xref: /openbmc/linux/net/ipv4/sysctl_net_ipv4.c (revision 9b5db89e)
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 tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
43 static int tcp_min_snd_mss_max = 65535;
44 static int ip_privileged_port_min;
45 static int ip_privileged_port_max = 65535;
46 static int ip_ttl_min = 1;
47 static int ip_ttl_max = 255;
48 static int tcp_syn_retries_min = 1;
49 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
50 static int ip_ping_group_range_min[] = { 0, 0 };
51 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
52 static int comp_sack_nr_max = 255;
53 static u32 u32_max_div_HZ = UINT_MAX / HZ;
54 static int one_day_secs = 24 * 3600;
55 
56 /* obsolete */
57 static int sysctl_tcp_low_latency __read_mostly;
58 
59 /* Update system visible IP port range */
60 static void set_local_port_range(struct net *net, int range[2])
61 {
62 	bool same_parity = !((range[0] ^ range[1]) & 1);
63 
64 	write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
65 	if (same_parity && !net->ipv4.ip_local_ports.warned) {
66 		net->ipv4.ip_local_ports.warned = true;
67 		pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
68 	}
69 	net->ipv4.ip_local_ports.range[0] = range[0];
70 	net->ipv4.ip_local_ports.range[1] = range[1];
71 	write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
72 }
73 
74 /* Validate changes from /proc interface. */
75 static int ipv4_local_port_range(struct ctl_table *table, int write,
76 				 void __user *buffer,
77 				 size_t *lenp, loff_t *ppos)
78 {
79 	struct net *net =
80 		container_of(table->data, struct net, ipv4.ip_local_ports.range);
81 	int ret;
82 	int range[2];
83 	struct ctl_table tmp = {
84 		.data = &range,
85 		.maxlen = sizeof(range),
86 		.mode = table->mode,
87 		.extra1 = &ip_local_port_range_min,
88 		.extra2 = &ip_local_port_range_max,
89 	};
90 
91 	inet_get_local_port_range(net, &range[0], &range[1]);
92 
93 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
94 
95 	if (write && ret == 0) {
96 		/* Ensure that the upper limit is not smaller than the lower,
97 		 * and that the lower does not encroach upon the privileged
98 		 * port limit.
99 		 */
100 		if ((range[1] < range[0]) ||
101 		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
102 			ret = -EINVAL;
103 		else
104 			set_local_port_range(net, range);
105 	}
106 
107 	return ret;
108 }
109 
110 /* Validate changes from /proc interface. */
111 static int ipv4_privileged_ports(struct ctl_table *table, int write,
112 				void __user *buffer, size_t *lenp, loff_t *ppos)
113 {
114 	struct net *net = container_of(table->data, struct net,
115 	    ipv4.sysctl_ip_prot_sock);
116 	int ret;
117 	int pports;
118 	int range[2];
119 	struct ctl_table tmp = {
120 		.data = &pports,
121 		.maxlen = sizeof(pports),
122 		.mode = table->mode,
123 		.extra1 = &ip_privileged_port_min,
124 		.extra2 = &ip_privileged_port_max,
125 	};
126 
127 	pports = net->ipv4.sysctl_ip_prot_sock;
128 
129 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
130 
131 	if (write && ret == 0) {
132 		inet_get_local_port_range(net, &range[0], &range[1]);
133 		/* Ensure that the local port range doesn't overlap with the
134 		 * privileged port range.
135 		 */
136 		if (range[0] < pports)
137 			ret = -EINVAL;
138 		else
139 			net->ipv4.sysctl_ip_prot_sock = pports;
140 	}
141 
142 	return ret;
143 }
144 
145 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
146 {
147 	kgid_t *data = table->data;
148 	struct net *net =
149 		container_of(table->data, struct net, ipv4.ping_group_range.range);
150 	unsigned int seq;
151 	do {
152 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
153 
154 		*low = data[0];
155 		*high = data[1];
156 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
157 }
158 
159 /* Update system visible IP port range */
160 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
161 {
162 	kgid_t *data = table->data;
163 	struct net *net =
164 		container_of(table->data, struct net, ipv4.ping_group_range.range);
165 	write_seqlock(&net->ipv4.ping_group_range.lock);
166 	data[0] = low;
167 	data[1] = high;
168 	write_sequnlock(&net->ipv4.ping_group_range.lock);
169 }
170 
171 /* Validate changes from /proc interface. */
172 static int ipv4_ping_group_range(struct ctl_table *table, int write,
173 				 void __user *buffer,
174 				 size_t *lenp, loff_t *ppos)
175 {
176 	struct user_namespace *user_ns = current_user_ns();
177 	int ret;
178 	gid_t urange[2];
179 	kgid_t low, high;
180 	struct ctl_table tmp = {
181 		.data = &urange,
182 		.maxlen = sizeof(urange),
183 		.mode = table->mode,
184 		.extra1 = &ip_ping_group_range_min,
185 		.extra2 = &ip_ping_group_range_max,
186 	};
187 
188 	inet_get_ping_group_range_table(table, &low, &high);
189 	urange[0] = from_kgid_munged(user_ns, low);
190 	urange[1] = from_kgid_munged(user_ns, high);
191 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
192 
193 	if (write && ret == 0) {
194 		low = make_kgid(user_ns, urange[0]);
195 		high = make_kgid(user_ns, urange[1]);
196 		if (!gid_valid(low) || !gid_valid(high))
197 			return -EINVAL;
198 		if (urange[1] < urange[0] || gid_lt(high, low)) {
199 			low = make_kgid(&init_user_ns, 1);
200 			high = make_kgid(&init_user_ns, 0);
201 		}
202 		set_ping_group_range(table, low, high);
203 	}
204 
205 	return ret;
206 }
207 
208 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
209 				    void __user *buffer,
210 				    size_t *lenp, loff_t *ppos)
211 {
212 	struct net *net;
213 	int ret;
214 
215 	net = container_of(table->data, struct net,
216 			   ipv4.sysctl_ip_fwd_update_priority);
217 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
218 	if (write && ret == 0)
219 		call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
220 					net);
221 
222 	return ret;
223 }
224 
225 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
226 				       void __user *buffer, size_t *lenp, loff_t *ppos)
227 {
228 	struct net *net = container_of(ctl->data, struct net,
229 				       ipv4.tcp_congestion_control);
230 	char val[TCP_CA_NAME_MAX];
231 	struct ctl_table tbl = {
232 		.data = val,
233 		.maxlen = TCP_CA_NAME_MAX,
234 	};
235 	int ret;
236 
237 	tcp_get_default_congestion_control(net, val);
238 
239 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
240 	if (write && ret == 0)
241 		ret = tcp_set_default_congestion_control(net, val);
242 	return ret;
243 }
244 
245 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
246 						 int write,
247 						 void __user *buffer, size_t *lenp,
248 						 loff_t *ppos)
249 {
250 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
251 	int ret;
252 
253 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
254 	if (!tbl.data)
255 		return -ENOMEM;
256 	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
257 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
258 	kfree(tbl.data);
259 	return ret;
260 }
261 
262 static int proc_allowed_congestion_control(struct ctl_table *ctl,
263 					   int write,
264 					   void __user *buffer, size_t *lenp,
265 					   loff_t *ppos)
266 {
267 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
268 	int ret;
269 
270 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
271 	if (!tbl.data)
272 		return -ENOMEM;
273 
274 	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
275 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
276 	if (write && ret == 0)
277 		ret = tcp_set_allowed_congestion_control(tbl.data);
278 	kfree(tbl.data);
279 	return ret;
280 }
281 
282 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
283 				 void __user *buffer, size_t *lenp,
284 				 loff_t *ppos)
285 {
286 	struct net *net = container_of(table->data, struct net,
287 	    ipv4.sysctl_tcp_fastopen);
288 	struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
289 	struct tcp_fastopen_context *ctxt;
290 	u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
291 	__le32 key[4];
292 	int ret, i;
293 
294 	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
295 	if (!tbl.data)
296 		return -ENOMEM;
297 
298 	rcu_read_lock();
299 	ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
300 	if (ctxt)
301 		memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
302 	else
303 		memset(key, 0, sizeof(key));
304 	rcu_read_unlock();
305 
306 	for (i = 0; i < ARRAY_SIZE(key); i++)
307 		user_key[i] = le32_to_cpu(key[i]);
308 
309 	snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
310 		user_key[0], user_key[1], user_key[2], user_key[3]);
311 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
312 
313 	if (write && ret == 0) {
314 		if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
315 			   user_key + 2, user_key + 3) != 4) {
316 			ret = -EINVAL;
317 			goto bad_key;
318 		}
319 
320 		for (i = 0; i < ARRAY_SIZE(user_key); i++)
321 			key[i] = cpu_to_le32(user_key[i]);
322 
323 		tcp_fastopen_reset_cipher(net, NULL, key,
324 					  TCP_FASTOPEN_KEY_LENGTH);
325 	}
326 
327 bad_key:
328 	pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
329 		user_key[0], user_key[1], user_key[2], user_key[3],
330 	       (char *)tbl.data, ret);
331 	kfree(tbl.data);
332 	return ret;
333 }
334 
335 static void proc_configure_early_demux(int enabled, int protocol)
336 {
337 	struct net_protocol *ipprot;
338 #if IS_ENABLED(CONFIG_IPV6)
339 	struct inet6_protocol *ip6prot;
340 #endif
341 
342 	rcu_read_lock();
343 
344 	ipprot = rcu_dereference(inet_protos[protocol]);
345 	if (ipprot)
346 		ipprot->early_demux = enabled ? ipprot->early_demux_handler :
347 						NULL;
348 
349 #if IS_ENABLED(CONFIG_IPV6)
350 	ip6prot = rcu_dereference(inet6_protos[protocol]);
351 	if (ip6prot)
352 		ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
353 						 NULL;
354 #endif
355 	rcu_read_unlock();
356 }
357 
358 static int proc_tcp_early_demux(struct ctl_table *table, int write,
359 				void __user *buffer, size_t *lenp, loff_t *ppos)
360 {
361 	int ret = 0;
362 
363 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
364 
365 	if (write && !ret) {
366 		int enabled = init_net.ipv4.sysctl_tcp_early_demux;
367 
368 		proc_configure_early_demux(enabled, IPPROTO_TCP);
369 	}
370 
371 	return ret;
372 }
373 
374 static int proc_udp_early_demux(struct ctl_table *table, int write,
375 				void __user *buffer, size_t *lenp, loff_t *ppos)
376 {
377 	int ret = 0;
378 
379 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
380 
381 	if (write && !ret) {
382 		int enabled = init_net.ipv4.sysctl_udp_early_demux;
383 
384 		proc_configure_early_demux(enabled, IPPROTO_UDP);
385 	}
386 
387 	return ret;
388 }
389 
390 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
391 					     int write,
392 					     void __user *buffer,
393 					     size_t *lenp, loff_t *ppos)
394 {
395 	struct net *net = container_of(table->data, struct net,
396 	    ipv4.sysctl_tcp_fastopen_blackhole_timeout);
397 	int ret;
398 
399 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
400 	if (write && ret == 0)
401 		atomic_set(&net->ipv4.tfo_active_disable_times, 0);
402 
403 	return ret;
404 }
405 
406 static int proc_tcp_available_ulp(struct ctl_table *ctl,
407 				  int write,
408 				  void __user *buffer, size_t *lenp,
409 				  loff_t *ppos)
410 {
411 	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
412 	int ret;
413 
414 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
415 	if (!tbl.data)
416 		return -ENOMEM;
417 	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
418 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
419 	kfree(tbl.data);
420 
421 	return ret;
422 }
423 
424 #ifdef CONFIG_IP_ROUTE_MULTIPATH
425 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
426 					  void __user *buffer, size_t *lenp,
427 					  loff_t *ppos)
428 {
429 	struct net *net = container_of(table->data, struct net,
430 	    ipv4.sysctl_fib_multipath_hash_policy);
431 	int ret;
432 
433 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
434 	if (write && ret == 0)
435 		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
436 
437 	return ret;
438 }
439 #endif
440 
441 static struct ctl_table ipv4_table[] = {
442 	{
443 		.procname	= "tcp_max_orphans",
444 		.data		= &sysctl_tcp_max_orphans,
445 		.maxlen		= sizeof(int),
446 		.mode		= 0644,
447 		.proc_handler	= proc_dointvec
448 	},
449 	{
450 		.procname	= "inet_peer_threshold",
451 		.data		= &inet_peer_threshold,
452 		.maxlen		= sizeof(int),
453 		.mode		= 0644,
454 		.proc_handler	= proc_dointvec
455 	},
456 	{
457 		.procname	= "inet_peer_minttl",
458 		.data		= &inet_peer_minttl,
459 		.maxlen		= sizeof(int),
460 		.mode		= 0644,
461 		.proc_handler	= proc_dointvec_jiffies,
462 	},
463 	{
464 		.procname	= "inet_peer_maxttl",
465 		.data		= &inet_peer_maxttl,
466 		.maxlen		= sizeof(int),
467 		.mode		= 0644,
468 		.proc_handler	= proc_dointvec_jiffies,
469 	},
470 	{
471 		.procname	= "tcp_mem",
472 		.maxlen		= sizeof(sysctl_tcp_mem),
473 		.data		= &sysctl_tcp_mem,
474 		.mode		= 0644,
475 		.proc_handler	= proc_doulongvec_minmax,
476 	},
477 	{
478 		.procname	= "tcp_low_latency",
479 		.data		= &sysctl_tcp_low_latency,
480 		.maxlen		= sizeof(int),
481 		.mode		= 0644,
482 		.proc_handler	= proc_dointvec
483 	},
484 #ifdef CONFIG_NETLABEL
485 	{
486 		.procname	= "cipso_cache_enable",
487 		.data		= &cipso_v4_cache_enabled,
488 		.maxlen		= sizeof(int),
489 		.mode		= 0644,
490 		.proc_handler	= proc_dointvec,
491 	},
492 	{
493 		.procname	= "cipso_cache_bucket_size",
494 		.data		= &cipso_v4_cache_bucketsize,
495 		.maxlen		= sizeof(int),
496 		.mode		= 0644,
497 		.proc_handler	= proc_dointvec,
498 	},
499 	{
500 		.procname	= "cipso_rbm_optfmt",
501 		.data		= &cipso_v4_rbm_optfmt,
502 		.maxlen		= sizeof(int),
503 		.mode		= 0644,
504 		.proc_handler	= proc_dointvec,
505 	},
506 	{
507 		.procname	= "cipso_rbm_strictvalid",
508 		.data		= &cipso_v4_rbm_strictvalid,
509 		.maxlen		= sizeof(int),
510 		.mode		= 0644,
511 		.proc_handler	= proc_dointvec,
512 	},
513 #endif /* CONFIG_NETLABEL */
514 	{
515 		.procname	= "tcp_available_congestion_control",
516 		.maxlen		= TCP_CA_BUF_MAX,
517 		.mode		= 0444,
518 		.proc_handler   = proc_tcp_available_congestion_control,
519 	},
520 	{
521 		.procname	= "tcp_allowed_congestion_control",
522 		.maxlen		= TCP_CA_BUF_MAX,
523 		.mode		= 0644,
524 		.proc_handler   = proc_allowed_congestion_control,
525 	},
526 	{
527 		.procname	= "tcp_available_ulp",
528 		.maxlen		= TCP_ULP_BUF_MAX,
529 		.mode		= 0444,
530 		.proc_handler   = proc_tcp_available_ulp,
531 	},
532 	{
533 		.procname	= "icmp_msgs_per_sec",
534 		.data		= &sysctl_icmp_msgs_per_sec,
535 		.maxlen		= sizeof(int),
536 		.mode		= 0644,
537 		.proc_handler	= proc_dointvec_minmax,
538 		.extra1		= &zero,
539 	},
540 	{
541 		.procname	= "icmp_msgs_burst",
542 		.data		= &sysctl_icmp_msgs_burst,
543 		.maxlen		= sizeof(int),
544 		.mode		= 0644,
545 		.proc_handler	= proc_dointvec_minmax,
546 		.extra1		= &zero,
547 	},
548 	{
549 		.procname	= "udp_mem",
550 		.data		= &sysctl_udp_mem,
551 		.maxlen		= sizeof(sysctl_udp_mem),
552 		.mode		= 0644,
553 		.proc_handler	= proc_doulongvec_minmax,
554 	},
555 	{
556 		.procname	= "fib_sync_mem",
557 		.data		= &sysctl_fib_sync_mem,
558 		.maxlen		= sizeof(sysctl_fib_sync_mem),
559 		.mode		= 0644,
560 		.proc_handler	= proc_douintvec_minmax,
561 		.extra1		= &sysctl_fib_sync_mem_min,
562 		.extra2		= &sysctl_fib_sync_mem_max,
563 	},
564 	{
565 		.procname	= "tcp_rx_skb_cache",
566 		.data		= &tcp_rx_skb_cache_key.key,
567 		.mode		= 0644,
568 		.proc_handler	= proc_do_static_key,
569 	},
570 	{
571 		.procname	= "tcp_tx_skb_cache",
572 		.data		= &tcp_tx_skb_cache_key.key,
573 		.mode		= 0644,
574 		.proc_handler	= proc_do_static_key,
575 	},
576 	{ }
577 };
578 
579 static struct ctl_table ipv4_net_table[] = {
580 	{
581 		.procname	= "icmp_echo_ignore_all",
582 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
583 		.maxlen		= sizeof(int),
584 		.mode		= 0644,
585 		.proc_handler	= proc_dointvec
586 	},
587 	{
588 		.procname	= "icmp_echo_ignore_broadcasts",
589 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
590 		.maxlen		= sizeof(int),
591 		.mode		= 0644,
592 		.proc_handler	= proc_dointvec
593 	},
594 	{
595 		.procname	= "icmp_ignore_bogus_error_responses",
596 		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
597 		.maxlen		= sizeof(int),
598 		.mode		= 0644,
599 		.proc_handler	= proc_dointvec
600 	},
601 	{
602 		.procname	= "icmp_errors_use_inbound_ifaddr",
603 		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
604 		.maxlen		= sizeof(int),
605 		.mode		= 0644,
606 		.proc_handler	= proc_dointvec
607 	},
608 	{
609 		.procname	= "icmp_ratelimit",
610 		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
611 		.maxlen		= sizeof(int),
612 		.mode		= 0644,
613 		.proc_handler	= proc_dointvec_ms_jiffies,
614 	},
615 	{
616 		.procname	= "icmp_ratemask",
617 		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
618 		.maxlen		= sizeof(int),
619 		.mode		= 0644,
620 		.proc_handler	= proc_dointvec
621 	},
622 	{
623 		.procname	= "ping_group_range",
624 		.data		= &init_net.ipv4.ping_group_range.range,
625 		.maxlen		= sizeof(gid_t)*2,
626 		.mode		= 0644,
627 		.proc_handler	= ipv4_ping_group_range,
628 	},
629 #ifdef CONFIG_NET_L3_MASTER_DEV
630 	{
631 		.procname	= "raw_l3mdev_accept",
632 		.data		= &init_net.ipv4.sysctl_raw_l3mdev_accept,
633 		.maxlen		= sizeof(int),
634 		.mode		= 0644,
635 		.proc_handler	= proc_dointvec_minmax,
636 		.extra1		= &zero,
637 		.extra2		= &one,
638 	},
639 #endif
640 	{
641 		.procname	= "tcp_ecn",
642 		.data		= &init_net.ipv4.sysctl_tcp_ecn,
643 		.maxlen		= sizeof(int),
644 		.mode		= 0644,
645 		.proc_handler	= proc_dointvec
646 	},
647 	{
648 		.procname	= "tcp_ecn_fallback",
649 		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
650 		.maxlen		= sizeof(int),
651 		.mode		= 0644,
652 		.proc_handler	= proc_dointvec
653 	},
654 	{
655 		.procname	= "ip_dynaddr",
656 		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
657 		.maxlen		= sizeof(int),
658 		.mode		= 0644,
659 		.proc_handler	= proc_dointvec
660 	},
661 	{
662 		.procname	= "ip_early_demux",
663 		.data		= &init_net.ipv4.sysctl_ip_early_demux,
664 		.maxlen		= sizeof(int),
665 		.mode		= 0644,
666 		.proc_handler	= proc_dointvec
667 	},
668 	{
669 		.procname       = "udp_early_demux",
670 		.data           = &init_net.ipv4.sysctl_udp_early_demux,
671 		.maxlen         = sizeof(int),
672 		.mode           = 0644,
673 		.proc_handler   = proc_udp_early_demux
674 	},
675 	{
676 		.procname       = "tcp_early_demux",
677 		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
678 		.maxlen         = sizeof(int),
679 		.mode           = 0644,
680 		.proc_handler   = proc_tcp_early_demux
681 	},
682 	{
683 		.procname	= "ip_default_ttl",
684 		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
685 		.maxlen		= sizeof(int),
686 		.mode		= 0644,
687 		.proc_handler	= proc_dointvec_minmax,
688 		.extra1		= &ip_ttl_min,
689 		.extra2		= &ip_ttl_max,
690 	},
691 	{
692 		.procname	= "ip_local_port_range",
693 		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
694 		.data		= &init_net.ipv4.ip_local_ports.range,
695 		.mode		= 0644,
696 		.proc_handler	= ipv4_local_port_range,
697 	},
698 	{
699 		.procname	= "ip_local_reserved_ports",
700 		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
701 		.maxlen		= 65536,
702 		.mode		= 0644,
703 		.proc_handler	= proc_do_large_bitmap,
704 	},
705 	{
706 		.procname	= "ip_no_pmtu_disc",
707 		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
708 		.maxlen		= sizeof(int),
709 		.mode		= 0644,
710 		.proc_handler	= proc_dointvec
711 	},
712 	{
713 		.procname	= "ip_forward_use_pmtu",
714 		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
715 		.maxlen		= sizeof(int),
716 		.mode		= 0644,
717 		.proc_handler	= proc_dointvec,
718 	},
719 	{
720 		.procname	= "ip_forward_update_priority",
721 		.data		= &init_net.ipv4.sysctl_ip_fwd_update_priority,
722 		.maxlen		= sizeof(int),
723 		.mode		= 0644,
724 		.proc_handler   = ipv4_fwd_update_priority,
725 		.extra1		= &zero,
726 		.extra2		= &one,
727 	},
728 	{
729 		.procname	= "ip_nonlocal_bind",
730 		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
731 		.maxlen		= sizeof(int),
732 		.mode		= 0644,
733 		.proc_handler	= proc_dointvec
734 	},
735 	{
736 		.procname	= "fwmark_reflect",
737 		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
738 		.maxlen		= sizeof(int),
739 		.mode		= 0644,
740 		.proc_handler	= proc_dointvec,
741 	},
742 	{
743 		.procname	= "tcp_fwmark_accept",
744 		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
745 		.maxlen		= sizeof(int),
746 		.mode		= 0644,
747 		.proc_handler	= proc_dointvec,
748 	},
749 #ifdef CONFIG_NET_L3_MASTER_DEV
750 	{
751 		.procname	= "tcp_l3mdev_accept",
752 		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
753 		.maxlen		= sizeof(int),
754 		.mode		= 0644,
755 		.proc_handler	= proc_dointvec_minmax,
756 		.extra1		= &zero,
757 		.extra2		= &one,
758 	},
759 #endif
760 	{
761 		.procname	= "tcp_mtu_probing",
762 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
763 		.maxlen		= sizeof(int),
764 		.mode		= 0644,
765 		.proc_handler	= proc_dointvec,
766 	},
767 	{
768 		.procname	= "tcp_base_mss",
769 		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
770 		.maxlen		= sizeof(int),
771 		.mode		= 0644,
772 		.proc_handler	= proc_dointvec,
773 	},
774 	{
775 		.procname	= "tcp_min_snd_mss",
776 		.data		= &init_net.ipv4.sysctl_tcp_min_snd_mss,
777 		.maxlen		= sizeof(int),
778 		.mode		= 0644,
779 		.proc_handler	= proc_dointvec_minmax,
780 		.extra1		= &tcp_min_snd_mss_min,
781 		.extra2		= &tcp_min_snd_mss_max,
782 	},
783 	{
784 		.procname	= "tcp_probe_threshold",
785 		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
786 		.maxlen		= sizeof(int),
787 		.mode		= 0644,
788 		.proc_handler	= proc_dointvec,
789 	},
790 	{
791 		.procname	= "tcp_probe_interval",
792 		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
793 		.maxlen		= sizeof(u32),
794 		.mode		= 0644,
795 		.proc_handler	= proc_douintvec_minmax,
796 		.extra2		= &u32_max_div_HZ,
797 	},
798 	{
799 		.procname	= "igmp_link_local_mcast_reports",
800 		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
801 		.maxlen		= sizeof(int),
802 		.mode		= 0644,
803 		.proc_handler	= proc_dointvec
804 	},
805 	{
806 		.procname	= "igmp_max_memberships",
807 		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
808 		.maxlen		= sizeof(int),
809 		.mode		= 0644,
810 		.proc_handler	= proc_dointvec
811 	},
812 	{
813 		.procname	= "igmp_max_msf",
814 		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
815 		.maxlen		= sizeof(int),
816 		.mode		= 0644,
817 		.proc_handler	= proc_dointvec
818 	},
819 #ifdef CONFIG_IP_MULTICAST
820 	{
821 		.procname	= "igmp_qrv",
822 		.data		= &init_net.ipv4.sysctl_igmp_qrv,
823 		.maxlen		= sizeof(int),
824 		.mode		= 0644,
825 		.proc_handler	= proc_dointvec_minmax,
826 		.extra1		= &one
827 	},
828 #endif
829 	{
830 		.procname	= "tcp_congestion_control",
831 		.data		= &init_net.ipv4.tcp_congestion_control,
832 		.mode		= 0644,
833 		.maxlen		= TCP_CA_NAME_MAX,
834 		.proc_handler	= proc_tcp_congestion_control,
835 	},
836 	{
837 		.procname	= "tcp_keepalive_time",
838 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
839 		.maxlen		= sizeof(int),
840 		.mode		= 0644,
841 		.proc_handler	= proc_dointvec_jiffies,
842 	},
843 	{
844 		.procname	= "tcp_keepalive_probes",
845 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
846 		.maxlen		= sizeof(int),
847 		.mode		= 0644,
848 		.proc_handler	= proc_dointvec
849 	},
850 	{
851 		.procname	= "tcp_keepalive_intvl",
852 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
853 		.maxlen		= sizeof(int),
854 		.mode		= 0644,
855 		.proc_handler	= proc_dointvec_jiffies,
856 	},
857 	{
858 		.procname	= "tcp_syn_retries",
859 		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
860 		.maxlen		= sizeof(int),
861 		.mode		= 0644,
862 		.proc_handler	= proc_dointvec_minmax,
863 		.extra1		= &tcp_syn_retries_min,
864 		.extra2		= &tcp_syn_retries_max
865 	},
866 	{
867 		.procname	= "tcp_synack_retries",
868 		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
869 		.maxlen		= sizeof(int),
870 		.mode		= 0644,
871 		.proc_handler	= proc_dointvec
872 	},
873 #ifdef CONFIG_SYN_COOKIES
874 	{
875 		.procname	= "tcp_syncookies",
876 		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
877 		.maxlen		= sizeof(int),
878 		.mode		= 0644,
879 		.proc_handler	= proc_dointvec
880 	},
881 #endif
882 	{
883 		.procname	= "tcp_reordering",
884 		.data		= &init_net.ipv4.sysctl_tcp_reordering,
885 		.maxlen		= sizeof(int),
886 		.mode		= 0644,
887 		.proc_handler	= proc_dointvec
888 	},
889 	{
890 		.procname	= "tcp_retries1",
891 		.data		= &init_net.ipv4.sysctl_tcp_retries1,
892 		.maxlen		= sizeof(int),
893 		.mode		= 0644,
894 		.proc_handler	= proc_dointvec_minmax,
895 		.extra2		= &tcp_retr1_max
896 	},
897 	{
898 		.procname	= "tcp_retries2",
899 		.data		= &init_net.ipv4.sysctl_tcp_retries2,
900 		.maxlen		= sizeof(int),
901 		.mode		= 0644,
902 		.proc_handler	= proc_dointvec
903 	},
904 	{
905 		.procname	= "tcp_orphan_retries",
906 		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
907 		.maxlen		= sizeof(int),
908 		.mode		= 0644,
909 		.proc_handler	= proc_dointvec
910 	},
911 	{
912 		.procname	= "tcp_fin_timeout",
913 		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
914 		.maxlen		= sizeof(int),
915 		.mode		= 0644,
916 		.proc_handler	= proc_dointvec_jiffies,
917 	},
918 	{
919 		.procname	= "tcp_notsent_lowat",
920 		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
921 		.maxlen		= sizeof(unsigned int),
922 		.mode		= 0644,
923 		.proc_handler	= proc_douintvec,
924 	},
925 	{
926 		.procname	= "tcp_tw_reuse",
927 		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
928 		.maxlen		= sizeof(int),
929 		.mode		= 0644,
930 		.proc_handler	= proc_dointvec_minmax,
931 		.extra1		= &zero,
932 		.extra2		= &two,
933 	},
934 	{
935 		.procname	= "tcp_max_tw_buckets",
936 		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
937 		.maxlen		= sizeof(int),
938 		.mode		= 0644,
939 		.proc_handler	= proc_dointvec
940 	},
941 	{
942 		.procname	= "tcp_max_syn_backlog",
943 		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
944 		.maxlen		= sizeof(int),
945 		.mode		= 0644,
946 		.proc_handler	= proc_dointvec
947 	},
948 	{
949 		.procname	= "tcp_fastopen",
950 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
951 		.maxlen		= sizeof(int),
952 		.mode		= 0644,
953 		.proc_handler	= proc_dointvec,
954 	},
955 	{
956 		.procname	= "tcp_fastopen_key",
957 		.mode		= 0600,
958 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
959 		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
960 		.proc_handler	= proc_tcp_fastopen_key,
961 	},
962 	{
963 		.procname	= "tcp_fastopen_blackhole_timeout_sec",
964 		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
965 		.maxlen		= sizeof(int),
966 		.mode		= 0644,
967 		.proc_handler	= proc_tfo_blackhole_detect_timeout,
968 		.extra1		= &zero,
969 	},
970 #ifdef CONFIG_IP_ROUTE_MULTIPATH
971 	{
972 		.procname	= "fib_multipath_use_neigh",
973 		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
974 		.maxlen		= sizeof(int),
975 		.mode		= 0644,
976 		.proc_handler	= proc_dointvec_minmax,
977 		.extra1		= &zero,
978 		.extra2		= &one,
979 	},
980 	{
981 		.procname	= "fib_multipath_hash_policy",
982 		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
983 		.maxlen		= sizeof(int),
984 		.mode		= 0644,
985 		.proc_handler	= proc_fib_multipath_hash_policy,
986 		.extra1		= &zero,
987 		.extra2		= &one,
988 	},
989 #endif
990 	{
991 		.procname	= "ip_unprivileged_port_start",
992 		.maxlen		= sizeof(int),
993 		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
994 		.mode		= 0644,
995 		.proc_handler	= ipv4_privileged_ports,
996 	},
997 #ifdef CONFIG_NET_L3_MASTER_DEV
998 	{
999 		.procname	= "udp_l3mdev_accept",
1000 		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
1001 		.maxlen		= sizeof(int),
1002 		.mode		= 0644,
1003 		.proc_handler	= proc_dointvec_minmax,
1004 		.extra1		= &zero,
1005 		.extra2		= &one,
1006 	},
1007 #endif
1008 	{
1009 		.procname	= "tcp_sack",
1010 		.data		= &init_net.ipv4.sysctl_tcp_sack,
1011 		.maxlen		= sizeof(int),
1012 		.mode		= 0644,
1013 		.proc_handler	= proc_dointvec
1014 	},
1015 	{
1016 		.procname	= "tcp_window_scaling",
1017 		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
1018 		.maxlen		= sizeof(int),
1019 		.mode		= 0644,
1020 		.proc_handler	= proc_dointvec
1021 	},
1022 	{
1023 		.procname	= "tcp_timestamps",
1024 		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
1025 		.maxlen		= sizeof(int),
1026 		.mode		= 0644,
1027 		.proc_handler	= proc_dointvec
1028 	},
1029 	{
1030 		.procname	= "tcp_early_retrans",
1031 		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
1032 		.maxlen		= sizeof(int),
1033 		.mode		= 0644,
1034 		.proc_handler	= proc_dointvec_minmax,
1035 		.extra1		= &zero,
1036 		.extra2		= &four,
1037 	},
1038 	{
1039 		.procname	= "tcp_recovery",
1040 		.data		= &init_net.ipv4.sysctl_tcp_recovery,
1041 		.maxlen		= sizeof(int),
1042 		.mode		= 0644,
1043 		.proc_handler	= proc_dointvec,
1044 	},
1045 	{
1046 		.procname       = "tcp_thin_linear_timeouts",
1047 		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1048 		.maxlen         = sizeof(int),
1049 		.mode           = 0644,
1050 		.proc_handler   = proc_dointvec
1051 	},
1052 	{
1053 		.procname	= "tcp_slow_start_after_idle",
1054 		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1055 		.maxlen		= sizeof(int),
1056 		.mode		= 0644,
1057 		.proc_handler	= proc_dointvec
1058 	},
1059 	{
1060 		.procname	= "tcp_retrans_collapse",
1061 		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
1062 		.maxlen		= sizeof(int),
1063 		.mode		= 0644,
1064 		.proc_handler	= proc_dointvec
1065 	},
1066 	{
1067 		.procname	= "tcp_stdurg",
1068 		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
1069 		.maxlen		= sizeof(int),
1070 		.mode		= 0644,
1071 		.proc_handler	= proc_dointvec
1072 	},
1073 	{
1074 		.procname	= "tcp_rfc1337",
1075 		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
1076 		.maxlen		= sizeof(int),
1077 		.mode		= 0644,
1078 		.proc_handler	= proc_dointvec
1079 	},
1080 	{
1081 		.procname	= "tcp_abort_on_overflow",
1082 		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1083 		.maxlen		= sizeof(int),
1084 		.mode		= 0644,
1085 		.proc_handler	= proc_dointvec
1086 	},
1087 	{
1088 		.procname	= "tcp_fack",
1089 		.data		= &init_net.ipv4.sysctl_tcp_fack,
1090 		.maxlen		= sizeof(int),
1091 		.mode		= 0644,
1092 		.proc_handler	= proc_dointvec
1093 	},
1094 	{
1095 		.procname	= "tcp_max_reordering",
1096 		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1097 		.maxlen		= sizeof(int),
1098 		.mode		= 0644,
1099 		.proc_handler	= proc_dointvec
1100 	},
1101 	{
1102 		.procname	= "tcp_dsack",
1103 		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1104 		.maxlen		= sizeof(int),
1105 		.mode		= 0644,
1106 		.proc_handler	= proc_dointvec
1107 	},
1108 	{
1109 		.procname	= "tcp_app_win",
1110 		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1111 		.maxlen		= sizeof(int),
1112 		.mode		= 0644,
1113 		.proc_handler	= proc_dointvec
1114 	},
1115 	{
1116 		.procname	= "tcp_adv_win_scale",
1117 		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1118 		.maxlen		= sizeof(int),
1119 		.mode		= 0644,
1120 		.proc_handler	= proc_dointvec_minmax,
1121 		.extra1		= &tcp_adv_win_scale_min,
1122 		.extra2		= &tcp_adv_win_scale_max,
1123 	},
1124 	{
1125 		.procname	= "tcp_frto",
1126 		.data		= &init_net.ipv4.sysctl_tcp_frto,
1127 		.maxlen		= sizeof(int),
1128 		.mode		= 0644,
1129 		.proc_handler	= proc_dointvec
1130 	},
1131 	{
1132 		.procname	= "tcp_no_metrics_save",
1133 		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1134 		.maxlen		= sizeof(int),
1135 		.mode		= 0644,
1136 		.proc_handler	= proc_dointvec,
1137 	},
1138 	{
1139 		.procname	= "tcp_moderate_rcvbuf",
1140 		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1141 		.maxlen		= sizeof(int),
1142 		.mode		= 0644,
1143 		.proc_handler	= proc_dointvec,
1144 	},
1145 	{
1146 		.procname	= "tcp_tso_win_divisor",
1147 		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1148 		.maxlen		= sizeof(int),
1149 		.mode		= 0644,
1150 		.proc_handler	= proc_dointvec,
1151 	},
1152 	{
1153 		.procname	= "tcp_workaround_signed_windows",
1154 		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1155 		.maxlen		= sizeof(int),
1156 		.mode		= 0644,
1157 		.proc_handler	= proc_dointvec
1158 	},
1159 	{
1160 		.procname	= "tcp_limit_output_bytes",
1161 		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1162 		.maxlen		= sizeof(int),
1163 		.mode		= 0644,
1164 		.proc_handler	= proc_dointvec
1165 	},
1166 	{
1167 		.procname	= "tcp_challenge_ack_limit",
1168 		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1169 		.maxlen		= sizeof(int),
1170 		.mode		= 0644,
1171 		.proc_handler	= proc_dointvec
1172 	},
1173 	{
1174 		.procname	= "tcp_min_tso_segs",
1175 		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1176 		.maxlen		= sizeof(int),
1177 		.mode		= 0644,
1178 		.proc_handler	= proc_dointvec_minmax,
1179 		.extra1		= &one,
1180 		.extra2		= &gso_max_segs,
1181 	},
1182 	{
1183 		.procname	= "tcp_min_rtt_wlen",
1184 		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1185 		.maxlen		= sizeof(int),
1186 		.mode		= 0644,
1187 		.proc_handler	= proc_dointvec_minmax,
1188 		.extra1		= &zero,
1189 		.extra2		= &one_day_secs
1190 	},
1191 	{
1192 		.procname	= "tcp_autocorking",
1193 		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1194 		.maxlen		= sizeof(int),
1195 		.mode		= 0644,
1196 		.proc_handler	= proc_dointvec_minmax,
1197 		.extra1		= &zero,
1198 		.extra2		= &one,
1199 	},
1200 	{
1201 		.procname	= "tcp_invalid_ratelimit",
1202 		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1203 		.maxlen		= sizeof(int),
1204 		.mode		= 0644,
1205 		.proc_handler	= proc_dointvec_ms_jiffies,
1206 	},
1207 	{
1208 		.procname	= "tcp_pacing_ss_ratio",
1209 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1210 		.maxlen		= sizeof(int),
1211 		.mode		= 0644,
1212 		.proc_handler	= proc_dointvec_minmax,
1213 		.extra1		= &zero,
1214 		.extra2		= &thousand,
1215 	},
1216 	{
1217 		.procname	= "tcp_pacing_ca_ratio",
1218 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1219 		.maxlen		= sizeof(int),
1220 		.mode		= 0644,
1221 		.proc_handler	= proc_dointvec_minmax,
1222 		.extra1		= &zero,
1223 		.extra2		= &thousand,
1224 	},
1225 	{
1226 		.procname	= "tcp_wmem",
1227 		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1228 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1229 		.mode		= 0644,
1230 		.proc_handler	= proc_dointvec_minmax,
1231 		.extra1		= &one,
1232 	},
1233 	{
1234 		.procname	= "tcp_rmem",
1235 		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1236 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1237 		.mode		= 0644,
1238 		.proc_handler	= proc_dointvec_minmax,
1239 		.extra1		= &one,
1240 	},
1241 	{
1242 		.procname	= "tcp_comp_sack_delay_ns",
1243 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1244 		.maxlen		= sizeof(unsigned long),
1245 		.mode		= 0644,
1246 		.proc_handler	= proc_doulongvec_minmax,
1247 	},
1248 	{
1249 		.procname	= "tcp_comp_sack_nr",
1250 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1251 		.maxlen		= sizeof(int),
1252 		.mode		= 0644,
1253 		.proc_handler	= proc_dointvec_minmax,
1254 		.extra1		= &zero,
1255 		.extra2		= &comp_sack_nr_max,
1256 	},
1257 	{
1258 		.procname	= "udp_rmem_min",
1259 		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1260 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1261 		.mode		= 0644,
1262 		.proc_handler	= proc_dointvec_minmax,
1263 		.extra1		= &one
1264 	},
1265 	{
1266 		.procname	= "udp_wmem_min",
1267 		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1268 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1269 		.mode		= 0644,
1270 		.proc_handler	= proc_dointvec_minmax,
1271 		.extra1		= &one
1272 	},
1273 	{ }
1274 };
1275 
1276 static __net_init int ipv4_sysctl_init_net(struct net *net)
1277 {
1278 	struct ctl_table *table;
1279 
1280 	table = ipv4_net_table;
1281 	if (!net_eq(net, &init_net)) {
1282 		int i;
1283 
1284 		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1285 		if (!table)
1286 			goto err_alloc;
1287 
1288 		/* Update the variables to point into the current struct net */
1289 		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1290 			table[i].data += (void *)net - (void *)&init_net;
1291 	}
1292 
1293 	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1294 	if (!net->ipv4.ipv4_hdr)
1295 		goto err_reg;
1296 
1297 	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1298 	if (!net->ipv4.sysctl_local_reserved_ports)
1299 		goto err_ports;
1300 
1301 	return 0;
1302 
1303 err_ports:
1304 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1305 err_reg:
1306 	if (!net_eq(net, &init_net))
1307 		kfree(table);
1308 err_alloc:
1309 	return -ENOMEM;
1310 }
1311 
1312 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1313 {
1314 	struct ctl_table *table;
1315 
1316 	kfree(net->ipv4.sysctl_local_reserved_ports);
1317 	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1318 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1319 	kfree(table);
1320 }
1321 
1322 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1323 	.init = ipv4_sysctl_init_net,
1324 	.exit = ipv4_sysctl_exit_net,
1325 };
1326 
1327 static __init int sysctl_ipv4_init(void)
1328 {
1329 	struct ctl_table_header *hdr;
1330 
1331 	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1332 	if (!hdr)
1333 		return -ENOMEM;
1334 
1335 	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1336 		unregister_net_sysctl_table(hdr);
1337 		return -ENOMEM;
1338 	}
1339 
1340 	return 0;
1341 }
1342 
1343 __initcall(sysctl_ipv4_init);
1344