sysctl_net_ipv4.c revision 0e24c4fc52b16f0a1102a933f636d2f350c41098 
1/*
2 * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
3 *
4 * Begun April 1, 1996, Mike Shaver.
5 * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
6 */
7
8#include <linux/mm.h>
9#include <linux/module.h>
10#include <linux/sysctl.h>
11#include <linux/igmp.h>
12#include <linux/inetdevice.h>
13#include <linux/seqlock.h>
14#include <linux/init.h>
15#include <linux/slab.h>
16#include <linux/nsproxy.h>
17#include <linux/swap.h>
18#include <net/snmp.h>
19#include <net/icmp.h>
20#include <net/ip.h>
21#include <net/route.h>
22#include <net/tcp.h>
23#include <net/udp.h>
24#include <net/cipso_ipv4.h>
25#include <net/inet_frag.h>
26#include <net/ping.h>
27#include <net/tcp_memcontrol.h>
28
29static int zero;
30static int two = 2;
31static int tcp_retr1_max = 255;
32static int ip_local_port_range_min[] = { 1, 1 };
33static int ip_local_port_range_max[] = { 65535, 65535 };
34static int tcp_adv_win_scale_min = -31;
35static int tcp_adv_win_scale_max = 31;
36static int ip_ttl_min = 1;
37static int ip_ttl_max = 255;
38static int ip_ping_group_range_min[] = { 0, 0 };
39static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
40
41/* Update system visible IP port range */
42static void set_local_port_range(int range[2])
43{
44	write_seqlock(&sysctl_local_ports.lock);
45	sysctl_local_ports.range[0] = range[0];
46	sysctl_local_ports.range[1] = range[1];
47	write_sequnlock(&sysctl_local_ports.lock);
48}
49
50/* Validate changes from /proc interface. */
51static int ipv4_local_port_range(ctl_table *table, int write,
52				 void __user *buffer,
53				 size_t *lenp, loff_t *ppos)
54{
55	int ret;
56	int range[2];
57	ctl_table tmp = {
58		.data = &range,
59		.maxlen = sizeof(range),
60		.mode = table->mode,
61		.extra1 = &ip_local_port_range_min,
62		.extra2 = &ip_local_port_range_max,
63	};
64
65	inet_get_local_port_range(range, range + 1);
66	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
67
68	if (write && ret == 0) {
69		if (range[1] < range[0])
70			ret = -EINVAL;
71		else
72			set_local_port_range(range);
73	}
74
75	return ret;
76}
77
78
79static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
80{
81	kgid_t *data = table->data;
82	unsigned int seq;
83	do {
84		seq = read_seqbegin(&sysctl_local_ports.lock);
85
86		*low = data[0];
87		*high = data[1];
88	} while (read_seqretry(&sysctl_local_ports.lock, seq));
89}
90
91/* Update system visible IP port range */
92static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
93{
94	kgid_t *data = table->data;
95	write_seqlock(&sysctl_local_ports.lock);
96	data[0] = low;
97	data[1] = high;
98	write_sequnlock(&sysctl_local_ports.lock);
99}
100
101/* Validate changes from /proc interface. */
102static int ipv4_ping_group_range(ctl_table *table, int write,
103				 void __user *buffer,
104				 size_t *lenp, loff_t *ppos)
105{
106	struct user_namespace *user_ns = current_user_ns();
107	int ret;
108	gid_t urange[2];
109	kgid_t low, high;
110	ctl_table tmp = {
111		.data = &urange,
112		.maxlen = sizeof(urange),
113		.mode = table->mode,
114		.extra1 = &ip_ping_group_range_min,
115		.extra2 = &ip_ping_group_range_max,
116	};
117
118	inet_get_ping_group_range_table(table, &low, &high);
119	urange[0] = from_kgid_munged(user_ns, low);
120	urange[1] = from_kgid_munged(user_ns, high);
121	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
122
123	if (write && ret == 0) {
124		low = make_kgid(user_ns, urange[0]);
125		high = make_kgid(user_ns, urange[1]);
126		if (!gid_valid(low) || !gid_valid(high) ||
127		    (urange[1] < urange[0]) || gid_lt(high, low)) {
128			low = make_kgid(&init_user_ns, 1);
129			high = make_kgid(&init_user_ns, 0);
130		}
131		set_ping_group_range(table, low, high);
132	}
133
134	return ret;
135}
136
137static int proc_tcp_congestion_control(ctl_table *ctl, int write,
138				       void __user *buffer, size_t *lenp, loff_t *ppos)
139{
140	char val[TCP_CA_NAME_MAX];
141	ctl_table tbl = {
142		.data = val,
143		.maxlen = TCP_CA_NAME_MAX,
144	};
145	int ret;
146
147	tcp_get_default_congestion_control(val);
148
149	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
150	if (write && ret == 0)
151		ret = tcp_set_default_congestion_control(val);
152	return ret;
153}
154
155static int proc_tcp_available_congestion_control(ctl_table *ctl,
156						 int write,
157						 void __user *buffer, size_t *lenp,
158						 loff_t *ppos)
159{
160	ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
161	int ret;
162
163	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
164	if (!tbl.data)
165		return -ENOMEM;
166	tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
167	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
168	kfree(tbl.data);
169	return ret;
170}
171
172static int proc_allowed_congestion_control(ctl_table *ctl,
173					   int write,
174					   void __user *buffer, size_t *lenp,
175					   loff_t *ppos)
176{
177	ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
178	int ret;
179
180	tbl.data = kmalloc(tbl.maxlen, GFP_USER);
181	if (!tbl.data)
182		return -ENOMEM;
183
184	tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
185	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
186	if (write && ret == 0)
187		ret = tcp_set_allowed_congestion_control(tbl.data);
188	kfree(tbl.data);
189	return ret;
190}
191
192static int ipv4_tcp_mem(ctl_table *ctl, int write,
193			   void __user *buffer, size_t *lenp,
194			   loff_t *ppos)
195{
196	int ret;
197	unsigned long vec[3];
198	struct net *net = current->nsproxy->net_ns;
199#ifdef CONFIG_MEMCG_KMEM
200	struct mem_cgroup *memcg;
201#endif
202
203	ctl_table tmp = {
204		.data = &vec,
205		.maxlen = sizeof(vec),
206		.mode = ctl->mode,
207	};
208
209	if (!write) {
210		ctl->data = &net->ipv4.sysctl_tcp_mem;
211		return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
212	}
213
214	ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
215	if (ret)
216		return ret;
217
218#ifdef CONFIG_MEMCG_KMEM
219	rcu_read_lock();
220	memcg = mem_cgroup_from_task(current);
221
222	tcp_prot_mem(memcg, vec[0], 0);
223	tcp_prot_mem(memcg, vec[1], 1);
224	tcp_prot_mem(memcg, vec[2], 2);
225	rcu_read_unlock();
226#endif
227
228	net->ipv4.sysctl_tcp_mem[0] = vec[0];
229	net->ipv4.sysctl_tcp_mem[1] = vec[1];
230	net->ipv4.sysctl_tcp_mem[2] = vec[2];
231
232	return 0;
233}
234
235int proc_tcp_fastopen_key(ctl_table *ctl, int write, void __user *buffer,
236			  size_t *lenp, loff_t *ppos)
237{
238	ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
239	struct tcp_fastopen_context *ctxt;
240	int ret;
241	u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
242
243	tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
244	if (!tbl.data)
245		return -ENOMEM;
246
247	rcu_read_lock();
248	ctxt = rcu_dereference(tcp_fastopen_ctx);
249	if (ctxt)
250		memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
251	else
252		memset(user_key, 0, sizeof(user_key));
253	rcu_read_unlock();
254
255	snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
256		user_key[0], user_key[1], user_key[2], user_key[3]);
257	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
258
259	if (write && ret == 0) {
260		if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
261			   user_key + 2, user_key + 3) != 4) {
262			ret = -EINVAL;
263			goto bad_key;
264		}
265		tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
266	}
267
268bad_key:
269	pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
270	       user_key[0], user_key[1], user_key[2], user_key[3],
271	       (char *)tbl.data, ret);
272	kfree(tbl.data);
273	return ret;
274}
275
276static struct ctl_table ipv4_table[] = {
277	{
278		.procname	= "tcp_timestamps",
279		.data		= &sysctl_tcp_timestamps,
280		.maxlen		= sizeof(int),
281		.mode		= 0644,
282		.proc_handler	= proc_dointvec
283	},
284	{
285		.procname	= "tcp_window_scaling",
286		.data		= &sysctl_tcp_window_scaling,
287		.maxlen		= sizeof(int),
288		.mode		= 0644,
289		.proc_handler	= proc_dointvec
290	},
291	{
292		.procname	= "tcp_sack",
293		.data		= &sysctl_tcp_sack,
294		.maxlen		= sizeof(int),
295		.mode		= 0644,
296		.proc_handler	= proc_dointvec
297	},
298	{
299		.procname	= "tcp_retrans_collapse",
300		.data		= &sysctl_tcp_retrans_collapse,
301		.maxlen		= sizeof(int),
302		.mode		= 0644,
303		.proc_handler	= proc_dointvec
304	},
305	{
306		.procname	= "ip_default_ttl",
307		.data		= &sysctl_ip_default_ttl,
308		.maxlen		= sizeof(int),
309		.mode		= 0644,
310		.proc_handler	= proc_dointvec_minmax,
311		.extra1		= &ip_ttl_min,
312		.extra2		= &ip_ttl_max,
313	},
314	{
315		.procname	= "ip_no_pmtu_disc",
316		.data		= &ipv4_config.no_pmtu_disc,
317		.maxlen		= sizeof(int),
318		.mode		= 0644,
319		.proc_handler	= proc_dointvec
320	},
321	{
322		.procname	= "ip_nonlocal_bind",
323		.data		= &sysctl_ip_nonlocal_bind,
324		.maxlen		= sizeof(int),
325		.mode		= 0644,
326		.proc_handler	= proc_dointvec
327	},
328	{
329		.procname	= "tcp_syn_retries",
330		.data		= &sysctl_tcp_syn_retries,
331		.maxlen		= sizeof(int),
332		.mode		= 0644,
333		.proc_handler	= proc_dointvec
334	},
335	{
336		.procname	= "tcp_synack_retries",
337		.data		= &sysctl_tcp_synack_retries,
338		.maxlen		= sizeof(int),
339		.mode		= 0644,
340		.proc_handler	= proc_dointvec
341	},
342	{
343		.procname	= "tcp_max_orphans",
344		.data		= &sysctl_tcp_max_orphans,
345		.maxlen		= sizeof(int),
346		.mode		= 0644,
347		.proc_handler	= proc_dointvec
348	},
349	{
350		.procname	= "tcp_max_tw_buckets",
351		.data		= &tcp_death_row.sysctl_max_tw_buckets,
352		.maxlen		= sizeof(int),
353		.mode		= 0644,
354		.proc_handler	= proc_dointvec
355	},
356	{
357		.procname	= "ip_early_demux",
358		.data		= &sysctl_ip_early_demux,
359		.maxlen		= sizeof(int),
360		.mode		= 0644,
361		.proc_handler	= proc_dointvec
362	},
363	{
364		.procname	= "ip_dynaddr",
365		.data		= &sysctl_ip_dynaddr,
366		.maxlen		= sizeof(int),
367		.mode		= 0644,
368		.proc_handler	= proc_dointvec
369	},
370	{
371		.procname	= "tcp_keepalive_time",
372		.data		= &sysctl_tcp_keepalive_time,
373		.maxlen		= sizeof(int),
374		.mode		= 0644,
375		.proc_handler	= proc_dointvec_jiffies,
376	},
377	{
378		.procname	= "tcp_keepalive_probes",
379		.data		= &sysctl_tcp_keepalive_probes,
380		.maxlen		= sizeof(int),
381		.mode		= 0644,
382		.proc_handler	= proc_dointvec
383	},
384	{
385		.procname	= "tcp_keepalive_intvl",
386		.data		= &sysctl_tcp_keepalive_intvl,
387		.maxlen		= sizeof(int),
388		.mode		= 0644,
389		.proc_handler	= proc_dointvec_jiffies,
390	},
391	{
392		.procname	= "tcp_retries1",
393		.data		= &sysctl_tcp_retries1,
394		.maxlen		= sizeof(int),
395		.mode		= 0644,
396		.proc_handler	= proc_dointvec_minmax,
397		.extra2		= &tcp_retr1_max
398	},
399	{
400		.procname	= "tcp_retries2",
401		.data		= &sysctl_tcp_retries2,
402		.maxlen		= sizeof(int),
403		.mode		= 0644,
404		.proc_handler	= proc_dointvec
405	},
406	{
407		.procname	= "tcp_fin_timeout",
408		.data		= &sysctl_tcp_fin_timeout,
409		.maxlen		= sizeof(int),
410		.mode		= 0644,
411		.proc_handler	= proc_dointvec_jiffies,
412	},
413#ifdef CONFIG_SYN_COOKIES
414	{
415		.procname	= "tcp_syncookies",
416		.data		= &sysctl_tcp_syncookies,
417		.maxlen		= sizeof(int),
418		.mode		= 0644,
419		.proc_handler	= proc_dointvec
420	},
421#endif
422	{
423		.procname	= "tcp_fastopen",
424		.data		= &sysctl_tcp_fastopen,
425		.maxlen		= sizeof(int),
426		.mode		= 0644,
427		.proc_handler	= proc_dointvec,
428	},
429	{
430		.procname	= "tcp_fastopen_key",
431		.mode		= 0600,
432		.maxlen		= ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
433		.proc_handler	= proc_tcp_fastopen_key,
434	},
435	{
436		.procname	= "tcp_tw_recycle",
437		.data		= &tcp_death_row.sysctl_tw_recycle,
438		.maxlen		= sizeof(int),
439		.mode		= 0644,
440		.proc_handler	= proc_dointvec
441	},
442	{
443		.procname	= "tcp_abort_on_overflow",
444		.data		= &sysctl_tcp_abort_on_overflow,
445		.maxlen		= sizeof(int),
446		.mode		= 0644,
447		.proc_handler	= proc_dointvec
448	},
449	{
450		.procname	= "tcp_stdurg",
451		.data		= &sysctl_tcp_stdurg,
452		.maxlen		= sizeof(int),
453		.mode		= 0644,
454		.proc_handler	= proc_dointvec
455	},
456	{
457		.procname	= "tcp_rfc1337",
458		.data		= &sysctl_tcp_rfc1337,
459		.maxlen		= sizeof(int),
460		.mode		= 0644,
461		.proc_handler	= proc_dointvec
462	},
463	{
464		.procname	= "tcp_max_syn_backlog",
465		.data		= &sysctl_max_syn_backlog,
466		.maxlen		= sizeof(int),
467		.mode		= 0644,
468		.proc_handler	= proc_dointvec
469	},
470	{
471		.procname	= "ip_local_port_range",
472		.data		= &sysctl_local_ports.range,
473		.maxlen		= sizeof(sysctl_local_ports.range),
474		.mode		= 0644,
475		.proc_handler	= ipv4_local_port_range,
476	},
477	{
478		.procname	= "ip_local_reserved_ports",
479		.data		= NULL, /* initialized in sysctl_ipv4_init */
480		.maxlen		= 65536,
481		.mode		= 0644,
482		.proc_handler	= proc_do_large_bitmap,
483	},
484	{
485		.procname	= "igmp_max_memberships",
486		.data		= &sysctl_igmp_max_memberships,
487		.maxlen		= sizeof(int),
488		.mode		= 0644,
489		.proc_handler	= proc_dointvec
490	},
491	{
492		.procname	= "igmp_max_msf",
493		.data		= &sysctl_igmp_max_msf,
494		.maxlen		= sizeof(int),
495		.mode		= 0644,
496		.proc_handler	= proc_dointvec
497	},
498	{
499		.procname	= "inet_peer_threshold",
500		.data		= &inet_peer_threshold,
501		.maxlen		= sizeof(int),
502		.mode		= 0644,
503		.proc_handler	= proc_dointvec
504	},
505	{
506		.procname	= "inet_peer_minttl",
507		.data		= &inet_peer_minttl,
508		.maxlen		= sizeof(int),
509		.mode		= 0644,
510		.proc_handler	= proc_dointvec_jiffies,
511	},
512	{
513		.procname	= "inet_peer_maxttl",
514		.data		= &inet_peer_maxttl,
515		.maxlen		= sizeof(int),
516		.mode		= 0644,
517		.proc_handler	= proc_dointvec_jiffies,
518	},
519	{
520		.procname	= "tcp_orphan_retries",
521		.data		= &sysctl_tcp_orphan_retries,
522		.maxlen		= sizeof(int),
523		.mode		= 0644,
524		.proc_handler	= proc_dointvec
525	},
526	{
527		.procname	= "tcp_fack",
528		.data		= &sysctl_tcp_fack,
529		.maxlen		= sizeof(int),
530		.mode		= 0644,
531		.proc_handler	= proc_dointvec
532	},
533	{
534		.procname	= "tcp_reordering",
535		.data		= &sysctl_tcp_reordering,
536		.maxlen		= sizeof(int),
537		.mode		= 0644,
538		.proc_handler	= proc_dointvec
539	},
540	{
541		.procname	= "tcp_ecn",
542		.data		= &sysctl_tcp_ecn,
543		.maxlen		= sizeof(int),
544		.mode		= 0644,
545		.proc_handler	= proc_dointvec
546	},
547	{
548		.procname	= "tcp_dsack",
549		.data		= &sysctl_tcp_dsack,
550		.maxlen		= sizeof(int),
551		.mode		= 0644,
552		.proc_handler	= proc_dointvec
553	},
554	{
555		.procname	= "tcp_wmem",
556		.data		= &sysctl_tcp_wmem,
557		.maxlen		= sizeof(sysctl_tcp_wmem),
558		.mode		= 0644,
559		.proc_handler	= proc_dointvec
560	},
561	{
562		.procname	= "tcp_rmem",
563		.data		= &sysctl_tcp_rmem,
564		.maxlen		= sizeof(sysctl_tcp_rmem),
565		.mode		= 0644,
566		.proc_handler	= proc_dointvec
567	},
568	{
569		.procname	= "tcp_app_win",
570		.data		= &sysctl_tcp_app_win,
571		.maxlen		= sizeof(int),
572		.mode		= 0644,
573		.proc_handler	= proc_dointvec
574	},
575	{
576		.procname	= "tcp_adv_win_scale",
577		.data		= &sysctl_tcp_adv_win_scale,
578		.maxlen		= sizeof(int),
579		.mode		= 0644,
580		.proc_handler	= proc_dointvec_minmax,
581		.extra1		= &tcp_adv_win_scale_min,
582		.extra2		= &tcp_adv_win_scale_max,
583	},
584	{
585		.procname	= "tcp_tw_reuse",
586		.data		= &sysctl_tcp_tw_reuse,
587		.maxlen		= sizeof(int),
588		.mode		= 0644,
589		.proc_handler	= proc_dointvec
590	},
591	{
592		.procname	= "tcp_frto",
593		.data		= &sysctl_tcp_frto,
594		.maxlen		= sizeof(int),
595		.mode		= 0644,
596		.proc_handler	= proc_dointvec
597	},
598	{
599		.procname	= "tcp_frto_response",
600		.data		= &sysctl_tcp_frto_response,
601		.maxlen		= sizeof(int),
602		.mode		= 0644,
603		.proc_handler	= proc_dointvec
604	},
605	{
606		.procname	= "tcp_low_latency",
607		.data		= &sysctl_tcp_low_latency,
608		.maxlen		= sizeof(int),
609		.mode		= 0644,
610		.proc_handler	= proc_dointvec
611	},
612	{
613		.procname	= "tcp_no_metrics_save",
614		.data		= &sysctl_tcp_nometrics_save,
615		.maxlen		= sizeof(int),
616		.mode		= 0644,
617		.proc_handler	= proc_dointvec,
618	},
619	{
620		.procname	= "tcp_moderate_rcvbuf",
621		.data		= &sysctl_tcp_moderate_rcvbuf,
622		.maxlen		= sizeof(int),
623		.mode		= 0644,
624		.proc_handler	= proc_dointvec,
625	},
626	{
627		.procname	= "tcp_tso_win_divisor",
628		.data		= &sysctl_tcp_tso_win_divisor,
629		.maxlen		= sizeof(int),
630		.mode		= 0644,
631		.proc_handler	= proc_dointvec,
632	},
633	{
634		.procname	= "tcp_congestion_control",
635		.mode		= 0644,
636		.maxlen		= TCP_CA_NAME_MAX,
637		.proc_handler	= proc_tcp_congestion_control,
638	},
639	{
640		.procname	= "tcp_abc",
641		.data		= &sysctl_tcp_abc,
642		.maxlen		= sizeof(int),
643		.mode		= 0644,
644		.proc_handler	= proc_dointvec,
645	},
646	{
647		.procname	= "tcp_mtu_probing",
648		.data		= &sysctl_tcp_mtu_probing,
649		.maxlen		= sizeof(int),
650		.mode		= 0644,
651		.proc_handler	= proc_dointvec,
652	},
653	{
654		.procname	= "tcp_base_mss",
655		.data		= &sysctl_tcp_base_mss,
656		.maxlen		= sizeof(int),
657		.mode		= 0644,
658		.proc_handler	= proc_dointvec,
659	},
660	{
661		.procname	= "tcp_workaround_signed_windows",
662		.data		= &sysctl_tcp_workaround_signed_windows,
663		.maxlen		= sizeof(int),
664		.mode		= 0644,
665		.proc_handler	= proc_dointvec
666	},
667	{
668		.procname	= "tcp_limit_output_bytes",
669		.data		= &sysctl_tcp_limit_output_bytes,
670		.maxlen		= sizeof(int),
671		.mode		= 0644,
672		.proc_handler	= proc_dointvec
673	},
674	{
675		.procname	= "tcp_challenge_ack_limit",
676		.data		= &sysctl_tcp_challenge_ack_limit,
677		.maxlen		= sizeof(int),
678		.mode		= 0644,
679		.proc_handler	= proc_dointvec
680	},
681#ifdef CONFIG_NET_DMA
682	{
683		.procname	= "tcp_dma_copybreak",
684		.data		= &sysctl_tcp_dma_copybreak,
685		.maxlen		= sizeof(int),
686		.mode		= 0644,
687		.proc_handler	= proc_dointvec
688	},
689#endif
690	{
691		.procname	= "tcp_slow_start_after_idle",
692		.data		= &sysctl_tcp_slow_start_after_idle,
693		.maxlen		= sizeof(int),
694		.mode		= 0644,
695		.proc_handler	= proc_dointvec
696	},
697#ifdef CONFIG_NETLABEL
698	{
699		.procname	= "cipso_cache_enable",
700		.data		= &cipso_v4_cache_enabled,
701		.maxlen		= sizeof(int),
702		.mode		= 0644,
703		.proc_handler	= proc_dointvec,
704	},
705	{
706		.procname	= "cipso_cache_bucket_size",
707		.data		= &cipso_v4_cache_bucketsize,
708		.maxlen		= sizeof(int),
709		.mode		= 0644,
710		.proc_handler	= proc_dointvec,
711	},
712	{
713		.procname	= "cipso_rbm_optfmt",
714		.data		= &cipso_v4_rbm_optfmt,
715		.maxlen		= sizeof(int),
716		.mode		= 0644,
717		.proc_handler	= proc_dointvec,
718	},
719	{
720		.procname	= "cipso_rbm_strictvalid",
721		.data		= &cipso_v4_rbm_strictvalid,
722		.maxlen		= sizeof(int),
723		.mode		= 0644,
724		.proc_handler	= proc_dointvec,
725	},
726#endif /* CONFIG_NETLABEL */
727	{
728		.procname	= "tcp_available_congestion_control",
729		.maxlen		= TCP_CA_BUF_MAX,
730		.mode		= 0444,
731		.proc_handler   = proc_tcp_available_congestion_control,
732	},
733	{
734		.procname	= "tcp_allowed_congestion_control",
735		.maxlen		= TCP_CA_BUF_MAX,
736		.mode		= 0644,
737		.proc_handler   = proc_allowed_congestion_control,
738	},
739	{
740		.procname	= "tcp_max_ssthresh",
741		.data		= &sysctl_tcp_max_ssthresh,
742		.maxlen		= sizeof(int),
743		.mode		= 0644,
744		.proc_handler	= proc_dointvec,
745	},
746	{
747		.procname	= "tcp_cookie_size",
748		.data		= &sysctl_tcp_cookie_size,
749		.maxlen		= sizeof(int),
750		.mode		= 0644,
751		.proc_handler	= proc_dointvec
752	},
753	{
754		.procname       = "tcp_thin_linear_timeouts",
755		.data           = &sysctl_tcp_thin_linear_timeouts,
756		.maxlen         = sizeof(int),
757		.mode           = 0644,
758		.proc_handler   = proc_dointvec
759	},
760        {
761		.procname       = "tcp_thin_dupack",
762		.data           = &sysctl_tcp_thin_dupack,
763		.maxlen         = sizeof(int),
764		.mode           = 0644,
765		.proc_handler   = proc_dointvec
766	},
767	{
768		.procname	= "tcp_early_retrans",
769		.data		= &sysctl_tcp_early_retrans,
770		.maxlen		= sizeof(int),
771		.mode		= 0644,
772		.proc_handler	= proc_dointvec_minmax,
773		.extra1		= &zero,
774		.extra2		= &two,
775	},
776	{
777		.procname	= "udp_mem",
778		.data		= &sysctl_udp_mem,
779		.maxlen		= sizeof(sysctl_udp_mem),
780		.mode		= 0644,
781		.proc_handler	= proc_doulongvec_minmax,
782	},
783	{
784		.procname	= "udp_rmem_min",
785		.data		= &sysctl_udp_rmem_min,
786		.maxlen		= sizeof(sysctl_udp_rmem_min),
787		.mode		= 0644,
788		.proc_handler	= proc_dointvec_minmax,
789		.extra1		= &zero
790	},
791	{
792		.procname	= "udp_wmem_min",
793		.data		= &sysctl_udp_wmem_min,
794		.maxlen		= sizeof(sysctl_udp_wmem_min),
795		.mode		= 0644,
796		.proc_handler	= proc_dointvec_minmax,
797		.extra1		= &zero
798	},
799	{ }
800};
801
802static struct ctl_table ipv4_net_table[] = {
803	{
804		.procname	= "icmp_echo_ignore_all",
805		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_all,
806		.maxlen		= sizeof(int),
807		.mode		= 0644,
808		.proc_handler	= proc_dointvec
809	},
810	{
811		.procname	= "icmp_echo_ignore_broadcasts",
812		.data		= &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
813		.maxlen		= sizeof(int),
814		.mode		= 0644,
815		.proc_handler	= proc_dointvec
816	},
817	{
818		.procname	= "icmp_ignore_bogus_error_responses",
819		.data		= &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
820		.maxlen		= sizeof(int),
821		.mode		= 0644,
822		.proc_handler	= proc_dointvec
823	},
824	{
825		.procname	= "icmp_errors_use_inbound_ifaddr",
826		.data		= &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
827		.maxlen		= sizeof(int),
828		.mode		= 0644,
829		.proc_handler	= proc_dointvec
830	},
831	{
832		.procname	= "icmp_ratelimit",
833		.data		= &init_net.ipv4.sysctl_icmp_ratelimit,
834		.maxlen		= sizeof(int),
835		.mode		= 0644,
836		.proc_handler	= proc_dointvec_ms_jiffies,
837	},
838	{
839		.procname	= "icmp_ratemask",
840		.data		= &init_net.ipv4.sysctl_icmp_ratemask,
841		.maxlen		= sizeof(int),
842		.mode		= 0644,
843		.proc_handler	= proc_dointvec
844	},
845	{
846		.procname	= "ping_group_range",
847		.data		= &init_net.ipv4.sysctl_ping_group_range,
848		.maxlen		= sizeof(gid_t)*2,
849		.mode		= 0644,
850		.proc_handler	= ipv4_ping_group_range,
851	},
852	{
853		.procname	= "tcp_mem",
854		.maxlen		= sizeof(init_net.ipv4.sysctl_tcp_mem),
855		.mode		= 0644,
856		.proc_handler	= ipv4_tcp_mem,
857	},
858	{ }
859};
860
861static __net_init int ipv4_sysctl_init_net(struct net *net)
862{
863	struct ctl_table *table;
864
865	table = ipv4_net_table;
866	if (!net_eq(net, &init_net)) {
867		table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
868		if (table == NULL)
869			goto err_alloc;
870
871		table[0].data =
872			&net->ipv4.sysctl_icmp_echo_ignore_all;
873		table[1].data =
874			&net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
875		table[2].data =
876			&net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
877		table[3].data =
878			&net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
879		table[4].data =
880			&net->ipv4.sysctl_icmp_ratelimit;
881		table[5].data =
882			&net->ipv4.sysctl_icmp_ratemask;
883		table[6].data =
884			&net->ipv4.sysctl_ping_group_range;
885
886	}
887
888	/*
889	 * Sane defaults - nobody may create ping sockets.
890	 * Boot scripts should set this to distro-specific group.
891	 */
892	net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
893	net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
894
895	tcp_init_mem(net);
896
897	net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
898	if (net->ipv4.ipv4_hdr == NULL)
899		goto err_reg;
900
901	return 0;
902
903err_reg:
904	if (!net_eq(net, &init_net))
905		kfree(table);
906err_alloc:
907	return -ENOMEM;
908}
909
910static __net_exit void ipv4_sysctl_exit_net(struct net *net)
911{
912	struct ctl_table *table;
913
914	table = net->ipv4.ipv4_hdr->ctl_table_arg;
915	unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
916	kfree(table);
917}
918
919static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
920	.init = ipv4_sysctl_init_net,
921	.exit = ipv4_sysctl_exit_net,
922};
923
924static __init int sysctl_ipv4_init(void)
925{
926	struct ctl_table_header *hdr;
927	struct ctl_table *i;
928
929	for (i = ipv4_table; i->procname; i++) {
930		if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
931			i->data = sysctl_local_reserved_ports;
932			break;
933		}
934	}
935	if (!i->procname)
936		return -EINVAL;
937
938	hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
939	if (hdr == NULL)
940		return -ENOMEM;
941
942	if (register_pernet_subsys(&ipv4_sysctl_ops)) {
943		unregister_net_sysctl_table(hdr);
944		return -ENOMEM;
945	}
946
947	return 0;
948}
949
950__initcall(sysctl_ipv4_init);
951