xref: /openbmc/linux/net/ipv4/sysctl_net_ipv4.c (revision 43ffe817)
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 gso_max_segs = GSO_MAX_SEGS;
35 static int tcp_retr1_max = 255;
36 static int ip_local_port_range_min[] = { 1, 1 };
37 static int ip_local_port_range_max[] = { 65535, 65535 };
38 static int tcp_adv_win_scale_min = -31;
39 static int tcp_adv_win_scale_max = 31;
40 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
41 static int tcp_min_snd_mss_max = 65535;
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 *buffer, size_t *lenp, loff_t *ppos)
75 {
76 	struct net *net =
77 		container_of(table->data, struct net, ipv4.ip_local_ports.range);
78 	int ret;
79 	int range[2];
80 	struct ctl_table tmp = {
81 		.data = &range,
82 		.maxlen = sizeof(range),
83 		.mode = table->mode,
84 		.extra1 = &ip_local_port_range_min,
85 		.extra2 = &ip_local_port_range_max,
86 	};
87 
88 	inet_get_local_port_range(net, &range[0], &range[1]);
89 
90 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
91 
92 	if (write && ret == 0) {
93 		/* Ensure that the upper limit is not smaller than the lower,
94 		 * and that the lower does not encroach upon the privileged
95 		 * port limit.
96 		 */
97 		if ((range[1] < range[0]) ||
98 		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
99 			ret = -EINVAL;
100 		else
101 			set_local_port_range(net, range);
102 	}
103 
104 	return ret;
105 }
106 
107 /* Validate changes from /proc interface. */
108 static int ipv4_privileged_ports(struct ctl_table *table, int write,
109 				void *buffer, size_t *lenp, loff_t *ppos)
110 {
111 	struct net *net = container_of(table->data, struct net,
112 	    ipv4.sysctl_ip_prot_sock);
113 	int ret;
114 	int pports;
115 	int range[2];
116 	struct ctl_table tmp = {
117 		.data = &pports,
118 		.maxlen = sizeof(pports),
119 		.mode = table->mode,
120 		.extra1 = &ip_privileged_port_min,
121 		.extra2 = &ip_privileged_port_max,
122 	};
123 
124 	pports = net->ipv4.sysctl_ip_prot_sock;
125 
126 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
127 
128 	if (write && ret == 0) {
129 		inet_get_local_port_range(net, &range[0], &range[1]);
130 		/* Ensure that the local port range doesn't overlap with the
131 		 * privileged port range.
132 		 */
133 		if (range[0] < pports)
134 			ret = -EINVAL;
135 		else
136 			net->ipv4.sysctl_ip_prot_sock = pports;
137 	}
138 
139 	return ret;
140 }
141 
142 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
143 {
144 	kgid_t *data = table->data;
145 	struct net *net =
146 		container_of(table->data, struct net, ipv4.ping_group_range.range);
147 	unsigned int seq;
148 	do {
149 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
150 
151 		*low = data[0];
152 		*high = data[1];
153 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
154 }
155 
156 /* Update system visible IP port range */
157 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
158 {
159 	kgid_t *data = table->data;
160 	struct net *net =
161 		container_of(table->data, struct net, ipv4.ping_group_range.range);
162 	write_seqlock(&net->ipv4.ping_group_range.lock);
163 	data[0] = low;
164 	data[1] = high;
165 	write_sequnlock(&net->ipv4.ping_group_range.lock);
166 }
167 
168 /* Validate changes from /proc interface. */
169 static int ipv4_ping_group_range(struct ctl_table *table, int write,
170 				 void *buffer, size_t *lenp, loff_t *ppos)
171 {
172 	struct user_namespace *user_ns = current_user_ns();
173 	int ret;
174 	gid_t urange[2];
175 	kgid_t low, high;
176 	struct ctl_table tmp = {
177 		.data = &urange,
178 		.maxlen = sizeof(urange),
179 		.mode = table->mode,
180 		.extra1 = &ip_ping_group_range_min,
181 		.extra2 = &ip_ping_group_range_max,
182 	};
183 
184 	inet_get_ping_group_range_table(table, &low, &high);
185 	urange[0] = from_kgid_munged(user_ns, low);
186 	urange[1] = from_kgid_munged(user_ns, high);
187 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
188 
189 	if (write && ret == 0) {
190 		low = make_kgid(user_ns, urange[0]);
191 		high = make_kgid(user_ns, urange[1]);
192 		if (!gid_valid(low) || !gid_valid(high))
193 			return -EINVAL;
194 		if (urange[1] < urange[0] || gid_lt(high, low)) {
195 			low = make_kgid(&init_user_ns, 1);
196 			high = make_kgid(&init_user_ns, 0);
197 		}
198 		set_ping_group_range(table, low, high);
199 	}
200 
201 	return ret;
202 }
203 
204 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
205 				    void *buffer, size_t *lenp, loff_t *ppos)
206 {
207 	struct net *net;
208 	int ret;
209 
210 	net = container_of(table->data, struct net,
211 			   ipv4.sysctl_ip_fwd_update_priority);
212 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
213 	if (write && ret == 0)
214 		call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
215 					net);
216 
217 	return ret;
218 }
219 
220 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
221 				       void *buffer, size_t *lenp, loff_t *ppos)
222 {
223 	struct net *net = container_of(ctl->data, struct net,
224 				       ipv4.tcp_congestion_control);
225 	char val[TCP_CA_NAME_MAX];
226 	struct ctl_table tbl = {
227 		.data = val,
228 		.maxlen = TCP_CA_NAME_MAX,
229 	};
230 	int ret;
231 
232 	tcp_get_default_congestion_control(net, val);
233 
234 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
235 	if (write && ret == 0)
236 		ret = tcp_set_default_congestion_control(net, val);
237 	return ret;
238 }
239 
240 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
241 						 int write, void *buffer,
242 						 size_t *lenp, loff_t *ppos)
243 {
244 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
245 	int ret;
246 
247 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
248 	if (!tbl.data)
249 		return -ENOMEM;
250 	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
251 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
252 	kfree(tbl.data);
253 	return ret;
254 }
255 
256 static int proc_allowed_congestion_control(struct ctl_table *ctl,
257 					   int write, void *buffer,
258 					   size_t *lenp, loff_t *ppos)
259 {
260 	struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
261 	int ret;
262 
263 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
264 	if (!tbl.data)
265 		return -ENOMEM;
266 
267 	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
268 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
269 	if (write && ret == 0)
270 		ret = tcp_set_allowed_congestion_control(tbl.data);
271 	kfree(tbl.data);
272 	return ret;
273 }
274 
275 static int sscanf_key(char *buf, __le32 *key)
276 {
277 	u32 user_key[4];
278 	int i, ret = 0;
279 
280 	if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
281 		   user_key + 2, user_key + 3) != 4) {
282 		ret = -EINVAL;
283 	} else {
284 		for (i = 0; i < ARRAY_SIZE(user_key); i++)
285 			key[i] = cpu_to_le32(user_key[i]);
286 	}
287 	pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
288 		 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
289 
290 	return ret;
291 }
292 
293 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
294 				 void *buffer, size_t *lenp, loff_t *ppos)
295 {
296 	struct net *net = container_of(table->data, struct net,
297 	    ipv4.sysctl_tcp_fastopen);
298 	/* maxlen to print the list of keys in hex (*2), with dashes
299 	 * separating doublewords and a comma in between keys.
300 	 */
301 	struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
302 					    2 * TCP_FASTOPEN_KEY_MAX) +
303 					    (TCP_FASTOPEN_KEY_MAX * 5)) };
304 	u32 user_key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(u32)];
305 	__le32 key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(__le32)];
306 	char *backup_data;
307 	int ret, i = 0, off = 0, n_keys;
308 
309 	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
310 	if (!tbl.data)
311 		return -ENOMEM;
312 
313 	n_keys = tcp_fastopen_get_cipher(net, NULL, (u64 *)key);
314 	if (!n_keys) {
315 		memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
316 		n_keys = 1;
317 	}
318 
319 	for (i = 0; i < n_keys * 4; i++)
320 		user_key[i] = le32_to_cpu(key[i]);
321 
322 	for (i = 0; i < n_keys; i++) {
323 		off += snprintf(tbl.data + off, tbl.maxlen - off,
324 				"%08x-%08x-%08x-%08x",
325 				user_key[i * 4],
326 				user_key[i * 4 + 1],
327 				user_key[i * 4 + 2],
328 				user_key[i * 4 + 3]);
329 
330 		if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
331 			break;
332 
333 		if (i + 1 < n_keys)
334 			off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
335 	}
336 
337 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
338 
339 	if (write && ret == 0) {
340 		backup_data = strchr(tbl.data, ',');
341 		if (backup_data) {
342 			*backup_data = '\0';
343 			backup_data++;
344 		}
345 		if (sscanf_key(tbl.data, key)) {
346 			ret = -EINVAL;
347 			goto bad_key;
348 		}
349 		if (backup_data) {
350 			if (sscanf_key(backup_data, key + 4)) {
351 				ret = -EINVAL;
352 				goto bad_key;
353 			}
354 		}
355 		tcp_fastopen_reset_cipher(net, NULL, key,
356 					  backup_data ? key + 4 : NULL);
357 	}
358 
359 bad_key:
360 	kfree(tbl.data);
361 	return ret;
362 }
363 
364 static void proc_configure_early_demux(int enabled, int protocol)
365 {
366 	struct net_protocol *ipprot;
367 #if IS_ENABLED(CONFIG_IPV6)
368 	struct inet6_protocol *ip6prot;
369 #endif
370 
371 	rcu_read_lock();
372 
373 	ipprot = rcu_dereference(inet_protos[protocol]);
374 	if (ipprot)
375 		ipprot->early_demux = enabled ? ipprot->early_demux_handler :
376 						NULL;
377 
378 #if IS_ENABLED(CONFIG_IPV6)
379 	ip6prot = rcu_dereference(inet6_protos[protocol]);
380 	if (ip6prot)
381 		ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
382 						 NULL;
383 #endif
384 	rcu_read_unlock();
385 }
386 
387 static int proc_tcp_early_demux(struct ctl_table *table, int write,
388 				void *buffer, size_t *lenp, loff_t *ppos)
389 {
390 	int ret = 0;
391 
392 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
393 
394 	if (write && !ret) {
395 		int enabled = init_net.ipv4.sysctl_tcp_early_demux;
396 
397 		proc_configure_early_demux(enabled, IPPROTO_TCP);
398 	}
399 
400 	return ret;
401 }
402 
403 static int proc_udp_early_demux(struct ctl_table *table, int write,
404 				void *buffer, size_t *lenp, loff_t *ppos)
405 {
406 	int ret = 0;
407 
408 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
409 
410 	if (write && !ret) {
411 		int enabled = init_net.ipv4.sysctl_udp_early_demux;
412 
413 		proc_configure_early_demux(enabled, IPPROTO_UDP);
414 	}
415 
416 	return ret;
417 }
418 
419 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
420 					     int write, void *buffer,
421 					     size_t *lenp, loff_t *ppos)
422 {
423 	struct net *net = container_of(table->data, struct net,
424 	    ipv4.sysctl_tcp_fastopen_blackhole_timeout);
425 	int ret;
426 
427 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
428 	if (write && ret == 0)
429 		atomic_set(&net->ipv4.tfo_active_disable_times, 0);
430 
431 	return ret;
432 }
433 
434 static int proc_tcp_available_ulp(struct ctl_table *ctl,
435 				  int write, void *buffer, size_t *lenp,
436 				  loff_t *ppos)
437 {
438 	struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
439 	int ret;
440 
441 	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
442 	if (!tbl.data)
443 		return -ENOMEM;
444 	tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
445 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
446 	kfree(tbl.data);
447 
448 	return ret;
449 }
450 
451 #ifdef CONFIG_IP_ROUTE_MULTIPATH
452 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
453 					  void *buffer, size_t *lenp,
454 					  loff_t *ppos)
455 {
456 	struct net *net = container_of(table->data, struct net,
457 	    ipv4.sysctl_fib_multipath_hash_policy);
458 	int ret;
459 
460 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
461 	if (write && ret == 0)
462 		call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
463 
464 	return ret;
465 }
466 #endif
467 
468 static struct ctl_table ipv4_table[] = {
469 	{
470 		.procname	= "tcp_max_orphans",
471 		.data		= &sysctl_tcp_max_orphans,
472 		.maxlen		= sizeof(int),
473 		.mode		= 0644,
474 		.proc_handler	= proc_dointvec
475 	},
476 	{
477 		.procname	= "inet_peer_threshold",
478 		.data		= &inet_peer_threshold,
479 		.maxlen		= sizeof(int),
480 		.mode		= 0644,
481 		.proc_handler	= proc_dointvec
482 	},
483 	{
484 		.procname	= "inet_peer_minttl",
485 		.data		= &inet_peer_minttl,
486 		.maxlen		= sizeof(int),
487 		.mode		= 0644,
488 		.proc_handler	= proc_dointvec_jiffies,
489 	},
490 	{
491 		.procname	= "inet_peer_maxttl",
492 		.data		= &inet_peer_maxttl,
493 		.maxlen		= sizeof(int),
494 		.mode		= 0644,
495 		.proc_handler	= proc_dointvec_jiffies,
496 	},
497 	{
498 		.procname	= "tcp_mem",
499 		.maxlen		= sizeof(sysctl_tcp_mem),
500 		.data		= &sysctl_tcp_mem,
501 		.mode		= 0644,
502 		.proc_handler	= proc_doulongvec_minmax,
503 	},
504 	{
505 		.procname	= "tcp_low_latency",
506 		.data		= &sysctl_tcp_low_latency,
507 		.maxlen		= sizeof(int),
508 		.mode		= 0644,
509 		.proc_handler	= proc_dointvec
510 	},
511 #ifdef CONFIG_NETLABEL
512 	{
513 		.procname	= "cipso_cache_enable",
514 		.data		= &cipso_v4_cache_enabled,
515 		.maxlen		= sizeof(int),
516 		.mode		= 0644,
517 		.proc_handler	= proc_dointvec,
518 	},
519 	{
520 		.procname	= "cipso_cache_bucket_size",
521 		.data		= &cipso_v4_cache_bucketsize,
522 		.maxlen		= sizeof(int),
523 		.mode		= 0644,
524 		.proc_handler	= proc_dointvec,
525 	},
526 	{
527 		.procname	= "cipso_rbm_optfmt",
528 		.data		= &cipso_v4_rbm_optfmt,
529 		.maxlen		= sizeof(int),
530 		.mode		= 0644,
531 		.proc_handler	= proc_dointvec,
532 	},
533 	{
534 		.procname	= "cipso_rbm_strictvalid",
535 		.data		= &cipso_v4_rbm_strictvalid,
536 		.maxlen		= sizeof(int),
537 		.mode		= 0644,
538 		.proc_handler	= proc_dointvec,
539 	},
540 #endif /* CONFIG_NETLABEL */
541 	{
542 		.procname	= "tcp_available_ulp",
543 		.maxlen		= TCP_ULP_BUF_MAX,
544 		.mode		= 0444,
545 		.proc_handler   = proc_tcp_available_ulp,
546 	},
547 	{
548 		.procname	= "icmp_msgs_per_sec",
549 		.data		= &sysctl_icmp_msgs_per_sec,
550 		.maxlen		= sizeof(int),
551 		.mode		= 0644,
552 		.proc_handler	= proc_dointvec_minmax,
553 		.extra1		= SYSCTL_ZERO,
554 	},
555 	{
556 		.procname	= "icmp_msgs_burst",
557 		.data		= &sysctl_icmp_msgs_burst,
558 		.maxlen		= sizeof(int),
559 		.mode		= 0644,
560 		.proc_handler	= proc_dointvec_minmax,
561 		.extra1		= SYSCTL_ZERO,
562 	},
563 	{
564 		.procname	= "udp_mem",
565 		.data		= &sysctl_udp_mem,
566 		.maxlen		= sizeof(sysctl_udp_mem),
567 		.mode		= 0644,
568 		.proc_handler	= proc_doulongvec_minmax,
569 	},
570 	{
571 		.procname	= "fib_sync_mem",
572 		.data		= &sysctl_fib_sync_mem,
573 		.maxlen		= sizeof(sysctl_fib_sync_mem),
574 		.mode		= 0644,
575 		.proc_handler	= proc_douintvec_minmax,
576 		.extra1		= &sysctl_fib_sync_mem_min,
577 		.extra2		= &sysctl_fib_sync_mem_max,
578 	},
579 	{
580 		.procname	= "tcp_rx_skb_cache",
581 		.data		= &tcp_rx_skb_cache_key.key,
582 		.mode		= 0644,
583 		.proc_handler	= proc_do_static_key,
584 	},
585 	{
586 		.procname	= "tcp_tx_skb_cache",
587 		.data		= &tcp_tx_skb_cache_key.key,
588 		.mode		= 0644,
589 		.proc_handler	= proc_do_static_key,
590 	},
591 	{ }
592 };
593 
594 static struct ctl_table ipv4_net_table[] = {
595 	{
596 		.procname	= "icmp_echo_ignore_all",
597 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
598 		.maxlen		= sizeof(int),
599 		.mode		= 0644,
600 		.proc_handler	= proc_dointvec
601 	},
602 	{
603 		.procname	= "icmp_echo_ignore_broadcasts",
604 		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
605 		.maxlen		= sizeof(int),
606 		.mode		= 0644,
607 		.proc_handler	= proc_dointvec
608 	},
609 	{
610 		.procname	= "icmp_ignore_bogus_error_responses",
611 		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
612 		.maxlen		= sizeof(int),
613 		.mode		= 0644,
614 		.proc_handler	= proc_dointvec
615 	},
616 	{
617 		.procname	= "icmp_errors_use_inbound_ifaddr",
618 		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
619 		.maxlen		= sizeof(int),
620 		.mode		= 0644,
621 		.proc_handler	= proc_dointvec
622 	},
623 	{
624 		.procname	= "icmp_ratelimit",
625 		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
626 		.maxlen		= sizeof(int),
627 		.mode		= 0644,
628 		.proc_handler	= proc_dointvec_ms_jiffies,
629 	},
630 	{
631 		.procname	= "icmp_ratemask",
632 		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
633 		.maxlen		= sizeof(int),
634 		.mode		= 0644,
635 		.proc_handler	= proc_dointvec
636 	},
637 	{
638 		.procname	= "ping_group_range",
639 		.data		= &init_net.ipv4.ping_group_range.range,
640 		.maxlen		= sizeof(gid_t)*2,
641 		.mode		= 0644,
642 		.proc_handler	= ipv4_ping_group_range,
643 	},
644 #ifdef CONFIG_NET_L3_MASTER_DEV
645 	{
646 		.procname	= "raw_l3mdev_accept",
647 		.data		= &init_net.ipv4.sysctl_raw_l3mdev_accept,
648 		.maxlen		= sizeof(int),
649 		.mode		= 0644,
650 		.proc_handler	= proc_dointvec_minmax,
651 		.extra1		= SYSCTL_ZERO,
652 		.extra2		= SYSCTL_ONE,
653 	},
654 #endif
655 	{
656 		.procname	= "tcp_ecn",
657 		.data		= &init_net.ipv4.sysctl_tcp_ecn,
658 		.maxlen		= sizeof(int),
659 		.mode		= 0644,
660 		.proc_handler	= proc_dointvec
661 	},
662 	{
663 		.procname	= "tcp_ecn_fallback",
664 		.data		= &init_net.ipv4.sysctl_tcp_ecn_fallback,
665 		.maxlen		= sizeof(int),
666 		.mode		= 0644,
667 		.proc_handler	= proc_dointvec
668 	},
669 	{
670 		.procname	= "ip_dynaddr",
671 		.data		= &init_net.ipv4.sysctl_ip_dynaddr,
672 		.maxlen		= sizeof(int),
673 		.mode		= 0644,
674 		.proc_handler	= proc_dointvec
675 	},
676 	{
677 		.procname	= "ip_early_demux",
678 		.data		= &init_net.ipv4.sysctl_ip_early_demux,
679 		.maxlen		= sizeof(int),
680 		.mode		= 0644,
681 		.proc_handler	= proc_dointvec
682 	},
683 	{
684 		.procname       = "udp_early_demux",
685 		.data           = &init_net.ipv4.sysctl_udp_early_demux,
686 		.maxlen         = sizeof(int),
687 		.mode           = 0644,
688 		.proc_handler   = proc_udp_early_demux
689 	},
690 	{
691 		.procname       = "tcp_early_demux",
692 		.data           = &init_net.ipv4.sysctl_tcp_early_demux,
693 		.maxlen         = sizeof(int),
694 		.mode           = 0644,
695 		.proc_handler   = proc_tcp_early_demux
696 	},
697 	{
698 		.procname       = "nexthop_compat_mode",
699 		.data           = &init_net.ipv4.sysctl_nexthop_compat_mode,
700 		.maxlen         = sizeof(int),
701 		.mode           = 0644,
702 		.proc_handler   = proc_dointvec_minmax,
703 		.extra1		= SYSCTL_ZERO,
704 		.extra2		= SYSCTL_ONE,
705 	},
706 	{
707 		.procname	= "ip_default_ttl",
708 		.data		= &init_net.ipv4.sysctl_ip_default_ttl,
709 		.maxlen		= sizeof(int),
710 		.mode		= 0644,
711 		.proc_handler	= proc_dointvec_minmax,
712 		.extra1		= &ip_ttl_min,
713 		.extra2		= &ip_ttl_max,
714 	},
715 	{
716 		.procname	= "ip_local_port_range",
717 		.maxlen		= sizeof(init_net.ipv4.ip_local_ports.range),
718 		.data		= &init_net.ipv4.ip_local_ports.range,
719 		.mode		= 0644,
720 		.proc_handler	= ipv4_local_port_range,
721 	},
722 	{
723 		.procname	= "ip_local_reserved_ports",
724 		.data		= &init_net.ipv4.sysctl_local_reserved_ports,
725 		.maxlen		= 65536,
726 		.mode		= 0644,
727 		.proc_handler	= proc_do_large_bitmap,
728 	},
729 	{
730 		.procname	= "ip_no_pmtu_disc",
731 		.data		= &init_net.ipv4.sysctl_ip_no_pmtu_disc,
732 		.maxlen		= sizeof(int),
733 		.mode		= 0644,
734 		.proc_handler	= proc_dointvec
735 	},
736 	{
737 		.procname	= "ip_forward_use_pmtu",
738 		.data		= &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
739 		.maxlen		= sizeof(int),
740 		.mode		= 0644,
741 		.proc_handler	= proc_dointvec,
742 	},
743 	{
744 		.procname	= "ip_forward_update_priority",
745 		.data		= &init_net.ipv4.sysctl_ip_fwd_update_priority,
746 		.maxlen		= sizeof(int),
747 		.mode		= 0644,
748 		.proc_handler   = ipv4_fwd_update_priority,
749 		.extra1		= SYSCTL_ZERO,
750 		.extra2		= SYSCTL_ONE,
751 	},
752 	{
753 		.procname	= "ip_nonlocal_bind",
754 		.data		= &init_net.ipv4.sysctl_ip_nonlocal_bind,
755 		.maxlen		= sizeof(int),
756 		.mode		= 0644,
757 		.proc_handler	= proc_dointvec
758 	},
759 	{
760 		.procname	= "ip_autobind_reuse",
761 		.data		= &init_net.ipv4.sysctl_ip_autobind_reuse,
762 		.maxlen		= sizeof(int),
763 		.mode		= 0644,
764 		.proc_handler	= proc_dointvec_minmax,
765 		.extra1         = SYSCTL_ZERO,
766 		.extra2         = SYSCTL_ONE,
767 	},
768 	{
769 		.procname	= "fwmark_reflect",
770 		.data		= &init_net.ipv4.sysctl_fwmark_reflect,
771 		.maxlen		= sizeof(int),
772 		.mode		= 0644,
773 		.proc_handler	= proc_dointvec,
774 	},
775 	{
776 		.procname	= "tcp_fwmark_accept",
777 		.data		= &init_net.ipv4.sysctl_tcp_fwmark_accept,
778 		.maxlen		= sizeof(int),
779 		.mode		= 0644,
780 		.proc_handler	= proc_dointvec,
781 	},
782 #ifdef CONFIG_NET_L3_MASTER_DEV
783 	{
784 		.procname	= "tcp_l3mdev_accept",
785 		.data		= &init_net.ipv4.sysctl_tcp_l3mdev_accept,
786 		.maxlen		= sizeof(int),
787 		.mode		= 0644,
788 		.proc_handler	= proc_dointvec_minmax,
789 		.extra1		= SYSCTL_ZERO,
790 		.extra2		= SYSCTL_ONE,
791 	},
792 #endif
793 	{
794 		.procname	= "tcp_mtu_probing",
795 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probing,
796 		.maxlen		= sizeof(int),
797 		.mode		= 0644,
798 		.proc_handler	= proc_dointvec,
799 	},
800 	{
801 		.procname	= "tcp_base_mss",
802 		.data		= &init_net.ipv4.sysctl_tcp_base_mss,
803 		.maxlen		= sizeof(int),
804 		.mode		= 0644,
805 		.proc_handler	= proc_dointvec,
806 	},
807 	{
808 		.procname	= "tcp_min_snd_mss",
809 		.data		= &init_net.ipv4.sysctl_tcp_min_snd_mss,
810 		.maxlen		= sizeof(int),
811 		.mode		= 0644,
812 		.proc_handler	= proc_dointvec_minmax,
813 		.extra1		= &tcp_min_snd_mss_min,
814 		.extra2		= &tcp_min_snd_mss_max,
815 	},
816 	{
817 		.procname	= "tcp_mtu_probe_floor",
818 		.data		= &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
819 		.maxlen		= sizeof(int),
820 		.mode		= 0644,
821 		.proc_handler	= proc_dointvec_minmax,
822 		.extra1		= &tcp_min_snd_mss_min,
823 		.extra2		= &tcp_min_snd_mss_max,
824 	},
825 	{
826 		.procname	= "tcp_probe_threshold",
827 		.data		= &init_net.ipv4.sysctl_tcp_probe_threshold,
828 		.maxlen		= sizeof(int),
829 		.mode		= 0644,
830 		.proc_handler	= proc_dointvec,
831 	},
832 	{
833 		.procname	= "tcp_probe_interval",
834 		.data		= &init_net.ipv4.sysctl_tcp_probe_interval,
835 		.maxlen		= sizeof(u32),
836 		.mode		= 0644,
837 		.proc_handler	= proc_douintvec_minmax,
838 		.extra2		= &u32_max_div_HZ,
839 	},
840 	{
841 		.procname	= "igmp_link_local_mcast_reports",
842 		.data		= &init_net.ipv4.sysctl_igmp_llm_reports,
843 		.maxlen		= sizeof(int),
844 		.mode		= 0644,
845 		.proc_handler	= proc_dointvec
846 	},
847 	{
848 		.procname	= "igmp_max_memberships",
849 		.data		= &init_net.ipv4.sysctl_igmp_max_memberships,
850 		.maxlen		= sizeof(int),
851 		.mode		= 0644,
852 		.proc_handler	= proc_dointvec
853 	},
854 	{
855 		.procname	= "igmp_max_msf",
856 		.data		= &init_net.ipv4.sysctl_igmp_max_msf,
857 		.maxlen		= sizeof(int),
858 		.mode		= 0644,
859 		.proc_handler	= proc_dointvec
860 	},
861 #ifdef CONFIG_IP_MULTICAST
862 	{
863 		.procname	= "igmp_qrv",
864 		.data		= &init_net.ipv4.sysctl_igmp_qrv,
865 		.maxlen		= sizeof(int),
866 		.mode		= 0644,
867 		.proc_handler	= proc_dointvec_minmax,
868 		.extra1		= SYSCTL_ONE
869 	},
870 #endif
871 	{
872 		.procname	= "tcp_congestion_control",
873 		.data		= &init_net.ipv4.tcp_congestion_control,
874 		.mode		= 0644,
875 		.maxlen		= TCP_CA_NAME_MAX,
876 		.proc_handler	= proc_tcp_congestion_control,
877 	},
878 	{
879 		.procname	= "tcp_available_congestion_control",
880 		.maxlen		= TCP_CA_BUF_MAX,
881 		.mode		= 0444,
882 		.proc_handler   = proc_tcp_available_congestion_control,
883 	},
884 	{
885 		.procname	= "tcp_allowed_congestion_control",
886 		.maxlen		= TCP_CA_BUF_MAX,
887 		.mode		= 0644,
888 		.proc_handler   = proc_allowed_congestion_control,
889 	},
890 	{
891 		.procname	= "tcp_keepalive_time",
892 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,
893 		.maxlen		= sizeof(int),
894 		.mode		= 0644,
895 		.proc_handler	= proc_dointvec_jiffies,
896 	},
897 	{
898 		.procname	= "tcp_keepalive_probes",
899 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_probes,
900 		.maxlen		= sizeof(int),
901 		.mode		= 0644,
902 		.proc_handler	= proc_dointvec
903 	},
904 	{
905 		.procname	= "tcp_keepalive_intvl",
906 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_intvl,
907 		.maxlen		= sizeof(int),
908 		.mode		= 0644,
909 		.proc_handler	= proc_dointvec_jiffies,
910 	},
911 	{
912 		.procname	= "tcp_syn_retries",
913 		.data		= &init_net.ipv4.sysctl_tcp_syn_retries,
914 		.maxlen		= sizeof(int),
915 		.mode		= 0644,
916 		.proc_handler	= proc_dointvec_minmax,
917 		.extra1		= &tcp_syn_retries_min,
918 		.extra2		= &tcp_syn_retries_max
919 	},
920 	{
921 		.procname	= "tcp_synack_retries",
922 		.data		= &init_net.ipv4.sysctl_tcp_synack_retries,
923 		.maxlen		= sizeof(int),
924 		.mode		= 0644,
925 		.proc_handler	= proc_dointvec
926 	},
927 #ifdef CONFIG_SYN_COOKIES
928 	{
929 		.procname	= "tcp_syncookies",
930 		.data		= &init_net.ipv4.sysctl_tcp_syncookies,
931 		.maxlen		= sizeof(int),
932 		.mode		= 0644,
933 		.proc_handler	= proc_dointvec
934 	},
935 #endif
936 	{
937 		.procname	= "tcp_reordering",
938 		.data		= &init_net.ipv4.sysctl_tcp_reordering,
939 		.maxlen		= sizeof(int),
940 		.mode		= 0644,
941 		.proc_handler	= proc_dointvec
942 	},
943 	{
944 		.procname	= "tcp_retries1",
945 		.data		= &init_net.ipv4.sysctl_tcp_retries1,
946 		.maxlen		= sizeof(int),
947 		.mode		= 0644,
948 		.proc_handler	= proc_dointvec_minmax,
949 		.extra2		= &tcp_retr1_max
950 	},
951 	{
952 		.procname	= "tcp_retries2",
953 		.data		= &init_net.ipv4.sysctl_tcp_retries2,
954 		.maxlen		= sizeof(int),
955 		.mode		= 0644,
956 		.proc_handler	= proc_dointvec
957 	},
958 	{
959 		.procname	= "tcp_orphan_retries",
960 		.data		= &init_net.ipv4.sysctl_tcp_orphan_retries,
961 		.maxlen		= sizeof(int),
962 		.mode		= 0644,
963 		.proc_handler	= proc_dointvec
964 	},
965 	{
966 		.procname	= "tcp_fin_timeout",
967 		.data		= &init_net.ipv4.sysctl_tcp_fin_timeout,
968 		.maxlen		= sizeof(int),
969 		.mode		= 0644,
970 		.proc_handler	= proc_dointvec_jiffies,
971 	},
972 	{
973 		.procname	= "tcp_notsent_lowat",
974 		.data		= &init_net.ipv4.sysctl_tcp_notsent_lowat,
975 		.maxlen		= sizeof(unsigned int),
976 		.mode		= 0644,
977 		.proc_handler	= proc_douintvec,
978 	},
979 	{
980 		.procname	= "tcp_tw_reuse",
981 		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
982 		.maxlen		= sizeof(int),
983 		.mode		= 0644,
984 		.proc_handler	= proc_dointvec_minmax,
985 		.extra1		= SYSCTL_ZERO,
986 		.extra2		= &two,
987 	},
988 	{
989 		.procname	= "tcp_max_tw_buckets",
990 		.data		= &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
991 		.maxlen		= sizeof(int),
992 		.mode		= 0644,
993 		.proc_handler	= proc_dointvec
994 	},
995 	{
996 		.procname	= "tcp_max_syn_backlog",
997 		.data		= &init_net.ipv4.sysctl_max_syn_backlog,
998 		.maxlen		= sizeof(int),
999 		.mode		= 0644,
1000 		.proc_handler	= proc_dointvec
1001 	},
1002 	{
1003 		.procname	= "tcp_fastopen",
1004 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1005 		.maxlen		= sizeof(int),
1006 		.mode		= 0644,
1007 		.proc_handler	= proc_dointvec,
1008 	},
1009 	{
1010 		.procname	= "tcp_fastopen_key",
1011 		.mode		= 0600,
1012 		.data		= &init_net.ipv4.sysctl_tcp_fastopen,
1013 		/* maxlen to print the list of keys in hex (*2), with dashes
1014 		 * separating doublewords and a comma in between keys.
1015 		 */
1016 		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH *
1017 				   2 * TCP_FASTOPEN_KEY_MAX) +
1018 				   (TCP_FASTOPEN_KEY_MAX * 5)),
1019 		.proc_handler	= proc_tcp_fastopen_key,
1020 	},
1021 	{
1022 		.procname	= "tcp_fastopen_blackhole_timeout_sec",
1023 		.data		= &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1024 		.maxlen		= sizeof(int),
1025 		.mode		= 0644,
1026 		.proc_handler	= proc_tfo_blackhole_detect_timeout,
1027 		.extra1		= SYSCTL_ZERO,
1028 	},
1029 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1030 	{
1031 		.procname	= "fib_multipath_use_neigh",
1032 		.data		= &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1033 		.maxlen		= sizeof(int),
1034 		.mode		= 0644,
1035 		.proc_handler	= proc_dointvec_minmax,
1036 		.extra1		= SYSCTL_ZERO,
1037 		.extra2		= SYSCTL_ONE,
1038 	},
1039 	{
1040 		.procname	= "fib_multipath_hash_policy",
1041 		.data		= &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1042 		.maxlen		= sizeof(int),
1043 		.mode		= 0644,
1044 		.proc_handler	= proc_fib_multipath_hash_policy,
1045 		.extra1		= SYSCTL_ZERO,
1046 		.extra2		= &two,
1047 	},
1048 #endif
1049 	{
1050 		.procname	= "ip_unprivileged_port_start",
1051 		.maxlen		= sizeof(int),
1052 		.data		= &init_net.ipv4.sysctl_ip_prot_sock,
1053 		.mode		= 0644,
1054 		.proc_handler	= ipv4_privileged_ports,
1055 	},
1056 #ifdef CONFIG_NET_L3_MASTER_DEV
1057 	{
1058 		.procname	= "udp_l3mdev_accept",
1059 		.data		= &init_net.ipv4.sysctl_udp_l3mdev_accept,
1060 		.maxlen		= sizeof(int),
1061 		.mode		= 0644,
1062 		.proc_handler	= proc_dointvec_minmax,
1063 		.extra1		= SYSCTL_ZERO,
1064 		.extra2		= SYSCTL_ONE,
1065 	},
1066 #endif
1067 	{
1068 		.procname	= "tcp_sack",
1069 		.data		= &init_net.ipv4.sysctl_tcp_sack,
1070 		.maxlen		= sizeof(int),
1071 		.mode		= 0644,
1072 		.proc_handler	= proc_dointvec
1073 	},
1074 	{
1075 		.procname	= "tcp_window_scaling",
1076 		.data		= &init_net.ipv4.sysctl_tcp_window_scaling,
1077 		.maxlen		= sizeof(int),
1078 		.mode		= 0644,
1079 		.proc_handler	= proc_dointvec
1080 	},
1081 	{
1082 		.procname	= "tcp_timestamps",
1083 		.data		= &init_net.ipv4.sysctl_tcp_timestamps,
1084 		.maxlen		= sizeof(int),
1085 		.mode		= 0644,
1086 		.proc_handler	= proc_dointvec
1087 	},
1088 	{
1089 		.procname	= "tcp_early_retrans",
1090 		.data		= &init_net.ipv4.sysctl_tcp_early_retrans,
1091 		.maxlen		= sizeof(int),
1092 		.mode		= 0644,
1093 		.proc_handler	= proc_dointvec_minmax,
1094 		.extra1		= SYSCTL_ZERO,
1095 		.extra2		= &four,
1096 	},
1097 	{
1098 		.procname	= "tcp_recovery",
1099 		.data		= &init_net.ipv4.sysctl_tcp_recovery,
1100 		.maxlen		= sizeof(int),
1101 		.mode		= 0644,
1102 		.proc_handler	= proc_dointvec,
1103 	},
1104 	{
1105 		.procname       = "tcp_thin_linear_timeouts",
1106 		.data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1107 		.maxlen         = sizeof(int),
1108 		.mode           = 0644,
1109 		.proc_handler   = proc_dointvec
1110 	},
1111 	{
1112 		.procname	= "tcp_slow_start_after_idle",
1113 		.data		= &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1114 		.maxlen		= sizeof(int),
1115 		.mode		= 0644,
1116 		.proc_handler	= proc_dointvec
1117 	},
1118 	{
1119 		.procname	= "tcp_retrans_collapse",
1120 		.data		= &init_net.ipv4.sysctl_tcp_retrans_collapse,
1121 		.maxlen		= sizeof(int),
1122 		.mode		= 0644,
1123 		.proc_handler	= proc_dointvec
1124 	},
1125 	{
1126 		.procname	= "tcp_stdurg",
1127 		.data		= &init_net.ipv4.sysctl_tcp_stdurg,
1128 		.maxlen		= sizeof(int),
1129 		.mode		= 0644,
1130 		.proc_handler	= proc_dointvec
1131 	},
1132 	{
1133 		.procname	= "tcp_rfc1337",
1134 		.data		= &init_net.ipv4.sysctl_tcp_rfc1337,
1135 		.maxlen		= sizeof(int),
1136 		.mode		= 0644,
1137 		.proc_handler	= proc_dointvec
1138 	},
1139 	{
1140 		.procname	= "tcp_abort_on_overflow",
1141 		.data		= &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1142 		.maxlen		= sizeof(int),
1143 		.mode		= 0644,
1144 		.proc_handler	= proc_dointvec
1145 	},
1146 	{
1147 		.procname	= "tcp_fack",
1148 		.data		= &init_net.ipv4.sysctl_tcp_fack,
1149 		.maxlen		= sizeof(int),
1150 		.mode		= 0644,
1151 		.proc_handler	= proc_dointvec
1152 	},
1153 	{
1154 		.procname	= "tcp_max_reordering",
1155 		.data		= &init_net.ipv4.sysctl_tcp_max_reordering,
1156 		.maxlen		= sizeof(int),
1157 		.mode		= 0644,
1158 		.proc_handler	= proc_dointvec
1159 	},
1160 	{
1161 		.procname	= "tcp_dsack",
1162 		.data		= &init_net.ipv4.sysctl_tcp_dsack,
1163 		.maxlen		= sizeof(int),
1164 		.mode		= 0644,
1165 		.proc_handler	= proc_dointvec
1166 	},
1167 	{
1168 		.procname	= "tcp_app_win",
1169 		.data		= &init_net.ipv4.sysctl_tcp_app_win,
1170 		.maxlen		= sizeof(int),
1171 		.mode		= 0644,
1172 		.proc_handler	= proc_dointvec
1173 	},
1174 	{
1175 		.procname	= "tcp_adv_win_scale",
1176 		.data		= &init_net.ipv4.sysctl_tcp_adv_win_scale,
1177 		.maxlen		= sizeof(int),
1178 		.mode		= 0644,
1179 		.proc_handler	= proc_dointvec_minmax,
1180 		.extra1		= &tcp_adv_win_scale_min,
1181 		.extra2		= &tcp_adv_win_scale_max,
1182 	},
1183 	{
1184 		.procname	= "tcp_frto",
1185 		.data		= &init_net.ipv4.sysctl_tcp_frto,
1186 		.maxlen		= sizeof(int),
1187 		.mode		= 0644,
1188 		.proc_handler	= proc_dointvec
1189 	},
1190 	{
1191 		.procname	= "tcp_no_metrics_save",
1192 		.data		= &init_net.ipv4.sysctl_tcp_nometrics_save,
1193 		.maxlen		= sizeof(int),
1194 		.mode		= 0644,
1195 		.proc_handler	= proc_dointvec,
1196 	},
1197 	{
1198 		.procname	= "tcp_no_ssthresh_metrics_save",
1199 		.data		= &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1200 		.maxlen		= sizeof(int),
1201 		.mode		= 0644,
1202 		.proc_handler	= proc_dointvec_minmax,
1203 		.extra1		= SYSCTL_ZERO,
1204 		.extra2		= SYSCTL_ONE,
1205 	},
1206 	{
1207 		.procname	= "tcp_moderate_rcvbuf",
1208 		.data		= &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1209 		.maxlen		= sizeof(int),
1210 		.mode		= 0644,
1211 		.proc_handler	= proc_dointvec,
1212 	},
1213 	{
1214 		.procname	= "tcp_tso_win_divisor",
1215 		.data		= &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1216 		.maxlen		= sizeof(int),
1217 		.mode		= 0644,
1218 		.proc_handler	= proc_dointvec,
1219 	},
1220 	{
1221 		.procname	= "tcp_workaround_signed_windows",
1222 		.data		= &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1223 		.maxlen		= sizeof(int),
1224 		.mode		= 0644,
1225 		.proc_handler	= proc_dointvec
1226 	},
1227 	{
1228 		.procname	= "tcp_limit_output_bytes",
1229 		.data		= &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1230 		.maxlen		= sizeof(int),
1231 		.mode		= 0644,
1232 		.proc_handler	= proc_dointvec
1233 	},
1234 	{
1235 		.procname	= "tcp_challenge_ack_limit",
1236 		.data		= &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1237 		.maxlen		= sizeof(int),
1238 		.mode		= 0644,
1239 		.proc_handler	= proc_dointvec
1240 	},
1241 	{
1242 		.procname	= "tcp_min_tso_segs",
1243 		.data		= &init_net.ipv4.sysctl_tcp_min_tso_segs,
1244 		.maxlen		= sizeof(int),
1245 		.mode		= 0644,
1246 		.proc_handler	= proc_dointvec_minmax,
1247 		.extra1		= SYSCTL_ONE,
1248 		.extra2		= &gso_max_segs,
1249 	},
1250 	{
1251 		.procname	= "tcp_min_rtt_wlen",
1252 		.data		= &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1253 		.maxlen		= sizeof(int),
1254 		.mode		= 0644,
1255 		.proc_handler	= proc_dointvec_minmax,
1256 		.extra1		= SYSCTL_ZERO,
1257 		.extra2		= &one_day_secs
1258 	},
1259 	{
1260 		.procname	= "tcp_autocorking",
1261 		.data		= &init_net.ipv4.sysctl_tcp_autocorking,
1262 		.maxlen		= sizeof(int),
1263 		.mode		= 0644,
1264 		.proc_handler	= proc_dointvec_minmax,
1265 		.extra1		= SYSCTL_ZERO,
1266 		.extra2		= SYSCTL_ONE,
1267 	},
1268 	{
1269 		.procname	= "tcp_invalid_ratelimit",
1270 		.data		= &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1271 		.maxlen		= sizeof(int),
1272 		.mode		= 0644,
1273 		.proc_handler	= proc_dointvec_ms_jiffies,
1274 	},
1275 	{
1276 		.procname	= "tcp_pacing_ss_ratio",
1277 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1278 		.maxlen		= sizeof(int),
1279 		.mode		= 0644,
1280 		.proc_handler	= proc_dointvec_minmax,
1281 		.extra1		= SYSCTL_ZERO,
1282 		.extra2		= &thousand,
1283 	},
1284 	{
1285 		.procname	= "tcp_pacing_ca_ratio",
1286 		.data		= &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1287 		.maxlen		= sizeof(int),
1288 		.mode		= 0644,
1289 		.proc_handler	= proc_dointvec_minmax,
1290 		.extra1		= SYSCTL_ZERO,
1291 		.extra2		= &thousand,
1292 	},
1293 	{
1294 		.procname	= "tcp_wmem",
1295 		.data		= &init_net.ipv4.sysctl_tcp_wmem,
1296 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_wmem),
1297 		.mode		= 0644,
1298 		.proc_handler	= proc_dointvec_minmax,
1299 		.extra1		= SYSCTL_ONE,
1300 	},
1301 	{
1302 		.procname	= "tcp_rmem",
1303 		.data		= &init_net.ipv4.sysctl_tcp_rmem,
1304 		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_rmem),
1305 		.mode		= 0644,
1306 		.proc_handler	= proc_dointvec_minmax,
1307 		.extra1		= SYSCTL_ONE,
1308 	},
1309 	{
1310 		.procname	= "tcp_comp_sack_delay_ns",
1311 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1312 		.maxlen		= sizeof(unsigned long),
1313 		.mode		= 0644,
1314 		.proc_handler	= proc_doulongvec_minmax,
1315 	},
1316 	{
1317 		.procname	= "tcp_comp_sack_slack_ns",
1318 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1319 		.maxlen		= sizeof(unsigned long),
1320 		.mode		= 0644,
1321 		.proc_handler	= proc_doulongvec_minmax,
1322 	},
1323 	{
1324 		.procname	= "tcp_comp_sack_nr",
1325 		.data		= &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1326 		.maxlen		= sizeof(int),
1327 		.mode		= 0644,
1328 		.proc_handler	= proc_dointvec_minmax,
1329 		.extra1		= SYSCTL_ZERO,
1330 		.extra2		= &comp_sack_nr_max,
1331 	},
1332 	{
1333 		.procname       = "tcp_reflect_tos",
1334 		.data           = &init_net.ipv4.sysctl_tcp_reflect_tos,
1335 		.maxlen         = sizeof(int),
1336 		.mode           = 0644,
1337 		.proc_handler   = proc_dointvec_minmax,
1338 		.extra1         = SYSCTL_ZERO,
1339 		.extra2         = SYSCTL_ONE,
1340 	},
1341 	{
1342 		.procname	= "udp_rmem_min",
1343 		.data		= &init_net.ipv4.sysctl_udp_rmem_min,
1344 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1345 		.mode		= 0644,
1346 		.proc_handler	= proc_dointvec_minmax,
1347 		.extra1		= SYSCTL_ONE
1348 	},
1349 	{
1350 		.procname	= "udp_wmem_min",
1351 		.data		= &init_net.ipv4.sysctl_udp_wmem_min,
1352 		.maxlen		= sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1353 		.mode		= 0644,
1354 		.proc_handler	= proc_dointvec_minmax,
1355 		.extra1		= SYSCTL_ONE
1356 	},
1357 	{ }
1358 };
1359 
1360 static __net_init int ipv4_sysctl_init_net(struct net *net)
1361 {
1362 	struct ctl_table *table;
1363 
1364 	table = ipv4_net_table;
1365 	if (!net_eq(net, &init_net)) {
1366 		int i;
1367 
1368 		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1369 		if (!table)
1370 			goto err_alloc;
1371 
1372 		/* Update the variables to point into the current struct net */
1373 		for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1374 			table[i].data += (void *)net - (void *)&init_net;
1375 	}
1376 
1377 	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1378 	if (!net->ipv4.ipv4_hdr)
1379 		goto err_reg;
1380 
1381 	net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1382 	if (!net->ipv4.sysctl_local_reserved_ports)
1383 		goto err_ports;
1384 
1385 	return 0;
1386 
1387 err_ports:
1388 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1389 err_reg:
1390 	if (!net_eq(net, &init_net))
1391 		kfree(table);
1392 err_alloc:
1393 	return -ENOMEM;
1394 }
1395 
1396 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1397 {
1398 	struct ctl_table *table;
1399 
1400 	kfree(net->ipv4.sysctl_local_reserved_ports);
1401 	table = net->ipv4.ipv4_hdr->ctl_table_arg;
1402 	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1403 	kfree(table);
1404 }
1405 
1406 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1407 	.init = ipv4_sysctl_init_net,
1408 	.exit = ipv4_sysctl_exit_net,
1409 };
1410 
1411 static __init int sysctl_ipv4_init(void)
1412 {
1413 	struct ctl_table_header *hdr;
1414 
1415 	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1416 	if (!hdr)
1417 		return -ENOMEM;
1418 
1419 	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1420 		unregister_net_sysctl_table(hdr);
1421 		return -ENOMEM;
1422 	}
1423 
1424 	return 0;
1425 }
1426 
1427 __initcall(sysctl_ipv4_init);
1428