loff_t *ppos)
387 {
388 	int ret = 0;
389 
390 	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
391 
392 	if (write && !ret) {
393 		int enabled = init_net.ipv4.sysctl_tcp_early_demux;
394 
395 		proc_configure_early_demux(enabled, IPPROTO_TCP);
396 	}
397 
398 	return ret;
399 }
400 
401 static int proc_udp_early_demux(struct ctl_table *table, int write,
402 				void *buffer, size_t *lenp, loff_t *ppos)
403 {
404 	int ret = 0;
405 
406 	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
407 
408 	if (write && !ret) {
409 		int enabled = init_net.ipv4.sysctl_udp_early_demux;
410 
411 		proc_configure_early_demux(enabled, IPPROTO_UDP);
412 	}
413 
414 	return ret;
415 }
416 
417 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
418 					     int write, void *buffer,
419 					     size_t *lenp, loff_t *ppos)
420 {
421 	struct net *net = container_of(table->data, struct net,
422 	    ipv4.sysctl_tcp_fastopen_blackhole_timeout);
423 	int ret;
424 
425 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
426 	if (write && ret == 0)
427 		atomic_set(&net->ipv4.tfo_active_disable_times, 0);
428 
429 	return ret;
430 }
431 
432 static int proc_tcp_available_ulp(struct ctl_table *ctl,
433 				  int write, void *buffer, size_t *lenp,
434 				  loff_t *ppos)
435 {
436 	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
437 	int ret;
438 
439 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
440 	if (!tbl.data)
441 		return -ENOMEM;
442 	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
443 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
444 	kfree(tbl.data);
445 
446 	return ret;
447 }
448 
449 #ifdef CONFIG_IP_ROUTE_MULTIPATH
450 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
451 					  void *buffer, size_t *lenp,
452 					  loff_t *ppos)
453 {
454 	struct net *net = container_of(table->data, struct net,
455 	    ipv4.sysctl_fib_multipath_hash_policy);
456 	int ret;
457 
458 	ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
459 	if (write && ret == 0)
460 		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
461 
462 	return ret;
463 }
464 #endif
465 
466 static struct ctl_table ipv4_table[] = {
467 	{
468 		.procname	= "tcp_max_orphans",
469 		.data		= &sysctl_tcp_max_orphans,
470 		.maxlen		= sizeof(int),
471 		.mode		= 0644,
472 		.proc_handler	= proc_dointvec
473 	},
474 	{
475 		.procname	= "inet_peer_threshold",
476 		.data		= &inet_peer_threshold,
477 		.maxlen		= sizeof(int),
478 		.mode		= 0644,
479 		.proc_handler	= proc_dointvec
480 	},
481 	{
482 		.procname	= "inet_peer_minttl",
483 		.data		= &inet_peer_minttl,
484 		.maxlen		= sizeof(int),
485 		.mode		= 0644,
486 		.proc_handler	= proc_dointvec_jiffies,
487 	},
488 	{
489 		.procname	= "inet_peer_maxttl",
490 		.data		= &inet_peer_maxttl,
491 		.maxlen		= sizeof(int),
492 		.mode		= 0644,
493 		.proc_handler	= proc_dointvec_jiffies,
494 	},
495 	{
496 		.procname	= "tcp_mem",
497 		.maxlen		= sizeof(sysctl_tcp_mem),
498 		.data		= &sysctl_tcp_mem,
499 		.mode		= 0644,
500 		.proc_handler	= proc_doulongvec_minmax,
501 	},
502 	{
503 		.procname	= "tcp_low_latency",
504 		.data		= &sysctl_tcp_low_latency,
505 		.maxlen		= sizeof(int),
506 		.mode		= 0644,
507 		.proc_handler	= proc_dointvec
508 	},
509 #ifdef CONFIG_NETLABEL
510 	{
511 		.procname	= "cipso_cache_enable",
512 		.data		= &cipso_v4_cache_enabled,
513 		.maxlen		= sizeof(int),
514 		.mode		= 0644,
515 		.proc_handler	= proc_dointvec,
516 	},
517 	{
518 		.procname	= "cipso_cache_bucket_size",
519 		.data		= &cipso_v4_cache_bucketsize,
520 		.maxlen		= sizeof(int),
521 		.mode		= 0644,
522 		.proc_handler	= proc_dointvec,
523 	},
524 	{
525 		.procname	= "cipso_rbm_optfmt",
526 		.data		= &cipso_v4_rbm_optfmt,
527 		.maxlen		= sizeof(int),
528 		.mode		= 0644,
529 		.proc_handler	= proc_dointvec,
530 	},
531 	{
532 		.procname	= "cipso_rbm_strictvalid",
533 		.data		= &cipso_v4_rbm_strictvalid,
534 		.maxlen		= sizeof(int),
535 		.mode		= 0644,
536 		.proc_handler	= proc_dointvec,
537 	},
538 #endif /* CONFIG_NETLABEL */
539 	{
540 		.procname	= "tcp_available_ulp",
541 		.maxlen		= TCP_ULP_BUF_MAX,
542 		.mode		= 0444,
543 		.proc_handler   = proc_tcp_available_ulp,
544 	},
545 	{
546 		.procname	= "icmp_msgs_per_sec",
547 		.data		= &sysctl_icmp_msgs_per_sec,
548 		.maxlen		= sizeof(int),
549 		.mode		= 0644,
550 		.proc_handler	= proc_dointvec_minmax,
551 		.extra1		= SYSCTL_ZERO,
552 	},
553 	{
554 		.procname	= "icmp_msgs_burst",
555 		.data		= &sysctl_icmp_msgs_burst,
556 		.maxlen		= sizeof(int),
557 		.mode		= 0644,
558 		.proc_handler	= proc_dointvec_minmax,
559 		.extra1		= SYSCTL_ZERO,
560 	},
561 	{
562 		.procname	= "udp_mem",
563 		.data		= &sysctl_udp_mem,
564 		.maxlen		= sizeof(sysctl_udp_mem),
565 		.mode		= 0644,
566 		.proc_handler	= proc_doulongvec_minmax,
567 	},
568 	{
569 		.procname	= "fib_sync_mem",
570 		.data		= &sysctl_fib_sync_mem,
571 		.maxlen		= sizeof(sysctl_fib_sync_mem),
572 		.mode		= 0644,
573 		.proc_handler	= proc_douintvec_minmax,
574 		.extra1		= &sysctl_fib_sync_mem_min,
575 		.extra2		= &sysctl_fib_sync_mem_max,
576 	},
577 	{
578 		.procname	= "tcp_rx_skb_cache",
579 		.data		= &tcp_rx_skb_cache_key.key,
580 		.mode		= 0644,
581 		.proc_handler	= proc_do_static_key,
582 	},
583 	{
584 		.procname	= "tcp_tx_skb_cache",
585 		.data		= &tcp_tx_skb_cache_key.key,
586 		.mode		= 0644,
587 		.proc_handler	= proc_do_static_key,
588 	},
589 	{ }
590 };
591 
592 static struct ctl_table ipv4_net_table[] = {
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 	},
600 	{
601 		.procname	= "icmp_echo_enable_probe",
602 		.data		= &init_net.ipv4.sysctl_icmp_echo_enable_probe,
603 		.maxlen		= sizeof(u8),
604 		.mode		= 0644,
605 		.proc_handler	= proc_dou8vec_minmax,
606 		.extra1		= SYSCTL_ZERO,
607 		.extra2		= SYSCTL_ONE
608 	},
609 	{
610 		.procname	= "icmp_echo_ignore_broadcasts",
611 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
612 		.maxlen		= sizeof(u8),
613 		.mode		= 0644,
614 		.proc_handler	= proc_dou8vec_minmax,
615 	},
616 	{
617 		.procname	= "icmp_ignore_bogus_error_responses",
618 		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
619 		.maxlen		= sizeof(u8),
620 		.mode		= 0644,
621 		.proc_handler	= proc_dou8vec_minmax,
622 	},
623 	{
624 		.procname	= "icmp_errors_use_inbound_ifaddr",
625 		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
626 		.maxlen		= sizeof(u8),
627 		.mode		= 0644,
628 		.proc_handler	= proc_dou8vec_minmax,
629 	},
630 	{
631 		.procname	= "icmp_ratelimit",
632 		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
633 		.maxlen		= sizeof(int),
634 		.mode		= 0644,
635 		.proc_handler	= proc_dointvec_ms_jiffies,
636 	},
637 	{
638 		.procname	= "icmp_ratemask",
639 		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
640 		.maxlen		= sizeof(int),
641 		.mode		= 0644,
642 		.proc_handler	= proc_dointvec
643 	},
644 	{
645 		.procname	= "ping_group_range",
646 		.data		= &init_net.ipv4.ping_group_range.range,
647 		.maxlen		= sizeof(gid_t)*2,
648 		.mode		= 0644,
649 		.proc_handler	= ipv4_ping_group_range,
650 	},
651 #ifdef CONFIG_NET_L3_MASTER_DEV
652 	{
653 		.procname	= "raw_l3mdev_accept",
654 		.data		= &init_net.ipv4.sysctl_raw_l3mdev_accept,
655 		.maxlen		= sizeof(u8),
656 		.mode		= 0644,
657 		.proc_handler	= proc_dou8vec_minmax,
658 		.extra1		= SYSCTL_ZERO,
659 		.extra2		= SYSCTL_ONE,
660 	},
661 #endif
662 	{
663 		.procname	= "tcp_ecn",
664 		.data		= &init_net.ipv4.sysctl_tcp_ecn,
665 		.maxlen		= sizeof(u8),
666 		.mode		= 0644,
667 		.proc_handler	= proc_dou8vec_minmax,
668 	},
669 	{
670 		.procname	= "tcp_ecn_fallback",
671 		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
672 		.maxlen		= sizeof(u8),
673 		.mode		= 0644,
674 		.proc_handler	= proc_dou8vec_minmax,
675 	},
676 	{
677 		.procname	= "ip_dynaddr",
678 		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
679 		.maxlen		= sizeof(u8),
680 		.mode		= 0644,
681 		.proc_handler	= proc_dou8vec_minmax,
682 	},
683 	{
684 		.procname	= "ip_early_demux",
685 		.data		= &init_net.ipv4.sysctl_ip_early_demux,
686 		.maxlen		= sizeof(u8),
687 		.mode		= 0644,
688 		.proc_handler	= proc_dou8vec_minmax,
689 	},
690 	{
691 		.procname       = "udp_early_demux",
692 		.data           = &init_net.ipv4.sysctl_udp_early_demux,
693 		.maxlen         = sizeof(u8),
694 		.mode           = 0644,
695 		.proc_handler   = proc_udp_early_demux
696 	},
697 	{
698 		.procname       = "tcp_early_demux",
699 		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
700 		.maxlen         = sizeof(u8),
701 		.mode           = 0644,
702 		.proc_handler   = proc_tcp_early_demux
703 	},
704 	{
705 		.procname       = "nexthop_compat_mode",
706 		.data           = &init_net.ipv4.sysctl_nexthop_compat_mode,
707 		.maxlen         = sizeof(u8),
708 		.mode           = 0644,
709 		.proc_handler   = proc_dou8vec_minmax,
710 		.extra1		= SYSCTL_ZERO,
711 		.extra2		= SYSCTL_ONE,
712 	},
713 	{
714 		.procname	= "ip_default_ttl",
715 		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
716 		.maxlen		= sizeof(u8),
717 		.mode		= 0644,
718 		.proc_handler	= proc_dou8vec_minmax,
719 		.extra1		= &ip_ttl_min,
720 		.extra2		= &ip_ttl_max,
721 	},
722 	{
723 		.procname	= "ip_local_port_range",
724 		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
725 		.data		= &init_net.ipv4.ip_local_ports.range,
726 		.mode		= 0644,
727 		.proc_handler	= ipv4_local_port_range,
728 	},
729 	{
730 		.procname	= "ip_local_reserved_ports",
731 		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
732 		.maxlen		= 65536,
733 		.mode		= 0644,
734 		.proc_handler	= proc_do_large_bitmap,
735 	},
736 	{
737 		.procname	= "ip_no_pmtu_disc",
738 		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
739 		.maxlen		= sizeof(u8),
740 		.mode		= 0644,
741 		.proc_handler	= proc_dou8vec_minmax,
742 	},
743 	{
744 		.procname	= "ip_forward_use_pmtu",
745 		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
746 		.maxlen		= sizeof(u8),
747 		.mode		= 0644,
748 		.proc_handler	= proc_dou8vec_minmax,
749 	},
750 	{
751 		.procname	= "ip_forward_update_priority",
752 		.data		= &init_net.ipv4.sysctl_ip_fwd_update_priority,
753 		.maxlen		= sizeof(u8),
754 		.mode		= 0644,
755 		.proc_handler   = ipv4_fwd_update_priority,
756 		.extra1		= SYSCTL_ZERO,
757 		.extra2		= SYSCTL_ONE,
758 	},
759 	{
760 		.procname	= "ip_nonlocal_bind",
761 		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
762 		.maxlen		= sizeof(u8),
763 		.mode		= 0644,
764 		.proc_handler	= proc_dou8vec_minmax,
765 	},
766 	{
767 		.procname	= "ip_autobind_reuse",
768 		.data		= &init_net.ipv4.sysctl_ip_autobind_reuse,
769 		.maxlen		= sizeof(u8),
770 		.mode		= 0644,
771 		.proc_handler	= proc_dou8vec_minmax,
772 		.extra1         = SYSCTL_ZERO,
773 		.extra2         = SYSCTL_ONE,
774 	},
775 	{
776 		.procname	= "fwmark_reflect",
777 		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
778 		.maxlen		= sizeof(u8),
779 		.mode		= 0644,
780 		.proc_handler	= proc_dou8vec_minmax,
781 	},
782 	{
783 		.procname	= "tcp_fwmark_accept",
784 		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
785 		.maxlen		= sizeof(u8),
786 		.mode		= 0644,
787 		.proc_handler	= proc_dou8vec_minmax,
788 	},
789 #ifdef CONFIG_NET_L3_MASTER_DEV
790 	{
791 		.procname	= "tcp_l3mdev_accept",
792 		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
793 		.maxlen		= sizeof(u8),
794 		.mode		= 0644,
795 		.proc_handler	= proc_dou8vec_minmax,
796 		.extra1		= SYSCTL_ZERO,
797 		.extra2		= SYSCTL_ONE,
798 	},
799 #endif
800 	{
801 		.procname	= "tcp_mtu_probing",
802 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
803 		.maxlen		= sizeof(u8),
804 		.mode		= 0644,
805 		.proc_handler	= proc_dou8vec_minmax,
806 	},
807 	{
808 		.procname	= "tcp_base_mss",
809 		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
810 		.maxlen		= sizeof(int),
811 		.mode		= 0644,
812 		.proc_handler	= proc_dointvec,
813 	},
814 	{
815 		.procname	= "tcp_min_snd_mss",
816 		.data		= &init_net.ipv4.sysctl_tcp_min_snd_mss,
817 		.maxlen		= sizeof(int),
818 		.mode		= 0644,
819 		.proc_handler	= proc_dointvec_minmax,
820 		.extra1		= &tcp_min_snd_mss_min,
821 		.extra2		= &tcp_min_snd_mss_max,
822 	},
823 	{
824 		.procname	= "tcp_mtu_probe_floor",
825 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
826 		.maxlen		= sizeof(int),
827 		.mode		= 0644,
828 		.proc_handler	= proc_dointvec_minmax,
829 		.extra1		= &tcp_min_snd_mss_min,
830 		.extra2		= &tcp_min_snd_mss_max,
831 	},
832 	{
833 		.procname	= "tcp_probe_threshold",
834 		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
835 		.maxlen		= sizeof(int),
836 		.mode		= 0644,
837 		.proc_handler	= proc_dointvec,
838 	},
839 	{
840 		.procname	= "tcp_probe_interval",
841 		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
842 		.maxlen		= sizeof(u32),
843 		.mode		= 0644,
844 		.proc_handler	= proc_douintvec_minmax,
845 		.extra2		= &u32_max_div_HZ,
846 	},
847 	{
848 		.procname	= "igmp_link_local_mcast_reports",
849 		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
850 		.maxlen		= sizeof(u8),
851 		.mode		= 0644,
852 		.proc_handler	= proc_dou8vec_minmax,
853 	},
854 	{
855 		.procname	= "igmp_max_memberships",
856 		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
857 		.maxlen		= sizeof(int),
858 		.mode		= 0644,
859 		.proc_handler	= proc_dointvec
860 	},
861 	{
862 		.procname	= "igmp_max_msf",
863 		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
864 		.maxlen		= sizeof(int),
865 		.mode		= 0644,
866 		.proc_handler	= proc_dointvec
867 	},
868 #ifdef CONFIG_IP_MULTICAST
869 	{
870 		.procname	= "igmp_qrv",
871 		.data		= &init_net.ipv4.sysctl_igmp_qrv,
872 		.maxlen		= sizeof(int),
873 		.mode		= 0644,
874 		.proc_handler	= proc_dointvec_minmax,
875 		.extra1		= SYSCTL_ONE
876 	},
877 #endif
878 	{
879 		.procname	= "tcp_congestion_control",
880 		.data		= &init_net.ipv4.tcp_congestion_control,
881 		.mode		= 0644,
882 		.maxlen		= TCP_CA_NAME_MAX,
883 		.proc_handler	= proc_tcp_congestion_control,
884 	},
885 	{
886 		.procname	= "tcp_available_congestion_control",
887 		.maxlen		= TCP_CA_BUF_MAX,
888 		.mode		= 0444,
889 		.proc_handler   = proc_tcp_available_congestion_control,
890 	},
891 	{
892 		.procname	= "tcp_allowed_congestion_control",
893 		.maxlen		= TCP_CA_BUF_MAX,
894 		.mode		= 0644,
895 		.proc_handler   = proc_allowed_congestion_control,
896 	},
897 	{
898 		.procname	= "tcp_keepalive_time",
899 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
900 		.maxlen		= sizeof(int),
901 		.mode		= 0644,
902 		.proc_handler	= proc_dointvec_jiffies,
903 	},
904 	{
905 		.procname	= "tcp_keepalive_probes",
906 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
907 		.maxlen		= sizeof(u8),
908 		.mode		= 0644,
909 		.proc_handler	= proc_dou8vec_minmax,
910 	},
911 	{
912 		.procname	= "tcp_keepalive_intvl",
913 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
914 		.maxlen		= sizeof(int),
915 		.mode		= 0644,
916 		.proc_handler	= proc_dointvec_jiffies,
917 	},
918 	{
919 		.procname	= "tcp_syn_retries",
920 		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
921 		.maxlen		= sizeof(u8),
922 		.mode		= 0644,
923 		.proc_handler	= proc_dou8vec_minmax,
924 		.extra1		= &tcp_syn_retries_min,
925 		.extra2		= &tcp_syn_retries_max
926 	},
927 	{
928 		.procname	= "tcp_synack_retries",
929 		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
930 		.maxlen		= sizeof(u8),
931 		.mode		= 0644,
932 		.proc_handler	= proc_dou8vec_minmax,
933 	},
934 #ifdef CONFIG_SYN_COOKIES
935 	{
936 		.procname	= "tcp_syncookies",
937 		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
938 		.maxlen		= sizeof(u8),
939 		.mode		= 0644,
940 		.proc_handler	= proc_dou8vec_minmax,
941 	},
942 #endif
943 	{
944 		.procname	= "tcp_reordering",
945 		.data		= &init_net.ipv4.sysctl_tcp_reordering,
946 		.maxlen		= sizeof(int),
947 		.mode		= 0644,
948 		.proc_handler	= proc_dointvec
949 	},
950 	{
951 		.procname	= "tcp_retries1",
952 		.data		= &init_net.ipv4.sysctl_tcp_retries1,
953 		.maxlen		= sizeof(u8),
954 		.mode		= 0644,
955 		.proc_handler	= proc_dou8vec_minmax,
956 		.extra2		= &tcp_retr1_max
957 	},
958 	{
959 		.procname	= "tcp_retries2",
960 		.data		= &init_net.ipv4.sysctl_tcp_retries2,
961 		.maxlen		= sizeof(u8),
962 		.mode		= 0644,
963 		.proc_handler	= proc_dou8vec_minmax,
964 	},
965 	{
966 		.procname	= "tcp_orphan_retries",
967 		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
968 		.maxlen		= sizeof(u8),
969 		.mode		= 0644,
970 		.proc_handler	= proc_dou8vec_minmax,
971 	},
972 	{
973 		.procname	= "tcp_fin_timeout",
974 		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
975 		.maxlen		= sizeof(int),
976 		.mode		= 0644,
977 		.proc_handler	= proc_dointvec_jiffies,
978 	},
979 	{
980 		.procname	= "tcp_notsent_lowat",
981 		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
982 		.maxlen		= sizeof(unsigned int),
983 		.mode		= 0644,
984 		.proc_handler	= proc_douintvec,
985 	},
986 	{
987 		.procname	= "tcp_tw_reuse",
988 		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
989 		.maxlen		= sizeof(u8),
990 		.mode		= 0644,
991 		.proc_handler	= proc_dou8vec_minmax,
992 		.extra1		= SYSCTL_ZERO,
993 		.extra2		= &two,
994 	},
995 	{
996 		.procname	= "tcp_max_tw_buckets",
997 		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
998 		.maxlen		= sizeof(int),
999 		.mode		= 0644,
1000 		.proc_handler	= proc_dointvec
1001 	},
1002 	{
1003 		.procname	= "tcp_max_syn_backlog",
1004 		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
1005 		.maxlen		= sizeof(int),
1006 		.mode		= 0644,
1007 		.proc_handler	= proc_dointvec
1008 	},
1009 	{
1010 		.procname	= "tcp_fastopen",
1011 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1012 		.maxlen		= sizeof(int),
1013 		.mode		= 0644,
1014 		.proc_handler	= proc_dointvec,
1015 	},
1016 	{
1017 		.procname	= "tcp_fastopen_key",
1018 		.mode		= 0600,
1019 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1020 		/* maxlen to print the list of keys in hex (*2), with dashes
1021 		 * separating doublewords and a comma in between keys.
1022 		 */
1023 		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH *
1024 				   2 * TCP_FASTOPEN_KEY_MAX) +
1025 				   (TCP_FASTOPEN_KEY_MAX * 5)),
1026 		.proc_handler	= proc_tcp_fastopen_key,
1027 	},
1028 	{
1029 		.procname	= "tcp_fastopen_blackhole_timeout_sec",
1030 		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1031 		.maxlen		= sizeof(int),
1032 		.mode		= 0644,
1033 		.proc_handler	= proc_tfo_blackhole_detect_timeout,
1034 		.extra1		= SYSCTL_ZERO,
1035 	},
1036 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1037 	{
1038 		.procname	= "fib_multipath_use_neigh",
1039 		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1040 		.maxlen		= sizeof(u8),
1041 		.mode		= 0644,
1042 		.proc_handler	= proc_dou8vec_minmax,
1043 		.extra1		= SYSCTL_ZERO,
1044 		.extra2		= SYSCTL_ONE,
1045 	},
1046 	{
1047 		.procname	= "fib_multipath_hash_policy",
1048 		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1049 		.maxlen		= sizeof(u8),
1050 		.mode		= 0644,
1051 		.proc_handler	= proc_fib_multipath_hash_policy,
1052 		.extra1		= SYSCTL_ZERO,
1053 		.extra2		= &two,
1054 	},
1055 #endif
1056 	{
1057 		.procname	= "ip_unprivileged_port_start",
1058 		.maxlen		= sizeof(int),
1059 		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
1060 		.mode		= 0644,
1061 		.proc_handler	= ipv4_privileged_ports,
1062 	},
1063 #ifdef CONFIG_NET_L3_MASTER_DEV
1064 	{
1065 		.procname	= "udp_l3mdev_accept",
1066 		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
1067 		.maxlen		= sizeof(u8),
1068 		.mode		= 0644,
1069 		.proc_handler	= proc_dou8vec_minmax,
1070 		.extra1		= SYSCTL_ZERO,
1071 		.extra2		= SYSCTL_ONE,
1072 	},
1073 #endif
1074 	{
1075 		.procname	= "tcp_sack",
1076 		.data		= &init_net.ipv4.sysctl_tcp_sack,
1077 		.maxlen		= sizeof(u8),
1078 		.mode		= 0644,
1079 		.proc_handler	= proc_dou8vec_minmax,
1080 	},
1081 	{
1082 		.procname	= "tcp_window_scaling",
1083 		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
1084 		.maxlen		= sizeof(u8),
1085 		.mode		= 0644,
1086 		.proc_handler	= proc_dou8vec_minmax,
1087 	},
1088 	{
1089 		.procname	= "tcp_timestamps",
1090 		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
1091 		.maxlen		= sizeof(u8),
1092 		.mode		= 0644,
1093 		.proc_handler	= proc_dou8vec_minmax,
1094 	},
1095 	{
1096 		.procname	= "tcp_early_retrans",
1097 		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
1098 		.maxlen		= sizeof(u8),
1099 		.mode		= 0644,
1100 		.proc_handler	= proc_dou8vec_minmax,
1101 		.extra1		= SYSCTL_ZERO,
1102 		.extra2		= &four,
1103 	},
1104 	{
1105 		.procname	= "tcp_recovery",
1106 		.data		= &init_net.ipv4.sysctl_tcp_recovery,
1107 		.maxlen		= sizeof(u8),
1108 		.mode		= 0644,
1109 		.proc_handler	= proc_dou8vec_minmax,
1110 	},
1111 	{
1112 		.procname       = "tcp_thin_linear_timeouts",
1113 		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1114 		.maxlen         = sizeof(u8),
1115 		.mode           = 0644,
1116 		.proc_handler   = proc_dou8vec_minmax,
1117 	},
1118 	{
1119 		.procname	= "tcp_slow_start_after_idle",
1120 		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1121 		.maxlen		= sizeof(u8),
1122 		.mode		= 0644,
1123 		.proc_handler	= proc_dou8vec_minmax,
1124 	},
1125 	{
1126 		.procname	= "tcp_retrans_collapse",
1127 		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
1128 		.maxlen		= sizeof(u8),
1129 		.mode		= 0644,
1130 		.proc_handler	= proc_dou8vec_minmax,
1131 	},
1132 	{
1133 		.procname	= "tcp_stdurg",
1134 		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
1135 		.maxlen		= sizeof(u8),
1136 		.mode		= 0644,
1137 		.proc_handler	= proc_dou8vec_minmax,
1138 	},
1139 	{
1140 		.procname	= "tcp_rfc1337",
1141 		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
1142 		.maxlen		= sizeof(u8),
1143 		.mode		= 0644,
1144 		.proc_handler	= proc_dou8vec_minmax,
1145 	},
1146 	{
1147 		.procname	= "tcp_abort_on_overflow",
1148 		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1149 		.maxlen		= sizeof(u8),
1150 		.mode		= 0644,
1151 		.proc_handler	= proc_dou8vec_minmax,
1152 	},
1153 	{
1154 		.procname	= "tcp_fack",
1155 		.data		= &init_net.ipv4.sysctl_tcp_fack,
1156 		.maxlen		= sizeof(u8),
1157 		.mode		= 0644,
1158 		.proc_handler	= proc_dou8vec_minmax,
1159 	},
1160 	{
1161 		.procname	= "tcp_max_reordering",
1162 		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1163 		.maxlen		= sizeof(int),
1164 		.mode		= 0644,
1165 		.proc_handler	= proc_dointvec
1166 	},
1167 	{
1168 		.procname	= "tcp_dsack",
1169 		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1170 		.maxlen		= sizeof(u8),
1171 		.mode		= 0644,
1172 		.proc_handler	= proc_dou8vec_minmax,
1173 	},
1174 	{
1175 		.procname	= "tcp_app_win",
1176 		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1177 		.maxlen		= sizeof(u8),
1178 		.mode		= 0644,
1179 		.proc_handler	= proc_dou8vec_minmax,
1180 	},
1181 	{
1182 		.procname	= "tcp_adv_win_scale",
1183 		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1184 		.maxlen		= sizeof(int),
1185 		.mode		= 0644,
1186 		.proc_handler	= proc_dointvec_minmax,
1187 		.extra1		= &tcp_adv_win_scale_min,
1188 		.extra2		= &tcp_adv_win_scale_max,
1189 	},
1190 	{
1191 		.procname	= "tcp_frto",
1192 		.data		= &init_net.ipv4.sysctl_tcp_frto,
1193 		.maxlen		= sizeof(u8),
1194 		.mode		= 0644,
1195 		.proc_handler	= proc_dou8vec_minmax,
1196 	},
1197 	{
1198 		.procname	= "tcp_no_metrics_save",
1199 		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1200 		.maxlen		= sizeof(u8),
1201 		.mode		= 0644,
1202 		.proc_handler	= proc_dou8vec_minmax,
1203 	},
1204 	{
1205 		.procname	= "tcp_no_ssthresh_metrics_save",
1206 		.data		= &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1207 		.maxlen		= sizeof(u8),
1208 		.mode		= 0644,
1209 		.proc_handler	= proc_dou8vec_minmax,
1210 		.extra1		= SYSCTL_ZERO,
1211 		.extra2		= SYSCTL_ONE,
1212 	},
1213 	{
1214 		.procname	= "tcp_moderate_rcvbuf",
1215 		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1216 		.maxlen		= sizeof(u8),
1217 		.mode		= 0644,
1218 		.proc_handler	= proc_dou8vec_minmax,
1219 	},
1220 	{
1221 		.procname	= "tcp_tso_win_divisor",
1222 		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1223 		.maxlen		= sizeof(u8),
1224 		.mode		= 0644,
1225 		.proc_handler	= proc_dou8vec_minmax,
1226 	},
1227 	{
1228 		.procname	= "tcp_workaround_signed_windows",
1229 		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1230 		.maxlen		= sizeof(u8),
1231 		.mode		= 0644,
1232 		.proc_handler	= proc_dou8vec_minmax,
1233 	},
1234 	{
1235 		.procname	= "tcp_limit_output_bytes",
1236 		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1237 		.maxlen		= sizeof(int),
1238 		.mode		= 0644,
1239 		.proc_handler	= proc_dointvec
1240 	},
1241 	{
1242 		.procname	= "tcp_challenge_ack_limit",
1243 		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1244 		.maxlen		= sizeof(int),
1245 		.mode		= 0644,
1246 		.proc_handler	= proc_dointvec
1247 	},
1248 	{
1249 		.procname	= "tcp_min_tso_segs",
1250 		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1251 		.maxlen		= sizeof(u8),
1252 		.mode		= 0644,
1253 		.proc_handler	= proc_dou8vec_minmax,
1254 		.extra1		= SYSCTL_ONE,
1255 	},
1256 	{
1257 		.procname	= "tcp_min_rtt_wlen",
1258 		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1259 		.maxlen		= sizeof(int),
1260 		.mode		= 0644,
1261 		.proc_handler	= proc_dointvec_minmax,
1262 		.extra1		= SYSCTL_ZERO,
1263 		.extra2		= &one_day_secs
1264 	},
1265 	{
1266 		.procname	= "tcp_autocorking",
1267 		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1268 		.maxlen		= sizeof(u8),
1269 		.mode		= 0644,
1270 		.proc_handler	= proc_dou8vec_minmax,
1271 		.extra1		= SYSCTL_ZERO,
1272 		.extra2		= SYSCTL_ONE,
1273 	},
1274 	{
1275 		.procname	= "tcp_invalid_ratelimit",
1276 		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1277 		.maxlen		= sizeof(int),
1278 		.mode		= 0644,
1279 		.proc_handler	= proc_dointvec_ms_jiffies,
1280 	},
1281 	{
1282 		.procname	= "tcp_pacing_ss_ratio",
1283 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1284 		.maxlen		= sizeof(int),
1285 		.mode		= 0644,
1286 		.proc_handler	= proc_dointvec_minmax,
1287 		.extra1		= SYSCTL_ZERO,
1288 		.extra2		= &thousand,
1289 	},
1290 	{
1291 		.procname	= "tcp_pacing_ca_ratio",
1292 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1293 		.maxlen		= sizeof(int),
1294 		.mode		= 0644,
1295 		.proc_handler	= proc_dointvec_minmax,
1296 		.extra1		= SYSCTL_ZERO,
1297 		.extra2		= &thousand,
1298 	},
1299 	{
1300 		.procname	= "tcp_wmem",
1301 		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1302 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1303 		.mode		= 0644,
1304 		.proc_handler	= proc_dointvec_minmax,
1305 		.extra1		= SYSCTL_ONE,
1306 	},
1307 	{
1308 		.procname	= "tcp_rmem",
1309 		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1310 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1311 		.mode		= 0644,
1312 		.proc_handler	= proc_dointvec_minmax,
1313 		.extra1		= SYSCTL_ONE,
1314 	},
1315 	{
1316 		.procname	= "tcp_comp_sack_delay_ns",
1317 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1318 		.maxlen		= sizeof(unsigned long),
1319 		.mode		= 0644,
1320 		.proc_handler	= proc_doulongvec_minmax,
1321 	},
1322 	{
1323 		.procname	= "tcp_comp_sack_slack_ns",
1324 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1325 		.maxlen		= sizeof(unsigned long),
1326 		.mode		= 0644,
1327 		.proc_handler	= proc_doulongvec_minmax,
1328 	},
1329 	{
1330 		.procname	= "tcp_comp_sack_nr",
1331 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1332 		.maxlen		= sizeof(u8),
1333 		.mode		= 0644,
1334 		.proc_handler	= proc_dou8vec_minmax,
1335 		.extra1		= SYSCTL_ZERO,
1336 	},
1337 	{
1338 		.procname       = "tcp_reflect_tos",
1339 		.data           = &init_net.ipv4.sysctl_tcp_reflect_tos,
1340 		.maxlen         = sizeof(u8),
1341 		.mode           = 0644,
1342 		.proc_handler   = proc_dou8vec_minmax,
1343 		.extra1         = SYSCTL_ZERO,
1344 		.extra2         = SYSCTL_ONE,
1345 	},
1346 	{
1347 		.procname	= "udp_rmem_min",
1348 		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1349 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1350 		.mode		= 0644,
1351 		.proc_handler	= proc_dointvec_minmax,
1352 		.extra1		= SYSCTL_ONE
1353 	},
1354 	{
1355 		.procname	= "udp_wmem_min",
1356 		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1357 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1358 		.mode		= 0644,
1359 		.proc_handler	= proc_dointvec_minmax,
1360 		.extra1		= SYSCTL_ONE
1361 	},
1362 	{
1363 		.procname	= "fib_notify_on_flag_change",
1364 		.data		= &init_net.ipv4.sysctl_fib_notify_on_flag_change,
1365 		.maxlen		= sizeof(u8),
1366 		.mode		= 0644,
1367 		.proc_handler	= proc_dou8vec_minmax,
1368 		.extra1		= SYSCTL_ZERO,
1369 		.extra2		= &two,
1370 	},
1371 	{ }
1372 };
1373 
1374 static __net_init int ipv4_sysctl_init_net(struct net *net)
1375 {
1376 	struct ctl_table *table;
1377 
1378 	table = ipv4_net_table;
1379 	if (!net_eq(net, &init_net)) {
1380 		int i;
1381 
1382 		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1383 		if (!table)
1384 			goto err_alloc;
1385 
1386 		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) {
1387 			if (table[i].data) {
1388 				/* Update the variables to point into
1389 				 * the current struct net
1390 				 */
1391 				table[i].data += (void *)net - (void *)&init_net;
1392 			} else {
1393 				/* Entries without data pointer are global;
1394 				 * Make them read-only in non-init_net ns
1395 				 */
1396 				table[i].mode &= ~0222;
1397 			}
1398 		}
1399 	}
1400 
1401 	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1402 	if (!net->ipv4.ipv4_hdr)
1403 		goto err_reg;
1404 
1405 	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1406 	if (!net->ipv4.sysctl_local_reserved_ports)
1407 		goto err_ports;
1408 
1409 	return 0;
1410 
1411 err_ports:
1412 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1413 err_reg:
1414 	if (!net_eq(net, &init_net))
1415 		kfree(table);
1416 err_alloc:
1417 	return -ENOMEM;
1418 }
1419 
1420 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1421 {
1422 	struct ctl_table *table;
1423 
1424 	kfree(net->ipv4.sysctl_local_reserved_ports);
1425 	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1426 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1427 	kfree(table);
1428 }
1429 
1430 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1431 	.init = ipv4_sysctl_init_net,
1432 	.exit = ipv4_sysctl_exit_net,
1433 };
1434 
1435 static __init int sysctl_ipv4_init(void)
1436 {
1437 	struct ctl_table_header *hdr;
1438 
1439 	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1440 	if (!hdr)
1441 		return -ENOMEM;
1442 
1443 	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1444 		unregister_net_sysctl_table(hdr);
1445 		return -ENOMEM;
1446 	}
1447 
1448 	return 0;
1449 }
1450 
1451 __initcall(sysctl_ipv4_init);
1452