xref: /linux/net/ipv4/sysctl_net_ipv4.c (revision a6cdeeb16bff89c8486324f53577db058cbe81ba)
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 zero;
32 static int one = 1;
33 static int two = 2;
34 static int four = 4;
35 static int thousand = 1000;
36 static int gso_max_segs = GSO_MAX_SEGS;
37 static int tcp_retr1_max = 255;
38 static int ip_local_port_range_min[] = { 1, 1 };
39 static int ip_local_port_range_max[] = { 65535, 65535 };
40 static int tcp_adv_win_scale_min = -31;
41 static int tcp_adv_win_scale_max = 31;
42 static int ip_privileged_port_min;
43 static int ip_privileged_port_max = 65535;
44 static int ip_ttl_min = 1;
45 static int ip_ttl_max = 255;
46 static int tcp_syn_retries_min = 1;
47 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
48 static int ip_ping_group_range_min[] = { 0, 0 };
49 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
50 static int comp_sack_nr_max = 255;
51 static u32 u32_max_div_HZ = UINT_MAX / HZ;
52 static int one_day_secs = 24 * 3600;
53 
54 /* obsolete */
55 static int sysctl_tcp_low_latency __read_mostly;
56 
57 /* Update system visible IP port range */
58 static void set_local_port_range(struct net *net, int range[2])
59 {
60 	bool same_parity = !((range[0] ^ range[1]) & 1);
61 
62 	write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
63 	if (same_parity && !net->ipv4.ip_local_ports.warned) {
64 		net->ipv4.ip_local_ports.warned = true;
65 		pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
66 	}
67 	net->ipv4.ip_local_ports.range[0] = range[0];
68 	net->ipv4.ip_local_ports.range[1] = range[1];
69 	write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
70 }
71 
72 /* Validate changes from /proc interface. */
73 static int ipv4_local_port_range(struct ctl_table *table, int write,
74 				 void __user *buffer,
75 				 size_t *lenp, loff_t *ppos)
76 {
77 	struct net *net =
78 		container_of(table->data, struct net, ipv4.ip_local_ports.range);
79 	int ret;
80 	int range[2];
81 	struct ctl_table tmp = {
82 		.data = &range,
83 		.maxlen = sizeof(range),
84 		.mode = table->mode,
85 		.extra1 = &ip_local_port_range_min,
86 		.extra2 = &ip_local_port_range_max,
87 	};
88 
89 	inet_get_local_port_range(net, &range[0], &range[1]);
90 
91 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
92 
93 	if (write && ret == 0) {
94 		/* Ensure that the upper limit is not smaller than the lower,
95 		 * and that the lower does not encroach upon the privileged
96 		 * port limit.
97 		 */
98 		if ((range[1] < range[0]) ||
99 		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
100 			ret = -EINVAL;
101 		else
102 			set_local_port_range(net, range);
103 	}
104 
105 	return ret;
106 }
107 
108 /* Validate changes from /proc interface. */
109 static int ipv4_privileged_ports(struct ctl_table *table, int write,
110 				void __user *buffer, size_t *lenp, loff_t *ppos)
111 {
112 	struct net *net = container_of(table->data, struct net,
113 	    ipv4.sysctl_ip_prot_sock);
114 	int ret;
115 	int pports;
116 	int range[2];
117 	struct ctl_table tmp = {
118 		.data = &pports,
119 		.maxlen = sizeof(pports),
120 		.mode = table->mode,
121 		.extra1 = &ip_privileged_port_min,
122 		.extra2 = &ip_privileged_port_max,
123 	};
124 
125 	pports = net->ipv4.sysctl_ip_prot_sock;
126 
127 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
128 
129 	if (write && ret == 0) {
130 		inet_get_local_port_range(net, &range[0], &range[1]);
131 		/* Ensure that the local port range doesn't overlap with the
132 		 * privileged port range.
133 		 */
134 		if (range[0] < pports)
135 			ret = -EINVAL;
136 		else
137 			net->ipv4.sysctl_ip_prot_sock = pports;
138 	}
139 
140 	return ret;
141 }
142 
143 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
144 {
145 	kgid_t *data = table->data;
146 	struct net *net =
147 		container_of(table->data, struct net, ipv4.ping_group_range.range);
148 	unsigned int seq;
149 	do {
150 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
151 
152 		*low = data[0];
153 		*high = data[1];
154 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
155 }
156 
157 /* Update system visible IP port range */
158 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
159 {
160 	kgid_t *data = table->data;
161 	struct net *net =
162 		container_of(table->data, struct net, ipv4.ping_group_range.range);
163 	write_seqlock(&net->ipv4.ping_group_range.lock);
164 	data[0] = low;
165 	data[1] = high;
166 	write_sequnlock(&net->ipv4.ping_group_range.lock);
167 }
168 
169 /* Validate changes from /proc interface. */
170 static int ipv4_ping_group_range(struct ctl_table *table, int write,
171 				 void __user *buffer,
172 				 size_t *lenp, loff_t *ppos)
173 {
174 	struct user_namespace *user_ns = current_user_ns();
175 	int ret;
176 	gid_t urange[2];
177 	kgid_t low, high;
178 	struct ctl_table tmp = {
179 		.data = &urange,
180 		.maxlen = sizeof(urange),
181 		.mode = table->mode,
182 		.extra1 = &ip_ping_group_range_min,
183 		.extra2 = &ip_ping_group_range_max,
184 	};
185 
186 	inet_get_ping_group_range_table(table, &low, &high);
187 	urange[0] = from_kgid_munged(user_ns, low);
188 	urange[1] = from_kgid_munged(user_ns, high);
189 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
190 
191 	if (write && ret == 0) {
192 		low = make_kgid(user_ns, urange[0]);
193 		high = make_kgid(user_ns, urange[1]);
194 		if (!gid_valid(low) || !gid_valid(high))
195 			return -EINVAL;
196 		if (urange[1] < urange[0] || gid_lt(high, low)) {
197 			low = make_kgid(&init_user_ns, 1);
198 			high = make_kgid(&init_user_ns, 0);
199 		}
200 		set_ping_group_range(table, low, high);
201 	}
202 
203 	return ret;
204 }
205 
206 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
207 				    void __user *buffer,
208 				    size_t *lenp, loff_t *ppos)
209 {
210 	struct net *net;
211 	int ret;
212 
213 	net = container_of(table->data, struct net,
214 			   ipv4.sysctl_ip_fwd_update_priority);
215 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
216 	if (write && ret == 0)
217 		call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
218 					net);
219 
220 	return ret;
221 }
222 
223 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
224 				       void __user *buffer, size_t *lenp, loff_t *ppos)
225 {
226 	struct net *net = container_of(ctl->data, struct net,
227 				       ipv4.tcp_congestion_control);
228 	char val[TCP_CA_NAME_MAX];
229 	struct ctl_table tbl = {
230 		.data = val,
231 		.maxlen = TCP_CA_NAME_MAX,
232 	};
233 	int ret;
234 
235 	tcp_get_default_congestion_control(net, val);
236 
237 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
238 	if (write && ret == 0)
239 		ret = tcp_set_default_congestion_control(net, val);
240 	return ret;
241 }
242 
243 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
244 						 int write,
245 						 void __user *buffer, size_t *lenp,
246 						 loff_t *ppos)
247 {
248 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
249 	int ret;
250 
251 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
252 	if (!tbl.data)
253 		return -ENOMEM;
254 	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
255 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
256 	kfree(tbl.data);
257 	return ret;
258 }
259 
260 static int proc_allowed_congestion_control(struct ctl_table *ctl,
261 					   int write,
262 					   void __user *buffer, size_t *lenp,
263 					   loff_t *ppos)
264 {
265 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
266 	int ret;
267 
268 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
269 	if (!tbl.data)
270 		return -ENOMEM;
271 
272 	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
273 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
274 	if (write && ret == 0)
275 		ret = tcp_set_allowed_congestion_control(tbl.data);
276 	kfree(tbl.data);
277 	return ret;
278 }
279 
280 static int sscanf_key(char *buf, __le32 *key)
281 {
282 	u32 user_key[4];
283 	int i, ret = 0;
284 
285 	if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
286 		   user_key + 2, user_key + 3) != 4) {
287 		ret = -EINVAL;
288 	} else {
289 		for (i = 0; i < ARRAY_SIZE(user_key); i++)
290 			key[i] = cpu_to_le32(user_key[i]);
291 	}
292 	pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
293 		 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
294 
295 	return ret;
296 }
297 
298 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
299 				 void __user *buffer, size_t *lenp,
300 				 loff_t *ppos)
301 {
302 	struct net *net = container_of(table->data, struct net,
303 	    ipv4.sysctl_tcp_fastopen);
304 	/* maxlen to print the list of keys in hex (*2), with dashes
305 	 * separating doublewords and a comma in between keys.
306 	 */
307 	struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
308 					    2 * TCP_FASTOPEN_KEY_MAX) +
309 					    (TCP_FASTOPEN_KEY_MAX * 5)) };
310 	struct tcp_fastopen_context *ctx;
311 	u32 user_key[TCP_FASTOPEN_KEY_MAX * 4];
312 	__le32 key[TCP_FASTOPEN_KEY_MAX * 4];
313 	char *backup_data;
314 	int ret, i = 0, off = 0, n_keys = 0;
315 
316 	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
317 	if (!tbl.data)
318 		return -ENOMEM;
319 
320 	rcu_read_lock();
321 	ctx = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
322 	if (ctx) {
323 		n_keys = tcp_fastopen_context_len(ctx);
324 		memcpy(&key[0], &ctx->key[0], TCP_FASTOPEN_KEY_LENGTH * n_keys);
325 	}
326 	rcu_read_unlock();
327 
328 	if (!n_keys) {
329 		memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
330 		n_keys = 1;
331 	}
332 
333 	for (i = 0; i < n_keys * 4; i++)
334 		user_key[i] = le32_to_cpu(key[i]);
335 
336 	for (i = 0; i < n_keys; i++) {
337 		off += snprintf(tbl.data + off, tbl.maxlen - off,
338 				"%08x-%08x-%08x-%08x",
339 				user_key[i * 4],
340 				user_key[i * 4 + 1],
341 				user_key[i * 4 + 2],
342 				user_key[i * 4 + 3]);
343 		if (i + 1 < n_keys)
344 			off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
345 	}
346 
347 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
348 
349 	if (write && ret == 0) {
350 		backup_data = strchr(tbl.data, ',');
351 		if (backup_data) {
352 			*backup_data = '\0';
353 			backup_data++;
354 		}
355 		if (sscanf_key(tbl.data, key)) {
356 			ret = -EINVAL;
357 			goto bad_key;
358 		}
359 		if (backup_data) {
360 			if (sscanf_key(backup_data, key + 4)) {
361 				ret = -EINVAL;
362 				goto bad_key;
363 			}
364 		}
365 		tcp_fastopen_reset_cipher(net, NULL, key,
366 					  backup_data ? key + 4 : NULL,
367 					  TCP_FASTOPEN_KEY_LENGTH);
368 	}
369 
370 bad_key:
371 	kfree(tbl.data);
372 	return ret;
373 }
374 
375 static void proc_configure_early_demux(int enabled, int protocol)
376 {
377 	struct net_protocol *ipprot;
378 #if IS_ENABLED(CONFIG_IPV6)
379 	struct inet6_protocol *ip6prot;
380 #endif
381 
382 	rcu_read_lock();
383 
384 	ipprot = rcu_dereference(inet_protos[protocol]);
385 	if (ipprot)
386 		ipprot->early_demux = enabled ? ipprot->early_demux_handler :
387 						NULL;
388 
389 #if IS_ENABLED(CONFIG_IPV6)
390 	ip6prot = rcu_dereference(inet6_protos[protocol]);
391 	if (ip6prot)
392 		ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
393 						 NULL;
394 #endif
395 	rcu_read_unlock();
396 }
397 
398 static int proc_tcp_early_demux(struct ctl_table *table, int write,
399 				void __user *buffer, size_t *lenp, loff_t *ppos)
400 {
401 	int ret = 0;
402 
403 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
404 
405 	if (write && !ret) {
406 		int enabled = init_net.ipv4.sysctl_tcp_early_demux;
407 
408 		proc_configure_early_demux(enabled, IPPROTO_TCP);
409 	}
410 
411 	return ret;
412 }
413 
414 static int proc_udp_early_demux(struct ctl_table *table, int write,
415 				void __user *buffer, size_t *lenp, loff_t *ppos)
416 {
417 	int ret = 0;
418 
419 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
420 
421 	if (write && !ret) {
422 		int enabled = init_net.ipv4.sysctl_udp_early_demux;
423 
424 		proc_configure_early_demux(enabled, IPPROTO_UDP);
425 	}
426 
427 	return ret;
428 }
429 
430 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
431 					     int write,
432 					     void __user *buffer,
433 					     size_t *lenp, loff_t *ppos)
434 {
435 	struct net *net = container_of(table->data, struct net,
436 	    ipv4.sysctl_tcp_fastopen_blackhole_timeout);
437 	int ret;
438 
439 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
440 	if (write && ret == 0)
441 		atomic_set(&net->ipv4.tfo_active_disable_times, 0);
442 
443 	return ret;
444 }
445 
446 static int proc_tcp_available_ulp(struct ctl_table *ctl,
447 				  int write,
448 				  void __user *buffer, size_t *lenp,
449 				  loff_t *ppos)
450 {
451 	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
452 	int ret;
453 
454 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
455 	if (!tbl.data)
456 		return -ENOMEM;
457 	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
458 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
459 	kfree(tbl.data);
460 
461 	return ret;
462 }
463 
464 #ifdef CONFIG_IP_ROUTE_MULTIPATH
465 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
466 					  void __user *buffer, size_t *lenp,
467 					  loff_t *ppos)
468 {
469 	struct net *net = container_of(table->data, struct net,
470 	    ipv4.sysctl_fib_multipath_hash_policy);
471 	int ret;
472 
473 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
474 	if (write && ret == 0)
475 		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
476 
477 	return ret;
478 }
479 #endif
480 
481 static struct ctl_table ipv4_table[] = {
482 	{
483 		.procname	= "tcp_max_orphans",
484 		.data		= &sysctl_tcp_max_orphans,
485 		.maxlen		= sizeof(int),
486 		.mode		= 0644,
487 		.proc_handler	= proc_dointvec
488 	},
489 	{
490 		.procname	= "inet_peer_threshold",
491 		.data		= &inet_peer_threshold,
492 		.maxlen		= sizeof(int),
493 		.mode		= 0644,
494 		.proc_handler	= proc_dointvec
495 	},
496 	{
497 		.procname	= "inet_peer_minttl",
498 		.data		= &inet_peer_minttl,
499 		.maxlen		= sizeof(int),
500 		.mode		= 0644,
501 		.proc_handler	= proc_dointvec_jiffies,
502 	},
503 	{
504 		.procname	= "inet_peer_maxttl",
505 		.data		= &inet_peer_maxttl,
506 		.maxlen		= sizeof(int),
507 		.mode		= 0644,
508 		.proc_handler	= proc_dointvec_jiffies,
509 	},
510 	{
511 		.procname	= "tcp_mem",
512 		.maxlen		= sizeof(sysctl_tcp_mem),
513 		.data		= &sysctl_tcp_mem,
514 		.mode		= 0644,
515 		.proc_handler	= proc_doulongvec_minmax,
516 	},
517 	{
518 		.procname	= "tcp_low_latency",
519 		.data		= &sysctl_tcp_low_latency,
520 		.maxlen		= sizeof(int),
521 		.mode		= 0644,
522 		.proc_handler	= proc_dointvec
523 	},
524 #ifdef CONFIG_NETLABEL
525 	{
526 		.procname	= "cipso_cache_enable",
527 		.data		= &cipso_v4_cache_enabled,
528 		.maxlen		= sizeof(int),
529 		.mode		= 0644,
530 		.proc_handler	= proc_dointvec,
531 	},
532 	{
533 		.procname	= "cipso_cache_bucket_size",
534 		.data		= &cipso_v4_cache_bucketsize,
535 		.maxlen		= sizeof(int),
536 		.mode		= 0644,
537 		.proc_handler	= proc_dointvec,
538 	},
539 	{
540 		.procname	= "cipso_rbm_optfmt",
541 		.data		= &cipso_v4_rbm_optfmt,
542 		.maxlen		= sizeof(int),
543 		.mode		= 0644,
544 		.proc_handler	= proc_dointvec,
545 	},
546 	{
547 		.procname	= "cipso_rbm_strictvalid",
548 		.data		= &cipso_v4_rbm_strictvalid,
549 		.maxlen		= sizeof(int),
550 		.mode		= 0644,
551 		.proc_handler	= proc_dointvec,
552 	},
553 #endif /* CONFIG_NETLABEL */
554 	{
555 		.procname	= "tcp_available_congestion_control",
556 		.maxlen		= TCP_CA_BUF_MAX,
557 		.mode		= 0444,
558 		.proc_handler   = proc_tcp_available_congestion_control,
559 	},
560 	{
561 		.procname	= "tcp_allowed_congestion_control",
562 		.maxlen		= TCP_CA_BUF_MAX,
563 		.mode		= 0644,
564 		.proc_handler   = proc_allowed_congestion_control,
565 	},
566 	{
567 		.procname	= "tcp_available_ulp",
568 		.maxlen		= TCP_ULP_BUF_MAX,
569 		.mode		= 0444,
570 		.proc_handler   = proc_tcp_available_ulp,
571 	},
572 	{
573 		.procname	= "icmp_msgs_per_sec",
574 		.data		= &sysctl_icmp_msgs_per_sec,
575 		.maxlen		= sizeof(int),
576 		.mode		= 0644,
577 		.proc_handler	= proc_dointvec_minmax,
578 		.extra1		= &zero,
579 	},
580 	{
581 		.procname	= "icmp_msgs_burst",
582 		.data		= &sysctl_icmp_msgs_burst,
583 		.maxlen		= sizeof(int),
584 		.mode		= 0644,
585 		.proc_handler	= proc_dointvec_minmax,
586 		.extra1		= &zero,
587 	},
588 	{
589 		.procname	= "udp_mem",
590 		.data		= &sysctl_udp_mem,
591 		.maxlen		= sizeof(sysctl_udp_mem),
592 		.mode		= 0644,
593 		.proc_handler	= proc_doulongvec_minmax,
594 	},
595 	{
596 		.procname	= "fib_sync_mem",
597 		.data		= &sysctl_fib_sync_mem,
598 		.maxlen		= sizeof(sysctl_fib_sync_mem),
599 		.mode		= 0644,
600 		.proc_handler	= proc_douintvec_minmax,
601 		.extra1		= &sysctl_fib_sync_mem_min,
602 		.extra2		= &sysctl_fib_sync_mem_max,
603 	},
604 	{ }
605 };
606 
607 static struct ctl_table ipv4_net_table[] = {
608 	{
609 		.procname	= "icmp_echo_ignore_all",
610 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
611 		.maxlen		= sizeof(int),
612 		.mode		= 0644,
613 		.proc_handler	= proc_dointvec
614 	},
615 	{
616 		.procname	= "icmp_echo_ignore_broadcasts",
617 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
618 		.maxlen		= sizeof(int),
619 		.mode		= 0644,
620 		.proc_handler	= proc_dointvec
621 	},
622 	{
623 		.procname	= "icmp_ignore_bogus_error_responses",
624 		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
625 		.maxlen		= sizeof(int),
626 		.mode		= 0644,
627 		.proc_handler	= proc_dointvec
628 	},
629 	{
630 		.procname	= "icmp_errors_use_inbound_ifaddr",
631 		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
632 		.maxlen		= sizeof(int),
633 		.mode		= 0644,
634 		.proc_handler	= proc_dointvec
635 	},
636 	{
637 		.procname	= "icmp_ratelimit",
638 		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
639 		.maxlen		= sizeof(int),
640 		.mode		= 0644,
641 		.proc_handler	= proc_dointvec_ms_jiffies,
642 	},
643 	{
644 		.procname	= "icmp_ratemask",
645 		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
646 		.maxlen		= sizeof(int),
647 		.mode		= 0644,
648 		.proc_handler	= proc_dointvec
649 	},
650 	{
651 		.procname	= "ping_group_range",
652 		.data		= &init_net.ipv4.ping_group_range.range,
653 		.maxlen		= sizeof(gid_t)*2,
654 		.mode		= 0644,
655 		.proc_handler	= ipv4_ping_group_range,
656 	},
657 #ifdef CONFIG_NET_L3_MASTER_DEV
658 	{
659 		.procname	= "raw_l3mdev_accept",
660 		.data		= &init_net.ipv4.sysctl_raw_l3mdev_accept,
661 		.maxlen		= sizeof(int),
662 		.mode		= 0644,
663 		.proc_handler	= proc_dointvec_minmax,
664 		.extra1		= &zero,
665 		.extra2		= &one,
666 	},
667 #endif
668 	{
669 		.procname	= "tcp_ecn",
670 		.data		= &init_net.ipv4.sysctl_tcp_ecn,
671 		.maxlen		= sizeof(int),
672 		.mode		= 0644,
673 		.proc_handler	= proc_dointvec
674 	},
675 	{
676 		.procname	= "tcp_ecn_fallback",
677 		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
678 		.maxlen		= sizeof(int),
679 		.mode		= 0644,
680 		.proc_handler	= proc_dointvec
681 	},
682 	{
683 		.procname	= "ip_dynaddr",
684 		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
685 		.maxlen		= sizeof(int),
686 		.mode		= 0644,
687 		.proc_handler	= proc_dointvec
688 	},
689 	{
690 		.procname	= "ip_early_demux",
691 		.data		= &init_net.ipv4.sysctl_ip_early_demux,
692 		.maxlen		= sizeof(int),
693 		.mode		= 0644,
694 		.proc_handler	= proc_dointvec
695 	},
696 	{
697 		.procname       = "udp_early_demux",
698 		.data           = &init_net.ipv4.sysctl_udp_early_demux,
699 		.maxlen         = sizeof(int),
700 		.mode           = 0644,
701 		.proc_handler   = proc_udp_early_demux
702 	},
703 	{
704 		.procname       = "tcp_early_demux",
705 		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
706 		.maxlen         = sizeof(int),
707 		.mode           = 0644,
708 		.proc_handler   = proc_tcp_early_demux
709 	},
710 	{
711 		.procname	= "ip_default_ttl",
712 		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
713 		.maxlen		= sizeof(int),
714 		.mode		= 0644,
715 		.proc_handler	= proc_dointvec_minmax,
716 		.extra1		= &ip_ttl_min,
717 		.extra2		= &ip_ttl_max,
718 	},
719 	{
720 		.procname	= "ip_local_port_range",
721 		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
722 		.data		= &init_net.ipv4.ip_local_ports.range,
723 		.mode		= 0644,
724 		.proc_handler	= ipv4_local_port_range,
725 	},
726 	{
727 		.procname	= "ip_local_reserved_ports",
728 		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
729 		.maxlen		= 65536,
730 		.mode		= 0644,
731 		.proc_handler	= proc_do_large_bitmap,
732 	},
733 	{
734 		.procname	= "ip_no_pmtu_disc",
735 		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
736 		.maxlen		= sizeof(int),
737 		.mode		= 0644,
738 		.proc_handler	= proc_dointvec
739 	},
740 	{
741 		.procname	= "ip_forward_use_pmtu",
742 		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
743 		.maxlen		= sizeof(int),
744 		.mode		= 0644,
745 		.proc_handler	= proc_dointvec,
746 	},
747 	{
748 		.procname	= "ip_forward_update_priority",
749 		.data		= &init_net.ipv4.sysctl_ip_fwd_update_priority,
750 		.maxlen		= sizeof(int),
751 		.mode		= 0644,
752 		.proc_handler   = ipv4_fwd_update_priority,
753 		.extra1		= &zero,
754 		.extra2		= &one,
755 	},
756 	{
757 		.procname	= "ip_nonlocal_bind",
758 		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
759 		.maxlen		= sizeof(int),
760 		.mode		= 0644,
761 		.proc_handler	= proc_dointvec
762 	},
763 	{
764 		.procname	= "fwmark_reflect",
765 		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
766 		.maxlen		= sizeof(int),
767 		.mode		= 0644,
768 		.proc_handler	= proc_dointvec,
769 	},
770 	{
771 		.procname	= "tcp_fwmark_accept",
772 		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
773 		.maxlen		= sizeof(int),
774 		.mode		= 0644,
775 		.proc_handler	= proc_dointvec,
776 	},
777 #ifdef CONFIG_NET_L3_MASTER_DEV
778 	{
779 		.procname	= "tcp_l3mdev_accept",
780 		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
781 		.maxlen		= sizeof(int),
782 		.mode		= 0644,
783 		.proc_handler	= proc_dointvec_minmax,
784 		.extra1		= &zero,
785 		.extra2		= &one,
786 	},
787 #endif
788 	{
789 		.procname	= "tcp_mtu_probing",
790 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
791 		.maxlen		= sizeof(int),
792 		.mode		= 0644,
793 		.proc_handler	= proc_dointvec,
794 	},
795 	{
796 		.procname	= "tcp_base_mss",
797 		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
798 		.maxlen		= sizeof(int),
799 		.mode		= 0644,
800 		.proc_handler	= proc_dointvec,
801 	},
802 	{
803 		.procname	= "tcp_probe_threshold",
804 		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
805 		.maxlen		= sizeof(int),
806 		.mode		= 0644,
807 		.proc_handler	= proc_dointvec,
808 	},
809 	{
810 		.procname	= "tcp_probe_interval",
811 		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
812 		.maxlen		= sizeof(u32),
813 		.mode		= 0644,
814 		.proc_handler	= proc_douintvec_minmax,
815 		.extra2		= &u32_max_div_HZ,
816 	},
817 	{
818 		.procname	= "igmp_link_local_mcast_reports",
819 		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
820 		.maxlen		= sizeof(int),
821 		.mode		= 0644,
822 		.proc_handler	= proc_dointvec
823 	},
824 	{
825 		.procname	= "igmp_max_memberships",
826 		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
827 		.maxlen		= sizeof(int),
828 		.mode		= 0644,
829 		.proc_handler	= proc_dointvec
830 	},
831 	{
832 		.procname	= "igmp_max_msf",
833 		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
834 		.maxlen		= sizeof(int),
835 		.mode		= 0644,
836 		.proc_handler	= proc_dointvec
837 	},
838 #ifdef CONFIG_IP_MULTICAST
839 	{
840 		.procname	= "igmp_qrv",
841 		.data		= &init_net.ipv4.sysctl_igmp_qrv,
842 		.maxlen		= sizeof(int),
843 		.mode		= 0644,
844 		.proc_handler	= proc_dointvec_minmax,
845 		.extra1		= &one
846 	},
847 #endif
848 	{
849 		.procname	= "tcp_congestion_control",
850 		.data		= &init_net.ipv4.tcp_congestion_control,
851 		.mode		= 0644,
852 		.maxlen		= TCP_CA_NAME_MAX,
853 		.proc_handler	= proc_tcp_congestion_control,
854 	},
855 	{
856 		.procname	= "tcp_keepalive_time",
857 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
858 		.maxlen		= sizeof(int),
859 		.mode		= 0644,
860 		.proc_handler	= proc_dointvec_jiffies,
861 	},
862 	{
863 		.procname	= "tcp_keepalive_probes",
864 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
865 		.maxlen		= sizeof(int),
866 		.mode		= 0644,
867 		.proc_handler	= proc_dointvec
868 	},
869 	{
870 		.procname	= "tcp_keepalive_intvl",
871 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
872 		.maxlen		= sizeof(int),
873 		.mode		= 0644,
874 		.proc_handler	= proc_dointvec_jiffies,
875 	},
876 	{
877 		.procname	= "tcp_syn_retries",
878 		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
879 		.maxlen		= sizeof(int),
880 		.mode		= 0644,
881 		.proc_handler	= proc_dointvec_minmax,
882 		.extra1		= &tcp_syn_retries_min,
883 		.extra2		= &tcp_syn_retries_max
884 	},
885 	{
886 		.procname	= "tcp_synack_retries",
887 		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
888 		.maxlen		= sizeof(int),
889 		.mode		= 0644,
890 		.proc_handler	= proc_dointvec
891 	},
892 #ifdef CONFIG_SYN_COOKIES
893 	{
894 		.procname	= "tcp_syncookies",
895 		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
896 		.maxlen		= sizeof(int),
897 		.mode		= 0644,
898 		.proc_handler	= proc_dointvec
899 	},
900 #endif
901 	{
902 		.procname	= "tcp_reordering",
903 		.data		= &init_net.ipv4.sysctl_tcp_reordering,
904 		.maxlen		= sizeof(int),
905 		.mode		= 0644,
906 		.proc_handler	= proc_dointvec
907 	},
908 	{
909 		.procname	= "tcp_retries1",
910 		.data		= &init_net.ipv4.sysctl_tcp_retries1,
911 		.maxlen		= sizeof(int),
912 		.mode		= 0644,
913 		.proc_handler	= proc_dointvec_minmax,
914 		.extra2		= &tcp_retr1_max
915 	},
916 	{
917 		.procname	= "tcp_retries2",
918 		.data		= &init_net.ipv4.sysctl_tcp_retries2,
919 		.maxlen		= sizeof(int),
920 		.mode		= 0644,
921 		.proc_handler	= proc_dointvec
922 	},
923 	{
924 		.procname	= "tcp_orphan_retries",
925 		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
926 		.maxlen		= sizeof(int),
927 		.mode		= 0644,
928 		.proc_handler	= proc_dointvec
929 	},
930 	{
931 		.procname	= "tcp_fin_timeout",
932 		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
933 		.maxlen		= sizeof(int),
934 		.mode		= 0644,
935 		.proc_handler	= proc_dointvec_jiffies,
936 	},
937 	{
938 		.procname	= "tcp_notsent_lowat",
939 		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
940 		.maxlen		= sizeof(unsigned int),
941 		.mode		= 0644,
942 		.proc_handler	= proc_douintvec,
943 	},
944 	{
945 		.procname	= "tcp_tw_reuse",
946 		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
947 		.maxlen		= sizeof(int),
948 		.mode		= 0644,
949 		.proc_handler	= proc_dointvec_minmax,
950 		.extra1		= &zero,
951 		.extra2		= &two,
952 	},
953 	{
954 		.procname	= "tcp_max_tw_buckets",
955 		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
956 		.maxlen		= sizeof(int),
957 		.mode		= 0644,
958 		.proc_handler	= proc_dointvec
959 	},
960 	{
961 		.procname	= "tcp_max_syn_backlog",
962 		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
963 		.maxlen		= sizeof(int),
964 		.mode		= 0644,
965 		.proc_handler	= proc_dointvec
966 	},
967 	{
968 		.procname	= "tcp_fastopen",
969 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
970 		.maxlen		= sizeof(int),
971 		.mode		= 0644,
972 		.proc_handler	= proc_dointvec,
973 	},
974 	{
975 		.procname	= "tcp_fastopen_key",
976 		.mode		= 0600,
977 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
978 		/* maxlen to print the list of keys in hex (*2), with dashes
979 		 * separating doublewords and a comma in between keys.
980 		 */
981 		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH *
982 				   2 * TCP_FASTOPEN_KEY_MAX) +
983 				   (TCP_FASTOPEN_KEY_MAX * 5)),
984 		.proc_handler	= proc_tcp_fastopen_key,
985 	},
986 	{
987 		.procname	= "tcp_fastopen_blackhole_timeout_sec",
988 		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
989 		.maxlen		= sizeof(int),
990 		.mode		= 0644,
991 		.proc_handler	= proc_tfo_blackhole_detect_timeout,
992 		.extra1		= &zero,
993 	},
994 #ifdef CONFIG_IP_ROUTE_MULTIPATH
995 	{
996 		.procname	= "fib_multipath_use_neigh",
997 		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
998 		.maxlen		= sizeof(int),
999 		.mode		= 0644,
1000 		.proc_handler	= proc_dointvec_minmax,
1001 		.extra1		= &zero,
1002 		.extra2		= &one,
1003 	},
1004 	{
1005 		.procname	= "fib_multipath_hash_policy",
1006 		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1007 		.maxlen		= sizeof(int),
1008 		.mode		= 0644,
1009 		.proc_handler	= proc_fib_multipath_hash_policy,
1010 		.extra1		= &zero,
1011 		.extra2		= &one,
1012 	},
1013 #endif
1014 	{
1015 		.procname	= "ip_unprivileged_port_start",
1016 		.maxlen		= sizeof(int),
1017 		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
1018 		.mode		= 0644,
1019 		.proc_handler	= ipv4_privileged_ports,
1020 	},
1021 #ifdef CONFIG_NET_L3_MASTER_DEV
1022 	{
1023 		.procname	= "udp_l3mdev_accept",
1024 		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
1025 		.maxlen		= sizeof(int),
1026 		.mode		= 0644,
1027 		.proc_handler	= proc_dointvec_minmax,
1028 		.extra1		= &zero,
1029 		.extra2		= &one,
1030 	},
1031 #endif
1032 	{
1033 		.procname	= "tcp_sack",
1034 		.data		= &init_net.ipv4.sysctl_tcp_sack,
1035 		.maxlen		= sizeof(int),
1036 		.mode		= 0644,
1037 		.proc_handler	= proc_dointvec
1038 	},
1039 	{
1040 		.procname	= "tcp_window_scaling",
1041 		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
1042 		.maxlen		= sizeof(int),
1043 		.mode		= 0644,
1044 		.proc_handler	= proc_dointvec
1045 	},
1046 	{
1047 		.procname	= "tcp_timestamps",
1048 		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
1049 		.maxlen		= sizeof(int),
1050 		.mode		= 0644,
1051 		.proc_handler	= proc_dointvec
1052 	},
1053 	{
1054 		.procname	= "tcp_early_retrans",
1055 		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
1056 		.maxlen		= sizeof(int),
1057 		.mode		= 0644,
1058 		.proc_handler	= proc_dointvec_minmax,
1059 		.extra1		= &zero,
1060 		.extra2		= &four,
1061 	},
1062 	{
1063 		.procname	= "tcp_recovery",
1064 		.data		= &init_net.ipv4.sysctl_tcp_recovery,
1065 		.maxlen		= sizeof(int),
1066 		.mode		= 0644,
1067 		.proc_handler	= proc_dointvec,
1068 	},
1069 	{
1070 		.procname       = "tcp_thin_linear_timeouts",
1071 		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1072 		.maxlen         = sizeof(int),
1073 		.mode           = 0644,
1074 		.proc_handler   = proc_dointvec
1075 	},
1076 	{
1077 		.procname	= "tcp_slow_start_after_idle",
1078 		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1079 		.maxlen		= sizeof(int),
1080 		.mode		= 0644,
1081 		.proc_handler	= proc_dointvec
1082 	},
1083 	{
1084 		.procname	= "tcp_retrans_collapse",
1085 		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
1086 		.maxlen		= sizeof(int),
1087 		.mode		= 0644,
1088 		.proc_handler	= proc_dointvec
1089 	},
1090 	{
1091 		.procname	= "tcp_stdurg",
1092 		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
1093 		.maxlen		= sizeof(int),
1094 		.mode		= 0644,
1095 		.proc_handler	= proc_dointvec
1096 	},
1097 	{
1098 		.procname	= "tcp_rfc1337",
1099 		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
1100 		.maxlen		= sizeof(int),
1101 		.mode		= 0644,
1102 		.proc_handler	= proc_dointvec
1103 	},
1104 	{
1105 		.procname	= "tcp_abort_on_overflow",
1106 		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1107 		.maxlen		= sizeof(int),
1108 		.mode		= 0644,
1109 		.proc_handler	= proc_dointvec
1110 	},
1111 	{
1112 		.procname	= "tcp_fack",
1113 		.data		= &init_net.ipv4.sysctl_tcp_fack,
1114 		.maxlen		= sizeof(int),
1115 		.mode		= 0644,
1116 		.proc_handler	= proc_dointvec
1117 	},
1118 	{
1119 		.procname	= "tcp_max_reordering",
1120 		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1121 		.maxlen		= sizeof(int),
1122 		.mode		= 0644,
1123 		.proc_handler	= proc_dointvec
1124 	},
1125 	{
1126 		.procname	= "tcp_dsack",
1127 		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1128 		.maxlen		= sizeof(int),
1129 		.mode		= 0644,
1130 		.proc_handler	= proc_dointvec
1131 	},
1132 	{
1133 		.procname	= "tcp_app_win",
1134 		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1135 		.maxlen		= sizeof(int),
1136 		.mode		= 0644,
1137 		.proc_handler	= proc_dointvec
1138 	},
1139 	{
1140 		.procname	= "tcp_adv_win_scale",
1141 		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1142 		.maxlen		= sizeof(int),
1143 		.mode		= 0644,
1144 		.proc_handler	= proc_dointvec_minmax,
1145 		.extra1		= &tcp_adv_win_scale_min,
1146 		.extra2		= &tcp_adv_win_scale_max,
1147 	},
1148 	{
1149 		.procname	= "tcp_frto",
1150 		.data		= &init_net.ipv4.sysctl_tcp_frto,
1151 		.maxlen		= sizeof(int),
1152 		.mode		= 0644,
1153 		.proc_handler	= proc_dointvec
1154 	},
1155 	{
1156 		.procname	= "tcp_no_metrics_save",
1157 		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1158 		.maxlen		= sizeof(int),
1159 		.mode		= 0644,
1160 		.proc_handler	= proc_dointvec,
1161 	},
1162 	{
1163 		.procname	= "tcp_moderate_rcvbuf",
1164 		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1165 		.maxlen		= sizeof(int),
1166 		.mode		= 0644,
1167 		.proc_handler	= proc_dointvec,
1168 	},
1169 	{
1170 		.procname	= "tcp_tso_win_divisor",
1171 		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1172 		.maxlen		= sizeof(int),
1173 		.mode		= 0644,
1174 		.proc_handler	= proc_dointvec,
1175 	},
1176 	{
1177 		.procname	= "tcp_workaround_signed_windows",
1178 		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1179 		.maxlen		= sizeof(int),
1180 		.mode		= 0644,
1181 		.proc_handler	= proc_dointvec
1182 	},
1183 	{
1184 		.procname	= "tcp_limit_output_bytes",
1185 		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1186 		.maxlen		= sizeof(int),
1187 		.mode		= 0644,
1188 		.proc_handler	= proc_dointvec
1189 	},
1190 	{
1191 		.procname	= "tcp_challenge_ack_limit",
1192 		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1193 		.maxlen		= sizeof(int),
1194 		.mode		= 0644,
1195 		.proc_handler	= proc_dointvec
1196 	},
1197 	{
1198 		.procname	= "tcp_min_tso_segs",
1199 		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1200 		.maxlen		= sizeof(int),
1201 		.mode		= 0644,
1202 		.proc_handler	= proc_dointvec_minmax,
1203 		.extra1		= &one,
1204 		.extra2		= &gso_max_segs,
1205 	},
1206 	{
1207 		.procname	= "tcp_min_rtt_wlen",
1208 		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1209 		.maxlen		= sizeof(int),
1210 		.mode		= 0644,
1211 		.proc_handler	= proc_dointvec_minmax,
1212 		.extra1		= &zero,
1213 		.extra2		= &one_day_secs
1214 	},
1215 	{
1216 		.procname	= "tcp_autocorking",
1217 		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1218 		.maxlen		= sizeof(int),
1219 		.mode		= 0644,
1220 		.proc_handler	= proc_dointvec_minmax,
1221 		.extra1		= &zero,
1222 		.extra2		= &one,
1223 	},
1224 	{
1225 		.procname	= "tcp_invalid_ratelimit",
1226 		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1227 		.maxlen		= sizeof(int),
1228 		.mode		= 0644,
1229 		.proc_handler	= proc_dointvec_ms_jiffies,
1230 	},
1231 	{
1232 		.procname	= "tcp_pacing_ss_ratio",
1233 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1234 		.maxlen		= sizeof(int),
1235 		.mode		= 0644,
1236 		.proc_handler	= proc_dointvec_minmax,
1237 		.extra1		= &zero,
1238 		.extra2		= &thousand,
1239 	},
1240 	{
1241 		.procname	= "tcp_pacing_ca_ratio",
1242 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1243 		.maxlen		= sizeof(int),
1244 		.mode		= 0644,
1245 		.proc_handler	= proc_dointvec_minmax,
1246 		.extra1		= &zero,
1247 		.extra2		= &thousand,
1248 	},
1249 	{
1250 		.procname	= "tcp_wmem",
1251 		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1252 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1253 		.mode		= 0644,
1254 		.proc_handler	= proc_dointvec_minmax,
1255 		.extra1		= &one,
1256 	},
1257 	{
1258 		.procname	= "tcp_rmem",
1259 		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1260 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1261 		.mode		= 0644,
1262 		.proc_handler	= proc_dointvec_minmax,
1263 		.extra1		= &one,
1264 	},
1265 	{
1266 		.procname	= "tcp_comp_sack_delay_ns",
1267 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1268 		.maxlen		= sizeof(unsigned long),
1269 		.mode		= 0644,
1270 		.proc_handler	= proc_doulongvec_minmax,
1271 	},
1272 	{
1273 		.procname	= "tcp_comp_sack_nr",
1274 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1275 		.maxlen		= sizeof(int),
1276 		.mode		= 0644,
1277 		.proc_handler	= proc_dointvec_minmax,
1278 		.extra1		= &zero,
1279 		.extra2		= &comp_sack_nr_max,
1280 	},
1281 	{
1282 		.procname	= "udp_rmem_min",
1283 		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1284 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1285 		.mode		= 0644,
1286 		.proc_handler	= proc_dointvec_minmax,
1287 		.extra1		= &one
1288 	},
1289 	{
1290 		.procname	= "udp_wmem_min",
1291 		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1292 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1293 		.mode		= 0644,
1294 		.proc_handler	= proc_dointvec_minmax,
1295 		.extra1		= &one
1296 	},
1297 	{ }
1298 };
1299 
1300 static __net_init int ipv4_sysctl_init_net(struct net *net)
1301 {
1302 	struct ctl_table *table;
1303 
1304 	table = ipv4_net_table;
1305 	if (!net_eq(net, &init_net)) {
1306 		int i;
1307 
1308 		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1309 		if (!table)
1310 			goto err_alloc;
1311 
1312 		/* Update the variables to point into the current struct net */
1313 		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1314 			table[i].data += (void *)net - (void *)&init_net;
1315 	}
1316 
1317 	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1318 	if (!net->ipv4.ipv4_hdr)
1319 		goto err_reg;
1320 
1321 	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1322 	if (!net->ipv4.sysctl_local_reserved_ports)
1323 		goto err_ports;
1324 
1325 	return 0;
1326 
1327 err_ports:
1328 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1329 err_reg:
1330 	if (!net_eq(net, &init_net))
1331 		kfree(table);
1332 err_alloc:
1333 	return -ENOMEM;
1334 }
1335 
1336 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1337 {
1338 	struct ctl_table *table;
1339 
1340 	kfree(net->ipv4.sysctl_local_reserved_ports);
1341 	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1342 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1343 	kfree(table);
1344 }
1345 
1346 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1347 	.init = ipv4_sysctl_init_net,
1348 	.exit = ipv4_sysctl_exit_net,
1349 };
1350 
1351 static __init int sysctl_ipv4_init(void)
1352 {
1353 	struct ctl_table_header *hdr;
1354 
1355 	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1356 	if (!hdr)
1357 		return -ENOMEM;
1358 
1359 	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1360 		unregister_net_sysctl_table(hdr);
1361 		return -ENOMEM;
1362 	}
1363 
1364 	return 0;
1365 }
1366 
1367 __initcall(sysctl_ipv4_init);
1368